feat(host): containerd snapshot rootfs service (#22621)

2026-08-30 17:13:08 +08:00 · 2025-06-04 11:00:27 +08:00
parent 76e0770b09
commit 4731240d25
646 changed files with 61549 additions and 7408 deletions
@@ -12,6 +12,7 @@ require (
 	github.com/bitly/go-simplejson v0.5.0
 	github.com/c-bata/go-prompt v0.2.4
 	github.com/cheggaaa/pb/v3 v3.0.8
+	github.com/containerd/containerd v1.4.1
 	github.com/coredns/coredns v1.3.0
 	github.com/coreos/go-iptables v0.6.0
 	github.com/creack/pty v1.1.18
@@ -59,7 +60,7 @@ require (
 	github.com/sevlyar/go-daemon v0.1.5
 	github.com/shirou/gopsutil v3.21.11+incompatible
 	github.com/shirou/gopsutil/v3 v3.22.10
-	github.com/sirupsen/logrus v1.9.0
+	github.com/sirupsen/logrus v1.9.3
 	github.com/skip2/go-qrcode v0.0.0-20190110000554-dc11ecdae0a9
 	github.com/smartystreets/goconvey v1.7.2
 	github.com/stretchr/testify v1.9.0
@@ -75,14 +76,14 @@ require (
 	go.etcd.io/etcd/client/v3 v3.5.0
 	golang.org/x/crypto v0.19.0
 	golang.org/x/net v0.21.0
-	golang.org/x/sync v0.6.0
-	golang.org/x/sys v0.17.0
+	golang.org/x/sync v0.8.0
+	golang.org/x/sys v0.26.0
 	golang.org/x/text v0.14.0
 	golang.org/x/time v0.5.0
 	golang.org/x/xerrors v0.0.0-20231012003039-104605ab7028
 	golang.zx2c4.com/wireguard/wgctrl v0.0.0-20230215201556-9c5414ab4bde
 	google.golang.org/grpc v1.62.0
-	google.golang.org/protobuf v1.32.0
+	google.golang.org/protobuf v1.35.1
 	gopkg.in/fatih/set.v0 v0.2.1
 	gopkg.in/ldap.v3 v3.0.3
 	gopkg.in/mail.v2 v2.3.1
@@ -126,7 +127,8 @@ require (
 	github.com/DataDog/dd-trace-go v0.6.1 // indirect
 	github.com/DataDog/zstd v1.3.4 // indirect
 	github.com/Microsoft/azure-vhd-utils v0.0.0-20181115010904-44cbada2ece3 // indirect
-	github.com/Microsoft/go-winio v0.4.15 // indirect
+	github.com/Microsoft/go-winio v0.6.2 // indirect
+	github.com/Microsoft/hcsshim v0.8.14 // indirect
 	github.com/RoaringBitmap/roaring v0.4.16 // indirect
 	github.com/Shopify/sarama v1.20.0 // indirect
 	github.com/Shopify/toxiproxy v2.1.4+incompatible // indirect
@@ -156,9 +158,12 @@ require (
 	github.com/checkpoint-restore/go-criu/v4 v4.1.0 // indirect
 	github.com/cilium/ebpf v0.0.0-20200702112145-1c8d4c9ef775 // indirect
 	github.com/cloudflare/golz4 v0.0.0-20150217214814-ef862a3cdc58 // indirect
+	github.com/containerd/cgroups v0.0.0-20200531161412-0dbf7f05ba59 // indirect
 	github.com/containerd/console v1.0.0 // indirect
-	github.com/containerd/containerd v1.4.1 // indirect
+	github.com/containerd/continuity v0.4.5 // indirect
+	github.com/containerd/log v0.1.0 // indirect
 	github.com/containerd/ttrpc v1.0.2 // indirect
+	github.com/containerd/typeurl v1.0.1 // indirect
 	github.com/coreos/go-semver v0.3.0 // indirect
 	github.com/coreos/go-systemd/v22 v22.3.2 // indirect
 	github.com/cyphar/filepath-securejoin v0.2.2 // indirect
@@ -168,6 +173,7 @@ require (
 	github.com/dnstap/golang-dnstap v0.0.0-20170829151710-2cf77a2b5e11 // indirect
 	github.com/docker/distribution v2.7.1-0.20190205005809-0d3efadf0154+incompatible // indirect
 	github.com/docker/go-connections v0.4.0 // indirect
+	github.com/docker/go-events v0.0.0-20250114142523-c867878c5e32 // indirect
 	github.com/docker/go-units v0.4.0 // indirect
 	github.com/dustin/go-humanize v1.0.0 // indirect
 	github.com/eapache/go-resiliency v1.1.0 // indirect
@@ -292,6 +298,7 @@ require (
 	github.com/xuri/efp v0.0.0-20220603152613-6918739fd470 // indirect
 	github.com/xuri/nfp v0.0.0-20220409054826-5e722a1d9e22 // indirect
 	github.com/yusufpapurcu/wmi v1.2.2 // indirect
+	go.etcd.io/bbolt v1.3.7 // indirect
 	go.etcd.io/etcd/client/pkg/v3 v3.5.0 // indirect
 	go.opencensus.io v0.24.0 // indirect
 	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.48.0 // indirect
@@ -93,8 +93,12 @@ github.com/LeeEirc/terminalparser v0.0.0-20240205084113-fbf78c8480f2 h1:XGB3B065
 github.com/LeeEirc/terminalparser v0.0.0-20240205084113-fbf78c8480f2/go.mod h1:tiLv6VBLH4Z3KdBSe2qIKRwQDGCVQ9/F5fOKpQGvyoA=
 github.com/Microsoft/azure-vhd-utils v0.0.0-20181115010904-44cbada2ece3 h1:gImoAO1xAcC1oDlYmD/X7dggsodGf2DFJOVE5m0ssms=
 github.com/Microsoft/azure-vhd-utils v0.0.0-20181115010904-44cbada2ece3/go.mod h1:u0H9gMieFLxkUy8RS0X8VbFWyPs2815qQAaitRbj6x0=
-github.com/Microsoft/go-winio v0.4.15 h1:qkLXKzb1QoVatRyd/YlXZ/Kg0m5K3SPuoD82jjSOaBc=
 github.com/Microsoft/go-winio v0.4.15/go.mod h1:tTuCMEN+UleMWgg9dVx4Hu52b1bJo+59jBh3ajtinzw=
+github.com/Microsoft/go-winio v0.4.16-0.20201130162521-d1ffc52c7331/go.mod h1:XB6nPKklQyQ7GC9LdcBEcBl8PF76WugXOPRXwdLnMv0=
+github.com/Microsoft/go-winio v0.6.2 h1:F2VQgta7ecxGYO8k3ZZz3RS8fVIXVxONVUPlNERoyfY=
+github.com/Microsoft/go-winio v0.6.2/go.mod h1:yd8OoFMLzJbo9gZq8j5qaps8bJ9aShtEA8Ipt1oGCvU=
+github.com/Microsoft/hcsshim v0.8.14 h1:lbPVK25c1cu5xTLITwpUcxoA9vKrKErASPYygvouJns=
+github.com/Microsoft/hcsshim v0.8.14/go.mod h1:NtVKoYxQuTLx6gEq0L96c9Ju4JbRJ4nY2ow3VK6a9Lg=
 github.com/NYTimes/gziphandler v0.0.0-20170623195520-56545f4a5d46/go.mod h1:3wb06e3pkSAbeQ52E9H9iFoQsEEwGN64994WTCIhntQ=
 github.com/OneOfOne/xxhash v1.2.2 h1:KMrpdQIwFcEqXDklaen+P1axHaj9BSKzvpUUfnHldSE=
 github.com/OneOfOne/xxhash v1.2.2/go.mod h1:HSdplMjZKSmBqAxg5vPj2TmRDmfkzw+cTzAElWljhcU=
@@ -207,6 +211,7 @@ github.com/cheggaaa/pb/v3 v3.0.8/go.mod h1:UICbiLec/XO6Hw6k+BHEtHeQFzzBH4i2/qk/o
 github.com/chzyer/logex v1.1.10/go.mod h1:+Ywpsq7O8HXn0nuIou7OrIPyXbp3wmkHB+jjWRnGsAI=
 github.com/chzyer/readline v0.0.0-20180603132655-2972be24d48e/go.mod h1:nSuG5e5PlCu98SY8svDHJxuZscDgtXS6KTTbou5AhLI=
 github.com/chzyer/test v0.0.0-20180213035817-a1ea475d72b1/go.mod h1:Q3SI9o4m/ZMnBNeIyt5eFwwo7qiLfzFZmjNmxjkiQlU=
+github.com/cilium/ebpf v0.0.0-20200110133405-4032b1d8aae3/go.mod h1:MA5e5Lr8slmEg9bt0VpxxWqJlO4iwu3FBdHUzV7wQVg=
 github.com/cilium/ebpf v0.0.0-20200702112145-1c8d4c9ef775 h1:cHzBGGVew0ezFsq2grfy2RsB8hO/eNyBgOLHBCqfR1U=
 github.com/cilium/ebpf v0.0.0-20200702112145-1c8d4c9ef775/go.mod h1:7cR51M8ViRLIdUjrmSXlK9pkrsDlLHbO8jiB8X8JnOc=
 github.com/client9/misspell v0.3.4/go.mod h1:qj6jICC3Q7zFZvVWo7KLAzC3yx5G7kyvSDkc90ppPyw=
@@ -216,12 +221,25 @@ github.com/cncf/udpa/go v0.0.0-20191209042840-269d4d468f6f/go.mod h1:M8M6+tZqaGX
 github.com/cncf/udpa/go v0.0.0-20201120205902-5459f2c99403/go.mod h1:WmhPx2Nbnhtbo57+VJT5O0JRkEi1Wbu0z5j0R8u5Hbk=
 github.com/cncf/xds/go v0.0.0-20231128003011-0fa0005c9caa h1:jQCWAUqqlij9Pgj2i/PB79y4KOPYVyFYdROxgaCwdTQ=
 github.com/cncf/xds/go v0.0.0-20231128003011-0fa0005c9caa/go.mod h1:x/1Gn8zydmfq8dk6e9PdstVsDgu9RuyIIJqAaF//0IM=
+github.com/containerd/cgroups v0.0.0-20200531161412-0dbf7f05ba59 h1:qWj4qVYZ95vLWwqyNJCQg7rDsG5wPdze0UaPolH7DUk=
+github.com/containerd/cgroups v0.0.0-20200531161412-0dbf7f05ba59/go.mod h1:pA0z1pT8KYB3TCXK/ocprsh7MAkoW8bZVzPdih9snmM=
+github.com/containerd/console v0.0.0-20180822173158-c12b1e7919c1/go.mod h1:Tj/on1eG8kiEhd0+fhSDzsPAFESxzBBvdyEgyryXffw=
 github.com/containerd/console v1.0.0 h1:fU3UuQapBs+zLJu82NhR11Rif1ny2zfMMAyPJzSN5tQ=
 github.com/containerd/console v1.0.0/go.mod h1:8Pf4gM6VEbTNRIT26AyyU7hxdQU3MvAvxVI0sc00XBE=
+github.com/containerd/containerd v1.3.2/go.mod h1:bC6axHOhabU15QhwfG7w5PipXdVtMXFTttgp+kVtyUA=
 github.com/containerd/containerd v1.4.1 h1:pASeJT3R3YyVn+94qEPk0SnU1OQ20Jd/T+SPKy9xehY=
 github.com/containerd/containerd v1.4.1/go.mod h1:bC6axHOhabU15QhwfG7w5PipXdVtMXFTttgp+kVtyUA=
+github.com/containerd/continuity v0.0.0-20190426062206-aaeac12a7ffc/go.mod h1:GL3xCUCBDV3CZiTSEKksMWbLE66hEyuu9qyDOOqM47Y=
+github.com/containerd/continuity v0.4.5 h1:ZRoN1sXq9u7V6QoHMcVWGhOwDFqZ4B9i5H6un1Wh0x4=
+github.com/containerd/continuity v0.4.5/go.mod h1:/lNJvtJKUQStBzpVQ1+rasXO1LAWtUQssk28EZvJ3nE=
+github.com/containerd/fifo v0.0.0-20190226154929-a9fb20d87448/go.mod h1:ODA38xgv3Kuk8dQz2ZQXpnv/UZZUHUCL7pnLehbXgQI=
+github.com/containerd/go-runc v0.0.0-20180907222934-5a6d9f37cfa3/go.mod h1:IV7qH3hrUgRmyYrtgEeGWJfWbgcHL9CSRruz2Vqcph0=
+github.com/containerd/log v0.1.0 h1:TCJt7ioM2cr/tfR8GPbGf9/VRAX8D2B4PjzCpfX540I=
+github.com/containerd/log v0.1.0/go.mod h1:VRRf09a7mHDIRezVKTRCrOq78v577GXq3bSa3EhrzVo=
+github.com/containerd/ttrpc v0.0.0-20190828154514-0e0f228740de/go.mod h1:PvCDdDGpgqzQIzDW1TphrGLssLDZp2GuS+X5DkEJB8o=
 github.com/containerd/ttrpc v1.0.2 h1:2/O3oTZN36q2xRolk0a2WWGgh7/Vf/liElg5hFYLX9U=
 github.com/containerd/ttrpc v1.0.2/go.mod h1:UAxOpgT9ziI0gJrmKvgcZivgxOp8iFPSk8httJEt98Y=
+github.com/containerd/typeurl v0.0.0-20180627222232-a93fcdb778cd/go.mod h1:Cm3kwCdlkCfMSHURc+r6fwoGH6/F1hH3S4sg0rLFWPc=
 github.com/containerd/typeurl v1.0.1 h1:PvuK4E3D5S5q6IqsPDCy928FhP0LUIGcmZ/Yhgp5Djw=
 github.com/containerd/typeurl v1.0.1/go.mod h1:TB1hUtrpaiO88KEK56ijojHS1+NeF0izUACaJW2mdXg=
 github.com/coredns/coredns v1.3.0 h1:Ubjr8ZrOpEj95/A9dqQqQAG2k5DWEFdg2ZB3tUXuULA=
@@ -230,10 +248,12 @@ github.com/coreos/go-iptables v0.6.0 h1:is9qnZMPYjLd8LYqmm/qlE+wwEgJIkTYdhV3rfZo
 github.com/coreos/go-iptables v0.6.0/go.mod h1:Qe8Bv2Xik5FyTXwgIbLAnv2sWSBmvWdFETJConOQ//Q=
 github.com/coreos/go-semver v0.3.0 h1:wkHLiw0WNATZnSG7epLsujiMCgPAc9xhjJ4tgnAxmfM=
 github.com/coreos/go-semver v0.3.0/go.mod h1:nnelYz7RCh+5ahJtPPxZlU+153eP4D4r3EedlOD2RNk=
+github.com/coreos/go-systemd/v22 v22.0.0/go.mod h1:xO0FLkIi5MaZafQlIrOotqXZ90ih+1atmu1JpKERPPk=
 github.com/coreos/go-systemd/v22 v22.1.0/go.mod h1:xO0FLkIi5MaZafQlIrOotqXZ90ih+1atmu1JpKERPPk=
 github.com/coreos/go-systemd/v22 v22.3.2 h1:D9/bQk5vlXQFZ6Kwuu6zaiXJ9oTPe68++AzAJc1DzSI=
 github.com/coreos/go-systemd/v22 v22.3.2/go.mod h1:Y58oyj3AT4RCenI/lSvhwexgC+NSVTIJ3seZv2GcEnc=
 github.com/cpuguy83/go-md2man/v2 v2.0.0-20190314233015-f79a8a8ca69d/go.mod h1:maD7wRr/U5Z6m/iR4s+kqSMx2CaBsrgA7czyZG/E6dU=
+github.com/cpuguy83/go-md2man/v2 v2.0.0/go.mod h1:maD7wRr/U5Z6m/iR4s+kqSMx2CaBsrgA7czyZG/E6dU=
 github.com/creack/pty v1.1.9/go.mod h1:oKZEueFk5CKHvIhNR5MUki03XCEU+Q6VDXinZuGJ33E=
 github.com/creack/pty v1.1.18 h1:n56/Zwd5o6whRC5PMGretI4IdRLlmBXYNjScPaBgsbY=
 github.com/creack/pty v1.1.18/go.mod h1:MOBLtS5ELjhRRrroQr9kyvTxUAFNvYEK993ew/Vr4O4=
@@ -256,6 +276,8 @@ github.com/docker/docker v20.10.27+incompatible h1:Id/ZooynV4ZlD6xX20RCd3SR0Ikn7
 github.com/docker/docker v20.10.27+incompatible/go.mod h1:eEKB0N0r5NX/I1kEveEz05bcu8tLC/8azJZsviup8Sk=
 github.com/docker/go-connections v0.4.0 h1:El9xVISelRB7BuFusrZozjnkIM5YnzCViNKohAFqRJQ=
 github.com/docker/go-connections v0.4.0/go.mod h1:Gbd7IOopHjR8Iph03tsViu4nIes5XhDvyHbTtUxmeec=
+github.com/docker/go-events v0.0.0-20250114142523-c867878c5e32 h1:EHZfspsnLAz8Hzccd67D5abwLiqoqym2jz/jOS39mCk=
+github.com/docker/go-events v0.0.0-20250114142523-c867878c5e32/go.mod h1:Uw6UezgYA44ePAFQYUehOuCzmy5zmg/+nl2ZfMWGkpA=
 github.com/docker/go-units v0.4.0 h1:3uh0PgVws3nIA0Q+MwDC8yjEPf9zjRfZZWXZYDct3Tw=
 github.com/docker/go-units v0.4.0/go.mod h1:fgPhTUdO+D/Jk86RDLlptpiXQzgHJF7gydDDbaIK4Dk=
 github.com/docker/spdystream v0.0.0-20160310174837-449fdfce4d96 h1:cenwrSVm+Z7QLSV/BsnenAOcDXdX4cMv4wP0B/5QbPg=
@@ -669,12 +691,15 @@ github.com/onsi/gomega v0.0.0-20170829124025-dcabb60a477c/go.mod h1:C1qb7wdrVGGV
 github.com/onsi/gomega v1.7.0/go.mod h1:ex+gbHU/CVuBBDIJjb2X0qEXbFg53c61hWP/1CpauHY=
 github.com/op/go-logging v0.0.0-20160315200505-970db520ece7 h1:lDH9UUVJtmYCjyT0CI4q8xvlXPxeZ0gYCVvWbmPlp88=
 github.com/op/go-logging v0.0.0-20160315200505-970db520ece7/go.mod h1:HzydrMdWErDVzsI23lYNej1Htcns9BCg93Dk0bBINWk=
+github.com/opencontainers/go-digest v0.0.0-20180430190053-c9281466c8b2/go.mod h1:cMLVZDEM3+U2I4VmLI6N8jQYUd2OVphdqWwCJHrFt2s=
 github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8Oi/yOhh5U=
 github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
 github.com/opencontainers/image-spec v1.0.1 h1:JMemWkRwHx4Zj+fVxWoMCFm/8sYGGrUVojFA6h/TRcI=
 github.com/opencontainers/image-spec v1.0.1/go.mod h1:BtxoFyWECRxE4U/7sNtV5W15zMzWCbyJoFRP3s7yZA0=
+github.com/opencontainers/runc v0.0.0-20190115041553-12f6a991201f/go.mod h1:qT5XzbpPznkRYVz/mWwUaVBUv2rmF59PVA73FjuZG0U=
 github.com/opencontainers/runc v1.0.0-rc92 h1:+IczUKCRzDzFDnw99O/PAqrcBBCoRp9xN3cB1SYSNS4=
 github.com/opencontainers/runc v1.0.0-rc92/go.mod h1:X1zlU4p7wOlX4+WRCz+hvlRv8phdL7UqbYD+vQwNMmE=
+github.com/opencontainers/runtime-spec v1.0.2/go.mod h1:jwyrGlmzljRJv/Fgzds9SsS/C5hL+LL3ko9hs6T5lQ0=
 github.com/opencontainers/runtime-spec v1.0.3-0.20200728170252-4d89ac9fbff6 h1:NhsM2gc769rVWDqJvapK37r+7+CBXI8xHhnfnt8uQsg=
 github.com/opencontainers/runtime-spec v1.0.3-0.20200728170252-4d89ac9fbff6/go.mod h1:jwyrGlmzljRJv/Fgzds9SsS/C5hL+LL3ko9hs6T5lQ0=
 github.com/opencontainers/selinux v1.6.0 h1:+bIAS/Za3q5FTwWym4fTB0vObnfCf3G/NC7K6Jx62mY=
@@ -727,6 +752,7 @@ github.com/prometheus/common v0.10.0/go.mod h1:Tlit/dnDKsSWFlCLTWaA1cyBgKHSMdTB8
 github.com/prometheus/common v0.26.0/go.mod h1:M7rCNAaPfAosfx8veZJCuw84e35h3Cfd9VFqTh1DIvc=
 github.com/prometheus/common v0.32.1 h1:hWIdL3N2HoUx3B8j3YN9mWor0qhY/NlEKZEaXxuIRh4=
 github.com/prometheus/common v0.32.1/go.mod h1:vu+V0TpY+O6vW9J44gczi3Ap/oXXR10b+M/gUGO4Hls=
+github.com/prometheus/procfs v0.0.0-20180125133057-cb4147076ac7/go.mod h1:c3At6R/oaqEKCNdg8wHV1ftS6bRYblBhIjjI8uT2IGk=
 github.com/prometheus/procfs v0.0.0-20181005140218-185b4288413d/go.mod h1:c3At6R/oaqEKCNdg8wHV1ftS6bRYblBhIjjI8uT2IGk=
 github.com/prometheus/procfs v0.0.0-20190522114515-bc1a522cf7b1/go.mod h1:TjEm7ze935MbeOT/UhFTIMYKhuLP4wbCsTZCD3I8kEA=
 github.com/prometheus/procfs v0.0.2/go.mod h1:TjEm7ze935MbeOT/UhFTIMYKhuLP4wbCsTZCD3I8kEA=
@@ -773,8 +799,9 @@ github.com/sirupsen/logrus v1.2.0/go.mod h1:LxeOpSwHxABJmUn/MG1IvRgCAasNZTLOkJPx
 github.com/sirupsen/logrus v1.4.1/go.mod h1:ni0Sbl8bgC9z8RoU9G6nDWqqs/fq4eDPysMBDgk/93Q=
 github.com/sirupsen/logrus v1.4.2/go.mod h1:tLMulIdttU9McNUspp0xgXVQah82FyeX6MwdIuYE2rE=
 github.com/sirupsen/logrus v1.6.0/go.mod h1:7uNnSEd1DgxDLC74fIahvMZmmYsHGZGEOFrfsX/uA88=
-github.com/sirupsen/logrus v1.9.0 h1:trlNQbNUG3OdDrDil03MCb1H2o9nJ1x4/5LYw7byDE0=
 github.com/sirupsen/logrus v1.9.0/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
+github.com/sirupsen/logrus v1.9.3 h1:dueUQJ1C2q9oE3F7wvmSGAaVtTmUizReu6fjN8uqzbQ=
+github.com/sirupsen/logrus v1.9.3/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
 github.com/skip2/go-qrcode v0.0.0-20190110000554-dc11ecdae0a9 h1:lpEzuenPuO1XNTeikEmvqYFcU37GVLl8SRNblzyvGBE=
 github.com/skip2/go-qrcode v0.0.0-20190110000554-dc11ecdae0a9/go.mod h1:PLPIyL7ikehBD1OAjmKKiOEhbvWyHGaNDjquXMcYABo=
 github.com/smartystreets/assertions v0.0.0-20180607162144-eb5b59917fa2/go.mod h1:OnSkiWE9lh6wB0YB77sQom3nweQdgAjqCqsofrRNTgc=
@@ -840,6 +867,7 @@ github.com/ugorji/go/codec v1.1.7/go.mod h1:Ax+UKWsSmolVDwsd+7N3ZtXu+yMGCf907BLY
 github.com/ulikunitz/xz v0.5.12 h1:37Nm15o69RwBkXM0J6A5OlE67RZTfzUxTj8fB3dfcsc=
 github.com/ulikunitz/xz v0.5.12/go.mod h1:nbz6k7qbPmH4IRqmfOplQw/tblSgqTqBwxkY0oWt/14=
 github.com/urfave/cli v1.22.1/go.mod h1:Gos4lmkARVdJ6EkW0WaNv/tZAAMe9V7XWyB60NtXRu0=
+github.com/urfave/cli v1.22.2/go.mod h1:Gos4lmkARVdJ6EkW0WaNv/tZAAMe9V7XWyB60NtXRu0=
 github.com/vishvananda/netlink v1.1.0 h1:1iyaYNBLmP6L0220aDnYQpo1QEV4t4hJ+xEEhhJH8j0=
 github.com/vishvananda/netlink v1.1.0/go.mod h1:cTgwzPIzzgDAYoQrMm0EdrjRUBkTqKYppBueQtXaqoE=
 github.com/vishvananda/netns v0.0.0-20191106174202-0a2b9b5464df/go.mod h1:JP3t17pCcGlemwknint6hfoeCVQrEMVwxRLRjXpq+BU=
@@ -879,6 +907,8 @@ github.com/zexi/influxql-to-metricsql v0.1.1 h1:dXnVzlxVLuo88zGkDca7erEnmvprGWjz
 github.com/zexi/influxql-to-metricsql v0.1.1/go.mod h1:JlC5FY+6De9ZPxG47G5DOgva8P9X1VaKS4ExzCmhSCc=
 github.com/zexi/promql/v2 v2.12.1 h1:crHKpULdLLsBZ9b78Rg6qQkugzlk6BHeCj93tw/F5RU=
 github.com/zexi/promql/v2 v2.12.1/go.mod h1:2UtzWZGmth95n2qdIZWpJ5yQ0cE5hEvz0dGsajI1Sqg=
+go.etcd.io/bbolt v1.3.7 h1:j+zJOnnEjF/kyHlDDgGnVL/AIqIJPq8UoB2GSNfkUfQ=
+go.etcd.io/bbolt v1.3.7/go.mod h1:N9Mkw9X8x5fupy0IKsmuqVtoGDyxsaDlbk4Rd05IAQw=
 go.etcd.io/etcd/api/v3 v3.5.0 h1:GsV3S+OfZEOCNXdtNkBSR7kgLobAa/SO6tCxRa0GAYw=
 go.etcd.io/etcd/api/v3 v3.5.0/go.mod h1:cbVKeC6lCfl7j/8jBhAK6aIYO9XOjdptoxU/nLQcPvs=
 go.etcd.io/etcd/client/pkg/v3 v3.5.0 h1:2aQv6F436YnN7I4VbI8PPYrBhu+SmrTaADcf8Mi/6PU=
@@ -984,6 +1014,7 @@ golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLL
 golang.org/x/net v0.0.0-20190628185345-da137c7871d7/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20190724013045-ca1201d0de80/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20190923162816-aa69164e4478/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
+golang.org/x/net v0.0.0-20191004110552-13f9640d40b9/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20191209160850-c0dbc17a3553/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200114155413-6afb5195e5aa/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200202094626-16171245cfb2/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
@@ -1032,8 +1063,8 @@ golang.org/x/sync v0.0.0-20201207232520-09787c993a3a/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20210220032951-036812b2e83c/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20220722155255-886fb9371eb4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.1.0/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.6.0 h1:5BMeUDZ7vkXGfEr1x9B4bRcTH4lpkTkpdh0T/J+qjbQ=
-golang.org/x/sync v0.6.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
+golang.org/x/sync v0.8.0 h1:3NFvSEYkUoMifnESzZl15y791HH1qU2xm6eCJU5ZPXQ=
+golang.org/x/sync v0.8.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
 golang.org/x/sys v0.0.0-20180724212812-e072cadbbdc8/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180830151530-49385e6e1522/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -1059,6 +1090,7 @@ golang.org/x/sys v0.0.0-20190922100055-0a153f010e69/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20190924154521-2837fb4f24fe/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191001151750-bb3f8db39f24/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191005200804-aed5e4c7ecf9/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20191022100944-742c48ecaeb7/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191115151921-52ab43148777/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191120155948-bd437916bb0e/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20191204072324-ce4227a45e2e/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
@@ -1089,6 +1121,7 @@ golang.org/x/sys v0.0.0-20200803210538-64077c9b5642/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201110211018-35f3e6cf4a65/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20201201145000-ef89a241ccb3/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201204225414-ed752295db88/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210124154548-22da62e12c0c/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210330210617-4fbd30eecc44/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
@@ -1107,8 +1140,8 @@ golang.org/x/sys v0.0.0-20220722155257-8c9f86f7a55f/go.mod h1:oPkhp1MJrh7nUepCBc
 golang.org/x/sys v0.1.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.5.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.7.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.17.0 h1:25cE3gD+tdBA7lp7QfhuV+rJiE9YXTcS3VG1SqssI/Y=
-golang.org/x/sys v0.17.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/sys v0.26.0 h1:KHjCJyddX0LoSTb3J+vWpupP9p0oznkqVk/IfjymZbo=
+golang.org/x/sys v0.26.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
 golang.org/x/term v0.0.0-20210927222741-03fcf44c2211/go.mod h1:jbD1KX2456YbFQfuXm/mYQcufACuNUgVhRMnK/tPxf8=
 golang.org/x/term v0.1.0/go.mod h1:jbD1KX2456YbFQfuXm/mYQcufACuNUgVhRMnK/tPxf8=
@@ -1269,6 +1302,7 @@ google.golang.org/grpc v1.20.1/go.mod h1:10oTOabMzJvdu6/UiuZezV6QK5dSlG84ov/aaiq
 google.golang.org/grpc v1.21.1/go.mod h1:oYelfM1adQP15Ek0mdvEgi9Df8B9CZIaU1084ijfRaM=
 google.golang.org/grpc v1.22.0/go.mod h1:Y5yQAOtifL1yxbo5wqy6BxZv8vAUGQwXBOALyacEbxg=
 google.golang.org/grpc v1.23.0/go.mod h1:Y5yQAOtifL1yxbo5wqy6BxZv8vAUGQwXBOALyacEbxg=
+google.golang.org/grpc v1.23.1/go.mod h1:Y5yQAOtifL1yxbo5wqy6BxZv8vAUGQwXBOALyacEbxg=
 google.golang.org/grpc v1.25.1/go.mod h1:c3i+UQWmh7LiEpx4sFZnkU36qjEYZ0imhYfXVyQciAY=
 google.golang.org/grpc v1.26.0/go.mod h1:qbnxyOmOxrQa7FizSgH+ReBfzJrCY1pSN7KXBS8abTk=
 google.golang.org/grpc v1.27.0/go.mod h1:qbnxyOmOxrQa7FizSgH+ReBfzJrCY1pSN7KXBS8abTk=
@@ -1294,8 +1328,8 @@ google.golang.org/protobuf v1.24.0/go.mod h1:r/3tXBNzIEhYS9I1OUVjXDlt8tc493IdKGj
 google.golang.org/protobuf v1.25.0/go.mod h1:9JNX74DMeImyA3h4bdi1ymwjUzf21/xIlbajtzgsN7c=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
 google.golang.org/protobuf v1.26.0/go.mod h1:9q0QmTI4eRPtz6boOQmLYwt+qCgq0jsYwAQnmE0givc=
-google.golang.org/protobuf v1.32.0 h1:pPC6BG5ex8PDFnkbrGU3EixyhKcQ2aDuBS36lqK/C7I=
-google.golang.org/protobuf v1.32.0/go.mod h1:c6P6GXX6sHbq/GpV6MGZEdwhWPcYBgnhAHhKbcUYpos=
+google.golang.org/protobuf v1.35.1 h1:m3LfL6/Ca+fqnjnlqQXNpFPABW1UD7mjh8KO2mKFytA=
+google.golang.org/protobuf v1.35.1/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
 gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
 gopkg.in/alexcesaro/quotedprintable.v3 v3.0.0-20150716171945-2caba252f4dc h1:2gGKlE2+asNV9m7xrywl36YYNnBG5ZQ0r/BOOxqPpmk=
 gopkg.in/alexcesaro/quotedprintable.v3 v3.0.0-20150716171945-2caba252f4dc/go.mod h1:m7x9LTH6d71AHyAX77c9yqWCCa3UKHcVEj9y7hAtKDk=
@@ -130,6 +130,7 @@ const (
 type ContainerSpec struct {
 	apis.ContainerSpec
 	// Volume mounts
+	RootFs       *apis.ContainerRootfs        `json:"rootfs"`
 	VolumeMounts []*apis.ContainerVolumeMount `json:"volume_mounts"`
 	Devices      []*ContainerDevice           `json:"devices"`
 }
@@ -346,3 +346,9 @@ type ContainerPullImageAuthConfig struct {
 	// RegistryToken is a bearer token to be sent to a registry
 	RegistryToken string `json:"registry_token,omitempty"`
 }
+
+type ContainerRootfs struct {
+	Type ContainerVolumeMountType  `json:"type"`
+	Disk *ContainerVolumeMountDisk `json:"disk"`
+	//CephFS *ContainerVolumeMountCephFS `json:"ceph_fs"`
+}
@@ -41,6 +41,12 @@ type ContainerVolumeMountCephFS struct {
 	Name    string `json:"name"`
 }

+type ContainerRootfs struct {
+	Type apis.ContainerVolumeMountType `json:"type"`
+	Disk *ContainerVolumeMountDisk     `json:"disk"`
+	// CephFS *ContainerVolumeMountCephFS   `json:"ceph_fs"`
+}
+
 type ContainerVolumeMount struct {
 	// 用于标识当前 pod volume mount 的唯一性
 	UniqueName string                             `json:"unique_name"`
@@ -65,6 +71,7 @@ type ContainerVolumeMount struct {
 type ContainerSpec struct {
 	apis.ContainerSpec
 	ImageCredentialToken string                  `json:"image_credential_token"`
+	Rootfs               *ContainerRootfs        `json:"rootfs"`
 	VolumeMounts         []*ContainerVolumeMount `json:"volume_mounts"`
 	Devices              []*ContainerDevice      `json:"devices"`
 }
@@ -23,6 +23,7 @@ import (

 	"yunion.io/x/onecloud/pkg/apis"
 	api "yunion.io/x/onecloud/pkg/apis/compute"
+	hostapi "yunion.io/x/onecloud/pkg/apis/host"
 	"yunion.io/x/onecloud/pkg/compute/models"
 	"yunion.io/x/onecloud/pkg/httperrors"
 	"yunion.io/x/onecloud/pkg/mcclient"
@@ -64,8 +65,7 @@ func (d disk) GetType() apis.ContainerVolumeMountType {
 	return apis.CONTAINER_VOLUME_MOUNT_TYPE_DISK
 }

-func (d disk) validateCreateData(ctx context.Context, userCred mcclient.TokenCredential, vm *apis.ContainerVolumeMount) (*apis.ContainerVolumeMount, error) {
-	disk := vm.Disk
+func (d disk) validateDiskCreateData(ctx context.Context, userCred mcclient.TokenCredential, disk *apis.ContainerVolumeMountDisk) (*apis.ContainerVolumeMountDisk, error) {
 	if disk == nil {
 		return nil, httperrors.NewNotEmptyError("disk is nil")
 	}
@@ -77,6 +77,15 @@ func (d disk) validateCreateData(ctx context.Context, userCred mcclient.TokenCre
 			return nil, httperrors.NewInputParameterError("index is less than 0")
 		}
 	}
+	return disk, nil
+}
+
+func (d disk) validateCreateData(ctx context.Context, userCred mcclient.TokenCredential, vm *apis.ContainerVolumeMount) (*apis.ContainerVolumeMount, error) {
+	disk, err := d.validateDiskCreateData(ctx, userCred, vm.Disk)
+	if err != nil {
+		return nil, err
+	}
+	vm.Disk = disk
 	return vm, nil
 }

@@ -102,31 +111,61 @@ func (d disk) validateCaseInsensitive(disk *models.SDisk, vm *apis.ContainerVolu
 	return nil
 }

-func (d disk) ValidateCreateData(ctx context.Context, userCred mcclient.TokenCredential, pod *models.SGuest, vm *apis.ContainerVolumeMount) (*apis.ContainerVolumeMount, error) {
-	if _, err := d.validateCreateData(ctx, userCred, vm); err != nil {
-		return nil, err
+func (d disk) ValidateRootFsCreateData(ctx context.Context, userCred mcclient.TokenCredential, pod *models.SGuest, rootFs *apis.ContainerRootfs) error {
+	disk, err := d.validateDiskCreateData(ctx, userCred, rootFs.Disk)
+	if err != nil {
+		return err
 	}
+	disk, _, err = d.validateDiskIndex(ctx, userCred, pod, disk)
+	if err != nil {
+		return err
+	}
+	rootFs.Disk = disk
+	if rootFs.Disk.Overlay != nil {
+		return httperrors.NewInputParameterError("can't use overlay and root_fs at the same time")
+	}
+	if len(rootFs.Disk.PostOverlay) > 0 {
+		return httperrors.NewInputParameterError("can't use post_overlay and root_fs at the same time")
+	}
+	return nil
+}
+
+func (d disk) ToHostRootFs(rootFs *apis.ContainerRootfs) (*hostapi.ContainerRootfs, error) {
+	diskObj := models.DiskManager.FetchDiskById(rootFs.Disk.Id)
+	if diskObj == nil {
+		return nil, errors.Wrapf(httperrors.ErrNotFound, "fetch disk %s", rootFs.Disk.Id)
+	}
+	disk := rootFs.Disk
+	return &hostapi.ContainerRootfs{
+		Type: rootFs.Type,
+		Disk: &hostapi.ContainerVolumeMountDisk{
+			Index: disk.Index,
+			Id:    disk.Id,
+		},
+	}, nil
+}
+
+func (d disk) validateDiskIndex(ctx context.Context, userCred mcclient.TokenCredential, pod *models.SGuest, disk *apis.ContainerVolumeMountDisk) (*apis.ContainerVolumeMountDisk, *models.SDisk, error) {
 	disks, err := pod.GetDisks()
 	if err != nil {
-		return nil, errors.Wrap(err, "get pod disks")
+		return nil, nil, errors.Wrap(err, "get pod disks")
 	}
-	disk := vm.Disk
 	var diskObj models.SDisk
 	if disk.Index != nil {
 		diskIndex := *disk.Index
 		if diskIndex >= len(disks) {
-			return nil, httperrors.NewInputParameterError("disk.index %d is large than disk size %d", diskIndex, len(disks))
+			return nil, nil, httperrors.NewInputParameterError("disk.index %d is large than disk size %d", diskIndex, len(disks))
 		}
 		diskObj = disks[diskIndex]
-		vm.Disk.Id = diskObj.GetId()
+		disk.Id = diskObj.GetId()
 		// remove index
-		vm.Disk.Index = nil
+		disk.Index = nil
 		if err := d.validateCaseInsensitive(&diskObj, disk); err != nil {
-			return nil, err
+			return nil, nil, err
 		}
 	} else {
 		if disk.Id == "" {
-			return nil, httperrors.NewNotEmptyError("disk.id is empty")
+			return nil, nil, httperrors.NewNotEmptyError("disk.id is empty")
 		}
 		foundDisk := false
 		for _, d := range disks {
@@ -138,13 +177,26 @@ func (d disk) ValidateCreateData(ctx context.Context, userCred mcclient.TokenCre
 			}
 		}
 		if !foundDisk {
-			return nil, httperrors.NewNotFoundError("not found pod disk by %s", disk.Id)
+			return nil, nil, httperrors.NewNotFoundError("not found pod disk by %s", disk.Id)
 		}
 		if err := d.validateCaseInsensitive(&diskObj, disk); err != nil {
-			return nil, err
+			return nil, nil, err
 		}
 	}
-	if err := d.validateOverlay(ctx, userCred, vm, &diskObj); err != nil {
+	return disk, &diskObj, nil
+}
+
+func (d disk) ValidateCreateData(ctx context.Context, userCred mcclient.TokenCredential, pod *models.SGuest, vm *apis.ContainerVolumeMount) (*apis.ContainerVolumeMount, error) {
+	if _, err := d.validateCreateData(ctx, userCred, vm); err != nil {
+		return nil, err
+	}
+	disk, diskObj, err := d.validateDiskIndex(ctx, userCred, pod, vm.Disk)
+	if err != nil {
+		return nil, err
+	}
+	vm.Disk = disk
+
+	if err := d.validateOverlay(ctx, userCred, vm, diskObj); err != nil {
 		return nil, errors.Wrapf(err, "validate overlay")
 	}
 	if err := d.ValidatePostOverlay(ctx, userCred, vm); err != nil {
@@ -89,12 +89,27 @@ func GetContainerVolumeMountDriverWithError(typ apis.ContainerVolumeMountType) (
 	return getContainerDriverWithError[apis.ContainerVolumeMountType, IContainerVolumeMountDriver](containerVolumeDrivers, typ)
 }

+func GetContainerRootFsDriverWithError(typ apis.ContainerVolumeMountType) (IContainerRootFsDriver, error) {
+	return getContainerDriverWithError[apis.ContainerVolumeMountType, IContainerRootFsDriver](containerVolumeDrivers, typ)
+}
+
+func GetContainerRootFsDriver(typ apis.ContainerVolumeMountType) IContainerRootFsDriver {
+	return getContainerDriver[apis.ContainerVolumeMountType, IContainerRootFsDriver](containerVolumeDrivers, typ)
+}
+
 type IContainerVolumeMountDriver interface {
 	GetType() apis.ContainerVolumeMountType
 	ValidatePodCreateData(ctx context.Context, userCred mcclient.TokenCredential, vm *apis.ContainerVolumeMount, input *api.ServerCreateInput) error
 	ValidateCreateData(ctx context.Context, userCred mcclient.TokenCredential, pod *SGuest, vm *apis.ContainerVolumeMount) (*apis.ContainerVolumeMount, error)
 }

+type IContainerRootFsDriver interface {
+	IContainerVolumeMountDriver
+
+	ToHostRootFs(rootFs *apis.ContainerRootfs) (*hostapi.ContainerRootfs, error)
+	ValidateRootFsCreateData(ctx context.Context, userCred mcclient.TokenCredential, pod *SGuest, rootFs *apis.ContainerRootfs) error
+}
+
 type IContainerVolumeMountDiskDriver interface {
 	IContainerVolumeMountDriver

@@ -178,6 +178,9 @@ func (m *SContainerManager) ValidateSpec(ctx context.Context, userCred mcclient.
 	}

 	if pod != nil {
+		if err := m.ValidateSpecRootFs(ctx, userCred, pod, spec, ctr); err != nil {
+			return errors.Wrap(err, "ValidateSpecRootFs")
+		}
 		if err := m.ValidateSpecVolumeMounts(ctx, userCred, pod, spec, ctr); err != nil {
 			return errors.Wrap(err, "ValidateSpecVolumeMounts")
 		}
@@ -231,6 +234,21 @@ func (m *SContainerManager) ValidateSpecDevice(ctx context.Context, userCred mcc
 	return drv.ValidateCreateData(ctx, userCred, pod, dev)
 }

+func (m *SContainerManager) ValidateSpecRootFs(ctx context.Context, userCred mcclient.TokenCredential, pod *SGuest, spec *api.ContainerSpec, ctr *SContainer) error {
+	if spec.RootFs == nil {
+		return nil
+	}
+	rootFs := spec.RootFs
+	drv, err := GetContainerRootFsDriverWithError(rootFs.Type)
+	if err != nil {
+		return errors.Wrapf(err, "get container volume mount driver %q", rootFs.Type)
+	}
+	if err := drv.ValidateRootFsCreateData(ctx, userCred, pod, rootFs); err != nil {
+		return errors.Wrapf(err, "validate %s create data", drv.GetType())
+	}
+	return nil
+}
+
 func (m *SContainerManager) ValidateSpecVolumeMounts(ctx context.Context, userCred mcclient.TokenCredential, pod *SGuest, spec *api.ContainerSpec, ctr *SContainer) error {
 	relation, err := m.GetVolumeMountRelations(pod, spec)
 	if err != nil {
@@ -461,6 +479,23 @@ func (c *SContainer) GetVolumeMounts() []*apis.ContainerVolumeMount {
 	return c.Spec.VolumeMounts
 }

+type ContainerRootFsRelation struct {
+	RootFs *apis.ContainerRootfs
+	pod    *SGuest
+}
+
+func (rr *ContainerRootFsRelation) ToHostRootFs() (*hostapi.ContainerRootfs, error) {
+	drv, err := GetContainerRootFsDriverWithError(rr.RootFs.Type)
+	if err != nil {
+		return nil, errors.Wrapf(err, "get container root fs driver %q", rr.RootFs.Type)
+	}
+	rootFs, err := drv.ToHostRootFs(rr.RootFs)
+	if err != nil {
+		return nil, errors.Wrapf(err, "to host root fs")
+	}
+	return rootFs, nil
+}
+
 type ContainerVolumeMountRelation struct {
 	VolumeMount *apis.ContainerVolumeMount

@@ -547,6 +582,13 @@ func (vm *ContainerVolumeMountRelation) ToHostMount(ctx context.Context, userCre
 	return ret, nil
 }

+func (m *SContainerManager) GetRootFsRelation(pod *SGuest, spec *api.ContainerSpec) (*ContainerRootFsRelation, error) {
+	return &ContainerRootFsRelation{
+		RootFs: spec.RootFs,
+		pod:    pod,
+	}, nil
+}
+
 func (m *SContainerManager) GetVolumeMountRelations(pod *SGuest, spec *api.ContainerSpec) ([]*ContainerVolumeMountRelation, error) {
 	relation := make([]*ContainerVolumeMountRelation, len(spec.VolumeMounts))
 	for idx, vm := range spec.VolumeMounts {
@@ -559,6 +601,10 @@ func (m *SContainerManager) GetVolumeMountRelations(pod *SGuest, spec *api.Conta
 	return relation, nil
 }

+func (c *SContainer) GetRootFsRelation() (*ContainerRootFsRelation, error) {
+	return GetContainerManager().GetRootFsRelation(c.GetPod(), c.Spec)
+}
+
 func (c *SContainer) GetVolumeMountRelations() ([]*ContainerVolumeMountRelation, error) {
 	return GetContainerManager().GetVolumeMountRelations(c.GetPod(), c.Spec)
 }
@@ -716,7 +762,27 @@ func (m *SContainerManager) ConvertVolumeMountRelationToSpec(ctx context.Context
 	return mounts, nil
 }

+func (m *SContainerManager) ConvertRootFsRelationToSpec(ctx context.Context, userCred mcclient.TokenCredential, relation *ContainerRootFsRelation) (*hostapi.ContainerRootfs, error) {
+	if relation.RootFs == nil {
+		return nil, nil
+	}
+	rootFs, err := relation.ToHostRootFs()
+	if err != nil {
+		return nil, errors.Wrap(err, "ToHostRootFs")
+	}
+	return rootFs, nil
+}
+
 func (c *SContainer) ToHostContainerSpec(ctx context.Context, userCred mcclient.TokenCredential) (*hostapi.ContainerSpec, error) {
+	rootFsRelation, err := c.GetRootFsRelation()
+	if err != nil {
+		return nil, errors.Wrap(err, "GetRootFsRelation")
+	}
+	rootFs, err := GetContainerManager().ConvertRootFsRelationToSpec(ctx, userCred, rootFsRelation)
+	if err != nil {
+		return nil, errors.Wrap(err, "ConvertRootFsRelationToSpec")
+	}
+
 	vmRelation, err := c.GetVolumeMountRelations()
 	if err != nil {
 		return nil, errors.Wrap(err, "GetVolumeMountRelations")
@@ -737,6 +803,7 @@ func (c *SContainer) ToHostContainerSpec(ctx context.Context, userCred mcclient.
 	spec := c.Spec.ContainerSpec
 	hSpec := &hostapi.ContainerSpec{
 		ContainerSpec: spec,
+		Rootfs:        rootFs,
 		VolumeMounts:  mounts,
 		Devices:       ctrDevs,
 	}
@@ -0,0 +1,8 @@
+package snapshot_service
+
+const (
+	DefaultRootDir   = "/opt/yunion/containerd/io.yunion.containerd.host-overlay-snapshotter"
+	SocksFileName    = "grpc.socks"
+	LabelServerId    = "containerd.io/snapshot/host-overlay.server-id"
+	LabelContainerId = "containerd.io/snapshot/host-overlay.container-id"
+)
@@ -0,0 +1 @@
+package snapshot_service // import "yunion.io/x/onecloud/pkg/hostman/container/snapshot_service"
@@ -0,0 +1,150 @@
+package snapshot_service
+
+import (
+	"context"
+	"net"
+	"os"
+	"path"
+	"strings"
+
+	snapshotsapi "github.com/containerd/containerd/api/services/snapshots/v1"
+	"github.com/containerd/containerd/contrib/snapshotservice"
+	"github.com/containerd/containerd/mount"
+	"github.com/containerd/containerd/snapshots"
+	"github.com/containerd/containerd/snapshots/overlay"
+	"google.golang.org/grpc"
+
+	"yunion.io/x/jsonutils"
+	"yunion.io/x/log"
+	"yunion.io/x/pkg/errors"
+)
+
+func StartService(guestMan IGuestManager, root string) error {
+	sn, err := NewSnapshotter(guestMan, root)
+	if err != nil {
+		return errors.Wrap(err, "NewSnapshotter")
+	}
+	svc := snapshotservice.FromSnapshotter(sn)
+	rpc := grpc.NewServer()
+	snapshotsapi.RegisterSnapshotsServer(rpc, svc)
+	socksPath := path.Join(root, SocksFileName)
+	if err := os.RemoveAll(socksPath); err != nil {
+		return errors.Wrapf(err, "RemoveAll %s", socksPath)
+	}
+	listener, err := net.Listen("unix", socksPath)
+	if err != nil {
+		return errors.Wrapf(err, "Listen %s", socksPath)
+	}
+	return rpc.Serve(listener)
+}
+
+type IGuestManager interface {
+	GetContainerManager(serverId string) (ISnapshotContainerManager, error)
+}
+
+type ISnapshotContainerManager interface {
+	GetRootFsMountPath(ctx context.Context, containerId string) (string, error)
+}
+
+func NewSnapshotter(guestMan IGuestManager, root string, opts ...overlay.Opt) (snapshots.Snapshotter, error) {
+	sn, err := overlay.NewSnapshotter(root, opts...)
+	if err != nil {
+		return nil, errors.Wrap(err, "NewSnapshotter")
+	}
+	return &overlayRootFsUpperSnapshotter{
+		Snapshotter: sn,
+		guestMan:    guestMan,
+	}, nil
+}
+
+type overlayRootFsUpperSnapshotter struct {
+	snapshots.Snapshotter
+	guestMan IGuestManager
+}
+
+func (s *overlayRootFsUpperSnapshotter) Mounts(ctx context.Context, key string) ([]mount.Mount, error) {
+	info, err := s.Snapshotter.Stat(ctx, key)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Stat with %s", key)
+	}
+	infoJson := jsonutils.Marshal(info)
+	log.Infof("mount request info: %s", infoJson.String())
+	mounts, err := s.Snapshotter.Mounts(ctx, key)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Mounts with %s", key)
+	}
+	mounts, err = s.changeUpper(ctx, key, mounts)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Mounts.changeUpper with %s", key)
+	}
+	return mounts, nil
+}
+
+func (s *overlayRootFsUpperSnapshotter) Prepare(ctx context.Context, key string, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) {
+	mounts, err := s.Snapshotter.Prepare(ctx, key, parent, opts...)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Prepare with %s", key)
+	}
+	mounts, err = s.changeUpper(ctx, key, mounts)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Prepare.changeUpper with %s", key)
+	}
+	return mounts, nil
+}
+
+func (s *overlayRootFsUpperSnapshotter) View(ctx context.Context, key string, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) {
+	mounts, err := s.Snapshotter.View(ctx, key, parent, opts...)
+	if err != nil {
+		return nil, errors.Wrapf(err, "View with %s", key)
+	}
+	mounts, err = s.changeUpper(ctx, key, mounts)
+	if err != nil {
+		return nil, errors.Wrapf(err, "View.changeUpper with %s", key)
+	}
+	return mounts, nil
+}
+
+func (s *overlayRootFsUpperSnapshotter) changeUpper(ctx context.Context, key string, mounts []mount.Mount) ([]mount.Mount, error) {
+	if len(mounts) != 1 || mounts[0].Type != "overlay" {
+		return mounts, nil
+	}
+	info, err := s.Snapshotter.Stat(ctx, key)
+	if err != nil {
+		return nil, errors.Wrapf(err, "Stat with %s", key)
+	}
+	serverId, ok := info.Labels[LabelServerId]
+	if !ok || serverId == "" {
+		return mounts, nil
+	}
+	containerId, ok := info.Labels[LabelContainerId]
+	if !ok || containerId == "" {
+		return mounts, nil
+	}
+	ctrMan, err := s.guestMan.GetContainerManager(serverId)
+	if err != nil {
+		return mounts, errors.Wrapf(err, "GetContainerManager with %s", serverId)
+	}
+	rootFsPath, err := ctrMan.GetRootFsMountPath(ctx, containerId)
+	if err != nil {
+		return mounts, errors.Wrapf(err, "GetRootFsMountPath with %s, %s", serverId, containerId)
+	}
+	upperPath := path.Join(rootFsPath, "upper")
+	workPath := path.Join(rootFsPath, "work")
+	for _, dir := range []string{upperPath, workPath} {
+		if err := os.MkdirAll(dir, 0755); err != nil {
+			return mounts, errors.Wrapf(err, "MkdirAll %s", dir)
+		}
+	}
+
+	upperDirKey := "upperdir="
+	workDirKey := "workdir="
+	for i, o := range mounts[0].Options {
+		if strings.HasPrefix(o, upperDirKey) {
+			mounts[0].Options[i] = upperDirKey + upperPath
+		}
+		if strings.HasPrefix(o, workDirKey) {
+			mounts[0].Options[i] = workDirKey + workPath
+		}
+	}
+	return mounts, nil
+}
@@ -41,6 +41,7 @@ import (
 	hostapi "yunion.io/x/onecloud/pkg/apis/host"
 	"yunion.io/x/onecloud/pkg/appsrv"
 	"yunion.io/x/onecloud/pkg/hostman/container/prober"
+	"yunion.io/x/onecloud/pkg/hostman/container/snapshot_service"
 	"yunion.io/x/onecloud/pkg/hostman/guestman/arch"
 	"yunion.io/x/onecloud/pkg/hostman/guestman/desc"
 	fwd "yunion.io/x/onecloud/pkg/hostman/guestman/forwarder"
@@ -155,10 +156,32 @@ func NewGuestManager(host hostutils.IHost, serversPath string, workerCnt int) (*
 		manager.containerRuntimeManager = runtimeMan
 		manager.pleg = pleg.NewGenericPLEG(runtimeMan, pleg.ChannelCapacity, pleg.RelistPeriod, manager.podCache, clock.RealClock{})
 		manager.pleg.Start()
+		go func() {
+			if err := manager.startContainerdSnapshotService(); err != nil {
+				log.Fatalf("start containerd snapshot service: %s", err)
+			}
+		}()
 	}
 	return manager, nil
 }

+func (h *SGuestManager) startContainerdSnapshotService() error {
+	root := filepath.Join(options.HostOptions.ServersPath, "containerd_snapshots")
+	err := snapshot_service.StartService(h, root)
+	if err != nil {
+		return errors.Wrap(err, "new snapshot service")
+	}
+	return nil
+}
+
+func (h *SGuestManager) GetContainerManager(serverId string) (snapshot_service.ISnapshotContainerManager, error) {
+	pod, ok := h.GetServer(serverId)
+	if !ok {
+		return nil, errors.Wrapf(httperrors.ErrNotFound, "server %s not found", serverId)
+	}
+	return pod.(snapshot_service.ISnapshotContainerManager), nil
+}
+
 func (m *SGuestManager) startContainerSyncLoop() {
 	if m.host.IsContainerHost() {
 		go func() {
@@ -44,6 +44,7 @@ import (
 	"yunion.io/x/onecloud/pkg/hostman/container/lifecycle"
 	"yunion.io/x/onecloud/pkg/hostman/container/prober"
 	proberesults "yunion.io/x/onecloud/pkg/hostman/container/prober/results"
+	"yunion.io/x/onecloud/pkg/hostman/container/snapshot_service"
 	"yunion.io/x/onecloud/pkg/hostman/container/status"
 	"yunion.io/x/onecloud/pkg/hostman/container/volume_mount"
 	"yunion.io/x/onecloud/pkg/hostman/container/volume_mount/disk"
@@ -627,7 +628,68 @@ func (s *sPodGuestInstance) GetDisks() []*desc.SGuestDisk {
 	return s.GetDesc().Disks
 }

+func (s *sPodGuestInstance) ConvertRootFsToVolumeMount(rootFs *hostapi.ContainerRootfs) *hostapi.ContainerVolumeMount {
+	return &hostapi.ContainerVolumeMount{
+		Type: rootFs.Type,
+		Disk: rootFs.Disk,
+	}
+}
+
+func (s *sPodGuestInstance) mountRootFs(ctrId string, rootFs *hostapi.ContainerRootfs) error {
+	drv := volume_mount.GetDriver(rootFs.Type)
+	vol := &hostapi.ContainerVolumeMount{
+		Type: rootFs.Type,
+		Disk: rootFs.Disk,
+	}
+	if err := drv.Mount(s, ctrId, vol); err != nil {
+		return errors.Wrapf(err, "mount root fs %s, ctrId %s", jsonutils.Marshal(rootFs), ctrId)
+	}
+	return nil
+}
+
+func (s *sPodGuestInstance) umountRootFs(ctrId string, rootFs *hostapi.ContainerRootfs) error {
+	drv := volume_mount.GetDriver(rootFs.Type)
+	vol := &hostapi.ContainerVolumeMount{
+		Type: rootFs.Type,
+		Disk: rootFs.Disk,
+	}
+	if err := drv.Unmount(s, ctrId, vol); err != nil {
+		return errors.Wrapf(err, "unmount root fs %s, ctrId %s", jsonutils.Marshal(rootFs), ctrId)
+	}
+	return nil
+}
+
+func (s *sPodGuestInstance) getRootFsMountPath(ctrId string) (string, error) {
+	ctr := s.GetContainerById(ctrId)
+	if ctr == nil {
+		return "", errors.Wrapf(httperrors.ErrNotFound, "not found container %s", ctrId)
+	}
+	rootFs := ctr.Spec.Rootfs
+	if rootFs == nil {
+		return "", errors.Wrapf(httperrors.ErrNotFound, "not found root fs %s", ctrId)
+	}
+	vol := s.ConvertRootFsToVolumeMount(rootFs)
+	drv := volume_mount.GetDriver(vol.Type)
+	hostPath, err := drv.GetRuntimeMountHostPath(s, ctrId, vol)
+	if err != nil {
+		return "", errors.Wrapf(err, "get runtime mount host path %s, ctrId %s", jsonutils.Marshal(vol), ctrId)
+	}
+	return hostPath, nil
+}
+
+func (s *sPodGuestInstance) GetRootFsMountPath(ctx context.Context, ctrId string) (string, error) {
+	return s.getRootFsMountPath(ctrId)
+}
+
 func (s *sPodGuestInstance) mountPodVolumes() error {
+	for _, ctr := range s.GetDesc().Containers {
+		if ctr.Spec.Rootfs == nil {
+			continue
+		}
+		if err := s.mountRootFs(ctr.Id, ctr.Spec.Rootfs); err != nil {
+			return errors.Wrapf(err, "mount root fs %s, ctrId %s", jsonutils.Marshal(ctr.Spec.Rootfs), ctr.Id)
+		}
+	}
 	for ctrId, vols := range s.getContainerVolumeMounts() {
 		for _, vol := range vols {
 			drv := volume_mount.GetDriver(vol.Type)
@@ -646,6 +708,14 @@ func (s *sPodGuestInstance) mountPodVolumes() error {
 }

 func (s *sPodGuestInstance) umountPodVolumes() error {
+	for _, ctr := range s.GetDesc().Containers {
+		if ctr.Spec.Rootfs == nil {
+			continue
+		}
+		if err := s.umountRootFs(ctr.Id, ctr.Spec.Rootfs); err != nil {
+			return errors.Wrapf(err, "umount root fs %s, ctrId %s", jsonutils.Marshal(ctr.Spec.Rootfs), ctr.Id)
+		}
+	}
 	for ctrId, vols := range s.getContainerVolumeMounts() {
 		for _, vol := range vols {
 			if err := volume_mount.GetDriver(vol.Type).Unmount(s, ctrId, vol); err != nil {
@@ -1764,6 +1834,10 @@ func (s *sPodGuestInstance) createContainer(ctx context.Context, userCred mcclie
 		ctrCfg.Linux.SecurityContext.NamespaceOptions.Pid = runtimeapi.NamespaceMode_TARGET
 		ctrCfg.Linux.SecurityContext.NamespaceOptions.TargetId = s.GetCRIId()
 	}*/
+	if spec.Rootfs != nil {
+		ctrCfg.Labels[snapshot_service.LabelServerId] = s.GetId()
+		ctrCfg.Labels[snapshot_service.LabelContainerId] = ctrId
+	}

 	// inherit security context
 	if spec.SecurityContext != nil {
@@ -54,6 +54,7 @@ type ContainerCreateCommonOptions struct {
 	Args              []string `help:"Args for the Command (i.e. command for docker)" json:"args"`
 	WorkingDir        string   `help:"Current working directory of the command" json:"working_dir"`
 	Env               []string `help:"List of environment variable to set in the container and the format is: <key>=<value>"`
+	RootFs            string   `help:"Root filesystem of the container, e.g.: disk_index=<disk_number>,disk_id=<disk_id>"`
 	VolumeMount       []string `help:"Volume mount of the container and the format is: name=<val>,mount=<container_path>,readonly=<true_or_false>,case_insensitive_paths=p1,p2,disk_index=<disk_number>,disk_id=<disk_id>"`
 	Device            []string `help:"Host device: <host_path>:<container_path>:<permissions>, e.g.: /dev/snd:/dev/snd:rwm"`
 	Privileged        bool     `help:"Privileged mode"`
@@ -125,6 +126,13 @@ func (o ContainerCreateCommonOptions) getCreateSpec() (*computeapi.ContainerSpec
 		}
 		req.Envs = append(req.Envs, e)
 	}
+	if len(o.RootFs) != 0 {
+		rootFs, err := parseContainerRootFs(o.RootFs)
+		if err != nil {
+			return nil, errors.Wrapf(err, "parseContainerRootFs %s", o.RootFs)
+		}
+		req.RootFs = rootFs
+	}
 	for _, vmStr := range o.VolumeMount {
 		vm, err := parseContainerVolumeMount(vmStr)
 		if err != nil {
@@ -174,6 +182,32 @@ func parseContainerEnv(env string) (*apis.ContainerKeyValue, error) {
 	}, nil
 }

+func parseContainerRootFs(rootFs string) (*apis.ContainerRootfs, error) {
+	out := &apis.ContainerRootfs{
+		Type: apis.CONTAINER_VOLUME_MOUNT_TYPE_DISK,
+		Disk: &apis.ContainerVolumeMountDisk{},
+	}
+	for _, seg := range strings.Split(rootFs, ",") {
+		info := strings.Split(seg, "=")
+		if len(info) != 2 {
+			return nil, errors.Errorf("invalid option %s", seg)
+		}
+		key := info[0]
+		val := info[1]
+		switch key {
+		case "disk_index":
+			index, err := strconv.Atoi(val)
+			if err != nil {
+				return nil, errors.Wrapf(err, "wrong disk_index %s", val)
+			}
+			out.Disk.Index = &index
+		case "disk_id":
+			out.Disk.Id = val
+		}
+	}
+	return out, nil
+}
+
 func parseContainerVolumeMount(vmStr string) (*apis.ContainerVolumeMount, error) {
 	vm := &apis.ContainerVolumeMount{}
 	for _, seg := range strings.Split(vmStr, ",") {
@@ -0,0 +1 @@
+* text=auto eol=lf
@@ -1 +1,10 @@
+.vscode/
+
 *.exe
+
+# testing
+testdata
+
+# go workspaces
+go.work
+go.work.sum
@@ -0,0 +1,147 @@
+linters:
+  enable:
+    # style
+    - containedctx # struct contains a context
+    - dupl # duplicate code
+    - errname # erorrs are named correctly
+    - nolintlint # "//nolint" directives are properly explained
+    - revive # golint replacement
+    - unconvert # unnecessary conversions
+    - wastedassign
+
+    # bugs, performance, unused, etc ...
+    - contextcheck # function uses a non-inherited context
+    - errorlint # errors not wrapped for 1.13
+    - exhaustive # check exhaustiveness of enum switch statements
+    - gofmt # files are gofmt'ed
+    - gosec # security
+    - nilerr # returns nil even with non-nil error
+    - thelper #  test helpers without t.Helper()
+    - unparam # unused function params
+
+issues:
+  exclude-dirs:
+    - pkg/etw/sample
+
+  exclude-rules:
+    # err is very often shadowed in nested scopes
+    - linters:
+        - govet
+      text: '^shadow: declaration of "err" shadows declaration'
+
+    # ignore long lines for skip autogen directives
+    - linters:
+        - revive
+      text: "^line-length-limit: "
+      source: "^//(go:generate|sys) "
+
+    #TODO: remove after upgrading to go1.18
+    # ignore comment spacing for nolint and sys directives
+    - linters:
+        - revive
+      text: "^comment-spacings: no space between comment delimiter and comment text"
+      source: "//(cspell:|nolint:|sys |todo)"
+
+    # not on go 1.18 yet, so no any
+    - linters:
+        - revive
+      text: "^use-any: since GO 1.18 'interface{}' can be replaced by 'any'"
+
+    # allow unjustified ignores of error checks in defer statements
+    - linters:
+        - nolintlint
+      text: "^directive `//nolint:errcheck` should provide explanation"
+      source: '^\s*defer '
+
+    # allow unjustified ignores of error lints for io.EOF
+    - linters:
+        - nolintlint
+      text: "^directive `//nolint:errorlint` should provide explanation"
+      source: '[=|!]= io.EOF'
+
+
+linters-settings:
+  exhaustive:
+    default-signifies-exhaustive: true
+  govet:
+    enable-all: true
+    disable:
+      # struct order is often for Win32 compat
+      # also, ignore pointer bytes/GC issues for now until performance becomes an issue
+      - fieldalignment
+  nolintlint:
+    require-explanation: true
+    require-specific: true
+  revive:
+    # revive is more configurable than static check, so likely the preferred alternative to static-check
+    # (once the perf issue is solved: https://github.com/golangci/golangci-lint/issues/2997)
+    enable-all-rules:
+      true
+      # https://github.com/mgechev/revive/blob/master/RULES_DESCRIPTIONS.md
+    rules:
+      # rules with required arguments
+      - name: argument-limit
+        disabled: true
+      - name: banned-characters
+        disabled: true
+      - name: cognitive-complexity
+        disabled: true
+      - name: cyclomatic
+        disabled: true
+      - name: file-header
+        disabled: true
+      - name: function-length
+        disabled: true
+      - name: function-result-limit
+        disabled: true
+      - name: max-public-structs
+        disabled: true
+      # geneally annoying rules
+      - name: add-constant # complains about any and all strings and integers
+        disabled: true
+      - name: confusing-naming # we frequently use "Foo()" and "foo()" together
+        disabled: true
+      - name: flag-parameter # excessive, and a common idiom we use
+        disabled: true
+      - name: unhandled-error # warns over common fmt.Print* and io.Close; rely on errcheck instead
+        disabled: true
+      # general config
+      - name: line-length-limit
+        arguments:
+          - 140
+      - name: var-naming
+        arguments:
+          - []
+          - - CID
+            - CRI
+            - CTRD
+            - DACL
+            - DLL
+            - DOS
+            - ETW
+            - FSCTL
+            - GCS
+            - GMSA
+            - HCS
+            - HV
+            - IO
+            - LCOW
+            - LDAP
+            - LPAC
+            - LTSC
+            - MMIO
+            - NT
+            - OCI
+            - PMEM
+            - PWSH
+            - RX
+            - SACl
+            - SID
+            - SMB
+            - TX
+            - VHD
+            - VHDX
+            - VMID
+            - VPCI
+            - WCOW
+            - WIM
@@ -0,0 +1 @@
+  * @microsoft/containerplat
@@ -1,4 +1,4 @@
-# go-winio
+# go-winio [![Build Status](https://github.com/microsoft/go-winio/actions/workflows/ci.yml/badge.svg)](https://github.com/microsoft/go-winio/actions/workflows/ci.yml)

 This repository contains utilities for efficiently performing Win32 IO operations in
 Go. Currently, this is focused on accessing named pipes and other file handles, and
@@ -11,12 +11,79 @@ package.

 Please see the LICENSE file for licensing information.

-This project has adopted the [Microsoft Open Source Code of
-Conduct](https://opensource.microsoft.com/codeofconduct/). For more information
-see the [Code of Conduct
-FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or contact
-[opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional
-questions or comments.
+## Contributing

-Thanks to natefinch for the inspiration for this library. See https://github.com/natefinch/npipe
-for another named pipe implementation.
+This project welcomes contributions and suggestions.
+Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that
+you have the right to, and actually do, grant us the rights to use your contribution.
+For details, visit [Microsoft CLA](https://cla.microsoft.com).
+
+When you submit a pull request, a CLA-bot will automatically determine whether you need to
+provide a CLA and decorate the PR appropriately (e.g., label, comment).
+Simply follow the instructions provided by the bot.
+You will only need to do this once across all repos using our CLA.
+
+Additionally, the pull request pipeline requires the following steps to be performed before
+mergining.
+
+### Code Sign-Off
+
+We require that contributors sign their commits using [`git commit --signoff`][git-commit-s]
+to certify they either authored the work themselves or otherwise have permission to use it in this project.
+
+A range of commits can be signed off using [`git rebase --signoff`][git-rebase-s].
+
+Please see [the developer certificate](https://developercertificate.org) for more info,
+as well as to make sure that you can attest to the rules listed.
+Our CI uses the DCO Github app to ensure that all commits in a given PR are signed-off.
+
+### Linting
+
+Code must pass a linting stage, which uses [`golangci-lint`][lint].
+The linting settings are stored in [`.golangci.yaml`](./.golangci.yaml), and can be run
+automatically with VSCode by adding the following to your workspace or folder settings:
+
+```json
+    "go.lintTool": "golangci-lint",
+    "go.lintOnSave": "package",
+```
+
+Additional editor [integrations options are also available][lint-ide].
+
+Alternatively, `golangci-lint` can be [installed locally][lint-install] and run from the repo root:
+
+```shell
+# use . or specify a path to only lint a package
+# to show all lint errors, use flags "--max-issues-per-linter=0 --max-same-issues=0"
+> golangci-lint run ./...
+```
+
+### Go Generate
+
+The pipeline checks that auto-generated code, via `go generate`, are up to date.
+
+This can be done for the entire repo:
+
+```shell
+> go generate ./...
+```
+
+## Code of Conduct
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
+For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or
+contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.
+
+## Special Thanks
+
+Thanks to [natefinch][natefinch] for the inspiration for this library.
+See [npipe](https://github.com/natefinch/npipe) for another named pipe implementation.
+
+[lint]: https://golangci-lint.run/
+[lint-ide]: https://golangci-lint.run/usage/integrations/#editor-integration
+[lint-install]: https://golangci-lint.run/usage/install/#local-installation
+
+[git-commit-s]: https://git-scm.com/docs/git-commit#Documentation/git-commit.txt--s
+[git-rebase-s]: https://git-scm.com/docs/git-rebase#Documentation/git-rebase.txt---signoff
+
+[natefinch]: https://github.com/natefinch
@@ -0,0 +1,41 @@
+<!-- BEGIN MICROSOFT SECURITY.MD V0.0.7 BLOCK -->
+
+## Security
+
+Microsoft takes the security of our software products and services seriously, which includes all source code repositories managed through our GitHub organizations, which include [Microsoft](https://github.com/Microsoft), [Azure](https://github.com/Azure), [DotNet](https://github.com/dotnet), [AspNet](https://github.com/aspnet), [Xamarin](https://github.com/xamarin), and [our GitHub organizations](https://opensource.microsoft.com/).
+
+If you believe you have found a security vulnerability in any Microsoft-owned repository that meets [Microsoft's definition of a security vulnerability](https://aka.ms/opensource/security/definition), please report it to us as described below.
+
+## Reporting Security Issues
+
+**Please do not report security vulnerabilities through public GitHub issues.**
+
+Instead, please report them to the Microsoft Security Response Center (MSRC) at [https://msrc.microsoft.com/create-report](https://aka.ms/opensource/security/create-report).
+
+If you prefer to submit without logging in, send email to [secure@microsoft.com](mailto:secure@microsoft.com).  If possible, encrypt your message with our PGP key; please download it from the [Microsoft Security Response Center PGP Key page](https://aka.ms/opensource/security/pgpkey).
+
+You should receive a response within 24 hours. If for some reason you do not, please follow up via email to ensure we received your original message. Additional information can be found at [microsoft.com/msrc](https://aka.ms/opensource/security/msrc). 
+
+Please include the requested information listed below (as much as you can provide) to help us better understand the nature and scope of the possible issue:
+
+  * Type of issue (e.g. buffer overflow, SQL injection, cross-site scripting, etc.)
+  * Full paths of source file(s) related to the manifestation of the issue
+  * The location of the affected source code (tag/branch/commit or direct URL)
+  * Any special configuration required to reproduce the issue
+  * Step-by-step instructions to reproduce the issue
+  * Proof-of-concept or exploit code (if possible)
+  * Impact of the issue, including how an attacker might exploit the issue
+
+This information will help us triage your report more quickly.
+
+If you are reporting for a bug bounty, more complete reports can contribute to a higher bounty award. Please visit our [Microsoft Bug Bounty Program](https://aka.ms/opensource/security/bounty) page for more details about our active programs.
+
+## Preferred Languages
+
+We prefer all communications to be in English.
+
+## Policy
+
+Microsoft follows the principle of [Coordinated Vulnerability Disclosure](https://aka.ms/opensource/security/cvd).
+
+<!-- END MICROSOFT SECURITY.MD BLOCK -->
@@ -1,3 +1,4 @@
+//go:build windows
 // +build windows

 package winio
@@ -7,15 +8,16 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"io/ioutil"
 	"os"
 	"runtime"
-	"syscall"
 	"unicode/utf16"
+
+	"github.com/Microsoft/go-winio/internal/fs"
+	"golang.org/x/sys/windows"
 )

-//sys backupRead(h syscall.Handle, b []byte, bytesRead *uint32, abort bool, processSecurity bool, context *uintptr) (err error) = BackupRead
-//sys backupWrite(h syscall.Handle, b []byte, bytesWritten *uint32, abort bool, processSecurity bool, context *uintptr) (err error) = BackupWrite
+//sys backupRead(h windows.Handle, b []byte, bytesRead *uint32, abort bool, processSecurity bool, context *uintptr) (err error) = BackupRead
+//sys backupWrite(h windows.Handle, b []byte, bytesWritten *uint32, abort bool, processSecurity bool, context *uintptr) (err error) = BackupWrite

 const (
 	BackupData = uint32(iota + 1)
@@ -24,7 +26,7 @@ const (
 	BackupAlternateData
 	BackupLink
 	BackupPropertyData
-	BackupObjectId
+	BackupObjectId //revive:disable-line:var-naming ID, not Id
 	BackupReparseData
 	BackupSparseBlock
 	BackupTxfsData
@@ -34,14 +36,16 @@ const (
 	StreamSparseAttributes = uint32(8)
 )

+//nolint:revive // var-naming: ALL_CAPS
 const (
-	WRITE_DAC              = 0x40000
-	WRITE_OWNER            = 0x80000
-	ACCESS_SYSTEM_SECURITY = 0x1000000
+	WRITE_DAC              = windows.WRITE_DAC
+	WRITE_OWNER            = windows.WRITE_OWNER
+	ACCESS_SYSTEM_SECURITY = windows.ACCESS_SYSTEM_SECURITY
 )

 // BackupHeader represents a backup stream of a file.
 type BackupHeader struct {
+	//revive:disable-next-line:var-naming ID, not Id
 	Id         uint32 // The backup stream ID
 	Attributes uint32 // Stream attributes
 	Size       int64  // The size of the stream in bytes
@@ -49,8 +53,8 @@ type BackupHeader struct {
 	Offset     int64  // The offset of the stream in the file (for BackupSparseBlock only).
 }

-type win32StreamId struct {
-	StreamId   uint32
+type win32StreamID struct {
+	StreamID   uint32
 	Attributes uint32
 	Size       uint64
 	NameSize   uint32
@@ -71,7 +75,7 @@ func NewBackupStreamReader(r io.Reader) *BackupStreamReader {
 // Next returns the next backup stream and prepares for calls to Read(). It skips the remainder of the current stream if
 // it was not completely read.
 func (r *BackupStreamReader) Next() (*BackupHeader, error) {
-	if r.bytesLeft > 0 {
+	if r.bytesLeft > 0 { //nolint:nestif // todo: flatten this
 		if s, ok := r.r.(io.Seeker); ok {
 			// Make sure Seek on io.SeekCurrent sometimes succeeds
 			// before trying the actual seek.
@@ -82,16 +86,16 @@ func (r *BackupStreamReader) Next() (*BackupHeader, error) {
 				r.bytesLeft = 0
 			}
 		}
-		if _, err := io.Copy(ioutil.Discard, r); err != nil {
+		if _, err := io.Copy(io.Discard, r); err != nil {
 			return nil, err
 		}
 	}
-	var wsi win32StreamId
+	var wsi win32StreamID
 	if err := binary.Read(r.r, binary.LittleEndian, &wsi); err != nil {
 		return nil, err
 	}
 	hdr := &BackupHeader{
-		Id:         wsi.StreamId,
+		Id:         wsi.StreamID,
 		Attributes: wsi.Attributes,
 		Size:       int64(wsi.Size),
 	}
@@ -100,9 +104,9 @@ func (r *BackupStreamReader) Next() (*BackupHeader, error) {
 		if err := binary.Read(r.r, binary.LittleEndian, name); err != nil {
 			return nil, err
 		}
-		hdr.Name = syscall.UTF16ToString(name)
+		hdr.Name = windows.UTF16ToString(name)
 	}
-	if wsi.StreamId == BackupSparseBlock {
+	if wsi.StreamID == BackupSparseBlock {
 		if err := binary.Read(r.r, binary.LittleEndian, &hdr.Offset); err != nil {
 			return nil, err
 		}
@@ -147,8 +151,8 @@ func (w *BackupStreamWriter) WriteHeader(hdr *BackupHeader) error {
 		return fmt.Errorf("missing %d bytes", w.bytesLeft)
 	}
 	name := utf16.Encode([]rune(hdr.Name))
-	wsi := win32StreamId{
-		StreamId:   hdr.Id,
+	wsi := win32StreamID{
+		StreamID:   hdr.Id,
 		Attributes: hdr.Attributes,
 		Size:       uint64(hdr.Size),
 		NameSize:   uint32(len(name) * 2),
@@ -201,9 +205,9 @@ func NewBackupFileReader(f *os.File, includeSecurity bool) *BackupFileReader {
 // Read reads a backup stream from the file by calling the Win32 API BackupRead().
 func (r *BackupFileReader) Read(b []byte) (int, error) {
 	var bytesRead uint32
-	err := backupRead(syscall.Handle(r.f.Fd()), b, &bytesRead, false, r.includeSecurity, &r.ctx)
+	err := backupRead(windows.Handle(r.f.Fd()), b, &bytesRead, false, r.includeSecurity, &r.ctx)
 	if err != nil {
-		return 0, &os.PathError{"BackupRead", r.f.Name(), err}
+		return 0, &os.PathError{Op: "BackupRead", Path: r.f.Name(), Err: err}
 	}
 	runtime.KeepAlive(r.f)
 	if bytesRead == 0 {
@@ -216,7 +220,7 @@ func (r *BackupFileReader) Read(b []byte) (int, error) {
 // the underlying file.
 func (r *BackupFileReader) Close() error {
 	if r.ctx != 0 {
-		backupRead(syscall.Handle(r.f.Fd()), nil, nil, true, false, &r.ctx)
+		_ = backupRead(windows.Handle(r.f.Fd()), nil, nil, true, false, &r.ctx)
 		runtime.KeepAlive(r.f)
 		r.ctx = 0
 	}
@@ -240,9 +244,9 @@ func NewBackupFileWriter(f *os.File, includeSecurity bool) *BackupFileWriter {
 // Write restores a portion of the file using the provided backup stream.
 func (w *BackupFileWriter) Write(b []byte) (int, error) {
 	var bytesWritten uint32
-	err := backupWrite(syscall.Handle(w.f.Fd()), b, &bytesWritten, false, w.includeSecurity, &w.ctx)
+	err := backupWrite(windows.Handle(w.f.Fd()), b, &bytesWritten, false, w.includeSecurity, &w.ctx)
 	if err != nil {
-		return 0, &os.PathError{"BackupWrite", w.f.Name(), err}
+		return 0, &os.PathError{Op: "BackupWrite", Path: w.f.Name(), Err: err}
 	}
 	runtime.KeepAlive(w.f)
 	if int(bytesWritten) != len(b) {
@@ -255,7 +259,7 @@ func (w *BackupFileWriter) Write(b []byte) (int, error) {
 // close the underlying file.
 func (w *BackupFileWriter) Close() error {
 	if w.ctx != 0 {
-		backupWrite(syscall.Handle(w.f.Fd()), nil, nil, true, false, &w.ctx)
+		_ = backupWrite(windows.Handle(w.f.Fd()), nil, nil, true, false, &w.ctx)
 		runtime.KeepAlive(w.f)
 		w.ctx = 0
 	}
@@ -267,11 +271,14 @@ func (w *BackupFileWriter) Close() error {
 //
 // If the file opened was a directory, it cannot be used with Readdir().
 func OpenForBackup(path string, access uint32, share uint32, createmode uint32) (*os.File, error) {
-	winPath, err := syscall.UTF16FromString(path)
-	if err != nil {
-		return nil, err
-	}
-	h, err := syscall.CreateFile(&winPath[0], access, share, nil, createmode, syscall.FILE_FLAG_BACKUP_SEMANTICS|syscall.FILE_FLAG_OPEN_REPARSE_POINT, 0)
+	h, err := fs.CreateFile(path,
+		fs.AccessMask(access),
+		fs.FileShareMode(share),
+		nil,
+		fs.FileCreationDisposition(createmode),
+		fs.FILE_FLAG_BACKUP_SEMANTICS|fs.FILE_FLAG_OPEN_REPARSE_POINT,
+		0,
+	)
 	if err != nil {
 		err = &os.PathError{Op: "open", Path: path, Err: err}
 		return nil, err
@@ -0,0 +1,22 @@
+// This package provides utilities for efficiently performing Win32 IO operations in Go.
+// Currently, this package is provides support for genreal IO and management of
+//   - named pipes
+//   - files
+//   - [Hyper-V sockets]
+//
+// This code is similar to Go's [net] package, and uses IO completion ports to avoid
+// blocking IO on system threads, allowing Go to reuse the thread to schedule other goroutines.
+//
+// This limits support to Windows Vista and newer operating systems.
+//
+// Additionally, this package provides support for:
+//   - creating and managing GUIDs
+//   - writing to [ETW]
+//   - opening and manageing VHDs
+//   - parsing [Windows Image files]
+//   - auto-generating Win32 API code
+//
+// [Hyper-V sockets]: https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-guide/make-integration-service
+// [ETW]: https://docs.microsoft.com/en-us/windows-hardware/drivers/devtest/event-tracing-for-windows--etw-
+// [Windows Image files]: https://docs.microsoft.com/en-us/windows-hardware/manufacture/desktop/work-with-windows-images
+package winio
@@ -33,7 +33,7 @@ func parseEa(b []byte) (ea ExtendedAttribute, nb []byte, err error) {
 	err = binary.Read(bytes.NewReader(b), binary.LittleEndian, &info)
 	if err != nil {
 		err = errInvalidEaBuffer
-		return
+		return ea, nb, err
 	}

 	nameOffset := fileFullEaInformationSize
@@ -43,7 +43,7 @@ func parseEa(b []byte) (ea ExtendedAttribute, nb []byte, err error) {
 	nextOffset := int(info.NextEntryOffset)
 	if valueLen+valueOffset > len(b) || nextOffset < 0 || nextOffset > len(b) {
 		err = errInvalidEaBuffer
-		return
+		return ea, nb, err
 	}

 	ea.Name = string(b[nameOffset : nameOffset+nameLen])
@@ -52,7 +52,7 @@ func parseEa(b []byte) (ea ExtendedAttribute, nb []byte, err error) {
 	if info.NextEntryOffset != 0 {
 		nb = b[info.NextEntryOffset:]
 	}
-	return
+	return ea, nb, err
 }

 // DecodeExtendedAttributes decodes a list of EAs from a FILE_FULL_EA_INFORMATION
@@ -67,7 +67,7 @@ func DecodeExtendedAttributes(b []byte) (eas []ExtendedAttribute, err error) {
 		eas = append(eas, ea)
 		b = nb
 	}
-	return
+	return eas, err
 }

 func writeEa(buf *bytes.Buffer, ea *ExtendedAttribute, last bool) error {
@@ -1,3 +1,4 @@
+//go:build windows
 // +build windows

 package winio
@@ -10,31 +11,15 @@ import (
 	"sync/atomic"
 	"syscall"
 	"time"
+
+	"golang.org/x/sys/windows"
 )

-//sys cancelIoEx(file syscall.Handle, o *syscall.Overlapped) (err error) = CancelIoEx
-//sys createIoCompletionPort(file syscall.Handle, port syscall.Handle, key uintptr, threadCount uint32) (newport syscall.Handle, err error) = CreateIoCompletionPort
-//sys getQueuedCompletionStatus(port syscall.Handle, bytes *uint32, key *uintptr, o **ioOperation, timeout uint32) (err error) = GetQueuedCompletionStatus
-//sys setFileCompletionNotificationModes(h syscall.Handle, flags uint8) (err error) = SetFileCompletionNotificationModes
-//sys wsaGetOverlappedResult(h syscall.Handle, o *syscall.Overlapped, bytes *uint32, wait bool, flags *uint32) (err error) = ws2_32.WSAGetOverlappedResult
-
-type atomicBool int32
-
-func (b *atomicBool) isSet() bool { return atomic.LoadInt32((*int32)(b)) != 0 }
-func (b *atomicBool) setFalse()   { atomic.StoreInt32((*int32)(b), 0) }
-func (b *atomicBool) setTrue()    { atomic.StoreInt32((*int32)(b), 1) }
-func (b *atomicBool) swap(new bool) bool {
-	var newInt int32
-	if new {
-		newInt = 1
-	}
-	return atomic.SwapInt32((*int32)(b), newInt) == 1
-}
-
-const (
-	cFILE_SKIP_COMPLETION_PORT_ON_SUCCESS = 1
-	cFILE_SKIP_SET_EVENT_ON_HANDLE        = 2
-)
+//sys cancelIoEx(file windows.Handle, o *windows.Overlapped) (err error) = CancelIoEx
+//sys createIoCompletionPort(file windows.Handle, port windows.Handle, key uintptr, threadCount uint32) (newport windows.Handle, err error) = CreateIoCompletionPort
+//sys getQueuedCompletionStatus(port windows.Handle, bytes *uint32, key *uintptr, o **ioOperation, timeout uint32) (err error) = GetQueuedCompletionStatus
+//sys setFileCompletionNotificationModes(h windows.Handle, flags uint8) (err error) = SetFileCompletionNotificationModes
+//sys wsaGetOverlappedResult(h windows.Handle, o *windows.Overlapped, bytes *uint32, wait bool, flags *uint32) (err error) = ws2_32.WSAGetOverlappedResult

 var (
 	ErrFileClosed = errors.New("file has already been closed")
@@ -43,29 +28,29 @@ var (

 type timeoutError struct{}

-func (e *timeoutError) Error() string   { return "i/o timeout" }
-func (e *timeoutError) Timeout() bool   { return true }
-func (e *timeoutError) Temporary() bool { return true }
+func (*timeoutError) Error() string   { return "i/o timeout" }
+func (*timeoutError) Timeout() bool   { return true }
+func (*timeoutError) Temporary() bool { return true }

 type timeoutChan chan struct{}

 var ioInitOnce sync.Once
-var ioCompletionPort syscall.Handle
+var ioCompletionPort windows.Handle

-// ioResult contains the result of an asynchronous IO operation
+// ioResult contains the result of an asynchronous IO operation.
 type ioResult struct {
 	bytes uint32
 	err   error
 }

-// ioOperation represents an outstanding asynchronous Win32 IO
+// ioOperation represents an outstanding asynchronous Win32 IO.
 type ioOperation struct {
-	o  syscall.Overlapped
+	o  windows.Overlapped
 	ch chan ioResult
 }

-func initIo() {
-	h, err := createIoCompletionPort(syscall.InvalidHandle, 0, 0, 0xffffffff)
+func initIO() {
+	h, err := createIoCompletionPort(windows.InvalidHandle, 0, 0, 0xffffffff)
 	if err != nil {
 		panic(err)
 	}
@@ -76,10 +61,10 @@ func initIo() {
 // win32File implements Reader, Writer, and Closer on a Win32 handle without blocking in a syscall.
 // It takes ownership of this handle and will close it if it is garbage collected.
 type win32File struct {
-	handle        syscall.Handle
+	handle        windows.Handle
 	wg            sync.WaitGroup
 	wgLock        sync.RWMutex
-	closing       atomicBool
+	closing       atomic.Bool
 	socket        bool
 	readDeadline  deadlineHandler
 	writeDeadline deadlineHandler
@@ -90,18 +75,18 @@ type deadlineHandler struct {
 	channel     timeoutChan
 	channelLock sync.RWMutex
 	timer       *time.Timer
-	timedout    atomicBool
+	timedout    atomic.Bool
 }

-// makeWin32File makes a new win32File from an existing file handle
-func makeWin32File(h syscall.Handle) (*win32File, error) {
+// makeWin32File makes a new win32File from an existing file handle.
+func makeWin32File(h windows.Handle) (*win32File, error) {
 	f := &win32File{handle: h}
-	ioInitOnce.Do(initIo)
+	ioInitOnce.Do(initIO)
 	_, err := createIoCompletionPort(h, ioCompletionPort, 0, 0xffffffff)
 	if err != nil {
 		return nil, err
 	}
-	err = setFileCompletionNotificationModes(h, cFILE_SKIP_COMPLETION_PORT_ON_SUCCESS|cFILE_SKIP_SET_EVENT_ON_HANDLE)
+	err = setFileCompletionNotificationModes(h, windows.FILE_SKIP_COMPLETION_PORT_ON_SUCCESS|windows.FILE_SKIP_SET_EVENT_ON_HANDLE)
 	if err != nil {
 		return nil, err
 	}
@@ -110,7 +95,12 @@ func makeWin32File(h syscall.Handle) (*win32File, error) {
 	return f, nil
 }

+// Deprecated: use NewOpenFile instead.
 func MakeOpenFile(h syscall.Handle) (io.ReadWriteCloser, error) {
+	return NewOpenFile(windows.Handle(h))
+}
+
+func NewOpenFile(h windows.Handle) (io.ReadWriteCloser, error) {
 	// If we return the result of makeWin32File directly, it can result in an
 	// interface-wrapped nil, rather than a nil interface value.
 	f, err := makeWin32File(h)
@@ -120,17 +110,17 @@ func MakeOpenFile(h syscall.Handle) (io.ReadWriteCloser, error) {
 	return f, nil
 }

-// closeHandle closes the resources associated with a Win32 handle
+// closeHandle closes the resources associated with a Win32 handle.
 func (f *win32File) closeHandle() {
 	f.wgLock.Lock()
 	// Atomically set that we are closing, releasing the resources only once.
-	if !f.closing.swap(true) {
+	if !f.closing.Swap(true) {
 		f.wgLock.Unlock()
 		// cancel all IO and wait for it to complete
-		cancelIoEx(f.handle, nil)
+		_ = cancelIoEx(f.handle, nil)
 		f.wg.Wait()
 		// at this point, no new IO can start
-		syscall.Close(f.handle)
+		windows.Close(f.handle)
 		f.handle = 0
 	} else {
 		f.wgLock.Unlock()
@@ -143,11 +133,16 @@ func (f *win32File) Close() error {
 	return nil
 }

-// prepareIo prepares for a new IO operation.
+// IsClosed checks if the file has been closed.
+func (f *win32File) IsClosed() bool {
+	return f.closing.Load()
+}
+
+// prepareIO prepares for a new IO operation.
 // The caller must call f.wg.Done() when the IO is finished, prior to Close() returning.
-func (f *win32File) prepareIo() (*ioOperation, error) {
+func (f *win32File) prepareIO() (*ioOperation, error) {
 	f.wgLock.RLock()
-	if f.closing.isSet() {
+	if f.closing.Load() {
 		f.wgLock.RUnlock()
 		return nil, ErrFileClosed
 	}
@@ -158,13 +153,13 @@ func (f *win32File) prepareIo() (*ioOperation, error) {
 	return c, nil
 }

-// ioCompletionProcessor processes completed async IOs forever
-func ioCompletionProcessor(h syscall.Handle) {
+// ioCompletionProcessor processes completed async IOs forever.
+func ioCompletionProcessor(h windows.Handle) {
 	for {
 		var bytes uint32
 		var key uintptr
 		var op *ioOperation
-		err := getQueuedCompletionStatus(h, &bytes, &key, &op, syscall.INFINITE)
+		err := getQueuedCompletionStatus(h, &bytes, &key, &op, windows.INFINITE)
 		if op == nil {
 			panic(err)
 		}
@@ -172,15 +167,17 @@ func ioCompletionProcessor(h syscall.Handle) {
 	}
 }

-// asyncIo processes the return value from ReadFile or WriteFile, blocking until
+// todo: helsaawy - create an asyncIO version that takes a context
+
+// asyncIO processes the return value from ReadFile or WriteFile, blocking until
 // the operation has actually completed.
-func (f *win32File) asyncIo(c *ioOperation, d *deadlineHandler, bytes uint32, err error) (int, error) {
-	if err != syscall.ERROR_IO_PENDING {
+func (f *win32File) asyncIO(c *ioOperation, d *deadlineHandler, bytes uint32, err error) (int, error) {
+	if err != windows.ERROR_IO_PENDING { //nolint:errorlint // err is Errno
 		return int(bytes), err
 	}

-	if f.closing.isSet() {
-		cancelIoEx(f.handle, &c.o)
+	if f.closing.Load() {
+		_ = cancelIoEx(f.handle, &c.o)
 	}

 	var timeout timeoutChan
@@ -194,8 +191,8 @@ func (f *win32File) asyncIo(c *ioOperation, d *deadlineHandler, bytes uint32, er
 	select {
 	case r = <-c.ch:
 		err = r.err
-		if err == syscall.ERROR_OPERATION_ABORTED {
-			if f.closing.isSet() {
+		if err == windows.ERROR_OPERATION_ABORTED { //nolint:errorlint // err is Errno
+			if f.closing.Load() {
 				err = ErrFileClosed
 			}
 		} else if err != nil && f.socket {
@@ -204,10 +201,10 @@ func (f *win32File) asyncIo(c *ioOperation, d *deadlineHandler, bytes uint32, er
 			err = wsaGetOverlappedResult(f.handle, &c.o, &bytes, false, &flags)
 		}
 	case <-timeout:
-		cancelIoEx(f.handle, &c.o)
+		_ = cancelIoEx(f.handle, &c.o)
 		r = <-c.ch
 		err = r.err
-		if err == syscall.ERROR_OPERATION_ABORTED {
+		if err == windows.ERROR_OPERATION_ABORTED { //nolint:errorlint // err is Errno
 			err = ErrTimeout
 		}
 	}
@@ -215,52 +212,52 @@ func (f *win32File) asyncIo(c *ioOperation, d *deadlineHandler, bytes uint32, er
 	// runtime.KeepAlive is needed, as c is passed via native
 	// code to ioCompletionProcessor, c must remain alive
 	// until the channel read is complete.
+	// todo: (de)allocate *ioOperation via win32 heap functions, instead of needing to KeepAlive?
 	runtime.KeepAlive(c)
 	return int(r.bytes), err
 }

 // Read reads from a file handle.
 func (f *win32File) Read(b []byte) (int, error) {
-	c, err := f.prepareIo()
+	c, err := f.prepareIO()
 	if err != nil {
 		return 0, err
 	}
 	defer f.wg.Done()

-	if f.readDeadline.timedout.isSet() {
+	if f.readDeadline.timedout.Load() {
 		return 0, ErrTimeout
 	}

 	var bytes uint32
-	err = syscall.ReadFile(f.handle, b, &bytes, &c.o)
-	n, err := f.asyncIo(c, &f.readDeadline, bytes, err)
+	err = windows.ReadFile(f.handle, b, &bytes, &c.o)
+	n, err := f.asyncIO(c, &f.readDeadline, bytes, err)
 	runtime.KeepAlive(b)

 	// Handle EOF conditions.
 	if err == nil && n == 0 && len(b) != 0 {
 		return 0, io.EOF
-	} else if err == syscall.ERROR_BROKEN_PIPE {
+	} else if err == windows.ERROR_BROKEN_PIPE { //nolint:errorlint // err is Errno
 		return 0, io.EOF
-	} else {
-		return n, err
 	}
+	return n, err
 }

 // Write writes to a file handle.
 func (f *win32File) Write(b []byte) (int, error) {
-	c, err := f.prepareIo()
+	c, err := f.prepareIO()
 	if err != nil {
 		return 0, err
 	}
 	defer f.wg.Done()

-	if f.writeDeadline.timedout.isSet() {
+	if f.writeDeadline.timedout.Load() {
 		return 0, ErrTimeout
 	}

 	var bytes uint32
-	err = syscall.WriteFile(f.handle, b, &bytes, &c.o)
-	n, err := f.asyncIo(c, &f.writeDeadline, bytes, err)
+	err = windows.WriteFile(f.handle, b, &bytes, &c.o)
+	n, err := f.asyncIO(c, &f.writeDeadline, bytes, err)
 	runtime.KeepAlive(b)
 	return n, err
 }
@@ -274,7 +271,7 @@ func (f *win32File) SetWriteDeadline(deadline time.Time) error {
 }

 func (f *win32File) Flush() error {
-	return syscall.FlushFileBuffers(f.handle)
+	return windows.FlushFileBuffers(f.handle)
 }

 func (f *win32File) Fd() uintptr {
@@ -291,7 +288,7 @@ func (d *deadlineHandler) set(deadline time.Time) error {
 		}
 		d.timer = nil
 	}
-	d.timedout.setFalse()
+	d.timedout.Store(false)

 	select {
 	case <-d.channel:
@@ -306,7 +303,7 @@ func (d *deadlineHandler) set(deadline time.Time) error {
 	}

 	timeoutIO := func() {
-		d.timedout.setTrue()
+		d.timedout.Store(true)
 		close(d.channel)
 	}

@@ -1,3 +1,4 @@
+//go:build windows
 // +build windows

 package winio
@@ -5,44 +6,83 @@ package winio
 import (
 	"os"
 	"runtime"
-	"syscall"
 	"unsafe"
-)

-//sys getFileInformationByHandleEx(h syscall.Handle, class uint32, buffer *byte, size uint32) (err error) = GetFileInformationByHandleEx
-//sys setFileInformationByHandle(h syscall.Handle, class uint32, buffer *byte, size uint32) (err error) = SetFileInformationByHandle
-
-const (
-	fileBasicInfo = 0
-	fileIDInfo    = 0x12
+	"golang.org/x/sys/windows"
 )

 // FileBasicInfo contains file access time and file attributes information.
 type FileBasicInfo struct {
-	CreationTime, LastAccessTime, LastWriteTime, ChangeTime syscall.Filetime
+	CreationTime, LastAccessTime, LastWriteTime, ChangeTime windows.Filetime
 	FileAttributes                                          uint32
-	pad                                                     uint32 // padding
+	_                                                       uint32 // padding
+}
+
+// alignedFileBasicInfo is a FileBasicInfo, but aligned to uint64 by containing
+// uint64 rather than windows.Filetime. Filetime contains two uint32s. uint64
+// alignment is necessary to pass this as FILE_BASIC_INFO.
+type alignedFileBasicInfo struct {
+	CreationTime, LastAccessTime, LastWriteTime, ChangeTime uint64
+	FileAttributes                                          uint32
+	_                                                       uint32 // padding
 }

 // GetFileBasicInfo retrieves times and attributes for a file.
 func GetFileBasicInfo(f *os.File) (*FileBasicInfo, error) {
-	bi := &FileBasicInfo{}
-	if err := getFileInformationByHandleEx(syscall.Handle(f.Fd()), fileBasicInfo, (*byte)(unsafe.Pointer(bi)), uint32(unsafe.Sizeof(*bi))); err != nil {
+	bi := &alignedFileBasicInfo{}
+	if err := windows.GetFileInformationByHandleEx(
+		windows.Handle(f.Fd()),
+		windows.FileBasicInfo,
+		(*byte)(unsafe.Pointer(bi)),
+		uint32(unsafe.Sizeof(*bi)),
+	); err != nil {
 		return nil, &os.PathError{Op: "GetFileInformationByHandleEx", Path: f.Name(), Err: err}
 	}
 	runtime.KeepAlive(f)
-	return bi, nil
+	// Reinterpret the alignedFileBasicInfo as a FileBasicInfo so it matches the
+	// public API of this module. The data may be unnecessarily aligned.
+	return (*FileBasicInfo)(unsafe.Pointer(bi)), nil
 }

 // SetFileBasicInfo sets times and attributes for a file.
 func SetFileBasicInfo(f *os.File, bi *FileBasicInfo) error {
-	if err := setFileInformationByHandle(syscall.Handle(f.Fd()), fileBasicInfo, (*byte)(unsafe.Pointer(bi)), uint32(unsafe.Sizeof(*bi))); err != nil {
+	// Create an alignedFileBasicInfo based on a FileBasicInfo. The copy is
+	// suitable to pass to GetFileInformationByHandleEx.
+	biAligned := *(*alignedFileBasicInfo)(unsafe.Pointer(bi))
+	if err := windows.SetFileInformationByHandle(
+		windows.Handle(f.Fd()),
+		windows.FileBasicInfo,
+		(*byte)(unsafe.Pointer(&biAligned)),
+		uint32(unsafe.Sizeof(biAligned)),
+	); err != nil {
 		return &os.PathError{Op: "SetFileInformationByHandle", Path: f.Name(), Err: err}
 	}
 	runtime.KeepAlive(f)
 	return nil
 }

+// FileStandardInfo contains extended information for the file.
+// FILE_STANDARD_INFO in WinBase.h
+// https://docs.microsoft.com/en-us/windows/win32/api/winbase/ns-winbase-file_standard_info
+type FileStandardInfo struct {
+	AllocationSize, EndOfFile int64
+	NumberOfLinks             uint32
+	DeletePending, Directory  bool
+}
+
+// GetFileStandardInfo retrieves ended information for the file.
+func GetFileStandardInfo(f *os.File) (*FileStandardInfo, error) {
+	si := &FileStandardInfo{}
+	if err := windows.GetFileInformationByHandleEx(windows.Handle(f.Fd()),
+		windows.FileStandardInfo,
+		(*byte)(unsafe.Pointer(si)),
+		uint32(unsafe.Sizeof(*si))); err != nil {
+		return nil, &os.PathError{Op: "GetFileInformationByHandleEx", Path: f.Name(), Err: err}
+	}
+	runtime.KeepAlive(f)
+	return si, nil
+}
+
 // FileIDInfo contains the volume serial number and file ID for a file. This pair should be
 // unique on a system.
 type FileIDInfo struct {
@@ -53,7 +93,12 @@ type FileIDInfo struct {
 // GetFileID retrieves the unique (volume, file ID) pair for a file.
 func GetFileID(f *os.File) (*FileIDInfo, error) {
 	fileID := &FileIDInfo{}
-	if err := getFileInformationByHandleEx(syscall.Handle(f.Fd()), fileIDInfo, (*byte)(unsafe.Pointer(fileID)), uint32(unsafe.Sizeof(*fileID))); err != nil {
+	if err := windows.GetFileInformationByHandleEx(
+		windows.Handle(f.Fd()),
+		windows.FileIdInfo,
+		(*byte)(unsafe.Pointer(fileID)),
+		uint32(unsafe.Sizeof(*fileID)),
+	); err != nil {
 		return nil, &os.PathError{Op: "GetFileInformationByHandleEx", Path: f.Name(), Err: err}
 	}
 	runtime.KeepAlive(f)
@@ -1,24 +1,99 @@
+//go:build windows
+// +build windows
+
 package winio

 import (
+	"context"
+	"errors"
 	"fmt"
 	"io"
 	"net"
 	"os"
-	"syscall"
 	"time"
 	"unsafe"

+	"golang.org/x/sys/windows"
+
+	"github.com/Microsoft/go-winio/internal/socket"
 	"github.com/Microsoft/go-winio/pkg/guid"
 )

-//sys bind(s syscall.Handle, name unsafe.Pointer, namelen int32) (err error) [failretval==socketError] = ws2_32.bind
+const afHVSock = 34 // AF_HYPERV

-const (
-	afHvSock = 34 // AF_HYPERV
+// Well known Service and VM IDs
+// https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-guide/make-integration-service#vmid-wildcards

-	socketError = ^uintptr(0)
-)
+// HvsockGUIDWildcard is the wildcard VmId for accepting connections from all partitions.
+func HvsockGUIDWildcard() guid.GUID { // 00000000-0000-0000-0000-000000000000
+	return guid.GUID{}
+}
+
+// HvsockGUIDBroadcast is the wildcard VmId for broadcasting sends to all partitions.
+func HvsockGUIDBroadcast() guid.GUID { // ffffffff-ffff-ffff-ffff-ffffffffffff
+	return guid.GUID{
+		Data1: 0xffffffff,
+		Data2: 0xffff,
+		Data3: 0xffff,
+		Data4: [8]uint8{0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
+	}
+}
+
+// HvsockGUIDLoopback is the Loopback VmId for accepting connections to the same partition as the connector.
+func HvsockGUIDLoopback() guid.GUID { // e0e16197-dd56-4a10-9195-5ee7a155a838
+	return guid.GUID{
+		Data1: 0xe0e16197,
+		Data2: 0xdd56,
+		Data3: 0x4a10,
+		Data4: [8]uint8{0x91, 0x95, 0x5e, 0xe7, 0xa1, 0x55, 0xa8, 0x38},
+	}
+}
+
+// HvsockGUIDSiloHost is the address of a silo's host partition:
+//   - The silo host of a hosted silo is the utility VM.
+//   - The silo host of a silo on a physical host is the physical host.
+func HvsockGUIDSiloHost() guid.GUID { // 36bd0c5c-7276-4223-88ba-7d03b654c568
+	return guid.GUID{
+		Data1: 0x36bd0c5c,
+		Data2: 0x7276,
+		Data3: 0x4223,
+		Data4: [8]byte{0x88, 0xba, 0x7d, 0x03, 0xb6, 0x54, 0xc5, 0x68},
+	}
+}
+
+// HvsockGUIDChildren is the wildcard VmId for accepting connections from the connector's child partitions.
+func HvsockGUIDChildren() guid.GUID { // 90db8b89-0d35-4f79-8ce9-49ea0ac8b7cd
+	return guid.GUID{
+		Data1: 0x90db8b89,
+		Data2: 0xd35,
+		Data3: 0x4f79,
+		Data4: [8]uint8{0x8c, 0xe9, 0x49, 0xea, 0xa, 0xc8, 0xb7, 0xcd},
+	}
+}
+
+// HvsockGUIDParent is the wildcard VmId for accepting connections from the connector's parent partition.
+// Listening on this VmId accepts connection from:
+//   - Inside silos: silo host partition.
+//   - Inside hosted silo: host of the VM.
+//   - Inside VM: VM host.
+//   - Physical host: Not supported.
+func HvsockGUIDParent() guid.GUID { // a42e7cda-d03f-480c-9cc2-a4de20abb878
+	return guid.GUID{
+		Data1: 0xa42e7cda,
+		Data2: 0xd03f,
+		Data3: 0x480c,
+		Data4: [8]uint8{0x9c, 0xc2, 0xa4, 0xde, 0x20, 0xab, 0xb8, 0x78},
+	}
+}
+
+// hvsockVsockServiceTemplate is the Service GUID used for the VSOCK protocol.
+func hvsockVsockServiceTemplate() guid.GUID { // 00000000-facb-11e6-bd58-64006a7986d3
+	return guid.GUID{
+		Data2: 0xfacb,
+		Data3: 0x11e6,
+		Data4: [8]uint8{0xbd, 0x58, 0x64, 0x00, 0x6a, 0x79, 0x86, 0xd3},
+	}
+}

 // An HvsockAddr is an address for a AF_HYPERV socket.
 type HvsockAddr struct {
@@ -33,8 +108,10 @@ type rawHvsockAddr struct {
 	ServiceID guid.GUID
 }

+var _ socket.RawSockaddr = &rawHvsockAddr{}
+
 // Network returns the address's network name, "hvsock".
-func (addr *HvsockAddr) Network() string {
+func (*HvsockAddr) Network() string {
 	return "hvsock"
 }

@@ -44,14 +121,14 @@ func (addr *HvsockAddr) String() string {

 // VsockServiceID returns an hvsock service ID corresponding to the specified AF_VSOCK port.
 func VsockServiceID(port uint32) guid.GUID {
-	g, _ := guid.FromString("00000000-facb-11e6-bd58-64006a7986d3")
+	g := hvsockVsockServiceTemplate() // make a copy
 	g.Data1 = port
 	return g
 }

 func (addr *HvsockAddr) raw() rawHvsockAddr {
 	return rawHvsockAddr{
-		Family:    afHvSock,
+		Family:    afHVSock,
 		VMID:      addr.VMID,
 		ServiceID: addr.ServiceID,
 	}
@@ -62,26 +139,54 @@ func (addr *HvsockAddr) fromRaw(raw *rawHvsockAddr) {
 	addr.ServiceID = raw.ServiceID
 }

+// Sockaddr returns a pointer to and the size of this struct.
+//
+// Implements the [socket.RawSockaddr] interface, and allows use in
+// [socket.Bind] and [socket.ConnectEx].
+func (r *rawHvsockAddr) Sockaddr() (unsafe.Pointer, int32, error) {
+	return unsafe.Pointer(r), int32(unsafe.Sizeof(rawHvsockAddr{})), nil
+}
+
+// Sockaddr interface allows use with `sockets.Bind()` and `.ConnectEx()`.
+func (r *rawHvsockAddr) FromBytes(b []byte) error {
+	n := int(unsafe.Sizeof(rawHvsockAddr{}))
+
+	if len(b) < n {
+		return fmt.Errorf("got %d, want %d: %w", len(b), n, socket.ErrBufferSize)
+	}
+
+	copy(unsafe.Slice((*byte)(unsafe.Pointer(r)), n), b[:n])
+	if r.Family != afHVSock {
+		return fmt.Errorf("got %d, want %d: %w", r.Family, afHVSock, socket.ErrAddrFamily)
+	}
+
+	return nil
+}
+
 // HvsockListener is a socket listener for the AF_HYPERV address family.
 type HvsockListener struct {
 	sock *win32File
 	addr HvsockAddr
 }

+var _ net.Listener = &HvsockListener{}
+
 // HvsockConn is a connected socket of the AF_HYPERV address family.
 type HvsockConn struct {
 	sock          *win32File
 	local, remote HvsockAddr
 }

-func newHvSocket() (*win32File, error) {
-	fd, err := syscall.Socket(afHvSock, syscall.SOCK_STREAM, 1)
+var _ net.Conn = &HvsockConn{}
+
+func newHVSocket() (*win32File, error) {
+	fd, err := windows.Socket(afHVSock, windows.SOCK_STREAM, 1)
 	if err != nil {
 		return nil, os.NewSyscallError("socket", err)
 	}
 	f, err := makeWin32File(fd)
 	if err != nil {
-		syscall.Close(fd)
+		windows.Close(fd)
 		return nil, err
 	}
 	f.socket = true
@@ -91,16 +196,24 @@ func newHvSocket() (*win32File, error) {
 // ListenHvsock listens for connections on the specified hvsock address.
 func ListenHvsock(addr *HvsockAddr) (_ *HvsockListener, err error) {
 	l := &HvsockListener{addr: *addr}
-	sock, err := newHvSocket()
+
+	var sock *win32File
+	sock, err = newHVSocket()
 	if err != nil {
 		return nil, l.opErr("listen", err)
 	}
+	defer func() {
+		if err != nil {
+			_ = sock.Close()
+		}
+	}()
+
 	sa := addr.raw()
-	err = bind(sock.handle, unsafe.Pointer(&sa), int32(unsafe.Sizeof(sa)))
+	err = socket.Bind(sock.handle, &sa)
 	if err != nil {
 		return nil, l.opErr("listen", os.NewSyscallError("socket", err))
 	}
-	err = syscall.Listen(sock.handle, 16)
+	err = windows.Listen(sock.handle, 16)
 	if err != nil {
 		return nil, l.opErr("listen", os.NewSyscallError("listen", err))
 	}
@@ -118,7 +231,7 @@ func (l *HvsockListener) Addr() net.Addr {

 // Accept waits for the next connection and returns it.
 func (l *HvsockListener) Accept() (_ net.Conn, err error) {
-	sock, err := newHvSocket()
+	sock, err := newHVSocket()
 	if err != nil {
 		return nil, l.opErr("accept", err)
 	}
@@ -127,27 +240,42 @@ func (l *HvsockListener) Accept() (_ net.Conn, err error) {
 			sock.Close()
 		}
 	}()
-	c, err := l.sock.prepareIo()
+	c, err := l.sock.prepareIO()
 	if err != nil {
 		return nil, l.opErr("accept", err)
 	}
 	defer l.sock.wg.Done()

 	// AcceptEx, per documentation, requires an extra 16 bytes per address.
+	//
+	// https://docs.microsoft.com/en-us/windows/win32/api/mswsock/nf-mswsock-acceptex
 	const addrlen = uint32(16 + unsafe.Sizeof(rawHvsockAddr{}))
 	var addrbuf [addrlen * 2]byte

 	var bytes uint32
-	err = syscall.AcceptEx(l.sock.handle, sock.handle, &addrbuf[0], 0, addrlen, addrlen, &bytes, &c.o)
-	_, err = l.sock.asyncIo(c, nil, bytes, err)
-	if err != nil {
+	err = windows.AcceptEx(l.sock.handle, sock.handle, &addrbuf[0], 0 /* rxdatalen */, addrlen, addrlen, &bytes, &c.o)
+	if _, err = l.sock.asyncIO(c, nil, bytes, err); err != nil {
 		return nil, l.opErr("accept", os.NewSyscallError("acceptex", err))
 	}
+
 	conn := &HvsockConn{
 		sock: sock,
 	}
+	// The local address returned in the AcceptEx buffer is the same as the Listener socket's
+	// address. However, the service GUID reported by GetSockName is different from the Listeners
+	// socket, and is sometimes the same as the local address of the socket that dialed the
+	// address, with the service GUID.Data1 incremented, but othertimes is different.
+	// todo: does the local address matter? is the listener's address or the actual address appropriate?
 	conn.local.fromRaw((*rawHvsockAddr)(unsafe.Pointer(&addrbuf[0])))
 	conn.remote.fromRaw((*rawHvsockAddr)(unsafe.Pointer(&addrbuf[addrlen])))
+
+	// initialize the accepted socket and update its properties with those of the listening socket
+	if err = windows.Setsockopt(sock.handle,
+		windows.SOL_SOCKET, windows.SO_UPDATE_ACCEPT_CONTEXT,
+		(*byte)(unsafe.Pointer(&l.sock.handle)), int32(unsafe.Sizeof(l.sock.handle))); err != nil {
+		return nil, conn.opErr("accept", os.NewSyscallError("setsockopt", err))
+	}
+
 	sock = nil
 	return conn, nil
 }
@@ -157,54 +285,183 @@ func (l *HvsockListener) Close() error {
 	return l.sock.Close()
 }

-/* Need to finish ConnectEx handling
-func DialHvsock(ctx context.Context, addr *HvsockAddr) (*HvsockConn, error) {
-	sock, err := newHvSocket()
+// HvsockDialer configures and dials a Hyper-V Socket (ie, [HvsockConn]).
+type HvsockDialer struct {
+	// Deadline is the time the Dial operation must connect before erroring.
+	Deadline time.Time
+
+	// Retries is the number of additional connects to try if the connection times out, is refused,
+	// or the host is unreachable
+	Retries uint
+
+	// RetryWait is the time to wait after a connection error to retry
+	RetryWait time.Duration
+
+	rt *time.Timer // redial wait timer
+}
+
+// Dial the Hyper-V socket at addr.
+//
+// See [HvsockDialer.Dial] for more information.
+func Dial(ctx context.Context, addr *HvsockAddr) (conn *HvsockConn, err error) {
+	return (&HvsockDialer{}).Dial(ctx, addr)
+}
+
+// Dial attempts to connect to the Hyper-V socket at addr, and returns a connection if successful.
+// Will attempt (HvsockDialer).Retries if dialing fails, waiting (HvsockDialer).RetryWait between
+// retries.
+//
+// Dialing can be cancelled either by providing (HvsockDialer).Deadline, or cancelling ctx.
+func (d *HvsockDialer) Dial(ctx context.Context, addr *HvsockAddr) (conn *HvsockConn, err error) {
+	op := "dial"
+	// create the conn early to use opErr()
+	conn = &HvsockConn{
+		remote: *addr,
+	}
+
+	if !d.Deadline.IsZero() {
+		var cancel context.CancelFunc
+		ctx, cancel = context.WithDeadline(ctx, d.Deadline)
+		defer cancel()
+	}
+
+	// preemptive timeout/cancellation check
+	if err = ctx.Err(); err != nil {
+		return nil, conn.opErr(op, err)
+	}
+
+	sock, err := newHVSocket()
 	if err != nil {
-		return nil, err
+		return nil, conn.opErr(op, err)
 	}
 	defer func() {
 		if sock != nil {
 			sock.Close()
 		}
 	}()
-	c, err := sock.prepareIo()
+
+	sa := addr.raw()
+	err = socket.Bind(sock.handle, &sa)
 	if err != nil {
-		return nil, err
+		return nil, conn.opErr(op, os.NewSyscallError("bind", err))
+	}
+
+	c, err := sock.prepareIO()
+	if err != nil {
+		return nil, conn.opErr(op, err)
 	}
 	defer sock.wg.Done()
 	var bytes uint32
-	err = windows.ConnectEx(windows.Handle(sock.handle), sa, nil, 0, &bytes, &c.o)
-	_, err = sock.asyncIo(ctx, c, nil, bytes, err)
+	for i := uint(0); i <= d.Retries; i++ {
+		err = socket.ConnectEx(
+			sock.handle,
+			&sa,
+			nil, // sendBuf
+			0,   // sendDataLen
+			&bytes,
+			(*windows.Overlapped)(unsafe.Pointer(&c.o)))
+		_, err = sock.asyncIO(c, nil, bytes, err)
+		if i < d.Retries && canRedial(err) {
+			if err = d.redialWait(ctx); err == nil {
+				continue
+			}
+		}
+		break
+	}
 	if err != nil {
-		return nil, err
+		return nil, conn.opErr(op, os.NewSyscallError("connectex", err))
 	}
-	conn := &HvsockConn{
-		sock:   sock,
-		remote: *addr,
+
+	// update the connection properties, so shutdown can be used
+	if err = windows.Setsockopt(
+		sock.handle,
+		windows.SOL_SOCKET,
+		windows.SO_UPDATE_CONNECT_CONTEXT,
+		nil, // optvalue
+		0,   // optlen
+	); err != nil {
+		return nil, conn.opErr(op, os.NewSyscallError("setsockopt", err))
 	}
+
+	// get the local name
+	var sal rawHvsockAddr
+	err = socket.GetSockName(sock.handle, &sal)
+	if err != nil {
+		return nil, conn.opErr(op, os.NewSyscallError("getsockname", err))
+	}
+	conn.local.fromRaw(&sal)
+
+	// one last check for timeout, since asyncIO doesn't check the context
+	if err = ctx.Err(); err != nil {
+		return nil, conn.opErr(op, err)
+	}
+
+	conn.sock = sock
 	sock = nil
+
 	return conn, nil
 }
-*/
+
+// redialWait waits before attempting to redial, resetting the timer as appropriate.
+func (d *HvsockDialer) redialWait(ctx context.Context) (err error) {
+	if d.RetryWait == 0 {
+		return nil
+	}
+
+	if d.rt == nil {
+		d.rt = time.NewTimer(d.RetryWait)
+	} else {
+		// should already be stopped and drained
+		d.rt.Reset(d.RetryWait)
+	}
+
+	select {
+	case <-ctx.Done():
+	case <-d.rt.C:
+		return nil
+	}
+
+	// stop and drain the timer
+	if !d.rt.Stop() {
+		<-d.rt.C
+	}
+	return ctx.Err()
+}
+
+// assumes error is a plain, unwrapped windows.Errno provided by direct syscall.
+func canRedial(err error) bool {
+	//nolint:errorlint // guaranteed to be an Errno
+	switch err {
+	case windows.WSAECONNREFUSED, windows.WSAENETUNREACH, windows.WSAETIMEDOUT,
+		windows.ERROR_CONNECTION_REFUSED, windows.ERROR_CONNECTION_UNAVAIL:
+		return true
+	default:
+		return false
+	}
+}

 func (conn *HvsockConn) opErr(op string, err error) error {
+	// translate from "file closed" to "socket closed"
+	if errors.Is(err, ErrFileClosed) {
+		err = socket.ErrSocketClosed
+	}
 	return &net.OpError{Op: op, Net: "hvsock", Source: &conn.local, Addr: &conn.remote, Err: err}
 }

 func (conn *HvsockConn) Read(b []byte) (int, error) {
-	c, err := conn.sock.prepareIo()
+	c, err := conn.sock.prepareIO()
 	if err != nil {
 		return 0, conn.opErr("read", err)
 	}
 	defer conn.sock.wg.Done()
-	buf := syscall.WSABuf{Buf: &b[0], Len: uint32(len(b))}
+	buf := windows.WSABuf{Buf: &b[0], Len: uint32(len(b))}
 	var flags, bytes uint32
-	err = syscall.WSARecv(conn.sock.handle, &buf, 1, &bytes, &flags, &c.o, nil)
-	n, err := conn.sock.asyncIo(c, &conn.sock.readDeadline, bytes, err)
+	err = windows.WSARecv(conn.sock.handle, &buf, 1, &bytes, &flags, &c.o, nil)
+	n, err := conn.sock.asyncIO(c, &conn.sock.readDeadline, bytes, err)
 	if err != nil {
-		if _, ok := err.(syscall.Errno); ok {
-			err = os.NewSyscallError("wsarecv", err)
+		var eno windows.Errno
+		if errors.As(err, &eno) {
+			err = os.NewSyscallError("wsarecv", eno)
 		}
 		return 0, conn.opErr("read", err)
 	} else if n == 0 {
@@ -227,18 +484,19 @@ func (conn *HvsockConn) Write(b []byte) (int, error) {
 }

 func (conn *HvsockConn) write(b []byte) (int, error) {
-	c, err := conn.sock.prepareIo()
+	c, err := conn.sock.prepareIO()
 	if err != nil {
 		return 0, conn.opErr("write", err)
 	}
 	defer conn.sock.wg.Done()
-	buf := syscall.WSABuf{Buf: &b[0], Len: uint32(len(b))}
+	buf := windows.WSABuf{Buf: &b[0], Len: uint32(len(b))}
 	var bytes uint32
-	err = syscall.WSASend(conn.sock.handle, &buf, 1, &bytes, 0, &c.o, nil)
-	n, err := conn.sock.asyncIo(c, &conn.sock.writeDeadline, bytes, err)
+	err = windows.WSASend(conn.sock.handle, &buf, 1, &bytes, 0, &c.o, nil)
+	n, err := conn.sock.asyncIO(c, &conn.sock.writeDeadline, bytes, err)
 	if err != nil {
-		if _, ok := err.(syscall.Errno); ok {
-			err = os.NewSyscallError("wsasend", err)
+		var eno windows.Errno
+		if errors.As(err, &eno) {
+			err = os.NewSyscallError("wsasend", eno)
 		}
 		return 0, conn.opErr("write", err)
 	}
@@ -250,29 +508,43 @@ func (conn *HvsockConn) Close() error {
 	return conn.sock.Close()
 }

+func (conn *HvsockConn) IsClosed() bool {
+	return conn.sock.IsClosed()
+}
+
+// shutdown disables sending or receiving on a socket.
 func (conn *HvsockConn) shutdown(how int) error {
-	err := syscall.Shutdown(conn.sock.handle, syscall.SHUT_RD)
+	if conn.IsClosed() {
+		return socket.ErrSocketClosed
+	}
+
+	err := windows.Shutdown(conn.sock.handle, how)
 	if err != nil {
+		// If the connection was closed, shutdowns fail with "not connected"
+		if errors.Is(err, windows.WSAENOTCONN) ||
+			errors.Is(err, windows.WSAESHUTDOWN) {
+			err = socket.ErrSocketClosed
+		}
 		return os.NewSyscallError("shutdown", err)
 	}
 	return nil
 }

-// CloseRead shuts down the read end of the socket.
+// CloseRead shuts down the read end of the socket, preventing future read operations.
 func (conn *HvsockConn) CloseRead() error {
-	err := conn.shutdown(syscall.SHUT_RD)
+	err := conn.shutdown(windows.SHUT_RD)
 	if err != nil {
-		return conn.opErr("close", err)
+		return conn.opErr("closeread", err)
 	}
 	return nil
 }

-// CloseWrite shuts down the write end of the socket, notifying the other endpoint that
-// no more data will be written.
+// CloseWrite shuts down the write end of the socket, preventing future write operations and
+// notifying the other endpoint that no more data will be written.
 func (conn *HvsockConn) CloseWrite() error {
-	err := conn.shutdown(syscall.SHUT_WR)
+	err := conn.shutdown(windows.SHUT_WR)
 	if err != nil {
-		return conn.opErr("close", err)
+		return conn.opErr("closewrite", err)
 	}
 	return nil
 }
@@ -289,8 +561,13 @@ func (conn *HvsockConn) RemoteAddr() net.Addr {

 // SetDeadline implements the net.Conn SetDeadline method.
 func (conn *HvsockConn) SetDeadline(t time.Time) error {
-	conn.SetReadDeadline(t)
-	conn.SetWriteDeadline(t)
+	// todo: implement `SetDeadline` for `win32File`
+	if err := conn.SetReadDeadline(t); err != nil {
+		return fmt.Errorf("set read deadline: %w", err)
+	}
+	if err := conn.SetWriteDeadline(t); err != nil {
+		return fmt.Errorf("set write deadline: %w", err)
+	}
 	return nil
 }

@@ -0,0 +1,2 @@
+// This package contains Win32 filesystem functionality.
+package fs
@@ -0,0 +1,262 @@
+//go:build windows
+
+package fs
+
+import (
+	"golang.org/x/sys/windows"
+
+	"github.com/Microsoft/go-winio/internal/stringbuffer"
+)
+
+//go:generate go run github.com/Microsoft/go-winio/tools/mkwinsyscall -output zsyscall_windows.go fs.go
+
+// https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-createfilew
+//sys CreateFile(name string, access AccessMask, mode FileShareMode, sa *windows.SecurityAttributes, createmode FileCreationDisposition, attrs FileFlagOrAttribute, templatefile windows.Handle) (handle windows.Handle, err error) [failretval==windows.InvalidHandle] = CreateFileW
+
+const NullHandle windows.Handle = 0
+
+// AccessMask defines standard, specific, and generic rights.
+//
+// Used with CreateFile and NtCreateFile (and co.).
+//
+//	Bitmask:
+//	 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
+//	 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+//	+---------------+---------------+-------------------------------+
+//	|G|G|G|G|Resvd|A| StandardRights|         SpecificRights        |
+//	|R|W|E|A|     |S|               |                               |
+//	+-+-------------+---------------+-------------------------------+
+//
+//	GR     Generic Read
+//	GW     Generic Write
+//	GE     Generic Exectue
+//	GA     Generic All
+//	Resvd  Reserved
+//	AS     Access Security System
+//
+// https://learn.microsoft.com/en-us/windows/win32/secauthz/access-mask
+//
+// https://learn.microsoft.com/en-us/windows/win32/secauthz/generic-access-rights
+//
+// https://learn.microsoft.com/en-us/windows/win32/fileio/file-access-rights-constants
+type AccessMask = windows.ACCESS_MASK
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// Not actually any.
+	//
+	// For CreateFile: "query certain metadata such as file, directory, or device attributes without accessing that file or device"
+	// https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-createfilew#parameters
+	FILE_ANY_ACCESS AccessMask = 0
+
+	GENERIC_READ           AccessMask = 0x8000_0000
+	GENERIC_WRITE          AccessMask = 0x4000_0000
+	GENERIC_EXECUTE        AccessMask = 0x2000_0000
+	GENERIC_ALL            AccessMask = 0x1000_0000
+	ACCESS_SYSTEM_SECURITY AccessMask = 0x0100_0000
+
+	// Specific Object Access
+	// from ntioapi.h
+
+	FILE_READ_DATA      AccessMask = (0x0001) // file & pipe
+	FILE_LIST_DIRECTORY AccessMask = (0x0001) // directory
+
+	FILE_WRITE_DATA AccessMask = (0x0002) // file & pipe
+	FILE_ADD_FILE   AccessMask = (0x0002) // directory
+
+	FILE_APPEND_DATA          AccessMask = (0x0004) // file
+	FILE_ADD_SUBDIRECTORY     AccessMask = (0x0004) // directory
+	FILE_CREATE_PIPE_INSTANCE AccessMask = (0x0004) // named pipe
+
+	FILE_READ_EA         AccessMask = (0x0008) // file & directory
+	FILE_READ_PROPERTIES AccessMask = FILE_READ_EA
+
+	FILE_WRITE_EA         AccessMask = (0x0010) // file & directory
+	FILE_WRITE_PROPERTIES AccessMask = FILE_WRITE_EA
+
+	FILE_EXECUTE  AccessMask = (0x0020) // file
+	FILE_TRAVERSE AccessMask = (0x0020) // directory
+
+	FILE_DELETE_CHILD AccessMask = (0x0040) // directory
+
+	FILE_READ_ATTRIBUTES AccessMask = (0x0080) // all
+
+	FILE_WRITE_ATTRIBUTES AccessMask = (0x0100) // all
+
+	FILE_ALL_ACCESS      AccessMask = (STANDARD_RIGHTS_REQUIRED | SYNCHRONIZE | 0x1FF)
+	FILE_GENERIC_READ    AccessMask = (STANDARD_RIGHTS_READ | FILE_READ_DATA | FILE_READ_ATTRIBUTES | FILE_READ_EA | SYNCHRONIZE)
+	FILE_GENERIC_WRITE   AccessMask = (STANDARD_RIGHTS_WRITE | FILE_WRITE_DATA | FILE_WRITE_ATTRIBUTES | FILE_WRITE_EA | FILE_APPEND_DATA | SYNCHRONIZE)
+	FILE_GENERIC_EXECUTE AccessMask = (STANDARD_RIGHTS_EXECUTE | FILE_READ_ATTRIBUTES | FILE_EXECUTE | SYNCHRONIZE)
+
+	SPECIFIC_RIGHTS_ALL AccessMask = 0x0000FFFF
+
+	// Standard Access
+	// from ntseapi.h
+
+	DELETE       AccessMask = 0x0001_0000
+	READ_CONTROL AccessMask = 0x0002_0000
+	WRITE_DAC    AccessMask = 0x0004_0000
+	WRITE_OWNER  AccessMask = 0x0008_0000
+	SYNCHRONIZE  AccessMask = 0x0010_0000
+
+	STANDARD_RIGHTS_REQUIRED AccessMask = 0x000F_0000
+
+	STANDARD_RIGHTS_READ    AccessMask = READ_CONTROL
+	STANDARD_RIGHTS_WRITE   AccessMask = READ_CONTROL
+	STANDARD_RIGHTS_EXECUTE AccessMask = READ_CONTROL
+
+	STANDARD_RIGHTS_ALL AccessMask = 0x001F_0000
+)
+
+type FileShareMode uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	FILE_SHARE_NONE        FileShareMode = 0x00
+	FILE_SHARE_READ        FileShareMode = 0x01
+	FILE_SHARE_WRITE       FileShareMode = 0x02
+	FILE_SHARE_DELETE      FileShareMode = 0x04
+	FILE_SHARE_VALID_FLAGS FileShareMode = 0x07
+)
+
+type FileCreationDisposition uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// from winbase.h
+
+	CREATE_NEW        FileCreationDisposition = 0x01
+	CREATE_ALWAYS     FileCreationDisposition = 0x02
+	OPEN_EXISTING     FileCreationDisposition = 0x03
+	OPEN_ALWAYS       FileCreationDisposition = 0x04
+	TRUNCATE_EXISTING FileCreationDisposition = 0x05
+)
+
+// Create disposition values for NtCreate*
+type NTFileCreationDisposition uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// From ntioapi.h
+
+	FILE_SUPERSEDE           NTFileCreationDisposition = 0x00
+	FILE_OPEN                NTFileCreationDisposition = 0x01
+	FILE_CREATE              NTFileCreationDisposition = 0x02
+	FILE_OPEN_IF             NTFileCreationDisposition = 0x03
+	FILE_OVERWRITE           NTFileCreationDisposition = 0x04
+	FILE_OVERWRITE_IF        NTFileCreationDisposition = 0x05
+	FILE_MAXIMUM_DISPOSITION NTFileCreationDisposition = 0x05
+)
+
+// CreateFile and co. take flags or attributes together as one parameter.
+// Define alias until we can use generics to allow both
+//
+// https://learn.microsoft.com/en-us/windows/win32/fileio/file-attribute-constants
+type FileFlagOrAttribute uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// from winnt.h
+
+	FILE_FLAG_WRITE_THROUGH       FileFlagOrAttribute = 0x8000_0000
+	FILE_FLAG_OVERLAPPED          FileFlagOrAttribute = 0x4000_0000
+	FILE_FLAG_NO_BUFFERING        FileFlagOrAttribute = 0x2000_0000
+	FILE_FLAG_RANDOM_ACCESS       FileFlagOrAttribute = 0x1000_0000
+	FILE_FLAG_SEQUENTIAL_SCAN     FileFlagOrAttribute = 0x0800_0000
+	FILE_FLAG_DELETE_ON_CLOSE     FileFlagOrAttribute = 0x0400_0000
+	FILE_FLAG_BACKUP_SEMANTICS    FileFlagOrAttribute = 0x0200_0000
+	FILE_FLAG_POSIX_SEMANTICS     FileFlagOrAttribute = 0x0100_0000
+	FILE_FLAG_OPEN_REPARSE_POINT  FileFlagOrAttribute = 0x0020_0000
+	FILE_FLAG_OPEN_NO_RECALL      FileFlagOrAttribute = 0x0010_0000
+	FILE_FLAG_FIRST_PIPE_INSTANCE FileFlagOrAttribute = 0x0008_0000
+)
+
+// NtCreate* functions take a dedicated CreateOptions parameter.
+//
+// https://learn.microsoft.com/en-us/windows/win32/api/Winternl/nf-winternl-ntcreatefile
+//
+// https://learn.microsoft.com/en-us/windows/win32/devnotes/nt-create-named-pipe-file
+type NTCreateOptions uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// From ntioapi.h
+
+	FILE_DIRECTORY_FILE            NTCreateOptions = 0x0000_0001
+	FILE_WRITE_THROUGH             NTCreateOptions = 0x0000_0002
+	FILE_SEQUENTIAL_ONLY           NTCreateOptions = 0x0000_0004
+	FILE_NO_INTERMEDIATE_BUFFERING NTCreateOptions = 0x0000_0008
+
+	FILE_SYNCHRONOUS_IO_ALERT    NTCreateOptions = 0x0000_0010
+	FILE_SYNCHRONOUS_IO_NONALERT NTCreateOptions = 0x0000_0020
+	FILE_NON_DIRECTORY_FILE      NTCreateOptions = 0x0000_0040
+	FILE_CREATE_TREE_CONNECTION  NTCreateOptions = 0x0000_0080
+
+	FILE_COMPLETE_IF_OPLOCKED NTCreateOptions = 0x0000_0100
+	FILE_NO_EA_KNOWLEDGE      NTCreateOptions = 0x0000_0200
+	FILE_DISABLE_TUNNELING    NTCreateOptions = 0x0000_0400
+	FILE_RANDOM_ACCESS        NTCreateOptions = 0x0000_0800
+
+	FILE_DELETE_ON_CLOSE        NTCreateOptions = 0x0000_1000
+	FILE_OPEN_BY_FILE_ID        NTCreateOptions = 0x0000_2000
+	FILE_OPEN_FOR_BACKUP_INTENT NTCreateOptions = 0x0000_4000
+	FILE_NO_COMPRESSION         NTCreateOptions = 0x0000_8000
+)
+
+type FileSQSFlag = FileFlagOrAttribute
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	// from winbase.h
+
+	SECURITY_ANONYMOUS      FileSQSFlag = FileSQSFlag(SecurityAnonymous << 16)
+	SECURITY_IDENTIFICATION FileSQSFlag = FileSQSFlag(SecurityIdentification << 16)
+	SECURITY_IMPERSONATION  FileSQSFlag = FileSQSFlag(SecurityImpersonation << 16)
+	SECURITY_DELEGATION     FileSQSFlag = FileSQSFlag(SecurityDelegation << 16)
+
+	SECURITY_SQOS_PRESENT     FileSQSFlag = 0x0010_0000
+	SECURITY_VALID_SQOS_FLAGS FileSQSFlag = 0x001F_0000
+)
+
+// GetFinalPathNameByHandle flags
+//
+// https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-getfinalpathnamebyhandlew#parameters
+type GetFinalPathFlag uint32
+
+//nolint:revive // SNAKE_CASE is not idiomatic in Go, but aligned with Win32 API.
+const (
+	GetFinalPathDefaultFlag GetFinalPathFlag = 0x0
+
+	FILE_NAME_NORMALIZED GetFinalPathFlag = 0x0
+	FILE_NAME_OPENED     GetFinalPathFlag = 0x8
+
+	VOLUME_NAME_DOS  GetFinalPathFlag = 0x0
+	VOLUME_NAME_GUID GetFinalPathFlag = 0x1
+	VOLUME_NAME_NT   GetFinalPathFlag = 0x2
+	VOLUME_NAME_NONE GetFinalPathFlag = 0x4
+)
+
+// getFinalPathNameByHandle facilitates calling the Windows API GetFinalPathNameByHandle
+// with the given handle and flags. It transparently takes care of creating a buffer of the
+// correct size for the call.
+//
+// https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-getfinalpathnamebyhandlew
+func GetFinalPathNameByHandle(h windows.Handle, flags GetFinalPathFlag) (string, error) {
+	b := stringbuffer.NewWString()
+	//TODO: can loop infinitely if Win32 keeps returning the same (or a larger) n?
+	for {
+		n, err := windows.GetFinalPathNameByHandle(h, b.Pointer(), b.Cap(), uint32(flags))
+		if err != nil {
+			return "", err
+		}
+		// If the buffer wasn't large enough, n will be the total size needed (including null terminator).
+		// Resize and try again.
+		if n > b.Cap() {
+			b.ResizeTo(n)
+			continue
+		}
+		// If the buffer is large enough, n will be the size not including the null terminator.
+		// Convert to a Go string and return.
+		return b.String(), nil
+	}
+}
@@ -0,0 +1,12 @@
+package fs
+
+// https://learn.microsoft.com/en-us/windows/win32/api/winnt/ne-winnt-security_impersonation_level
+type SecurityImpersonationLevel int32 // C default enums underlying type is `int`, which is Go `int32`
+
+// Impersonation levels
+const (
+	SecurityAnonymous      SecurityImpersonationLevel = 0
+	SecurityIdentification SecurityImpersonationLevel = 1
+	SecurityImpersonation  SecurityImpersonationLevel = 2
+	SecurityDelegation     SecurityImpersonationLevel = 3
+)
@@ -0,0 +1,61 @@
+//go:build windows
+
+// Code generated by 'go generate' using "github.com/Microsoft/go-winio/tools/mkwinsyscall"; DO NOT EDIT.
+
+package fs
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+	errERROR_EINVAL     error = syscall.EINVAL
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return errERROR_EINVAL
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	return e
+}
+
+var (
+	modkernel32 = windows.NewLazySystemDLL("kernel32.dll")
+
+	procCreateFileW = modkernel32.NewProc("CreateFileW")
+)
+
+func CreateFile(name string, access AccessMask, mode FileShareMode, sa *windows.SecurityAttributes, createmode FileCreationDisposition, attrs FileFlagOrAttribute, templatefile windows.Handle) (handle windows.Handle, err error) {
+	var _p0 *uint16
+	_p0, err = syscall.UTF16PtrFromString(name)
+	if err != nil {
+		return
+	}
+	return _CreateFile(_p0, access, mode, sa, createmode, attrs, templatefile)
+}
+
+func _CreateFile(name *uint16, access AccessMask, mode FileShareMode, sa *windows.SecurityAttributes, createmode FileCreationDisposition, attrs FileFlagOrAttribute, templatefile windows.Handle) (handle windows.Handle, err error) {
+	r0, _, e1 := syscall.SyscallN(procCreateFileW.Addr(), uintptr(unsafe.Pointer(name)), uintptr(access), uintptr(mode), uintptr(unsafe.Pointer(sa)), uintptr(createmode), uintptr(attrs), uintptr(templatefile))
+	handle = windows.Handle(r0)
+	if handle == windows.InvalidHandle {
+		err = errnoErr(e1)
+	}
+	return
+}
@@ -0,0 +1,20 @@
+package socket
+
+import (
+	"unsafe"
+)
+
+// RawSockaddr allows structs to be used with [Bind] and [ConnectEx]. The
+// struct must meet the Win32 sockaddr requirements specified here:
+// https://docs.microsoft.com/en-us/windows/win32/winsock/sockaddr-2
+//
+// Specifically, the struct size must be least larger than an int16 (unsigned short)
+// for the address family.
+type RawSockaddr interface {
+	// Sockaddr returns a pointer to the RawSockaddr and its struct size, allowing
+	// for the RawSockaddr's data to be overwritten by syscalls (if necessary).
+	//
+	// It is the callers responsibility to validate that the values are valid; invalid
+	// pointers or size can cause a panic.
+	Sockaddr() (unsafe.Pointer, int32, error)
+}
@@ -0,0 +1,177 @@
+//go:build windows
+
+package socket
+
+import (
+	"errors"
+	"fmt"
+	"net"
+	"sync"
+	"syscall"
+	"unsafe"
+
+	"github.com/Microsoft/go-winio/pkg/guid"
+	"golang.org/x/sys/windows"
+)
+
+//go:generate go run github.com/Microsoft/go-winio/tools/mkwinsyscall -output zsyscall_windows.go socket.go
+
+//sys getsockname(s windows.Handle, name unsafe.Pointer, namelen *int32) (err error) [failretval==socketError] = ws2_32.getsockname
+//sys getpeername(s windows.Handle, name unsafe.Pointer, namelen *int32) (err error) [failretval==socketError] = ws2_32.getpeername
+//sys bind(s windows.Handle, name unsafe.Pointer, namelen int32) (err error) [failretval==socketError] = ws2_32.bind
+
+const socketError = uintptr(^uint32(0))
+
+var (
+	// todo(helsaawy): create custom error types to store the desired vs actual size and addr family?
+
+	ErrBufferSize     = errors.New("buffer size")
+	ErrAddrFamily     = errors.New("address family")
+	ErrInvalidPointer = errors.New("invalid pointer")
+	ErrSocketClosed   = fmt.Errorf("socket closed: %w", net.ErrClosed)
+)
+
+// todo(helsaawy): replace these with generics, ie: GetSockName[S RawSockaddr](s windows.Handle) (S, error)
+
+// GetSockName writes the local address of socket s to the [RawSockaddr] rsa.
+// If rsa is not large enough, the [windows.WSAEFAULT] is returned.
+func GetSockName(s windows.Handle, rsa RawSockaddr) error {
+	ptr, l, err := rsa.Sockaddr()
+	if err != nil {
+		return fmt.Errorf("could not retrieve socket pointer and size: %w", err)
+	}
+
+	// although getsockname returns WSAEFAULT if the buffer is too small, it does not set
+	// &l to the correct size, so--apart from doubling the buffer repeatedly--there is no remedy
+	return getsockname(s, ptr, &l)
+}
+
+// GetPeerName returns the remote address the socket is connected to.
+//
+// See [GetSockName] for more information.
+func GetPeerName(s windows.Handle, rsa RawSockaddr) error {
+	ptr, l, err := rsa.Sockaddr()
+	if err != nil {
+		return fmt.Errorf("could not retrieve socket pointer and size: %w", err)
+	}
+
+	return getpeername(s, ptr, &l)
+}
+
+func Bind(s windows.Handle, rsa RawSockaddr) (err error) {
+	ptr, l, err := rsa.Sockaddr()
+	if err != nil {
+		return fmt.Errorf("could not retrieve socket pointer and size: %w", err)
+	}
+
+	return bind(s, ptr, l)
+}
+
+// "golang.org/x/sys/windows".ConnectEx and .Bind only accept internal implementations of the
+// their sockaddr interface, so they cannot be used with HvsockAddr
+// Replicate functionality here from
+// https://cs.opensource.google/go/x/sys/+/master:windows/syscall_windows.go
+
+// The function pointers to `AcceptEx`, `ConnectEx` and `GetAcceptExSockaddrs` must be loaded at
+// runtime via a WSAIoctl call:
+// https://docs.microsoft.com/en-us/windows/win32/api/Mswsock/nc-mswsock-lpfn_connectex#remarks
+
+type runtimeFunc struct {
+	id   guid.GUID
+	once sync.Once
+	addr uintptr
+	err  error
+}
+
+func (f *runtimeFunc) Load() error {
+	f.once.Do(func() {
+		var s windows.Handle
+		s, f.err = windows.Socket(windows.AF_INET, windows.SOCK_STREAM, windows.IPPROTO_TCP)
+		if f.err != nil {
+			return
+		}
+		defer windows.CloseHandle(s) //nolint:errcheck
+
+		var n uint32
+		f.err = windows.WSAIoctl(s,
+			windows.SIO_GET_EXTENSION_FUNCTION_POINTER,
+			(*byte)(unsafe.Pointer(&f.id)),
+			uint32(unsafe.Sizeof(f.id)),
+			(*byte)(unsafe.Pointer(&f.addr)),
+			uint32(unsafe.Sizeof(f.addr)),
+			&n,
+			nil, // overlapped
+			0,   // completionRoutine
+		)
+	})
+	return f.err
+}
+
+var (
+	// todo: add `AcceptEx` and `GetAcceptExSockaddrs`
+	WSAID_CONNECTEX = guid.GUID{ //revive:disable-line:var-naming ALL_CAPS
+		Data1: 0x25a207b9,
+		Data2: 0xddf3,
+		Data3: 0x4660,
+		Data4: [8]byte{0x8e, 0xe9, 0x76, 0xe5, 0x8c, 0x74, 0x06, 0x3e},
+	}
+
+	connectExFunc = runtimeFunc{id: WSAID_CONNECTEX}
+)
+
+func ConnectEx(
+	fd windows.Handle,
+	rsa RawSockaddr,
+	sendBuf *byte,
+	sendDataLen uint32,
+	bytesSent *uint32,
+	overlapped *windows.Overlapped,
+) error {
+	if err := connectExFunc.Load(); err != nil {
+		return fmt.Errorf("failed to load ConnectEx function pointer: %w", err)
+	}
+	ptr, n, err := rsa.Sockaddr()
+	if err != nil {
+		return err
+	}
+	return connectEx(fd, ptr, n, sendBuf, sendDataLen, bytesSent, overlapped)
+}
+
+// BOOL LpfnConnectex(
+//   [in]           SOCKET s,
+//   [in]           const sockaddr *name,
+//   [in]           int namelen,
+//   [in, optional] PVOID lpSendBuffer,
+//   [in]           DWORD dwSendDataLength,
+//   [out]          LPDWORD lpdwBytesSent,
+//   [in]           LPOVERLAPPED lpOverlapped
+// )
+
+func connectEx(
+	s windows.Handle,
+	name unsafe.Pointer,
+	namelen int32,
+	sendBuf *byte,
+	sendDataLen uint32,
+	bytesSent *uint32,
+	overlapped *windows.Overlapped,
+) (err error) {
+	r1, _, e1 := syscall.SyscallN(connectExFunc.addr,
+		uintptr(s),
+		uintptr(name),
+		uintptr(namelen),
+		uintptr(unsafe.Pointer(sendBuf)),
+		uintptr(sendDataLen),
+		uintptr(unsafe.Pointer(bytesSent)),
+		uintptr(unsafe.Pointer(overlapped)),
+	)
+
+	if r1 == 0 {
+		if e1 != 0 {
+			err = error(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return err
+}
@@ -0,0 +1,69 @@
+//go:build windows
+
+// Code generated by 'go generate' using "github.com/Microsoft/go-winio/tools/mkwinsyscall"; DO NOT EDIT.
+
+package socket
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+	errERROR_EINVAL     error = syscall.EINVAL
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return errERROR_EINVAL
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	return e
+}
+
+var (
+	modws2_32 = windows.NewLazySystemDLL("ws2_32.dll")
+
+	procbind        = modws2_32.NewProc("bind")
+	procgetpeername = modws2_32.NewProc("getpeername")
+	procgetsockname = modws2_32.NewProc("getsockname")
+)
+
+func bind(s windows.Handle, name unsafe.Pointer, namelen int32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procbind.Addr(), uintptr(s), uintptr(name), uintptr(namelen))
+	if r1 == socketError {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func getpeername(s windows.Handle, name unsafe.Pointer, namelen *int32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procgetpeername.Addr(), uintptr(s), uintptr(name), uintptr(unsafe.Pointer(namelen)))
+	if r1 == socketError {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func getsockname(s windows.Handle, name unsafe.Pointer, namelen *int32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procgetsockname.Addr(), uintptr(s), uintptr(name), uintptr(unsafe.Pointer(namelen)))
+	if r1 == socketError {
+		err = errnoErr(e1)
+	}
+	return
+}
@@ -0,0 +1,132 @@
+package stringbuffer
+
+import (
+	"sync"
+	"unicode/utf16"
+)
+
+// TODO: worth exporting and using in mkwinsyscall?
+
+// Uint16BufferSize is the buffer size in the pool, chosen somewhat arbitrarily to accommodate
+// large path strings:
+// MAX_PATH (260) + size of volume GUID prefix (49) + null terminator = 310.
+const MinWStringCap = 310
+
+// use *[]uint16 since []uint16 creates an extra allocation where the slice header
+// is copied to heap and then referenced via pointer in the interface header that sync.Pool
+// stores.
+var pathPool = sync.Pool{ // if go1.18+ adds Pool[T], use that to store []uint16 directly
+	New: func() interface{} {
+		b := make([]uint16, MinWStringCap)
+		return &b
+	},
+}
+
+func newBuffer() []uint16 { return *(pathPool.Get().(*[]uint16)) }
+
+// freeBuffer copies the slice header data, and puts a pointer to that in the pool.
+// This avoids taking a pointer to the slice header in WString, which can be set to nil.
+func freeBuffer(b []uint16) { pathPool.Put(&b) }
+
+// WString is a wide string buffer ([]uint16) meant for storing UTF-16 encoded strings
+// for interacting with Win32 APIs.
+// Sizes are specified as uint32 and not int.
+//
+// It is not thread safe.
+type WString struct {
+	// type-def allows casting to []uint16 directly, use struct to prevent that and allow adding fields in the future.
+
+	// raw buffer
+	b []uint16
+}
+
+// NewWString returns a [WString] allocated from a shared pool with an
+// initial capacity of at least [MinWStringCap].
+// Since the buffer may have been previously used, its contents are not guaranteed to be empty.
+//
+// The buffer should be freed via [WString.Free]
+func NewWString() *WString {
+	return &WString{
+		b: newBuffer(),
+	}
+}
+
+func (b *WString) Free() {
+	if b.empty() {
+		return
+	}
+	freeBuffer(b.b)
+	b.b = nil
+}
+
+// ResizeTo grows the buffer to at least c and returns the new capacity, freeing the
+// previous buffer back into pool.
+func (b *WString) ResizeTo(c uint32) uint32 {
+	// already sufficient (or n is 0)
+	if c <= b.Cap() {
+		return b.Cap()
+	}
+
+	if c <= MinWStringCap {
+		c = MinWStringCap
+	}
+	// allocate at-least double buffer size, as is done in [bytes.Buffer] and other places
+	if c <= 2*b.Cap() {
+		c = 2 * b.Cap()
+	}
+
+	b2 := make([]uint16, c)
+	if !b.empty() {
+		copy(b2, b.b)
+		freeBuffer(b.b)
+	}
+	b.b = b2
+	return c
+}
+
+// Buffer returns the underlying []uint16 buffer.
+func (b *WString) Buffer() []uint16 {
+	if b.empty() {
+		return nil
+	}
+	return b.b
+}
+
+// Pointer returns a pointer to the first uint16 in the buffer.
+// If the [WString.Free] has already been called, the pointer will be nil.
+func (b *WString) Pointer() *uint16 {
+	if b.empty() {
+		return nil
+	}
+	return &b.b[0]
+}
+
+// String returns the returns the UTF-8 encoding of the UTF-16 string in the buffer.
+//
+// It assumes that the data is null-terminated.
+func (b *WString) String() string {
+	// Using [windows.UTF16ToString] would require importing "golang.org/x/sys/windows"
+	// and would make this code Windows-only, which makes no sense.
+	// So copy UTF16ToString code into here.
+	// If other windows-specific code is added, switch to [windows.UTF16ToString]
+
+	s := b.b
+	for i, v := range s {
+		if v == 0 {
+			s = s[:i]
+			break
+		}
+	}
+	return string(utf16.Decode(s))
+}
+
+// Cap returns the underlying buffer capacity.
+func (b *WString) Cap() uint32 {
+	if b.empty() {
+		return 0
+	}
+	return b.cap()
+}
+
+func (b *WString) cap() uint32 { return uint32(cap(b.b)) }
+func (b *WString) empty() bool { return b == nil || b.cap() == 0 }
@@ -1,3 +1,4 @@
+//go:build windows
 // +build windows

 package winio
@@ -10,26 +11,52 @@ import (
 	"net"
 	"os"
 	"runtime"
-	"syscall"
 	"time"
 	"unsafe"
+
+	"golang.org/x/sys/windows"
+
+	"github.com/Microsoft/go-winio/internal/fs"
 )

-//sys connectNamedPipe(pipe syscall.Handle, o *syscall.Overlapped) (err error) = ConnectNamedPipe
-//sys createNamedPipe(name string, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *syscall.SecurityAttributes) (handle syscall.Handle, err error)  [failretval==syscall.InvalidHandle] = CreateNamedPipeW
-//sys createFile(name string, access uint32, mode uint32, sa *syscall.SecurityAttributes, createmode uint32, attrs uint32, templatefile syscall.Handle) (handle syscall.Handle, err error) [failretval==syscall.InvalidHandle] = CreateFileW
-//sys getNamedPipeInfo(pipe syscall.Handle, flags *uint32, outSize *uint32, inSize *uint32, maxInstances *uint32) (err error) = GetNamedPipeInfo
-//sys getNamedPipeHandleState(pipe syscall.Handle, state *uint32, curInstances *uint32, maxCollectionCount *uint32, collectDataTimeout *uint32, userName *uint16, maxUserNameSize uint32) (err error) = GetNamedPipeHandleStateW
-//sys localAlloc(uFlags uint32, length uint32) (ptr uintptr) = LocalAlloc
-//sys ntCreateNamedPipeFile(pipe *syscall.Handle, access uint32, oa *objectAttributes, iosb *ioStatusBlock, share uint32, disposition uint32, options uint32, typ uint32, readMode uint32, completionMode uint32, maxInstances uint32, inboundQuota uint32, outputQuota uint32, timeout *int64) (status ntstatus) = ntdll.NtCreateNamedPipeFile
-//sys rtlNtStatusToDosError(status ntstatus) (winerr error) = ntdll.RtlNtStatusToDosErrorNoTeb
-//sys rtlDosPathNameToNtPathName(name *uint16, ntName *unicodeString, filePart uintptr, reserved uintptr) (status ntstatus) = ntdll.RtlDosPathNameToNtPathName_U
-//sys rtlDefaultNpAcl(dacl *uintptr) (status ntstatus) = ntdll.RtlDefaultNpAcl
+//sys connectNamedPipe(pipe windows.Handle, o *windows.Overlapped) (err error) = ConnectNamedPipe
+//sys createNamedPipe(name string, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *windows.SecurityAttributes) (handle windows.Handle, err error)  [failretval==windows.InvalidHandle] = CreateNamedPipeW
+//sys disconnectNamedPipe(pipe windows.Handle) (err error) = DisconnectNamedPipe
+//sys getNamedPipeInfo(pipe windows.Handle, flags *uint32, outSize *uint32, inSize *uint32, maxInstances *uint32) (err error) = GetNamedPipeInfo
+//sys getNamedPipeHandleState(pipe windows.Handle, state *uint32, curInstances *uint32, maxCollectionCount *uint32, collectDataTimeout *uint32, userName *uint16, maxUserNameSize uint32) (err error) = GetNamedPipeHandleStateW
+//sys ntCreateNamedPipeFile(pipe *windows.Handle, access ntAccessMask, oa *objectAttributes, iosb *ioStatusBlock, share ntFileShareMode, disposition ntFileCreationDisposition, options ntFileOptions, typ uint32, readMode uint32, completionMode uint32, maxInstances uint32, inboundQuota uint32, outputQuota uint32, timeout *int64) (status ntStatus) = ntdll.NtCreateNamedPipeFile
+//sys rtlNtStatusToDosError(status ntStatus) (winerr error) = ntdll.RtlNtStatusToDosErrorNoTeb
+//sys rtlDosPathNameToNtPathName(name *uint16, ntName *unicodeString, filePart uintptr, reserved uintptr) (status ntStatus) = ntdll.RtlDosPathNameToNtPathName_U
+//sys rtlDefaultNpAcl(dacl *uintptr) (status ntStatus) = ntdll.RtlDefaultNpAcl
+
+type PipeConn interface {
+	net.Conn
+	Disconnect() error
+	Flush() error
+}
+
+// type aliases for mkwinsyscall code
+type (
+	ntAccessMask              = fs.AccessMask
+	ntFileShareMode           = fs.FileShareMode
+	ntFileCreationDisposition = fs.NTFileCreationDisposition
+	ntFileOptions             = fs.NTCreateOptions
+)

 type ioStatusBlock struct {
 	Status, Information uintptr
 }

+//	typedef struct _OBJECT_ATTRIBUTES {
+//	  ULONG           Length;
+//	  HANDLE          RootDirectory;
+//	  PUNICODE_STRING ObjectName;
+//	  ULONG           Attributes;
+//	  PVOID           SecurityDescriptor;
+//	  PVOID           SecurityQualityOfService;
+//	} OBJECT_ATTRIBUTES;
+//
+// https://learn.microsoft.com/en-us/windows/win32/api/ntdef/ns-ntdef-_object_attributes
 type objectAttributes struct {
 	Length             uintptr
 	RootDirectory      uintptr
@@ -45,51 +72,39 @@ type unicodeString struct {
 	Buffer        uintptr
 }

+//	typedef struct _SECURITY_DESCRIPTOR {
+//	  BYTE                        Revision;
+//	  BYTE                        Sbz1;
+//	  SECURITY_DESCRIPTOR_CONTROL Control;
+//	  PSID                        Owner;
+//	  PSID                        Group;
+//	  PACL                        Sacl;
+//	  PACL                        Dacl;
+//	} SECURITY_DESCRIPTOR, *PISECURITY_DESCRIPTOR;
+//
+// https://learn.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-security_descriptor
 type securityDescriptor struct {
 	Revision byte
 	Sbz1     byte
 	Control  uint16
 	Owner    uintptr
 	Group    uintptr
-	Sacl     uintptr
-	Dacl     uintptr
+	Sacl     uintptr //revive:disable-line:var-naming SACL, not Sacl
+	Dacl     uintptr //revive:disable-line:var-naming DACL, not Dacl
 }

-type ntstatus int32
+type ntStatus int32

-func (status ntstatus) Err() error {
+func (status ntStatus) Err() error {
 	if status >= 0 {
 		return nil
 	}
 	return rtlNtStatusToDosError(status)
 }

-const (
-	cERROR_PIPE_BUSY      = syscall.Errno(231)
-	cERROR_NO_DATA        = syscall.Errno(232)
-	cERROR_PIPE_CONNECTED = syscall.Errno(535)
-	cERROR_SEM_TIMEOUT    = syscall.Errno(121)
-
-	cSECURITY_SQOS_PRESENT = 0x100000
-	cSECURITY_ANONYMOUS    = 0
-
-	cPIPE_TYPE_MESSAGE = 4
-
-	cPIPE_READMODE_MESSAGE = 2
-
-	cFILE_OPEN   = 1
-	cFILE_CREATE = 2
-
-	cFILE_PIPE_MESSAGE_TYPE          = 1
-	cFILE_PIPE_REJECT_REMOTE_CLIENTS = 2
-
-	cSE_DACL_PRESENT = 4
-)
-
 var (
 	// ErrPipeListenerClosed is returned for pipe operations on listeners that have been closed.
-	// This error should match net.errClosing since docker takes a dependency on its text.
-	ErrPipeListenerClosed = errors.New("use of closed network connection")
+	ErrPipeListenerClosed = net.ErrClosed

 	errPipeWriteClosed = errors.New("pipe has been closed for write")
 )
@@ -99,6 +114,8 @@ type win32Pipe struct {
 	path string
 }

+var _ PipeConn = (*win32Pipe)(nil)
+
 type win32MessageBytePipe struct {
 	win32Pipe
 	writeClosed bool
@@ -116,9 +133,14 @@ func (f *win32Pipe) RemoteAddr() net.Addr {
 }

 func (f *win32Pipe) SetDeadline(t time.Time) error {
-	f.SetReadDeadline(t)
-	f.SetWriteDeadline(t)
-	return nil
+	if err := f.SetReadDeadline(t); err != nil {
+		return err
+	}
+	return f.SetWriteDeadline(t)
+}
+
+func (f *win32Pipe) Disconnect() error {
+	return disconnectNamedPipe(f.win32File.handle)
 }

 // CloseWrite closes the write side of a message pipe in byte mode.
@@ -157,14 +179,14 @@ func (f *win32MessageBytePipe) Read(b []byte) (int, error) {
 		return 0, io.EOF
 	}
 	n, err := f.win32File.Read(b)
-	if err == io.EOF {
+	if err == io.EOF { //nolint:errorlint
 		// If this was the result of a zero-byte read, then
 		// it is possible that the read was due to a zero-size
 		// message. Since we are simulating CloseWrite with a
 		// zero-byte message, ensure that all future Read() calls
 		// also return EOF.
 		f.readEOF = true
-	} else if err == syscall.ERROR_MORE_DATA {
+	} else if err == windows.ERROR_MORE_DATA { //nolint:errorlint // err is Errno
 		// ERROR_MORE_DATA indicates that the pipe's read mode is message mode
 		// and the message still has more bytes. Treat this as a success, since
 		// this package presents all named pipes as byte streams.
@@ -173,7 +195,7 @@ func (f *win32MessageBytePipe) Read(b []byte) (int, error) {
 	return n, err
 }

-func (s pipeAddress) Network() string {
+func (pipeAddress) Network() string {
 	return "pipe"
 }

@@ -182,18 +204,24 @@ func (s pipeAddress) String() string {
 }

 // tryDialPipe attempts to dial the pipe at `path` until `ctx` cancellation or timeout.
-func tryDialPipe(ctx context.Context, path *string, access uint32) (syscall.Handle, error) {
+func tryDialPipe(ctx context.Context, path *string, access fs.AccessMask, impLevel PipeImpLevel) (windows.Handle, error) {
 	for {
-
 		select {
 		case <-ctx.Done():
-			return syscall.Handle(0), ctx.Err()
+			return windows.Handle(0), ctx.Err()
 		default:
-			h, err := createFile(*path, access, 0, nil, syscall.OPEN_EXISTING, syscall.FILE_FLAG_OVERLAPPED|cSECURITY_SQOS_PRESENT|cSECURITY_ANONYMOUS, 0)
+			h, err := fs.CreateFile(*path,
+				access,
+				0,   // mode
+				nil, // security attributes
+				fs.OPEN_EXISTING,
+				fs.FILE_FLAG_OVERLAPPED|fs.SECURITY_SQOS_PRESENT|fs.FileSQSFlag(impLevel),
+				0, // template file handle
+			)
 			if err == nil {
 				return h, nil
 			}
-			if err != cERROR_PIPE_BUSY {
+			if err != windows.ERROR_PIPE_BUSY { //nolint:errorlint // err is Errno
 				return h, &os.PathError{Err: err, Op: "open", Path: *path}
 			}
 			// Wait 10 msec and try again. This is a rather simplistic
@@ -213,9 +241,10 @@ func DialPipe(path string, timeout *time.Duration) (net.Conn, error) {
 	} else {
 		absTimeout = time.Now().Add(2 * time.Second)
 	}
-	ctx, _ := context.WithDeadline(context.Background(), absTimeout)
+	ctx, cancel := context.WithDeadline(context.Background(), absTimeout)
+	defer cancel()
 	conn, err := DialPipeContext(ctx, path)
-	if err == context.DeadlineExceeded {
+	if errors.Is(err, context.DeadlineExceeded) {
 		return nil, ErrTimeout
 	}
 	return conn, err
@@ -224,15 +253,33 @@ func DialPipe(path string, timeout *time.Duration) (net.Conn, error) {
 // DialPipeContext attempts to connect to a named pipe by `path` until `ctx`
 // cancellation or timeout.
 func DialPipeContext(ctx context.Context, path string) (net.Conn, error) {
-	return DialPipeAccess(ctx, path, syscall.GENERIC_READ|syscall.GENERIC_WRITE)
+	return DialPipeAccess(ctx, path, uint32(fs.GENERIC_READ|fs.GENERIC_WRITE))
 }

+// PipeImpLevel is an enumeration of impersonation levels that may be set
+// when calling DialPipeAccessImpersonation.
+type PipeImpLevel uint32
+
+const (
+	PipeImpLevelAnonymous      = PipeImpLevel(fs.SECURITY_ANONYMOUS)
+	PipeImpLevelIdentification = PipeImpLevel(fs.SECURITY_IDENTIFICATION)
+	PipeImpLevelImpersonation  = PipeImpLevel(fs.SECURITY_IMPERSONATION)
+	PipeImpLevelDelegation     = PipeImpLevel(fs.SECURITY_DELEGATION)
+)
+
 // DialPipeAccess attempts to connect to a named pipe by `path` with `access` until `ctx`
 // cancellation or timeout.
 func DialPipeAccess(ctx context.Context, path string, access uint32) (net.Conn, error) {
+	return DialPipeAccessImpLevel(ctx, path, access, PipeImpLevelAnonymous)
+}
+
+// DialPipeAccessImpLevel attempts to connect to a named pipe by `path` with
+// `access` at `impLevel` until `ctx` cancellation or timeout. The other
+// DialPipe* implementations use PipeImpLevelAnonymous.
+func DialPipeAccessImpLevel(ctx context.Context, path string, access uint32, impLevel PipeImpLevel) (net.Conn, error) {
 	var err error
-	var h syscall.Handle
-	h, err = tryDialPipe(ctx, &path, access)
+	var h windows.Handle
+	h, err = tryDialPipe(ctx, &path, fs.AccessMask(access), impLevel)
 	if err != nil {
 		return nil, err
 	}
@@ -245,13 +292,13 @@ func DialPipeAccess(ctx context.Context, path string, access uint32) (net.Conn,

 	f, err := makeWin32File(h)
 	if err != nil {
-		syscall.Close(h)
+		windows.Close(h)
 		return nil, err
 	}

 	// If the pipe is in message mode, return a message byte pipe, which
 	// supports CloseWrite().
-	if flags&cPIPE_TYPE_MESSAGE != 0 {
+	if flags&windows.PIPE_TYPE_MESSAGE != 0 {
 		return &win32MessageBytePipe{
 			win32Pipe: win32Pipe{win32File: f, path: path},
 		}, nil
@@ -265,7 +312,7 @@ type acceptResponse struct {
 }

 type win32PipeListener struct {
-	firstHandle syscall.Handle
+	firstHandle windows.Handle
 	path        string
 	config      PipeConfig
 	acceptCh    chan (chan acceptResponse)
@@ -273,8 +320,8 @@ type win32PipeListener struct {
 	doneCh      chan int
 }

-func makeServerPipeHandle(path string, sd []byte, c *PipeConfig, first bool) (syscall.Handle, error) {
-	path16, err := syscall.UTF16FromString(path)
+func makeServerPipeHandle(path string, sd []byte, c *PipeConfig, first bool) (windows.Handle, error) {
+	path16, err := windows.UTF16FromString(path)
 	if err != nil {
 		return 0, &os.PathError{Op: "open", Path: path, Err: err}
 	}
@@ -283,59 +330,81 @@ func makeServerPipeHandle(path string, sd []byte, c *PipeConfig, first bool) (sy
 	oa.Length = unsafe.Sizeof(oa)

 	var ntPath unicodeString
-	if err := rtlDosPathNameToNtPathName(&path16[0], &ntPath, 0, 0).Err(); err != nil {
+	if err := rtlDosPathNameToNtPathName(&path16[0],
+		&ntPath,
+		0,
+		0,
+	).Err(); err != nil {
 		return 0, &os.PathError{Op: "open", Path: path, Err: err}
 	}
-	defer localFree(ntPath.Buffer)
+	defer windows.LocalFree(windows.Handle(ntPath.Buffer)) //nolint:errcheck
 	oa.ObjectName = &ntPath
+	oa.Attributes = windows.OBJ_CASE_INSENSITIVE

 	// The security descriptor is only needed for the first pipe.
 	if first {
 		if sd != nil {
-			len := uint32(len(sd))
-			sdb := localAlloc(0, len)
-			defer localFree(sdb)
+			//todo: does `sdb` need to be allocated on the heap, or can go allocate it?
+			l := uint32(len(sd))
+			sdb, err := windows.LocalAlloc(0, l)
+			if err != nil {
+				return 0, fmt.Errorf("LocalAlloc for security descriptor with of length %d: %w", l, err)
+			}
+			defer windows.LocalFree(windows.Handle(sdb)) //nolint:errcheck
 			copy((*[0xffff]byte)(unsafe.Pointer(sdb))[:], sd)
 			oa.SecurityDescriptor = (*securityDescriptor)(unsafe.Pointer(sdb))
 		} else {
 			// Construct the default named pipe security descriptor.
 			var dacl uintptr
 			if err := rtlDefaultNpAcl(&dacl).Err(); err != nil {
-				return 0, fmt.Errorf("getting default named pipe ACL: %s", err)
+				return 0, fmt.Errorf("getting default named pipe ACL: %w", err)
 			}
-			defer localFree(dacl)
+			defer windows.LocalFree(windows.Handle(dacl)) //nolint:errcheck

 			sdb := &securityDescriptor{
 				Revision: 1,
-				Control:  cSE_DACL_PRESENT,
+				Control:  windows.SE_DACL_PRESENT,
 				Dacl:     dacl,
 			}
 			oa.SecurityDescriptor = sdb
 		}
 	}

-	typ := uint32(cFILE_PIPE_REJECT_REMOTE_CLIENTS)
+	typ := uint32(windows.FILE_PIPE_REJECT_REMOTE_CLIENTS)
 	if c.MessageMode {
-		typ |= cFILE_PIPE_MESSAGE_TYPE
+		typ |= windows.FILE_PIPE_MESSAGE_TYPE
 	}

-	disposition := uint32(cFILE_OPEN)
-	access := uint32(syscall.GENERIC_READ | syscall.GENERIC_WRITE | syscall.SYNCHRONIZE)
+	disposition := fs.FILE_OPEN
+	access := fs.GENERIC_READ | fs.GENERIC_WRITE | fs.SYNCHRONIZE
 	if first {
-		disposition = cFILE_CREATE
+		disposition = fs.FILE_CREATE
 		// By not asking for read or write access, the named pipe file system
 		// will put this pipe into an initially disconnected state, blocking
 		// client connections until the next call with first == false.
-		access = syscall.SYNCHRONIZE
+		access = fs.SYNCHRONIZE
 	}

 	timeout := int64(-50 * 10000) // 50ms

 	var (
-		h    syscall.Handle
+		h    windows.Handle
 		iosb ioStatusBlock
 	)
-	err = ntCreateNamedPipeFile(&h, access, &oa, &iosb, syscall.FILE_SHARE_READ|syscall.FILE_SHARE_WRITE, disposition, 0, typ, 0, 0, 0xffffffff, uint32(c.InputBufferSize), uint32(c.OutputBufferSize), &timeout).Err()
+	err = ntCreateNamedPipeFile(&h,
+		access,
+		&oa,
+		&iosb,
+		fs.FILE_SHARE_READ|fs.FILE_SHARE_WRITE,
+		disposition,
+		0,
+		typ,
+		0,
+		0,
+		0xffffffff,
+		uint32(c.InputBufferSize),
+		uint32(c.OutputBufferSize),
+		&timeout).Err()
 	if err != nil {
 		return 0, &os.PathError{Op: "open", Path: path, Err: err}
 	}
@@ -351,7 +420,7 @@ func (l *win32PipeListener) makeServerPipe() (*win32File, error) {
 	}
 	f, err := makeWin32File(h)
 	if err != nil {
-		syscall.Close(h)
+		windows.Close(h)
 		return nil, err
 	}
 	return f, nil
@@ -380,7 +449,7 @@ func (l *win32PipeListener) makeConnectedServerPipe() (*win32File, error) {
 		p.Close()
 		p = nil
 		err = <-ch
-		if err == nil || err == ErrFileClosed {
+		if err == nil || err == ErrFileClosed { //nolint:errorlint // err is Errno
 			err = ErrPipeListenerClosed
 		}
 	}
@@ -402,15 +471,15 @@ func (l *win32PipeListener) listenerRoutine() {
 				p, err = l.makeConnectedServerPipe()
 				// If the connection was immediately closed by the client, try
 				// again.
-				if err != cERROR_NO_DATA {
+				if err != windows.ERROR_NO_DATA { //nolint:errorlint // err is Errno
 					break
 				}
 			}
 			responseCh <- acceptResponse{p, err}
-			closed = err == ErrPipeListenerClosed
+			closed = err == ErrPipeListenerClosed //nolint:errorlint // err is Errno
 		}
 	}
-	syscall.Close(l.firstHandle)
+	windows.Close(l.firstHandle)
 	l.firstHandle = 0
 	// Notify Close() and Accept() callers that the handle has been closed.
 	close(l.doneCh)
@@ -429,10 +498,10 @@ type PipeConfig struct {
 	// when the pipe is in message mode.
 	MessageMode bool

-	// InputBufferSize specifies the size the input buffer, in bytes.
+	// InputBufferSize specifies the size of the input buffer, in bytes.
 	InputBufferSize int32

-	// OutputBufferSize specifies the size the input buffer, in bytes.
+	// OutputBufferSize specifies the size of the output buffer, in bytes.
 	OutputBufferSize int32
 }

@@ -469,15 +538,15 @@ func ListenPipe(path string, c *PipeConfig) (net.Listener, error) {
 }

 func connectPipe(p *win32File) error {
-	c, err := p.prepareIo()
+	c, err := p.prepareIO()
 	if err != nil {
 		return err
 	}
 	defer p.wg.Done()

 	err = connectNamedPipe(p.handle, &c.o)
-	_, err = p.asyncIo(c, nil, 0, err)
-	if err != nil && err != cERROR_PIPE_CONNECTED {
+	_, err = p.asyncIO(c, nil, 0, err)
+	if err != nil && err != windows.ERROR_PIPE_CONNECTED { //nolint:errorlint // err is Errno
 		return err
 	}
 	return nil
@@ -7,26 +7,26 @@ package guid

 import (
 	"crypto/rand"
-	"crypto/sha1"
+	"crypto/sha1" //nolint:gosec // not used for secure application
 	"encoding"
 	"encoding/binary"
 	"fmt"
 	"strconv"
-
-	"golang.org/x/sys/windows"
 )

+//go:generate go run golang.org/x/tools/cmd/stringer -type=Variant -trimprefix=Variant -linecomment
+
 // Variant specifies which GUID variant (or "type") of the GUID. It determines
 // how the entirety of the rest of the GUID is interpreted.
 type Variant uint8

-// The variants specified by RFC 4122.
+// The variants specified by RFC 4122 section 4.1.1.
 const (
 	// VariantUnknown specifies a GUID variant which does not conform to one of
 	// the variant encodings specified in RFC 4122.
 	VariantUnknown Variant = iota
 	VariantNCS
-	VariantRFC4122
+	VariantRFC4122 // RFC 4122
 	VariantMicrosoft
 	VariantFuture
 )
@@ -36,16 +36,13 @@ const (
 // hash of an input string.
 type Version uint8

+func (v Version) String() string {
+	return strconv.FormatUint(uint64(v), 10)
+}
+
 var _ = (encoding.TextMarshaler)(GUID{})
 var _ = (encoding.TextUnmarshaler)(&GUID{})

-// GUID represents a GUID/UUID. It has the same structure as
-// golang.org/x/sys/windows.GUID so that it can be used with functions expecting
-// that type. It is defined as its own type so that stringification and
-// marshaling can be supported. The representation matches that used by native
-// Windows code.
-type GUID windows.GUID
-
 // NewV4 returns a new version 4 (pseudorandom) GUID, as defined by RFC 4122.
 func NewV4() (GUID, error) {
 	var b [16]byte
@@ -68,7 +65,7 @@ func NewV4() (GUID, error) {
 // big-endian UTF16 stream of bytes. If that is desired, the string can be
 // encoded as such before being passed to this function.
 func NewV5(namespace GUID, name []byte) (GUID, error) {
-	b := sha1.New()
+	b := sha1.New() //nolint:gosec // not used for secure application
 	namespaceBytes := namespace.ToArray()
 	b.Write(namespaceBytes[:])
 	b.Write(name)
@@ -0,0 +1,16 @@
+//go:build !windows
+// +build !windows
+
+package guid
+
+// GUID represents a GUID/UUID. It has the same structure as
+// golang.org/x/sys/windows.GUID so that it can be used with functions expecting
+// that type. It is defined as its own type as that is only available to builds
+// targeted at `windows`. The representation matches that used by native Windows
+// code.
+type GUID struct {
+	Data1 uint32
+	Data2 uint16
+	Data3 uint16
+	Data4 [8]byte
+}
@@ -0,0 +1,13 @@
+//go:build windows
+// +build windows
+
+package guid
+
+import "golang.org/x/sys/windows"
+
+// GUID represents a GUID/UUID. It has the same structure as
+// golang.org/x/sys/windows.GUID so that it can be used with functions expecting
+// that type. It is defined as its own type so that stringification and
+// marshaling can be supported. The representation matches that used by native
+// Windows code.
+type GUID windows.GUID
@@ -0,0 +1,27 @@
+// Code generated by "stringer -type=Variant -trimprefix=Variant -linecomment"; DO NOT EDIT.
+
+package guid
+
+import "strconv"
+
+func _() {
+	// An "invalid array index" compiler error signifies that the constant values have changed.
+	// Re-run the stringer command to generate them again.
+	var x [1]struct{}
+	_ = x[VariantUnknown-0]
+	_ = x[VariantNCS-1]
+	_ = x[VariantRFC4122-2]
+	_ = x[VariantMicrosoft-3]
+	_ = x[VariantFuture-4]
+}
+
+const _Variant_name = "UnknownNCSRFC 4122MicrosoftFuture"
+
+var _Variant_index = [...]uint8{0, 7, 10, 18, 27, 33}
+
+func (i Variant) String() string {
+	if i >= Variant(len(_Variant_index)-1) {
+		return "Variant(" + strconv.FormatInt(int64(i), 10) + ")"
+	}
+	return _Variant_name[_Variant_index[i]:_Variant_index[i+1]]
+}
@@ -0,0 +1,168 @@
+//go:build windows
+// +build windows
+
+package security
+
+import (
+	"fmt"
+	"os"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+type (
+	accessMask          uint32
+	accessMode          uint32
+	desiredAccess       uint32
+	inheritMode         uint32
+	objectType          uint32
+	shareMode           uint32
+	securityInformation uint32
+	trusteeForm         uint32
+	trusteeType         uint32
+
+	explicitAccess struct {
+		accessPermissions accessMask
+		accessMode        accessMode
+		inheritance       inheritMode
+		trustee           trustee
+	}
+
+	trustee struct {
+		multipleTrustee          *trustee
+		multipleTrusteeOperation int32
+		trusteeForm              trusteeForm
+		trusteeType              trusteeType
+		name                     uintptr
+	}
+)
+
+const (
+	accessMaskDesiredPermission accessMask = 1 << 31 // GENERIC_READ
+
+	accessModeGrant accessMode = 1
+
+	desiredAccessReadControl desiredAccess = 0x20000
+	desiredAccessWriteDac    desiredAccess = 0x40000
+
+	//cspell:disable-next-line
+	gvmga = "GrantVmGroupAccess:"
+
+	inheritModeNoInheritance                  inheritMode = 0x0
+	inheritModeSubContainersAndObjectsInherit inheritMode = 0x3
+
+	objectTypeFileObject objectType = 0x1
+
+	securityInformationDACL securityInformation = 0x4
+
+	shareModeRead  shareMode = 0x1
+	shareModeWrite shareMode = 0x2
+
+	sidVMGroup = "S-1-5-83-0"
+
+	trusteeFormIsSID trusteeForm = 0
+
+	trusteeTypeWellKnownGroup trusteeType = 5
+)
+
+// GrantVMGroupAccess sets the DACL for a specified file or directory to
+// include Grant ACE entries for the VM Group SID. This is a golang re-
+// implementation of the same function in vmcompute, just not exported in
+// RS5. Which kind of sucks. Sucks a lot :/
+//
+//revive:disable-next-line:var-naming VM, not Vm
+func GrantVmGroupAccess(name string) error {
+	// Stat (to determine if `name` is a directory).
+	s, err := os.Stat(name)
+	if err != nil {
+		return fmt.Errorf("%s os.Stat %s: %w", gvmga, name, err)
+	}
+
+	// Get a handle to the file/directory. Must defer Close on success.
+	fd, err := createFile(name, s.IsDir())
+	if err != nil {
+		return err // Already wrapped
+	}
+	defer windows.CloseHandle(fd) //nolint:errcheck
+
+	// Get the current DACL and Security Descriptor. Must defer LocalFree on success.
+	ot := objectTypeFileObject
+	si := securityInformationDACL
+	sd := uintptr(0)
+	origDACL := uintptr(0)
+	if err := getSecurityInfo(fd, uint32(ot), uint32(si), nil, nil, &origDACL, nil, &sd); err != nil {
+		return fmt.Errorf("%s GetSecurityInfo %s: %w", gvmga, name, err)
+	}
+	defer windows.LocalFree(windows.Handle(sd)) //nolint:errcheck
+
+	// Generate a new DACL which is the current DACL with the required ACEs added.
+	// Must defer LocalFree on success.
+	newDACL, err := generateDACLWithAcesAdded(name, s.IsDir(), origDACL)
+	if err != nil {
+		return err // Already wrapped
+	}
+	defer windows.LocalFree(windows.Handle(newDACL)) //nolint:errcheck
+
+	// And finally use SetSecurityInfo to apply the updated DACL.
+	if err := setSecurityInfo(fd, uint32(ot), uint32(si), uintptr(0), uintptr(0), newDACL, uintptr(0)); err != nil {
+		return fmt.Errorf("%s SetSecurityInfo %s: %w", gvmga, name, err)
+	}
+
+	return nil
+}
+
+// createFile is a helper function to call [Nt]CreateFile to get a handle to
+// the file or directory.
+func createFile(name string, isDir bool) (windows.Handle, error) {
+	namep, err := windows.UTF16FromString(name)
+	if err != nil {
+		return windows.InvalidHandle, fmt.Errorf("could not convernt name to UTF-16: %w", err)
+	}
+	da := uint32(desiredAccessReadControl | desiredAccessWriteDac)
+	sm := uint32(shareModeRead | shareModeWrite)
+	fa := uint32(windows.FILE_ATTRIBUTE_NORMAL)
+	if isDir {
+		fa |= windows.FILE_FLAG_BACKUP_SEMANTICS
+	}
+	fd, err := windows.CreateFile(&namep[0], da, sm, nil, windows.OPEN_EXISTING, fa, 0)
+	if err != nil {
+		return windows.InvalidHandle, fmt.Errorf("%s windows.CreateFile %s: %w", gvmga, name, err)
+	}
+	return fd, nil
+}
+
+// generateDACLWithAcesAdded generates a new DACL with the two needed ACEs added.
+// The caller is responsible for LocalFree of the returned DACL on success.
+func generateDACLWithAcesAdded(name string, isDir bool, origDACL uintptr) (uintptr, error) {
+	// Generate pointers to the SIDs based on the string SIDs
+	sid, err := windows.StringToSid(sidVMGroup)
+	if err != nil {
+		return 0, fmt.Errorf("%s windows.StringToSid %s %s: %w", gvmga, name, sidVMGroup, err)
+	}
+
+	inheritance := inheritModeNoInheritance
+	if isDir {
+		inheritance = inheritModeSubContainersAndObjectsInherit
+	}
+
+	eaArray := []explicitAccess{
+		{
+			accessPermissions: accessMaskDesiredPermission,
+			accessMode:        accessModeGrant,
+			inheritance:       inheritance,
+			trustee: trustee{
+				trusteeForm: trusteeFormIsSID,
+				trusteeType: trusteeTypeWellKnownGroup,
+				name:        uintptr(unsafe.Pointer(sid)),
+			},
+		},
+	}
+
+	modifiedDACL := uintptr(0)
+	if err := setEntriesInAcl(uintptr(uint32(1)), uintptr(unsafe.Pointer(&eaArray[0])), origDACL, &modifiedDACL); err != nil {
+		return 0, fmt.Errorf("%s SetEntriesInAcl %s: %w", gvmga, name, err)
+	}
+
+	return modifiedDACL, nil
+}
@@ -0,0 +1,7 @@
+package security
+
+//go:generate go run github.com/Microsoft/go-winio/tools/mkwinsyscall -output zsyscall_windows.go syscall_windows.go
+
+//sys getSecurityInfo(handle windows.Handle, objectType uint32, si uint32, ppsidOwner **uintptr, ppsidGroup **uintptr, ppDacl *uintptr, ppSacl *uintptr, ppSecurityDescriptor *uintptr) (win32err error) = advapi32.GetSecurityInfo
+//sys setSecurityInfo(handle windows.Handle, objectType uint32, si uint32, psidOwner uintptr, psidGroup uintptr, pDacl uintptr, pSacl uintptr) (win32err error) = advapi32.SetSecurityInfo
+//sys setEntriesInAcl(count uintptr, pListOfEEs uintptr, oldAcl uintptr, newAcl *uintptr) (win32err error) = advapi32.SetEntriesInAclW
@@ -0,0 +1,69 @@
+//go:build windows
+
+// Code generated by 'go generate' using "github.com/Microsoft/go-winio/tools/mkwinsyscall"; DO NOT EDIT.
+
+package security
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+	errERROR_EINVAL     error = syscall.EINVAL
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return errERROR_EINVAL
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	return e
+}
+
+var (
+	modadvapi32 = windows.NewLazySystemDLL("advapi32.dll")
+
+	procGetSecurityInfo  = modadvapi32.NewProc("GetSecurityInfo")
+	procSetEntriesInAclW = modadvapi32.NewProc("SetEntriesInAclW")
+	procSetSecurityInfo  = modadvapi32.NewProc("SetSecurityInfo")
+)
+
+func getSecurityInfo(handle windows.Handle, objectType uint32, si uint32, ppsidOwner **uintptr, ppsidGroup **uintptr, ppDacl *uintptr, ppSacl *uintptr, ppSecurityDescriptor *uintptr) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procGetSecurityInfo.Addr(), uintptr(handle), uintptr(objectType), uintptr(si), uintptr(unsafe.Pointer(ppsidOwner)), uintptr(unsafe.Pointer(ppsidGroup)), uintptr(unsafe.Pointer(ppDacl)), uintptr(unsafe.Pointer(ppSacl)), uintptr(unsafe.Pointer(ppSecurityDescriptor)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func setEntriesInAcl(count uintptr, pListOfEEs uintptr, oldAcl uintptr, newAcl *uintptr) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procSetEntriesInAclW.Addr(), uintptr(count), uintptr(pListOfEEs), uintptr(oldAcl), uintptr(unsafe.Pointer(newAcl)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func setSecurityInfo(handle windows.Handle, objectType uint32, si uint32, psidOwner uintptr, psidGroup uintptr, pDacl uintptr, pSacl uintptr) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procSetSecurityInfo.Addr(), uintptr(handle), uintptr(objectType), uintptr(si), uintptr(psidOwner), uintptr(psidGroup), uintptr(pDacl), uintptr(pSacl))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
@@ -1,3 +1,4 @@
+//go:build windows
 // +build windows

 package winio
@@ -8,7 +9,6 @@ import (
 	"fmt"
 	"runtime"
 	"sync"
-	"syscall"
 	"unicode/utf16"

 	"golang.org/x/sys/windows"
@@ -17,26 +17,22 @@ import (
 //sys adjustTokenPrivileges(token windows.Token, releaseAll bool, input *byte, outputSize uint32, output *byte, requiredSize *uint32) (success bool, err error) [true] = advapi32.AdjustTokenPrivileges
 //sys impersonateSelf(level uint32) (err error) = advapi32.ImpersonateSelf
 //sys revertToSelf() (err error) = advapi32.RevertToSelf
-//sys openThreadToken(thread syscall.Handle, accessMask uint32, openAsSelf bool, token *windows.Token) (err error) = advapi32.OpenThreadToken
-//sys getCurrentThread() (h syscall.Handle) = GetCurrentThread
+//sys openThreadToken(thread windows.Handle, accessMask uint32, openAsSelf bool, token *windows.Token) (err error) = advapi32.OpenThreadToken
+//sys getCurrentThread() (h windows.Handle) = GetCurrentThread
 //sys lookupPrivilegeValue(systemName string, name string, luid *uint64) (err error) = advapi32.LookupPrivilegeValueW
 //sys lookupPrivilegeName(systemName string, luid *uint64, buffer *uint16, size *uint32) (err error) = advapi32.LookupPrivilegeNameW
 //sys lookupPrivilegeDisplayName(systemName string, name *uint16, buffer *uint16, size *uint32, languageId *uint32) (err error) = advapi32.LookupPrivilegeDisplayNameW

 const (
-	SE_PRIVILEGE_ENABLED = 2
+	//revive:disable-next-line:var-naming ALL_CAPS
+	SE_PRIVILEGE_ENABLED = windows.SE_PRIVILEGE_ENABLED

-	ERROR_NOT_ALL_ASSIGNED syscall.Errno = 1300
+	//revive:disable-next-line:var-naming ALL_CAPS
+	ERROR_NOT_ALL_ASSIGNED windows.Errno = windows.ERROR_NOT_ALL_ASSIGNED

-	SeBackupPrivilege  = "SeBackupPrivilege"
-	SeRestorePrivilege = "SeRestorePrivilege"
-)
-
-const (
-	securityAnonymous = iota
-	securityIdentification
-	securityImpersonation
-	securityDelegation
+	SeBackupPrivilege   = "SeBackupPrivilege"
+	SeRestorePrivilege  = "SeRestorePrivilege"
+	SeSecurityPrivilege = "SeSecurityPrivilege"
 )

 var (
@@ -50,11 +46,9 @@ type PrivilegeError struct {
 }

 func (e *PrivilegeError) Error() string {
-	s := ""
+	s := "Could not enable privilege "
 	if len(e.privileges) > 1 {
 		s = "Could not enable privileges "
-	} else {
-		s = "Could not enable privilege "
 	}
 	for i, p := range e.privileges {
 		if i != 0 {
@@ -93,7 +87,7 @@ func RunWithPrivileges(names []string, fn func() error) error {
 }

 func mapPrivileges(names []string) ([]uint64, error) {
-	var privileges []uint64
+	privileges := make([]uint64, 0, len(names))
 	privNameMutex.Lock()
 	defer privNameMutex.Unlock()
 	for _, name := range names {
@@ -126,7 +120,7 @@ func enableDisableProcessPrivilege(names []string, action uint32) error {
 		return err
 	}

-	p, _ := windows.GetCurrentProcess()
+	p := windows.CurrentProcess()
 	var token windows.Token
 	err = windows.OpenProcessToken(p, windows.TOKEN_ADJUST_PRIVILEGES|windows.TOKEN_QUERY, &token)
 	if err != nil {
@@ -139,10 +133,10 @@ func enableDisableProcessPrivilege(names []string, action uint32) error {

 func adjustPrivileges(token windows.Token, privileges []uint64, action uint32) error {
 	var b bytes.Buffer
-	binary.Write(&b, binary.LittleEndian, uint32(len(privileges)))
+	_ = binary.Write(&b, binary.LittleEndian, uint32(len(privileges)))
 	for _, p := range privileges {
-		binary.Write(&b, binary.LittleEndian, p)
-		binary.Write(&b, binary.LittleEndian, action)
+		_ = binary.Write(&b, binary.LittleEndian, p)
+		_ = binary.Write(&b, binary.LittleEndian, action)
 	}
 	prevState := make([]byte, b.Len())
 	reqSize := uint32(0)
@@ -150,7 +144,7 @@ func adjustPrivileges(token windows.Token, privileges []uint64, action uint32) e
 	if !success {
 		return err
 	}
-	if err == ERROR_NOT_ALL_ASSIGNED {
+	if err == ERROR_NOT_ALL_ASSIGNED { //nolint:errorlint // err is Errno
 		return &PrivilegeError{privileges}
 	}
 	return nil
@@ -176,13 +170,13 @@ func getPrivilegeName(luid uint64) string {
 }

 func newThreadToken() (windows.Token, error) {
-	err := impersonateSelf(securityImpersonation)
+	err := impersonateSelf(windows.SecurityImpersonation)
 	if err != nil {
 		return 0, err
 	}

 	var token windows.Token
-	err = openThreadToken(getCurrentThread(), syscall.TOKEN_ADJUST_PRIVILEGES|syscall.TOKEN_QUERY, false, &token)
+	err = openThreadToken(getCurrentThread(), windows.TOKEN_ADJUST_PRIVILEGES|windows.TOKEN_QUERY, false, &token)
 	if err != nil {
 		rerr := revertToSelf()
 		if rerr != nil {
@@ -1,3 +1,6 @@
+//go:build windows
+// +build windows
+
 package winio

 import (
@@ -113,16 +116,16 @@ func EncodeReparsePoint(rp *ReparsePoint) []byte {
 	}

 	var b bytes.Buffer
-	binary.Write(&b, binary.LittleEndian, &data)
+	_ = binary.Write(&b, binary.LittleEndian, &data)
 	if !rp.IsMountPoint {
 		flags := uint32(0)
 		if relative {
 			flags |= 1
 		}
-		binary.Write(&b, binary.LittleEndian, flags)
+		_ = binary.Write(&b, binary.LittleEndian, flags)
 	}

-	binary.Write(&b, binary.LittleEndian, ntTarget16)
-	binary.Write(&b, binary.LittleEndian, target16)
+	_ = binary.Write(&b, binary.LittleEndian, ntTarget16)
+	_ = binary.Write(&b, binary.LittleEndian, target16)
 	return b.Bytes()
 }
@@ -1,22 +1,20 @@
+//go:build windows
 // +build windows

 package winio

 import (
-	"syscall"
+	"errors"
+	"fmt"
 	"unsafe"
+
+	"golang.org/x/sys/windows"
 )

 //sys lookupAccountName(systemName *uint16, accountName string, sid *byte, sidSize *uint32, refDomain *uint16, refDomainSize *uint32, sidNameUse *uint32) (err error) = advapi32.LookupAccountNameW
+//sys lookupAccountSid(systemName *uint16, sid *byte, name *uint16, nameSize *uint32, refDomain *uint16, refDomainSize *uint32, sidNameUse *uint32) (err error) = advapi32.LookupAccountSidW
 //sys convertSidToStringSid(sid *byte, str **uint16) (err error) = advapi32.ConvertSidToStringSidW
-//sys convertStringSecurityDescriptorToSecurityDescriptor(str string, revision uint32, sd *uintptr, size *uint32) (err error) = advapi32.ConvertStringSecurityDescriptorToSecurityDescriptorW
-//sys convertSecurityDescriptorToStringSecurityDescriptor(sd *byte, revision uint32, secInfo uint32, sddl **uint16, sddlSize *uint32) (err error) = advapi32.ConvertSecurityDescriptorToStringSecurityDescriptorW
-//sys localFree(mem uintptr) = LocalFree
-//sys getSecurityDescriptorLength(sd uintptr) (len uint32) = advapi32.GetSecurityDescriptorLength
-
-const (
-	cERROR_NONE_MAPPED = syscall.Errno(1332)
-)
+//sys convertStringSidToSid(str *uint16, sid **byte) (err error) = advapi32.ConvertStringSidToSidW

 type AccountLookupError struct {
 	Name string
@@ -28,8 +26,10 @@ func (e *AccountLookupError) Error() string {
 		return "lookup account: empty account name specified"
 	}
 	var s string
-	switch e.Err {
-	case cERROR_NONE_MAPPED:
+	switch {
+	case errors.Is(e.Err, windows.ERROR_INVALID_SID):
+		s = "the security ID structure is invalid"
+	case errors.Is(e.Err, windows.ERROR_NONE_MAPPED):
 		s = "not found"
 	default:
 		s = e.Err.Error()
@@ -37,6 +37,8 @@ func (e *AccountLookupError) Error() string {
 	return "lookup account " + e.Name + ": " + s
 }

+func (e *AccountLookupError) Unwrap() error { return e.Err }
+
 type SddlConversionError struct {
 	Sddl string
 	Err  error
@@ -46,15 +48,19 @@ func (e *SddlConversionError) Error() string {
 	return "convert " + e.Sddl + ": " + e.Err.Error()
 }

+func (e *SddlConversionError) Unwrap() error { return e.Err }
+
 // LookupSidByName looks up the SID of an account by name
+//
+//revive:disable-next-line:var-naming SID, not Sid
 func LookupSidByName(name string) (sid string, err error) {
 	if name == "" {
-		return "", &AccountLookupError{name, cERROR_NONE_MAPPED}
+		return "", &AccountLookupError{name, windows.ERROR_NONE_MAPPED}
 	}

 	var sidSize, sidNameUse, refDomainSize uint32
 	err = lookupAccountName(nil, name, nil, &sidSize, nil, &refDomainSize, &sidNameUse)
-	if err != nil && err != syscall.ERROR_INSUFFICIENT_BUFFER {
+	if err != nil && err != windows.ERROR_INSUFFICIENT_BUFFER { //nolint:errorlint // err is Errno
 		return "", &AccountLookupError{name, err}
 	}
 	sidBuffer := make([]byte, sidSize)
@@ -68,31 +74,60 @@ func LookupSidByName(name string) (sid string, err error) {
 	if err != nil {
 		return "", &AccountLookupError{name, err}
 	}
-	sid = syscall.UTF16ToString((*[0xffff]uint16)(unsafe.Pointer(strBuffer))[:])
-	localFree(uintptr(unsafe.Pointer(strBuffer)))
+	sid = windows.UTF16ToString((*[0xffff]uint16)(unsafe.Pointer(strBuffer))[:])
+	_, _ = windows.LocalFree(windows.Handle(unsafe.Pointer(strBuffer)))
 	return sid, nil
 }

-func SddlToSecurityDescriptor(sddl string) ([]byte, error) {
-	var sdBuffer uintptr
-	err := convertStringSecurityDescriptorToSecurityDescriptor(sddl, 1, &sdBuffer, nil)
-	if err != nil {
-		return nil, &SddlConversionError{sddl, err}
+// LookupNameBySid looks up the name of an account by SID
+//
+//revive:disable-next-line:var-naming SID, not Sid
+func LookupNameBySid(sid string) (name string, err error) {
+	if sid == "" {
+		return "", &AccountLookupError{sid, windows.ERROR_NONE_MAPPED}
 	}
-	defer localFree(sdBuffer)
-	sd := make([]byte, getSecurityDescriptorLength(sdBuffer))
-	copy(sd, (*[0xffff]byte)(unsafe.Pointer(sdBuffer))[:len(sd)])
-	return sd, nil
+
+	sidBuffer, err := windows.UTF16PtrFromString(sid)
+	if err != nil {
+		return "", &AccountLookupError{sid, err}
+	}
+
+	var sidPtr *byte
+	if err = convertStringSidToSid(sidBuffer, &sidPtr); err != nil {
+		return "", &AccountLookupError{sid, err}
+	}
+	defer windows.LocalFree(windows.Handle(unsafe.Pointer(sidPtr))) //nolint:errcheck
+
+	var nameSize, refDomainSize, sidNameUse uint32
+	err = lookupAccountSid(nil, sidPtr, nil, &nameSize, nil, &refDomainSize, &sidNameUse)
+	if err != nil && err != windows.ERROR_INSUFFICIENT_BUFFER { //nolint:errorlint // err is Errno
+		return "", &AccountLookupError{sid, err}
+	}
+
+	nameBuffer := make([]uint16, nameSize)
+	refDomainBuffer := make([]uint16, refDomainSize)
+	err = lookupAccountSid(nil, sidPtr, &nameBuffer[0], &nameSize, &refDomainBuffer[0], &refDomainSize, &sidNameUse)
+	if err != nil {
+		return "", &AccountLookupError{sid, err}
+	}
+
+	name = windows.UTF16ToString(nameBuffer)
+	return name, nil
+}
+
+func SddlToSecurityDescriptor(sddl string) ([]byte, error) {
+	sd, err := windows.SecurityDescriptorFromString(sddl)
+	if err != nil {
+		return nil, &SddlConversionError{Sddl: sddl, Err: err}
+	}
+	b := unsafe.Slice((*byte)(unsafe.Pointer(sd)), sd.Length())
+	return b, nil
 }

 func SecurityDescriptorToSddl(sd []byte) (string, error) {
-	var sddl *uint16
-	// The returned string length seems to including an aribtrary number of terminating NULs.
-	// Don't use it.
-	err := convertSecurityDescriptorToStringSecurityDescriptor(&sd[0], 1, 0xff, &sddl, nil)
-	if err != nil {
-		return "", err
+	if l := int(unsafe.Sizeof(windows.SECURITY_DESCRIPTOR{})); len(sd) < l {
+		return "", fmt.Errorf("SecurityDescriptor (%d) smaller than expected (%d): %w", len(sd), l, windows.ERROR_INCORRECT_SIZE)
 	}
-	defer localFree(uintptr(unsafe.Pointer(sddl)))
-	return syscall.UTF16ToString((*[0xffff]uint16)(unsafe.Pointer(sddl))[:]), nil
+	s := (*windows.SECURITY_DESCRIPTOR)(unsafe.Pointer(&sd[0]))
+	return s.String(), nil
 }
@@ -1,3 +1,5 @@
+//go:build windows
+
 package winio

-//go:generate go run $GOROOT/src/syscall/mksyscall_windows.go -output zsyscall_windows.go file.go pipe.go sd.go fileinfo.go privilege.go backup.go hvsock.go
+//go:generate go run github.com/Microsoft/go-winio/tools/mkwinsyscall -output zsyscall_windows.go ./*.go
@@ -0,0 +1,377 @@
+//go:build windows
+// +build windows
+
+package vhd
+
+import (
+	"fmt"
+	"syscall"
+
+	"github.com/Microsoft/go-winio/pkg/guid"
+	"golang.org/x/sys/windows"
+)
+
+//go:generate go run github.com/Microsoft/go-winio/tools/mkwinsyscall -output zvhd_windows.go vhd.go
+
+//sys createVirtualDisk(virtualStorageType *VirtualStorageType, path string, virtualDiskAccessMask uint32, securityDescriptor *uintptr, createVirtualDiskFlags uint32, providerSpecificFlags uint32, parameters *CreateVirtualDiskParameters, overlapped *syscall.Overlapped, handle *syscall.Handle) (win32err error) = virtdisk.CreateVirtualDisk
+//sys openVirtualDisk(virtualStorageType *VirtualStorageType, path string, virtualDiskAccessMask uint32, openVirtualDiskFlags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (win32err error) = virtdisk.OpenVirtualDisk
+//sys attachVirtualDisk(handle syscall.Handle, securityDescriptor *uintptr, attachVirtualDiskFlag uint32, providerSpecificFlags uint32, parameters *AttachVirtualDiskParameters, overlapped *syscall.Overlapped) (win32err error) = virtdisk.AttachVirtualDisk
+//sys detachVirtualDisk(handle syscall.Handle, detachVirtualDiskFlags uint32, providerSpecificFlags uint32) (win32err error) = virtdisk.DetachVirtualDisk
+//sys getVirtualDiskPhysicalPath(handle syscall.Handle, diskPathSizeInBytes *uint32, buffer *uint16) (win32err error) = virtdisk.GetVirtualDiskPhysicalPath
+
+type (
+	CreateVirtualDiskFlag uint32
+	VirtualDiskFlag       uint32
+	AttachVirtualDiskFlag uint32
+	DetachVirtualDiskFlag uint32
+	VirtualDiskAccessMask uint32
+)
+
+type VirtualStorageType struct {
+	DeviceID uint32
+	VendorID guid.GUID
+}
+
+type CreateVersion2 struct {
+	UniqueID                 guid.GUID
+	MaximumSize              uint64
+	BlockSizeInBytes         uint32
+	SectorSizeInBytes        uint32
+	PhysicalSectorSizeInByte uint32
+	ParentPath               *uint16 // string
+	SourcePath               *uint16 // string
+	OpenFlags                uint32
+	ParentVirtualStorageType VirtualStorageType
+	SourceVirtualStorageType VirtualStorageType
+	ResiliencyGUID           guid.GUID
+}
+
+type CreateVirtualDiskParameters struct {
+	Version  uint32 // Must always be set to 2
+	Version2 CreateVersion2
+}
+
+type OpenVersion2 struct {
+	GetInfoOnly    bool
+	ReadOnly       bool
+	ResiliencyGUID guid.GUID
+}
+
+type OpenVirtualDiskParameters struct {
+	Version  uint32 // Must always be set to 2
+	Version2 OpenVersion2
+}
+
+// The higher level `OpenVersion2` struct uses `bool`s to refer to `GetInfoOnly` and `ReadOnly` for ease of use. However,
+// the internal windows structure uses `BOOL`s aka int32s for these types. `openVersion2` is used for translating
+// `OpenVersion2` fields to the correct windows internal field types on the `Open____` methods.
+type openVersion2 struct {
+	getInfoOnly    int32
+	readOnly       int32
+	resiliencyGUID guid.GUID
+}
+
+type openVirtualDiskParameters struct {
+	version  uint32
+	version2 openVersion2
+}
+
+type AttachVersion2 struct {
+	RestrictedOffset uint64
+	RestrictedLength uint64
+}
+
+type AttachVirtualDiskParameters struct {
+	Version  uint32
+	Version2 AttachVersion2
+}
+
+const (
+	//revive:disable-next-line:var-naming ALL_CAPS
+	VIRTUAL_STORAGE_TYPE_DEVICE_VHDX = 0x3
+
+	// Access Mask for opening a VHD.
+	VirtualDiskAccessNone     VirtualDiskAccessMask = 0x00000000
+	VirtualDiskAccessAttachRO VirtualDiskAccessMask = 0x00010000
+	VirtualDiskAccessAttachRW VirtualDiskAccessMask = 0x00020000
+	VirtualDiskAccessDetach   VirtualDiskAccessMask = 0x00040000
+	VirtualDiskAccessGetInfo  VirtualDiskAccessMask = 0x00080000
+	VirtualDiskAccessCreate   VirtualDiskAccessMask = 0x00100000
+	VirtualDiskAccessMetaOps  VirtualDiskAccessMask = 0x00200000
+	VirtualDiskAccessRead     VirtualDiskAccessMask = 0x000d0000
+	VirtualDiskAccessAll      VirtualDiskAccessMask = 0x003f0000
+	VirtualDiskAccessWritable VirtualDiskAccessMask = 0x00320000
+
+	// Flags for creating a VHD.
+	CreateVirtualDiskFlagNone                              CreateVirtualDiskFlag = 0x0
+	CreateVirtualDiskFlagFullPhysicalAllocation            CreateVirtualDiskFlag = 0x1
+	CreateVirtualDiskFlagPreventWritesToSourceDisk         CreateVirtualDiskFlag = 0x2
+	CreateVirtualDiskFlagDoNotCopyMetadataFromParent       CreateVirtualDiskFlag = 0x4
+	CreateVirtualDiskFlagCreateBackingStorage              CreateVirtualDiskFlag = 0x8
+	CreateVirtualDiskFlagUseChangeTrackingSourceLimit      CreateVirtualDiskFlag = 0x10
+	CreateVirtualDiskFlagPreserveParentChangeTrackingState CreateVirtualDiskFlag = 0x20
+	CreateVirtualDiskFlagVhdSetUseOriginalBackingStorage   CreateVirtualDiskFlag = 0x40 //revive:disable-line:var-naming VHD, not Vhd
+	CreateVirtualDiskFlagSparseFile                        CreateVirtualDiskFlag = 0x80
+	CreateVirtualDiskFlagPmemCompatible                    CreateVirtualDiskFlag = 0x100 //revive:disable-line:var-naming PMEM, not Pmem
+	CreateVirtualDiskFlagSupportCompressedVolumes          CreateVirtualDiskFlag = 0x200
+
+	// Flags for opening a VHD.
+	OpenVirtualDiskFlagNone                        VirtualDiskFlag = 0x00000000
+	OpenVirtualDiskFlagNoParents                   VirtualDiskFlag = 0x00000001
+	OpenVirtualDiskFlagBlankFile                   VirtualDiskFlag = 0x00000002
+	OpenVirtualDiskFlagBootDrive                   VirtualDiskFlag = 0x00000004
+	OpenVirtualDiskFlagCachedIO                    VirtualDiskFlag = 0x00000008
+	OpenVirtualDiskFlagCustomDiffChain             VirtualDiskFlag = 0x00000010
+	OpenVirtualDiskFlagParentCachedIO              VirtualDiskFlag = 0x00000020
+	OpenVirtualDiskFlagVhdsetFileOnly              VirtualDiskFlag = 0x00000040
+	OpenVirtualDiskFlagIgnoreRelativeParentLocator VirtualDiskFlag = 0x00000080
+	OpenVirtualDiskFlagNoWriteHardening            VirtualDiskFlag = 0x00000100
+	OpenVirtualDiskFlagSupportCompressedVolumes    VirtualDiskFlag = 0x00000200
+
+	// Flags for attaching a VHD.
+	AttachVirtualDiskFlagNone                          AttachVirtualDiskFlag = 0x00000000
+	AttachVirtualDiskFlagReadOnly                      AttachVirtualDiskFlag = 0x00000001
+	AttachVirtualDiskFlagNoDriveLetter                 AttachVirtualDiskFlag = 0x00000002
+	AttachVirtualDiskFlagPermanentLifetime             AttachVirtualDiskFlag = 0x00000004
+	AttachVirtualDiskFlagNoLocalHost                   AttachVirtualDiskFlag = 0x00000008
+	AttachVirtualDiskFlagNoSecurityDescriptor          AttachVirtualDiskFlag = 0x00000010
+	AttachVirtualDiskFlagBypassDefaultEncryptionPolicy AttachVirtualDiskFlag = 0x00000020
+	AttachVirtualDiskFlagNonPnp                        AttachVirtualDiskFlag = 0x00000040
+	AttachVirtualDiskFlagRestrictedRange               AttachVirtualDiskFlag = 0x00000080
+	AttachVirtualDiskFlagSinglePartition               AttachVirtualDiskFlag = 0x00000100
+	AttachVirtualDiskFlagRegisterVolume                AttachVirtualDiskFlag = 0x00000200
+
+	// Flags for detaching a VHD.
+	DetachVirtualDiskFlagNone DetachVirtualDiskFlag = 0x0
+)
+
+// CreateVhdx is a helper function to create a simple vhdx file at the given path using
+// default values.
+//
+//revive:disable-next-line:var-naming VHDX, not Vhdx
+func CreateVhdx(path string, maxSizeInGb, blockSizeInMb uint32) error {
+	params := CreateVirtualDiskParameters{
+		Version: 2,
+		Version2: CreateVersion2{
+			MaximumSize:      uint64(maxSizeInGb) * 1024 * 1024 * 1024,
+			BlockSizeInBytes: blockSizeInMb * 1024 * 1024,
+		},
+	}
+
+	handle, err := CreateVirtualDisk(path, VirtualDiskAccessNone, CreateVirtualDiskFlagNone, &params)
+	if err != nil {
+		return err
+	}
+
+	return syscall.CloseHandle(handle)
+}
+
+// DetachVirtualDisk detaches a virtual hard disk by handle.
+func DetachVirtualDisk(handle syscall.Handle) (err error) {
+	if err := detachVirtualDisk(handle, 0, 0); err != nil {
+		return fmt.Errorf("failed to detach virtual disk: %w", err)
+	}
+	return nil
+}
+
+// DetachVhd detaches a vhd found at `path`.
+//
+//revive:disable-next-line:var-naming VHD, not Vhd
+func DetachVhd(path string) error {
+	handle, err := OpenVirtualDisk(
+		path,
+		VirtualDiskAccessNone,
+		OpenVirtualDiskFlagCachedIO|OpenVirtualDiskFlagIgnoreRelativeParentLocator,
+	)
+	if err != nil {
+		return err
+	}
+	defer syscall.CloseHandle(handle) //nolint:errcheck
+	return DetachVirtualDisk(handle)
+}
+
+// AttachVirtualDisk attaches a virtual hard disk for use.
+func AttachVirtualDisk(
+	handle syscall.Handle,
+	attachVirtualDiskFlag AttachVirtualDiskFlag,
+	parameters *AttachVirtualDiskParameters,
+) (err error) {
+	// Supports both version 1 and 2 of the attach parameters as version 2 wasn't present in RS5.
+	if err := attachVirtualDisk(
+		handle,
+		nil,
+		uint32(attachVirtualDiskFlag),
+		0,
+		parameters,
+		nil,
+	); err != nil {
+		return fmt.Errorf("failed to attach virtual disk: %w", err)
+	}
+	return nil
+}
+
+// AttachVhd attaches a virtual hard disk at `path` for use. Attaches using version 2
+// of the ATTACH_VIRTUAL_DISK_PARAMETERS.
+//
+//revive:disable-next-line:var-naming VHD, not Vhd
+func AttachVhd(path string) (err error) {
+	handle, err := OpenVirtualDisk(
+		path,
+		VirtualDiskAccessNone,
+		OpenVirtualDiskFlagCachedIO|OpenVirtualDiskFlagIgnoreRelativeParentLocator,
+	)
+	if err != nil {
+		return err
+	}
+
+	defer syscall.CloseHandle(handle) //nolint:errcheck
+	params := AttachVirtualDiskParameters{Version: 2}
+	if err := AttachVirtualDisk(
+		handle,
+		AttachVirtualDiskFlagNone,
+		&params,
+	); err != nil {
+		return fmt.Errorf("failed to attach virtual disk: %w", err)
+	}
+	return nil
+}
+
+// OpenVirtualDisk obtains a handle to a VHD opened with supplied access mask and flags.
+func OpenVirtualDisk(
+	vhdPath string,
+	virtualDiskAccessMask VirtualDiskAccessMask,
+	openVirtualDiskFlags VirtualDiskFlag,
+) (syscall.Handle, error) {
+	parameters := OpenVirtualDiskParameters{Version: 2}
+	handle, err := OpenVirtualDiskWithParameters(
+		vhdPath,
+		virtualDiskAccessMask,
+		openVirtualDiskFlags,
+		&parameters,
+	)
+	if err != nil {
+		return 0, err
+	}
+	return handle, nil
+}
+
+// OpenVirtualDiskWithParameters obtains a handle to a VHD opened with supplied access mask, flags and parameters.
+func OpenVirtualDiskWithParameters(
+	vhdPath string,
+	virtualDiskAccessMask VirtualDiskAccessMask,
+	openVirtualDiskFlags VirtualDiskFlag,
+	parameters *OpenVirtualDiskParameters,
+) (syscall.Handle, error) {
+	var (
+		handle      syscall.Handle
+		defaultType VirtualStorageType
+		getInfoOnly int32
+		readOnly    int32
+	)
+	if parameters.Version != 2 {
+		return handle, fmt.Errorf("only version 2 VHDs are supported, found version: %d", parameters.Version)
+	}
+	if parameters.Version2.GetInfoOnly {
+		getInfoOnly = 1
+	}
+	if parameters.Version2.ReadOnly {
+		readOnly = 1
+	}
+	params := &openVirtualDiskParameters{
+		version: parameters.Version,
+		version2: openVersion2{
+			getInfoOnly,
+			readOnly,
+			parameters.Version2.ResiliencyGUID,
+		},
+	}
+	if err := openVirtualDisk(
+		&defaultType,
+		vhdPath,
+		uint32(virtualDiskAccessMask),
+		uint32(openVirtualDiskFlags),
+		params,
+		&handle,
+	); err != nil {
+		return 0, fmt.Errorf("failed to open virtual disk: %w", err)
+	}
+	return handle, nil
+}
+
+// CreateVirtualDisk creates a virtual harddisk and returns a handle to the disk.
+func CreateVirtualDisk(
+	path string,
+	virtualDiskAccessMask VirtualDiskAccessMask,
+	createVirtualDiskFlags CreateVirtualDiskFlag,
+	parameters *CreateVirtualDiskParameters,
+) (syscall.Handle, error) {
+	var (
+		handle      syscall.Handle
+		defaultType VirtualStorageType
+	)
+	if parameters.Version != 2 {
+		return handle, fmt.Errorf("only version 2 VHDs are supported, found version: %d", parameters.Version)
+	}
+
+	if err := createVirtualDisk(
+		&defaultType,
+		path,
+		uint32(virtualDiskAccessMask),
+		nil,
+		uint32(createVirtualDiskFlags),
+		0,
+		parameters,
+		nil,
+		&handle,
+	); err != nil {
+		return handle, fmt.Errorf("failed to create virtual disk: %w", err)
+	}
+	return handle, nil
+}
+
+// GetVirtualDiskPhysicalPath takes a handle to a virtual hard disk and returns the physical
+// path of the disk on the machine. This path is in the form \\.\PhysicalDriveX where X is an integer
+// that represents the particular enumeration of the physical disk on the caller's system.
+func GetVirtualDiskPhysicalPath(handle syscall.Handle) (_ string, err error) {
+	var (
+		diskPathSizeInBytes uint32 = 256 * 2 // max path length 256 wide chars
+		diskPhysicalPathBuf [256]uint16
+	)
+	if err := getVirtualDiskPhysicalPath(
+		handle,
+		&diskPathSizeInBytes,
+		&diskPhysicalPathBuf[0],
+	); err != nil {
+		return "", fmt.Errorf("failed to get disk physical path: %w", err)
+	}
+	return windows.UTF16ToString(diskPhysicalPathBuf[:]), nil
+}
+
+// CreateDiffVhd is a helper function to create a differencing virtual disk.
+//
+//revive:disable-next-line:var-naming VHD, not Vhd
+func CreateDiffVhd(diffVhdPath, baseVhdPath string, blockSizeInMB uint32) error {
+	// Setting `ParentPath` is how to signal to create a differencing disk.
+	createParams := &CreateVirtualDiskParameters{
+		Version: 2,
+		Version2: CreateVersion2{
+			ParentPath:       windows.StringToUTF16Ptr(baseVhdPath),
+			BlockSizeInBytes: blockSizeInMB * 1024 * 1024,
+			OpenFlags:        uint32(OpenVirtualDiskFlagCachedIO),
+		},
+	}
+
+	vhdHandle, err := CreateVirtualDisk(
+		diffVhdPath,
+		VirtualDiskAccessNone,
+		CreateVirtualDiskFlagNone,
+		createParams,
+	)
+	if err != nil {
+		return fmt.Errorf("failed to create differencing vhd: %w", err)
+	}
+	if err := syscall.CloseHandle(vhdHandle); err != nil {
+		return fmt.Errorf("failed to close differencing vhd handle: %w", err)
+	}
+	return nil
+}
@@ -0,0 +1,105 @@
+//go:build windows
+
+// Code generated by 'go generate' using "github.com/Microsoft/go-winio/tools/mkwinsyscall"; DO NOT EDIT.
+
+package vhd
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+	errERROR_EINVAL     error = syscall.EINVAL
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return errERROR_EINVAL
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	return e
+}
+
+var (
+	modvirtdisk = windows.NewLazySystemDLL("virtdisk.dll")
+
+	procAttachVirtualDisk          = modvirtdisk.NewProc("AttachVirtualDisk")
+	procCreateVirtualDisk          = modvirtdisk.NewProc("CreateVirtualDisk")
+	procDetachVirtualDisk          = modvirtdisk.NewProc("DetachVirtualDisk")
+	procGetVirtualDiskPhysicalPath = modvirtdisk.NewProc("GetVirtualDiskPhysicalPath")
+	procOpenVirtualDisk            = modvirtdisk.NewProc("OpenVirtualDisk")
+)
+
+func attachVirtualDisk(handle syscall.Handle, securityDescriptor *uintptr, attachVirtualDiskFlag uint32, providerSpecificFlags uint32, parameters *AttachVirtualDiskParameters, overlapped *syscall.Overlapped) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procAttachVirtualDisk.Addr(), uintptr(handle), uintptr(unsafe.Pointer(securityDescriptor)), uintptr(attachVirtualDiskFlag), uintptr(providerSpecificFlags), uintptr(unsafe.Pointer(parameters)), uintptr(unsafe.Pointer(overlapped)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func createVirtualDisk(virtualStorageType *VirtualStorageType, path string, virtualDiskAccessMask uint32, securityDescriptor *uintptr, createVirtualDiskFlags uint32, providerSpecificFlags uint32, parameters *CreateVirtualDiskParameters, overlapped *syscall.Overlapped, handle *syscall.Handle) (win32err error) {
+	var _p0 *uint16
+	_p0, win32err = syscall.UTF16PtrFromString(path)
+	if win32err != nil {
+		return
+	}
+	return _createVirtualDisk(virtualStorageType, _p0, virtualDiskAccessMask, securityDescriptor, createVirtualDiskFlags, providerSpecificFlags, parameters, overlapped, handle)
+}
+
+func _createVirtualDisk(virtualStorageType *VirtualStorageType, path *uint16, virtualDiskAccessMask uint32, securityDescriptor *uintptr, createVirtualDiskFlags uint32, providerSpecificFlags uint32, parameters *CreateVirtualDiskParameters, overlapped *syscall.Overlapped, handle *syscall.Handle) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procCreateVirtualDisk.Addr(), uintptr(unsafe.Pointer(virtualStorageType)), uintptr(unsafe.Pointer(path)), uintptr(virtualDiskAccessMask), uintptr(unsafe.Pointer(securityDescriptor)), uintptr(createVirtualDiskFlags), uintptr(providerSpecificFlags), uintptr(unsafe.Pointer(parameters)), uintptr(unsafe.Pointer(overlapped)), uintptr(unsafe.Pointer(handle)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func detachVirtualDisk(handle syscall.Handle, detachVirtualDiskFlags uint32, providerSpecificFlags uint32) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procDetachVirtualDisk.Addr(), uintptr(handle), uintptr(detachVirtualDiskFlags), uintptr(providerSpecificFlags))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func getVirtualDiskPhysicalPath(handle syscall.Handle, diskPathSizeInBytes *uint32, buffer *uint16) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procGetVirtualDiskPhysicalPath.Addr(), uintptr(handle), uintptr(unsafe.Pointer(diskPathSizeInBytes)), uintptr(unsafe.Pointer(buffer)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
+
+func openVirtualDisk(virtualStorageType *VirtualStorageType, path string, virtualDiskAccessMask uint32, openVirtualDiskFlags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (win32err error) {
+	var _p0 *uint16
+	_p0, win32err = syscall.UTF16PtrFromString(path)
+	if win32err != nil {
+		return
+	}
+	return _openVirtualDisk(virtualStorageType, _p0, virtualDiskAccessMask, openVirtualDiskFlags, parameters, handle)
+}
+
+func _openVirtualDisk(virtualStorageType *VirtualStorageType, path *uint16, virtualDiskAccessMask uint32, openVirtualDiskFlags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (win32err error) {
+	r0, _, _ := syscall.SyscallN(procOpenVirtualDisk.Addr(), uintptr(unsafe.Pointer(virtualStorageType)), uintptr(unsafe.Pointer(path)), uintptr(virtualDiskAccessMask), uintptr(openVirtualDiskFlags), uintptr(unsafe.Pointer(parameters)), uintptr(unsafe.Pointer(handle)))
+	if r0 != 0 {
+		win32err = syscall.Errno(r0)
+	}
+	return
+}
@@ -1,4 +1,6 @@
-// Code generated by 'go generate'; DO NOT EDIT.
+//go:build windows
+
+// Code generated by 'go generate' using "github.com/Microsoft/go-winio/tools/mkwinsyscall"; DO NOT EDIT.

 package winio

@@ -19,6 +21,7 @@ const (

 var (
 	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+	errERROR_EINVAL     error = syscall.EINVAL
 )

 // errnoErr returns common boxed Errno values, to prevent
@@ -26,237 +29,86 @@ var (
 func errnoErr(e syscall.Errno) error {
 	switch e {
 	case 0:
-		return nil
+		return errERROR_EINVAL
 	case errnoERROR_IO_PENDING:
 		return errERROR_IO_PENDING
 	}
-	// TODO: add more here, after collecting data on the common
-	// error values see on Windows. (perhaps when running
-	// all.bat?)
 	return e
 }

 var (
-	modkernel32 = windows.NewLazySystemDLL("kernel32.dll")
-	modws2_32   = windows.NewLazySystemDLL("ws2_32.dll")
-	modntdll    = windows.NewLazySystemDLL("ntdll.dll")
 	modadvapi32 = windows.NewLazySystemDLL("advapi32.dll")
+	modkernel32 = windows.NewLazySystemDLL("kernel32.dll")
+	modntdll    = windows.NewLazySystemDLL("ntdll.dll")
+	modws2_32   = windows.NewLazySystemDLL("ws2_32.dll")

-	procCancelIoEx                                           = modkernel32.NewProc("CancelIoEx")
-	procCreateIoCompletionPort                               = modkernel32.NewProc("CreateIoCompletionPort")
-	procGetQueuedCompletionStatus                            = modkernel32.NewProc("GetQueuedCompletionStatus")
-	procSetFileCompletionNotificationModes                   = modkernel32.NewProc("SetFileCompletionNotificationModes")
-	procWSAGetOverlappedResult                               = modws2_32.NewProc("WSAGetOverlappedResult")
-	procConnectNamedPipe                                     = modkernel32.NewProc("ConnectNamedPipe")
-	procCreateNamedPipeW                                     = modkernel32.NewProc("CreateNamedPipeW")
-	procCreateFileW                                          = modkernel32.NewProc("CreateFileW")
-	procGetNamedPipeInfo                                     = modkernel32.NewProc("GetNamedPipeInfo")
-	procGetNamedPipeHandleStateW                             = modkernel32.NewProc("GetNamedPipeHandleStateW")
-	procLocalAlloc                                           = modkernel32.NewProc("LocalAlloc")
-	procNtCreateNamedPipeFile                                = modntdll.NewProc("NtCreateNamedPipeFile")
-	procRtlNtStatusToDosErrorNoTeb                           = modntdll.NewProc("RtlNtStatusToDosErrorNoTeb")
-	procRtlDosPathNameToNtPathName_U                         = modntdll.NewProc("RtlDosPathNameToNtPathName_U")
-	procRtlDefaultNpAcl                                      = modntdll.NewProc("RtlDefaultNpAcl")
-	procLookupAccountNameW                                   = modadvapi32.NewProc("LookupAccountNameW")
-	procConvertSidToStringSidW                               = modadvapi32.NewProc("ConvertSidToStringSidW")
-	procConvertStringSecurityDescriptorToSecurityDescriptorW = modadvapi32.NewProc("ConvertStringSecurityDescriptorToSecurityDescriptorW")
-	procConvertSecurityDescriptorToStringSecurityDescriptorW = modadvapi32.NewProc("ConvertSecurityDescriptorToStringSecurityDescriptorW")
-	procLocalFree                                            = modkernel32.NewProc("LocalFree")
-	procGetSecurityDescriptorLength                          = modadvapi32.NewProc("GetSecurityDescriptorLength")
-	procGetFileInformationByHandleEx                         = modkernel32.NewProc("GetFileInformationByHandleEx")
-	procSetFileInformationByHandle                           = modkernel32.NewProc("SetFileInformationByHandle")
-	procAdjustTokenPrivileges                                = modadvapi32.NewProc("AdjustTokenPrivileges")
-	procImpersonateSelf                                      = modadvapi32.NewProc("ImpersonateSelf")
-	procRevertToSelf                                         = modadvapi32.NewProc("RevertToSelf")
-	procOpenThreadToken                                      = modadvapi32.NewProc("OpenThreadToken")
-	procGetCurrentThread                                     = modkernel32.NewProc("GetCurrentThread")
-	procLookupPrivilegeValueW                                = modadvapi32.NewProc("LookupPrivilegeValueW")
-	procLookupPrivilegeNameW                                 = modadvapi32.NewProc("LookupPrivilegeNameW")
-	procLookupPrivilegeDisplayNameW                          = modadvapi32.NewProc("LookupPrivilegeDisplayNameW")
-	procBackupRead                                           = modkernel32.NewProc("BackupRead")
-	procBackupWrite                                          = modkernel32.NewProc("BackupWrite")
-	procbind                                                 = modws2_32.NewProc("bind")
+	procAdjustTokenPrivileges              = modadvapi32.NewProc("AdjustTokenPrivileges")
+	procConvertSidToStringSidW             = modadvapi32.NewProc("ConvertSidToStringSidW")
+	procConvertStringSidToSidW             = modadvapi32.NewProc("ConvertStringSidToSidW")
+	procImpersonateSelf                    = modadvapi32.NewProc("ImpersonateSelf")
+	procLookupAccountNameW                 = modadvapi32.NewProc("LookupAccountNameW")
+	procLookupAccountSidW                  = modadvapi32.NewProc("LookupAccountSidW")
+	procLookupPrivilegeDisplayNameW        = modadvapi32.NewProc("LookupPrivilegeDisplayNameW")
+	procLookupPrivilegeNameW               = modadvapi32.NewProc("LookupPrivilegeNameW")
+	procLookupPrivilegeValueW              = modadvapi32.NewProc("LookupPrivilegeValueW")
+	procOpenThreadToken                    = modadvapi32.NewProc("OpenThreadToken")
+	procRevertToSelf                       = modadvapi32.NewProc("RevertToSelf")
+	procBackupRead                         = modkernel32.NewProc("BackupRead")
+	procBackupWrite                        = modkernel32.NewProc("BackupWrite")
+	procCancelIoEx                         = modkernel32.NewProc("CancelIoEx")
+	procConnectNamedPipe                   = modkernel32.NewProc("ConnectNamedPipe")
+	procCreateIoCompletionPort             = modkernel32.NewProc("CreateIoCompletionPort")
+	procCreateNamedPipeW                   = modkernel32.NewProc("CreateNamedPipeW")
+	procDisconnectNamedPipe                = modkernel32.NewProc("DisconnectNamedPipe")
+	procGetCurrentThread                   = modkernel32.NewProc("GetCurrentThread")
+	procGetNamedPipeHandleStateW           = modkernel32.NewProc("GetNamedPipeHandleStateW")
+	procGetNamedPipeInfo                   = modkernel32.NewProc("GetNamedPipeInfo")
+	procGetQueuedCompletionStatus          = modkernel32.NewProc("GetQueuedCompletionStatus")
+	procSetFileCompletionNotificationModes = modkernel32.NewProc("SetFileCompletionNotificationModes")
+	procNtCreateNamedPipeFile              = modntdll.NewProc("NtCreateNamedPipeFile")
+	procRtlDefaultNpAcl                    = modntdll.NewProc("RtlDefaultNpAcl")
+	procRtlDosPathNameToNtPathName_U       = modntdll.NewProc("RtlDosPathNameToNtPathName_U")
+	procRtlNtStatusToDosErrorNoTeb         = modntdll.NewProc("RtlNtStatusToDosErrorNoTeb")
+	procWSAGetOverlappedResult             = modws2_32.NewProc("WSAGetOverlappedResult")
 )

-func cancelIoEx(file syscall.Handle, o *syscall.Overlapped) (err error) {
-	r1, _, e1 := syscall.Syscall(procCancelIoEx.Addr(), 2, uintptr(file), uintptr(unsafe.Pointer(o)), 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func createIoCompletionPort(file syscall.Handle, port syscall.Handle, key uintptr, threadCount uint32) (newport syscall.Handle, err error) {
-	r0, _, e1 := syscall.Syscall6(procCreateIoCompletionPort.Addr(), 4, uintptr(file), uintptr(port), uintptr(key), uintptr(threadCount), 0, 0)
-	newport = syscall.Handle(r0)
-	if newport == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func getQueuedCompletionStatus(port syscall.Handle, bytes *uint32, key *uintptr, o **ioOperation, timeout uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procGetQueuedCompletionStatus.Addr(), 5, uintptr(port), uintptr(unsafe.Pointer(bytes)), uintptr(unsafe.Pointer(key)), uintptr(unsafe.Pointer(o)), uintptr(timeout), 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func setFileCompletionNotificationModes(h syscall.Handle, flags uint8) (err error) {
-	r1, _, e1 := syscall.Syscall(procSetFileCompletionNotificationModes.Addr(), 2, uintptr(h), uintptr(flags), 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func wsaGetOverlappedResult(h syscall.Handle, o *syscall.Overlapped, bytes *uint32, wait bool, flags *uint32) (err error) {
+func adjustTokenPrivileges(token windows.Token, releaseAll bool, input *byte, outputSize uint32, output *byte, requiredSize *uint32) (success bool, err error) {
 	var _p0 uint32
-	if wait {
+	if releaseAll {
 		_p0 = 1
-	} else {
-		_p0 = 0
 	}
-	r1, _, e1 := syscall.Syscall6(procWSAGetOverlappedResult.Addr(), 5, uintptr(h), uintptr(unsafe.Pointer(o)), uintptr(unsafe.Pointer(bytes)), uintptr(_p0), uintptr(unsafe.Pointer(flags)), 0)
+	r0, _, e1 := syscall.SyscallN(procAdjustTokenPrivileges.Addr(), uintptr(token), uintptr(_p0), uintptr(unsafe.Pointer(input)), uintptr(outputSize), uintptr(unsafe.Pointer(output)), uintptr(unsafe.Pointer(requiredSize)))
+	success = r0 != 0
+	if true {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func convertSidToStringSid(sid *byte, str **uint16) (err error) {
+	r1, _, e1 := syscall.SyscallN(procConvertSidToStringSidW.Addr(), uintptr(unsafe.Pointer(sid)), uintptr(unsafe.Pointer(str)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func connectNamedPipe(pipe syscall.Handle, o *syscall.Overlapped) (err error) {
-	r1, _, e1 := syscall.Syscall(procConnectNamedPipe.Addr(), 2, uintptr(pipe), uintptr(unsafe.Pointer(o)), 0)
+func convertStringSidToSid(str *uint16, sid **byte) (err error) {
+	r1, _, e1 := syscall.SyscallN(procConvertStringSidToSidW.Addr(), uintptr(unsafe.Pointer(str)), uintptr(unsafe.Pointer(sid)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func createNamedPipe(name string, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *syscall.SecurityAttributes) (handle syscall.Handle, err error) {
-	var _p0 *uint16
-	_p0, err = syscall.UTF16PtrFromString(name)
-	if err != nil {
-		return
-	}
-	return _createNamedPipe(_p0, flags, pipeMode, maxInstances, outSize, inSize, defaultTimeout, sa)
-}
-
-func _createNamedPipe(name *uint16, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *syscall.SecurityAttributes) (handle syscall.Handle, err error) {
-	r0, _, e1 := syscall.Syscall9(procCreateNamedPipeW.Addr(), 8, uintptr(unsafe.Pointer(name)), uintptr(flags), uintptr(pipeMode), uintptr(maxInstances), uintptr(outSize), uintptr(inSize), uintptr(defaultTimeout), uintptr(unsafe.Pointer(sa)), 0)
-	handle = syscall.Handle(r0)
-	if handle == syscall.InvalidHandle {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func createFile(name string, access uint32, mode uint32, sa *syscall.SecurityAttributes, createmode uint32, attrs uint32, templatefile syscall.Handle) (handle syscall.Handle, err error) {
-	var _p0 *uint16
-	_p0, err = syscall.UTF16PtrFromString(name)
-	if err != nil {
-		return
-	}
-	return _createFile(_p0, access, mode, sa, createmode, attrs, templatefile)
-}
-
-func _createFile(name *uint16, access uint32, mode uint32, sa *syscall.SecurityAttributes, createmode uint32, attrs uint32, templatefile syscall.Handle) (handle syscall.Handle, err error) {
-	r0, _, e1 := syscall.Syscall9(procCreateFileW.Addr(), 7, uintptr(unsafe.Pointer(name)), uintptr(access), uintptr(mode), uintptr(unsafe.Pointer(sa)), uintptr(createmode), uintptr(attrs), uintptr(templatefile), 0, 0)
-	handle = syscall.Handle(r0)
-	if handle == syscall.InvalidHandle {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func getNamedPipeInfo(pipe syscall.Handle, flags *uint32, outSize *uint32, inSize *uint32, maxInstances *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procGetNamedPipeInfo.Addr(), 5, uintptr(pipe), uintptr(unsafe.Pointer(flags)), uintptr(unsafe.Pointer(outSize)), uintptr(unsafe.Pointer(inSize)), uintptr(unsafe.Pointer(maxInstances)), 0)
+func impersonateSelf(level uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procImpersonateSelf.Addr(), uintptr(level))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func getNamedPipeHandleState(pipe syscall.Handle, state *uint32, curInstances *uint32, maxCollectionCount *uint32, collectDataTimeout *uint32, userName *uint16, maxUserNameSize uint32) (err error) {
-	r1, _, e1 := syscall.Syscall9(procGetNamedPipeHandleStateW.Addr(), 7, uintptr(pipe), uintptr(unsafe.Pointer(state)), uintptr(unsafe.Pointer(curInstances)), uintptr(unsafe.Pointer(maxCollectionCount)), uintptr(unsafe.Pointer(collectDataTimeout)), uintptr(unsafe.Pointer(userName)), uintptr(maxUserNameSize), 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func localAlloc(uFlags uint32, length uint32) (ptr uintptr) {
-	r0, _, _ := syscall.Syscall(procLocalAlloc.Addr(), 2, uintptr(uFlags), uintptr(length), 0)
-	ptr = uintptr(r0)
-	return
-}
-
-func ntCreateNamedPipeFile(pipe *syscall.Handle, access uint32, oa *objectAttributes, iosb *ioStatusBlock, share uint32, disposition uint32, options uint32, typ uint32, readMode uint32, completionMode uint32, maxInstances uint32, inboundQuota uint32, outputQuota uint32, timeout *int64) (status ntstatus) {
-	r0, _, _ := syscall.Syscall15(procNtCreateNamedPipeFile.Addr(), 14, uintptr(unsafe.Pointer(pipe)), uintptr(access), uintptr(unsafe.Pointer(oa)), uintptr(unsafe.Pointer(iosb)), uintptr(share), uintptr(disposition), uintptr(options), uintptr(typ), uintptr(readMode), uintptr(completionMode), uintptr(maxInstances), uintptr(inboundQuota), uintptr(outputQuota), uintptr(unsafe.Pointer(timeout)), 0)
-	status = ntstatus(r0)
-	return
-}
-
-func rtlNtStatusToDosError(status ntstatus) (winerr error) {
-	r0, _, _ := syscall.Syscall(procRtlNtStatusToDosErrorNoTeb.Addr(), 1, uintptr(status), 0, 0)
-	if r0 != 0 {
-		winerr = syscall.Errno(r0)
-	}
-	return
-}
-
-func rtlDosPathNameToNtPathName(name *uint16, ntName *unicodeString, filePart uintptr, reserved uintptr) (status ntstatus) {
-	r0, _, _ := syscall.Syscall6(procRtlDosPathNameToNtPathName_U.Addr(), 4, uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(ntName)), uintptr(filePart), uintptr(reserved), 0, 0)
-	status = ntstatus(r0)
-	return
-}
-
-func rtlDefaultNpAcl(dacl *uintptr) (status ntstatus) {
-	r0, _, _ := syscall.Syscall(procRtlDefaultNpAcl.Addr(), 1, uintptr(unsafe.Pointer(dacl)), 0, 0)
-	status = ntstatus(r0)
-	return
-}
-
 func lookupAccountName(systemName *uint16, accountName string, sid *byte, sidSize *uint32, refDomain *uint16, refDomainSize *uint32, sidNameUse *uint32) (err error) {
 	var _p0 *uint16
 	_p0, err = syscall.UTF16PtrFromString(accountName)
@@ -267,164 +119,55 @@ func lookupAccountName(systemName *uint16, accountName string, sid *byte, sidSiz
 }

 func _lookupAccountName(systemName *uint16, accountName *uint16, sid *byte, sidSize *uint32, refDomain *uint16, refDomainSize *uint32, sidNameUse *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall9(procLookupAccountNameW.Addr(), 7, uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(accountName)), uintptr(unsafe.Pointer(sid)), uintptr(unsafe.Pointer(sidSize)), uintptr(unsafe.Pointer(refDomain)), uintptr(unsafe.Pointer(refDomainSize)), uintptr(unsafe.Pointer(sidNameUse)), 0, 0)
+	r1, _, e1 := syscall.SyscallN(procLookupAccountNameW.Addr(), uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(accountName)), uintptr(unsafe.Pointer(sid)), uintptr(unsafe.Pointer(sidSize)), uintptr(unsafe.Pointer(refDomain)), uintptr(unsafe.Pointer(refDomainSize)), uintptr(unsafe.Pointer(sidNameUse)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func convertSidToStringSid(sid *byte, str **uint16) (err error) {
-	r1, _, e1 := syscall.Syscall(procConvertSidToStringSidW.Addr(), 2, uintptr(unsafe.Pointer(sid)), uintptr(unsafe.Pointer(str)), 0)
+func lookupAccountSid(systemName *uint16, sid *byte, name *uint16, nameSize *uint32, refDomain *uint16, refDomainSize *uint32, sidNameUse *uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procLookupAccountSidW.Addr(), uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(sid)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(nameSize)), uintptr(unsafe.Pointer(refDomain)), uintptr(unsafe.Pointer(refDomainSize)), uintptr(unsafe.Pointer(sidNameUse)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func convertStringSecurityDescriptorToSecurityDescriptor(str string, revision uint32, sd *uintptr, size *uint32) (err error) {
+func lookupPrivilegeDisplayName(systemName string, name *uint16, buffer *uint16, size *uint32, languageId *uint32) (err error) {
 	var _p0 *uint16
-	_p0, err = syscall.UTF16PtrFromString(str)
+	_p0, err = syscall.UTF16PtrFromString(systemName)
 	if err != nil {
 		return
 	}
-	return _convertStringSecurityDescriptorToSecurityDescriptor(_p0, revision, sd, size)
+	return _lookupPrivilegeDisplayName(_p0, name, buffer, size, languageId)
 }

-func _convertStringSecurityDescriptorToSecurityDescriptor(str *uint16, revision uint32, sd *uintptr, size *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procConvertStringSecurityDescriptorToSecurityDescriptorW.Addr(), 4, uintptr(unsafe.Pointer(str)), uintptr(revision), uintptr(unsafe.Pointer(sd)), uintptr(unsafe.Pointer(size)), 0, 0)
+func _lookupPrivilegeDisplayName(systemName *uint16, name *uint16, buffer *uint16, size *uint32, languageId *uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procLookupPrivilegeDisplayNameW.Addr(), uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(buffer)), uintptr(unsafe.Pointer(size)), uintptr(unsafe.Pointer(languageId)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func convertSecurityDescriptorToStringSecurityDescriptor(sd *byte, revision uint32, secInfo uint32, sddl **uint16, sddlSize *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procConvertSecurityDescriptorToStringSecurityDescriptorW.Addr(), 5, uintptr(unsafe.Pointer(sd)), uintptr(revision), uintptr(secInfo), uintptr(unsafe.Pointer(sddl)), uintptr(unsafe.Pointer(sddlSize)), 0)
+func lookupPrivilegeName(systemName string, luid *uint64, buffer *uint16, size *uint32) (err error) {
+	var _p0 *uint16
+	_p0, err = syscall.UTF16PtrFromString(systemName)
+	if err != nil {
+		return
+	}
+	return _lookupPrivilegeName(_p0, luid, buffer, size)
+}
+
+func _lookupPrivilegeName(systemName *uint16, luid *uint64, buffer *uint16, size *uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procLookupPrivilegeNameW.Addr(), uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(luid)), uintptr(unsafe.Pointer(buffer)), uintptr(unsafe.Pointer(size)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func localFree(mem uintptr) {
-	syscall.Syscall(procLocalFree.Addr(), 1, uintptr(mem), 0, 0)
-	return
-}
-
-func getSecurityDescriptorLength(sd uintptr) (len uint32) {
-	r0, _, _ := syscall.Syscall(procGetSecurityDescriptorLength.Addr(), 1, uintptr(sd), 0, 0)
-	len = uint32(r0)
-	return
-}
-
-func getFileInformationByHandleEx(h syscall.Handle, class uint32, buffer *byte, size uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procGetFileInformationByHandleEx.Addr(), 4, uintptr(h), uintptr(class), uintptr(unsafe.Pointer(buffer)), uintptr(size), 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func setFileInformationByHandle(h syscall.Handle, class uint32, buffer *byte, size uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procSetFileInformationByHandle.Addr(), 4, uintptr(h), uintptr(class), uintptr(unsafe.Pointer(buffer)), uintptr(size), 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func adjustTokenPrivileges(token windows.Token, releaseAll bool, input *byte, outputSize uint32, output *byte, requiredSize *uint32) (success bool, err error) {
-	var _p0 uint32
-	if releaseAll {
-		_p0 = 1
-	} else {
-		_p0 = 0
-	}
-	r0, _, e1 := syscall.Syscall6(procAdjustTokenPrivileges.Addr(), 6, uintptr(token), uintptr(_p0), uintptr(unsafe.Pointer(input)), uintptr(outputSize), uintptr(unsafe.Pointer(output)), uintptr(unsafe.Pointer(requiredSize)))
-	success = r0 != 0
-	if true {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func impersonateSelf(level uint32) (err error) {
-	r1, _, e1 := syscall.Syscall(procImpersonateSelf.Addr(), 1, uintptr(level), 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func revertToSelf() (err error) {
-	r1, _, e1 := syscall.Syscall(procRevertToSelf.Addr(), 0, 0, 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func openThreadToken(thread syscall.Handle, accessMask uint32, openAsSelf bool, token *windows.Token) (err error) {
-	var _p0 uint32
-	if openAsSelf {
-		_p0 = 1
-	} else {
-		_p0 = 0
-	}
-	r1, _, e1 := syscall.Syscall6(procOpenThreadToken.Addr(), 4, uintptr(thread), uintptr(accessMask), uintptr(_p0), uintptr(unsafe.Pointer(token)), 0, 0)
-	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
-	}
-	return
-}
-
-func getCurrentThread() (h syscall.Handle) {
-	r0, _, _ := syscall.Syscall(procGetCurrentThread.Addr(), 0, 0, 0, 0)
-	h = syscall.Handle(r0)
-	return
-}
-
 func lookupPrivilegeValue(systemName string, name string, luid *uint64) (err error) {
 	var _p0 *uint16
 	_p0, err = syscall.UTF16PtrFromString(systemName)
@@ -440,60 +183,34 @@ func lookupPrivilegeValue(systemName string, name string, luid *uint64) (err err
 }

 func _lookupPrivilegeValue(systemName *uint16, name *uint16, luid *uint64) (err error) {
-	r1, _, e1 := syscall.Syscall(procLookupPrivilegeValueW.Addr(), 3, uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(luid)))
+	r1, _, e1 := syscall.SyscallN(procLookupPrivilegeValueW.Addr(), uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(luid)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func lookupPrivilegeName(systemName string, luid *uint64, buffer *uint16, size *uint32) (err error) {
-	var _p0 *uint16
-	_p0, err = syscall.UTF16PtrFromString(systemName)
-	if err != nil {
-		return
+func openThreadToken(thread windows.Handle, accessMask uint32, openAsSelf bool, token *windows.Token) (err error) {
+	var _p0 uint32
+	if openAsSelf {
+		_p0 = 1
 	}
-	return _lookupPrivilegeName(_p0, luid, buffer, size)
-}
-
-func _lookupPrivilegeName(systemName *uint16, luid *uint64, buffer *uint16, size *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procLookupPrivilegeNameW.Addr(), 4, uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(luid)), uintptr(unsafe.Pointer(buffer)), uintptr(unsafe.Pointer(size)), 0, 0)
+	r1, _, e1 := syscall.SyscallN(procOpenThreadToken.Addr(), uintptr(thread), uintptr(accessMask), uintptr(_p0), uintptr(unsafe.Pointer(token)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func lookupPrivilegeDisplayName(systemName string, name *uint16, buffer *uint16, size *uint32, languageId *uint32) (err error) {
-	var _p0 *uint16
-	_p0, err = syscall.UTF16PtrFromString(systemName)
-	if err != nil {
-		return
-	}
-	return _lookupPrivilegeDisplayName(_p0, name, buffer, size, languageId)
-}
-
-func _lookupPrivilegeDisplayName(systemName *uint16, name *uint16, buffer *uint16, size *uint32, languageId *uint32) (err error) {
-	r1, _, e1 := syscall.Syscall6(procLookupPrivilegeDisplayNameW.Addr(), 5, uintptr(unsafe.Pointer(systemName)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(buffer)), uintptr(unsafe.Pointer(size)), uintptr(unsafe.Pointer(languageId)), 0)
+func revertToSelf() (err error) {
+	r1, _, e1 := syscall.SyscallN(procRevertToSelf.Addr())
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func backupRead(h syscall.Handle, b []byte, bytesRead *uint32, abort bool, processSecurity bool, context *uintptr) (err error) {
+func backupRead(h windows.Handle, b []byte, bytesRead *uint32, abort bool, processSecurity bool, context *uintptr) (err error) {
 	var _p0 *byte
 	if len(b) > 0 {
 		_p0 = &b[0]
@@ -501,27 +218,19 @@ func backupRead(h syscall.Handle, b []byte, bytesRead *uint32, abort bool, proce
 	var _p1 uint32
 	if abort {
 		_p1 = 1
-	} else {
-		_p1 = 0
 	}
 	var _p2 uint32
 	if processSecurity {
 		_p2 = 1
-	} else {
-		_p2 = 0
 	}
-	r1, _, e1 := syscall.Syscall9(procBackupRead.Addr(), 7, uintptr(h), uintptr(unsafe.Pointer(_p0)), uintptr(len(b)), uintptr(unsafe.Pointer(bytesRead)), uintptr(_p1), uintptr(_p2), uintptr(unsafe.Pointer(context)), 0, 0)
+	r1, _, e1 := syscall.SyscallN(procBackupRead.Addr(), uintptr(h), uintptr(unsafe.Pointer(_p0)), uintptr(len(b)), uintptr(unsafe.Pointer(bytesRead)), uintptr(_p1), uintptr(_p2), uintptr(unsafe.Pointer(context)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func backupWrite(h syscall.Handle, b []byte, bytesWritten *uint32, abort bool, processSecurity bool, context *uintptr) (err error) {
+func backupWrite(h windows.Handle, b []byte, bytesWritten *uint32, abort bool, processSecurity bool, context *uintptr) (err error) {
 	var _p0 *byte
 	if len(b) > 0 {
 		_p0 = &b[0]
@@ -529,34 +238,141 @@ func backupWrite(h syscall.Handle, b []byte, bytesWritten *uint32, abort bool, p
 	var _p1 uint32
 	if abort {
 		_p1 = 1
-	} else {
-		_p1 = 0
 	}
 	var _p2 uint32
 	if processSecurity {
 		_p2 = 1
-	} else {
-		_p2 = 0
 	}
-	r1, _, e1 := syscall.Syscall9(procBackupWrite.Addr(), 7, uintptr(h), uintptr(unsafe.Pointer(_p0)), uintptr(len(b)), uintptr(unsafe.Pointer(bytesWritten)), uintptr(_p1), uintptr(_p2), uintptr(unsafe.Pointer(context)), 0, 0)
+	r1, _, e1 := syscall.SyscallN(procBackupWrite.Addr(), uintptr(h), uintptr(unsafe.Pointer(_p0)), uintptr(len(b)), uintptr(unsafe.Pointer(bytesWritten)), uintptr(_p1), uintptr(_p2), uintptr(unsafe.Pointer(context)))
 	if r1 == 0 {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+		err = errnoErr(e1)
 	}
 	return
 }

-func bind(s syscall.Handle, name unsafe.Pointer, namelen int32) (err error) {
-	r1, _, e1 := syscall.Syscall(procbind.Addr(), 3, uintptr(s), uintptr(name), uintptr(namelen))
-	if r1 == socketError {
-		if e1 != 0 {
-			err = errnoErr(e1)
-		} else {
-			err = syscall.EINVAL
-		}
+func cancelIoEx(file windows.Handle, o *windows.Overlapped) (err error) {
+	r1, _, e1 := syscall.SyscallN(procCancelIoEx.Addr(), uintptr(file), uintptr(unsafe.Pointer(o)))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func connectNamedPipe(pipe windows.Handle, o *windows.Overlapped) (err error) {
+	r1, _, e1 := syscall.SyscallN(procConnectNamedPipe.Addr(), uintptr(pipe), uintptr(unsafe.Pointer(o)))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func createIoCompletionPort(file windows.Handle, port windows.Handle, key uintptr, threadCount uint32) (newport windows.Handle, err error) {
+	r0, _, e1 := syscall.SyscallN(procCreateIoCompletionPort.Addr(), uintptr(file), uintptr(port), uintptr(key), uintptr(threadCount))
+	newport = windows.Handle(r0)
+	if newport == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func createNamedPipe(name string, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *windows.SecurityAttributes) (handle windows.Handle, err error) {
+	var _p0 *uint16
+	_p0, err = syscall.UTF16PtrFromString(name)
+	if err != nil {
+		return
+	}
+	return _createNamedPipe(_p0, flags, pipeMode, maxInstances, outSize, inSize, defaultTimeout, sa)
+}
+
+func _createNamedPipe(name *uint16, flags uint32, pipeMode uint32, maxInstances uint32, outSize uint32, inSize uint32, defaultTimeout uint32, sa *windows.SecurityAttributes) (handle windows.Handle, err error) {
+	r0, _, e1 := syscall.SyscallN(procCreateNamedPipeW.Addr(), uintptr(unsafe.Pointer(name)), uintptr(flags), uintptr(pipeMode), uintptr(maxInstances), uintptr(outSize), uintptr(inSize), uintptr(defaultTimeout), uintptr(unsafe.Pointer(sa)))
+	handle = windows.Handle(r0)
+	if handle == windows.InvalidHandle {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func disconnectNamedPipe(pipe windows.Handle) (err error) {
+	r1, _, e1 := syscall.SyscallN(procDisconnectNamedPipe.Addr(), uintptr(pipe))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func getCurrentThread() (h windows.Handle) {
+	r0, _, _ := syscall.SyscallN(procGetCurrentThread.Addr())
+	h = windows.Handle(r0)
+	return
+}
+
+func getNamedPipeHandleState(pipe windows.Handle, state *uint32, curInstances *uint32, maxCollectionCount *uint32, collectDataTimeout *uint32, userName *uint16, maxUserNameSize uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procGetNamedPipeHandleStateW.Addr(), uintptr(pipe), uintptr(unsafe.Pointer(state)), uintptr(unsafe.Pointer(curInstances)), uintptr(unsafe.Pointer(maxCollectionCount)), uintptr(unsafe.Pointer(collectDataTimeout)), uintptr(unsafe.Pointer(userName)), uintptr(maxUserNameSize))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func getNamedPipeInfo(pipe windows.Handle, flags *uint32, outSize *uint32, inSize *uint32, maxInstances *uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procGetNamedPipeInfo.Addr(), uintptr(pipe), uintptr(unsafe.Pointer(flags)), uintptr(unsafe.Pointer(outSize)), uintptr(unsafe.Pointer(inSize)), uintptr(unsafe.Pointer(maxInstances)))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func getQueuedCompletionStatus(port windows.Handle, bytes *uint32, key *uintptr, o **ioOperation, timeout uint32) (err error) {
+	r1, _, e1 := syscall.SyscallN(procGetQueuedCompletionStatus.Addr(), uintptr(port), uintptr(unsafe.Pointer(bytes)), uintptr(unsafe.Pointer(key)), uintptr(unsafe.Pointer(o)), uintptr(timeout))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func setFileCompletionNotificationModes(h windows.Handle, flags uint8) (err error) {
+	r1, _, e1 := syscall.SyscallN(procSetFileCompletionNotificationModes.Addr(), uintptr(h), uintptr(flags))
+	if r1 == 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+func ntCreateNamedPipeFile(pipe *windows.Handle, access ntAccessMask, oa *objectAttributes, iosb *ioStatusBlock, share ntFileShareMode, disposition ntFileCreationDisposition, options ntFileOptions, typ uint32, readMode uint32, completionMode uint32, maxInstances uint32, inboundQuota uint32, outputQuota uint32, timeout *int64) (status ntStatus) {
+	r0, _, _ := syscall.SyscallN(procNtCreateNamedPipeFile.Addr(), uintptr(unsafe.Pointer(pipe)), uintptr(access), uintptr(unsafe.Pointer(oa)), uintptr(unsafe.Pointer(iosb)), uintptr(share), uintptr(disposition), uintptr(options), uintptr(typ), uintptr(readMode), uintptr(completionMode), uintptr(maxInstances), uintptr(inboundQuota), uintptr(outputQuota), uintptr(unsafe.Pointer(timeout)))
+	status = ntStatus(r0)
+	return
+}
+
+func rtlDefaultNpAcl(dacl *uintptr) (status ntStatus) {
+	r0, _, _ := syscall.SyscallN(procRtlDefaultNpAcl.Addr(), uintptr(unsafe.Pointer(dacl)))
+	status = ntStatus(r0)
+	return
+}
+
+func rtlDosPathNameToNtPathName(name *uint16, ntName *unicodeString, filePart uintptr, reserved uintptr) (status ntStatus) {
+	r0, _, _ := syscall.SyscallN(procRtlDosPathNameToNtPathName_U.Addr(), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(ntName)), uintptr(filePart), uintptr(reserved))
+	status = ntStatus(r0)
+	return
+}
+
+func rtlNtStatusToDosError(status ntStatus) (winerr error) {
+	r0, _, _ := syscall.SyscallN(procRtlNtStatusToDosErrorNoTeb.Addr(), uintptr(status))
+	if r0 != 0 {
+		winerr = syscall.Errno(r0)
+	}
+	return
+}
+
+func wsaGetOverlappedResult(h windows.Handle, o *windows.Overlapped, bytes *uint32, wait bool, flags *uint32) (err error) {
+	var _p0 uint32
+	if wait {
+		_p0 = 1
+	}
+	r1, _, e1 := syscall.SyscallN(procWSAGetOverlappedResult.Addr(), uintptr(h), uintptr(unsafe.Pointer(o)), uintptr(unsafe.Pointer(bytes)), uintptr(_p0), uintptr(unsafe.Pointer(flags)))
+	if r1 == 0 {
+		err = errnoErr(e1)
 	}
 	return
 }
@@ -0,0 +1 @@
+* text=auto eol=lf
@@ -0,0 +1 @@
+*.exe
@@ -0,0 +1,17 @@
+{
+    "Vendor": true,
+    "Deadline": "2m",
+    "Sort": [
+        "linter",
+        "severity",
+        "path",
+        "line"
+    ],
+    "Skip": [
+        "internal\\schema2"
+    ],
+    "EnableGC": true,
+    "Enable": [
+        "gofmt"
+    ]
+}
@@ -0,0 +1,3 @@
+* @microsoft/containerplat
+
+/hcn/* @nagiesek
@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 Microsoft
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
@@ -0,0 +1,54 @@
+version = "unstable"
+generator = "gogoctrd"
+plugins = ["grpc", "fieldpath"]
+
+# Control protoc include paths. Below are usually some good defaults, but feel
+# free to try it without them if it works for your project.
+[includes]
+  # Include paths that will be added before all others. Typically, you want to
+  # treat the root of the project as an include, but this may not be necessary.
+  before = ["./protobuf"]
+
+  # Paths that should be treated as include roots in relation to the vendor
+  # directory. These will be calculated with the vendor directory nearest the
+  # target package.
+  packages = ["github.com/gogo/protobuf"]
+
+  # Paths that will be added untouched to the end of the includes. We use
+  # `/usr/local/include` to pickup the common install location of protobuf.
+  # This is the default.
+  after = ["/usr/local/include"]
+
+# This section maps protobuf imports to Go packages. These will become
+# `-M` directives in the call to the go protobuf generator.
+[packages]
+  "gogoproto/gogo.proto" = "github.com/gogo/protobuf/gogoproto"
+  "google/protobuf/any.proto" = "github.com/gogo/protobuf/types"
+  "google/protobuf/empty.proto" = "github.com/gogo/protobuf/types"
+  "google/protobuf/descriptor.proto" = "github.com/gogo/protobuf/protoc-gen-gogo/descriptor"
+  "google/protobuf/field_mask.proto" = "github.com/gogo/protobuf/types"
+  "google/protobuf/timestamp.proto" = "github.com/gogo/protobuf/types"
+  "google/protobuf/duration.proto" = "github.com/gogo/protobuf/types"
+  "github/containerd/cgroups/stats/v1/metrics.proto" = "github.com/containerd/cgroups/stats/v1"
+
+[[overrides]]
+prefixes = ["github.com/Microsoft/hcsshim/internal/shimdiag"]
+plugins = ["ttrpc"]
+
+# Lock down runhcs config
+
+[[descriptors]]
+prefix = "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/options"
+target = "cmd/containerd-shim-runhcs-v1/options/next.pb.txt"
+ignore_files = [
+	"google/protobuf/descriptor.proto",
+	"gogoproto/gogo.proto"
+]
+
+[[descriptors]]
+prefix = "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/stats"
+target = "cmd/containerd-shim-runhcs-v1/stats/next.pb.txt"
+ignore_files = [
+	"google/protobuf/descriptor.proto",
+	"gogoproto/gogo.proto"
+]
@@ -0,0 +1,46 @@
+# hcsshim
+
+[![Build status](https://ci.appveyor.com/api/projects/status/nbcw28mnkqml0loa/branch/master?svg=true)](https://ci.appveyor.com/project/WindowsVirtualization/hcsshim/branch/master)
+
+This package contains the Golang interface for using the Windows [Host Compute Service](https://techcommunity.microsoft.com/t5/containers/introducing-the-host-compute-service-hcs/ba-p/382332) (HCS) to launch and manage [Windows Containers](https://docs.microsoft.com/en-us/virtualization/windowscontainers/about/). It also contains other helpers and functions for managing Windows Containers such as the Golang interface for the Host Network Service (HNS).
+
+It is primarily used in the [Moby Project](https://github.com/moby/moby), but it can be freely used by other projects as well.
+
+## Contributing
+
+This project welcomes contributions and suggestions.  Most contributions require you to agree to a
+Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us
+the rights to use your contribution. For details, visit https://cla.microsoft.com.
+
+When you submit a pull request, a CLA-bot will automatically determine whether you need to provide
+a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions
+provided by the bot. You will only need to do this once across all repos using our CLA.
+
+We also ask that contributors [sign their commits](https://git-scm.com/docs/git-commit) using `git commit -s` or `git commit --signoff` to certify they either authored the work themselves or otherwise have permission to use it in this project. 
+
+
+## Code of Conduct
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
+For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or
+contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.
+
+## Dependencies
+
+This project requires Golang 1.9 or newer to build.
+
+For system requirements to run this project, see the Microsoft docs on [Windows Container requirements](https://docs.microsoft.com/en-us/virtualization/windowscontainers/deploy-containers/system-requirements).
+
+## Reporting Security Issues
+
+Security issues and bugs should be reported privately, via email, to the Microsoft Security
+Response Center (MSRC) at [secure@microsoft.com](mailto:secure@microsoft.com). You should
+receive a response within 24 hours. If for some reason you do not, please follow up via
+email to ensure we received your original message. Further information, including the
+[MSRC PGP](https://technet.microsoft.com/en-us/security/dn606155) key, can be found in
+the [Security TechCenter](https://technet.microsoft.com/en-us/security/default).
+
+For additional details, see [Report a Computer Security Vulnerability](https://technet.microsoft.com/en-us/security/ff852094.aspx) on Technet
+
+---------------
+Copyright (c) 2018 Microsoft Corp.  All rights reserved.
@@ -0,0 +1,45 @@
+version: 0.1.{build}
+
+image: Visual Studio 2017
+
+clone_folder: c:\gopath\src\github.com\Microsoft\hcsshim
+
+environment:
+  GOPATH: c:\gopath
+  PATH: "%GOPATH%\\bin;C:\\gometalinter-2.0.12-windows-amd64;%PATH%"
+
+stack: go 1.13.4
+
+build_script:
+  - appveyor DownloadFile https://github.com/alecthomas/gometalinter/releases/download/v2.0.12/gometalinter-2.0.12-windows-amd64.zip
+  - 7z x gometalinter-2.0.12-windows-amd64.zip -y -oC:\ > NUL
+  - gometalinter.exe --config .gometalinter.json ./...
+  - go build ./cmd/containerd-shim-runhcs-v1
+  - go build ./cmd/runhcs
+  - go build ./cmd/tar2ext4
+  - go build ./cmd/wclayer
+  - go build ./cmd/device-util
+  - go build ./internal/tools/grantvmgroupaccess 
+  - go build ./internal/tools/uvmboot
+  - go build ./internal/tools/zapdir
+  - go test -v ./... -tags admin
+  - cd test
+  - go test -v ./internal -tags admin
+  - go test -c ./containerd-shim-runhcs-v1/ -tags functional
+  - go test -c ./cri-containerd/ -tags functional
+  - go test -c ./functional/ -tags functional
+  - go test -c ./runhcs/ -tags functional
+
+artifacts:
+  - path: 'containerd-shim-runhcs-v1.exe'
+  - path: 'runhcs.exe'
+  - path: 'tar2ext4.exe'
+  - path: 'device-util.exe'
+  - path: 'wclayer.exe'
+  - path: 'grantvmgroupaccess.exe'  
+  - path: 'uvmboot.exe'
+  - path: 'zapdir.exe'
+  - path: './test/containerd-shim-runhcs-v1.test.exe'
+  - path: './test/cri-containerd.test.exe'
+  - path: './test/functional.test.exe'
+  - path: './test/runhcs.test.exe'
@@ -0,0 +1,38 @@
+package computestorage
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// AttachLayerStorageFilter sets up the layer storage filter on a writable
+// container layer.
+//
+// `layerPath` is a path to a directory the writable layer is mounted. If the
+// path does not end in a `\` the platform will append it automatically.
+//
+// `layerData` is the parent read-only layer data.
+func AttachLayerStorageFilter(ctx context.Context, layerPath string, layerData LayerData) (err error) {
+	title := "hcsshim.AttachLayerStorageFilter"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+	)
+
+	bytes, err := json.Marshal(layerData)
+	if err != nil {
+		return err
+	}
+
+	err = hcsAttachLayerStorageFilter(layerPath, string(bytes))
+	if err != nil {
+		return fmt.Errorf("failed to attach layer storage filter: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,26 @@
+package computestorage
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// DestroyLayer deletes a container layer.
+//
+// `layerPath` is a path to a directory containing the layer to export.
+func DestroyLayer(ctx context.Context, layerPath string) (err error) {
+	title := "hcsshim.DestroyLayer"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("layerPath", layerPath))
+
+	err = hcsDestroyLayer(layerPath)
+	if err != nil {
+		return fmt.Errorf("failed to destroy layer: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,26 @@
+package computestorage
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// DetachLayerStorageFilter detaches the layer storage filter on a writable container layer.
+//
+// `layerPath` is a path to a directory containing the layer to export.
+func DetachLayerStorageFilter(ctx context.Context, layerPath string) (err error) {
+	title := "hcsshim.DetachLayerStorageFilter"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("layerPath", layerPath))
+
+	err = hcsDetachLayerStorageFilter(layerPath)
+	if err != nil {
+		return fmt.Errorf("failed to detach layer storage filter: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,46 @@
+package computestorage
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// ExportLayer exports a container layer.
+//
+// `layerPath` is a path to a directory containing the layer to export.
+//
+// `exportFolderPath` is a pre-existing folder to export the layer to.
+//
+// `layerData` is the parent layer data.
+//
+// `options` are the export options applied to the exported layer.
+func ExportLayer(ctx context.Context, layerPath, exportFolderPath string, layerData LayerData, options ExportLayerOptions) (err error) {
+	title := "hcsshim.ExportLayer"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+		trace.StringAttribute("exportFolderPath", exportFolderPath),
+	)
+
+	ldbytes, err := json.Marshal(layerData)
+	if err != nil {
+		return err
+	}
+
+	obytes, err := json.Marshal(options)
+	if err != nil {
+		return err
+	}
+
+	err = hcsExportLayer(layerPath, exportFolderPath, string(ldbytes), string(obytes))
+	if err != nil {
+		return fmt.Errorf("failed to export layer: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,26 @@
+package computestorage
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+	"golang.org/x/sys/windows"
+)
+
+// FormatWritableLayerVhd formats a virtual disk for use as a writable container layer.
+//
+// If the VHD is not mounted it will be temporarily mounted.
+func FormatWritableLayerVhd(ctx context.Context, vhdHandle windows.Handle) (err error) {
+	title := "hcsshim.FormatWritableLayerVhd"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+
+	err = hcsFormatWritableLayerVhd(vhdHandle)
+	if err != nil {
+		return fmt.Errorf("failed to format writable layer vhd: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,197 @@
+package computestorage
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"syscall"
+
+	"github.com/Microsoft/go-winio/pkg/security"
+	"github.com/Microsoft/go-winio/vhd"
+	"github.com/pkg/errors"
+	"golang.org/x/sys/windows"
+)
+
+const defaultVHDXBlockSizeInMB = 1
+
+// SetupContainerBaseLayer is a helper to setup a containers scratch. It
+// will create and format the vhdx's inside and the size is configurable with the sizeInGB
+// parameter.
+//
+// `layerPath` is the path to the base container layer on disk.
+//
+// `baseVhdPath` is the path to where the base vhdx for the base layer should be created.
+//
+// `diffVhdPath` is the path where the differencing disk for the base layer should be created.
+//
+// `sizeInGB` is the size in gigabytes to make the base vhdx.
+func SetupContainerBaseLayer(ctx context.Context, layerPath, baseVhdPath, diffVhdPath string, sizeInGB uint64) (err error) {
+	var (
+		hivesPath  = filepath.Join(layerPath, "Hives")
+		layoutPath = filepath.Join(layerPath, "Layout")
+	)
+
+	// We need to remove the hives directory and layout file as `SetupBaseOSLayer` fails if these files
+	// already exist. `SetupBaseOSLayer` will create these files internally. We also remove the base and
+	// differencing disks if they exist in case we're asking for a different size.
+	if _, err := os.Stat(hivesPath); err == nil {
+		if err := os.RemoveAll(hivesPath); err != nil {
+			return errors.Wrap(err, "failed to remove prexisting hives directory")
+		}
+	}
+	if _, err := os.Stat(layoutPath); err == nil {
+		if err := os.RemoveAll(layoutPath); err != nil {
+			return errors.Wrap(err, "failed to remove prexisting layout file")
+		}
+	}
+
+	if _, err := os.Stat(baseVhdPath); err == nil {
+		if err := os.RemoveAll(baseVhdPath); err != nil {
+			return errors.Wrap(err, "failed to remove base vhdx path")
+		}
+	}
+	if _, err := os.Stat(diffVhdPath); err == nil {
+		if err := os.RemoveAll(diffVhdPath); err != nil {
+			return errors.Wrap(err, "failed to remove differencing vhdx")
+		}
+	}
+
+	createParams := &vhd.CreateVirtualDiskParameters{
+		Version: 2,
+		Version2: vhd.CreateVersion2{
+			MaximumSize:      sizeInGB * 1024 * 1024 * 1024,
+			BlockSizeInBytes: defaultVHDXBlockSizeInMB * 1024 * 1024,
+		},
+	}
+	handle, err := vhd.CreateVirtualDisk(baseVhdPath, vhd.VirtualDiskAccessNone, vhd.CreateVirtualDiskFlagNone, createParams)
+	if err != nil {
+		return errors.Wrap(err, "failed to create vhdx")
+	}
+
+	defer func() {
+		if err != nil {
+			syscall.CloseHandle(handle)
+			os.RemoveAll(baseVhdPath)
+			if os.Stat(diffVhdPath); err == nil {
+				os.RemoveAll(diffVhdPath)
+			}
+		}
+	}()
+
+	if err = FormatWritableLayerVhd(ctx, windows.Handle(handle)); err != nil {
+		return err
+	}
+	// Base vhd handle must be closed before calling SetupBaseLayer in case of Container layer
+	if err = syscall.CloseHandle(handle); err != nil {
+		return errors.Wrap(err, "failed to close vhdx handle")
+	}
+
+	options := OsLayerOptions{
+		Type: OsLayerTypeContainer,
+	}
+
+	// SetupBaseOSLayer expects an empty vhd handle for a container layer and will
+	// error out otherwise.
+	if err = SetupBaseOSLayer(ctx, layerPath, 0, options); err != nil {
+		return err
+	}
+	// Create the differencing disk that will be what's copied for the final rw layer
+	// for a container.
+	if err = vhd.CreateDiffVhd(diffVhdPath, baseVhdPath, defaultVHDXBlockSizeInMB); err != nil {
+		return errors.Wrap(err, "failed to create differencing disk")
+	}
+
+	if err = security.GrantVmGroupAccess(baseVhdPath); err != nil {
+		return errors.Wrapf(err, "failed to grant vm group access to %s", baseVhdPath)
+	}
+	if err = security.GrantVmGroupAccess(diffVhdPath); err != nil {
+		return errors.Wrapf(err, "failed to grant vm group access to %s", diffVhdPath)
+	}
+	return nil
+}
+
+// SetupUtilityVMBaseLayer is a helper to setup a UVMs scratch space. It will create and format
+// the vhdx inside and the size is configurable by the sizeInGB parameter.
+//
+// `uvmPath` is the path to the UtilityVM filesystem.
+//
+// `baseVhdPath` is the path to where the base vhdx for the UVM should be created.
+//
+// `diffVhdPath` is the path where the differencing disk for the UVM should be created.
+//
+// `sizeInGB` specifies the size in gigabytes to make the base vhdx.
+func SetupUtilityVMBaseLayer(ctx context.Context, uvmPath, baseVhdPath, diffVhdPath string, sizeInGB uint64) (err error) {
+	// Remove the base and differencing disks if they exist in case we're asking for a different size.
+	if _, err := os.Stat(baseVhdPath); err == nil {
+		if err := os.RemoveAll(baseVhdPath); err != nil {
+			return errors.Wrap(err, "failed to remove base vhdx")
+		}
+	}
+	if _, err := os.Stat(diffVhdPath); err == nil {
+		if err := os.RemoveAll(diffVhdPath); err != nil {
+			return errors.Wrap(err, "failed to remove differencing vhdx")
+		}
+	}
+
+	// Just create the vhdx for utilityVM layer, no need to format it.
+	createParams := &vhd.CreateVirtualDiskParameters{
+		Version: 2,
+		Version2: vhd.CreateVersion2{
+			MaximumSize:      sizeInGB * 1024 * 1024 * 1024,
+			BlockSizeInBytes: defaultVHDXBlockSizeInMB * 1024 * 1024,
+		},
+	}
+	handle, err := vhd.CreateVirtualDisk(baseVhdPath, vhd.VirtualDiskAccessNone, vhd.CreateVirtualDiskFlagNone, createParams)
+	if err != nil {
+		return errors.Wrap(err, "failed to create vhdx")
+	}
+
+	defer func() {
+		if err != nil {
+			syscall.CloseHandle(handle)
+			os.RemoveAll(baseVhdPath)
+			if os.Stat(diffVhdPath); err == nil {
+				os.RemoveAll(diffVhdPath)
+			}
+		}
+	}()
+
+	// If it is a UtilityVM layer then the base vhdx must be attached when calling
+	// `SetupBaseOSLayer`
+	attachParams := &vhd.AttachVirtualDiskParameters{
+		Version: 2,
+	}
+	if err := vhd.AttachVirtualDisk(handle, vhd.AttachVirtualDiskFlagNone, attachParams); err != nil {
+		return errors.Wrapf(err, "failed to attach virtual disk")
+	}
+
+	options := OsLayerOptions{
+		Type: OsLayerTypeVM,
+	}
+	if err := SetupBaseOSLayer(ctx, uvmPath, windows.Handle(handle), options); err != nil {
+		return err
+	}
+
+	// Detach and close the handle after setting up the layer as we don't need the handle
+	// for anything else and we no longer need to be attached either.
+	if err = vhd.DetachVirtualDisk(handle); err != nil {
+		return errors.Wrap(err, "failed to detach vhdx")
+	}
+	if err = syscall.CloseHandle(handle); err != nil {
+		return errors.Wrap(err, "failed to close vhdx handle")
+	}
+
+	// Create the differencing disk that will be what's copied for the final rw layer
+	// for a container.
+	if err = vhd.CreateDiffVhd(diffVhdPath, baseVhdPath, defaultVHDXBlockSizeInMB); err != nil {
+		return errors.Wrap(err, "failed to create differencing disk")
+	}
+
+	if err := security.GrantVmGroupAccess(baseVhdPath); err != nil {
+		return errors.Wrapf(err, "failed to grant vm group access to %s", baseVhdPath)
+	}
+	if err := security.GrantVmGroupAccess(diffVhdPath); err != nil {
+		return errors.Wrapf(err, "failed to grant vm group access to %s", diffVhdPath)
+	}
+	return nil
+}
@@ -0,0 +1,41 @@
+package computestorage
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// ImportLayer imports a container layer.
+//
+// `layerPath` is a path to a directory to import the layer to. If the directory
+// does not exist it will be automatically created.
+//
+// `sourceFolderpath` is a pre-existing folder that contains the layer to
+// import.
+//
+// `layerData` is the parent layer data.
+func ImportLayer(ctx context.Context, layerPath, sourceFolderPath string, layerData LayerData) (err error) {
+	title := "hcsshim.ImportLayer"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+		trace.StringAttribute("sourceFolderPath", sourceFolderPath),
+	)
+
+	bytes, err := json.Marshal(layerData)
+	if err != nil {
+		return err
+	}
+
+	err = hcsImportLayer(layerPath, sourceFolderPath, string(bytes))
+	if err != nil {
+		return fmt.Errorf("failed to import layer: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,38 @@
+package computestorage
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+)
+
+// InitializeWritableLayer initializes a writable layer for a container.
+//
+// `layerPath` is a path to a directory the layer is mounted. If the
+// path does not end in a `\` the platform will append it automatically.
+//
+// `layerData` is the parent read-only layer data.
+func InitializeWritableLayer(ctx context.Context, layerPath string, layerData LayerData) (err error) {
+	title := "hcsshim.InitializeWritableLayer"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+	)
+
+	bytes, err := json.Marshal(layerData)
+	if err != nil {
+		return err
+	}
+
+	// Options are not used in the platform as of RS5
+	err = hcsInitializeWritableLayer(layerPath, string(bytes), "")
+	if err != nil {
+		return fmt.Errorf("failed to intitialize container layer: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,27 @@
+package computestorage
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/interop"
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"go.opencensus.io/trace"
+	"golang.org/x/sys/windows"
+)
+
+// GetLayerVhdMountPath returns the volume path for a virtual disk of a writable container layer.
+func GetLayerVhdMountPath(ctx context.Context, vhdHandle windows.Handle) (path string, err error) {
+	title := "hcsshim.GetLayerVhdMountPath"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+
+	var mountPath *uint16
+	err = hcsGetLayerVhdMountPath(vhdHandle, &mountPath)
+	if err != nil {
+		return "", fmt.Errorf("failed to get vhd mount path: %s", err)
+	}
+	path = interop.ConvertAndFreeCoTaskMemString(mountPath)
+	return path, nil
+}
@@ -0,0 +1,75 @@
+package computestorage
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"github.com/Microsoft/hcsshim/osversion"
+	"go.opencensus.io/trace"
+	"golang.org/x/sys/windows"
+)
+
+// SetupBaseOSLayer sets up a layer that contains a base OS for a container.
+//
+// `layerPath` is a path to a directory containing the layer.
+//
+// `vhdHandle` is an empty file handle of `options.Type == OsLayerTypeContainer`
+// or else it is a file handle to the 'SystemTemplateBase.vhdx' if `options.Type
+// == OsLayerTypeVm`.
+//
+// `options` are the options applied while processing the layer.
+func SetupBaseOSLayer(ctx context.Context, layerPath string, vhdHandle windows.Handle, options OsLayerOptions) (err error) {
+	title := "hcsshim.SetupBaseOSLayer"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+	)
+
+	bytes, err := json.Marshal(options)
+	if err != nil {
+		return err
+	}
+
+	err = hcsSetupBaseOSLayer(layerPath, vhdHandle, string(bytes))
+	if err != nil {
+		return fmt.Errorf("failed to setup base OS layer: %s", err)
+	}
+	return nil
+}
+
+// SetupBaseOSVolume sets up a volume that contains a base OS for a container.
+//
+// `layerPath` is a path to a directory containing the layer.
+//
+// `volumePath` is the path to the volume to be used for setup.
+//
+// `options` are the options applied while processing the layer.
+func SetupBaseOSVolume(ctx context.Context, layerPath, volumePath string, options OsLayerOptions) (err error) {
+	if osversion.Get().Build < 19645 {
+		return errors.New("SetupBaseOSVolume is not present on builds older than 19645")
+	}
+	title := "hcsshim.SetupBaseOSVolume"
+	ctx, span := trace.StartSpan(ctx, title)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("layerPath", layerPath),
+		trace.StringAttribute("volumePath", volumePath),
+	)
+
+	bytes, err := json.Marshal(options)
+	if err != nil {
+		return err
+	}
+
+	err = hcsSetupBaseOSVolume(layerPath, volumePath, string(bytes))
+	if err != nil {
+		return fmt.Errorf("failed to setup base OS layer: %s", err)
+	}
+	return nil
+}
@@ -0,0 +1,50 @@
+// Package computestorage is a wrapper around the HCS storage APIs. These are new storage APIs introduced
+// separate from the original graphdriver calls intended to give more freedom around creating
+// and managing container layers and scratch spaces.
+package computestorage
+
+import (
+	hcsschema "github.com/Microsoft/hcsshim/internal/schema2"
+)
+
+//go:generate go run ../mksyscall_windows.go -output zsyscall_windows.go storage.go
+
+//sys hcsImportLayer(layerPath string, sourceFolderPath string, layerData string) (hr error) = computestorage.HcsImportLayer?
+//sys hcsExportLayer(layerPath string, exportFolderPath string, layerData string, options string) (hr error) = computestorage.HcsExportLayer?
+//sys hcsDestroyLayer(layerPath string) (hr error) = computestorage.HcsDestoryLayer?
+//sys hcsSetupBaseOSLayer(layerPath string, handle windows.Handle, options string) (hr error) = computestorage.HcsSetupBaseOSLayer?
+//sys hcsInitializeWritableLayer(writableLayerPath string, layerData string, options string) (hr error) = computestorage.HcsInitializeWritableLayer?
+//sys hcsAttachLayerStorageFilter(layerPath string, layerData string) (hr error) = computestorage.HcsAttachLayerStorageFilter?
+//sys hcsDetachLayerStorageFilter(layerPath string) (hr error) = computestorage.HcsDetachLayerStorageFilter?
+//sys hcsFormatWritableLayerVhd(handle windows.Handle) (hr error) = computestorage.HcsFormatWritableLayerVhd?
+//sys hcsGetLayerVhdMountPath(vhdHandle windows.Handle, mountPath **uint16) (hr error) = computestorage.HcsGetLayerVhdMountPath?
+//sys hcsSetupBaseOSVolume(layerPath string, volumePath string, options string) (hr error) = computestorage.HcsSetupBaseOSVolume?
+
+// LayerData is the data used to describe parent layer information.
+type LayerData struct {
+	SchemaVersion hcsschema.Version `json:"SchemaVersion,omitempty"`
+	Layers        []hcsschema.Layer `json:"Layers,omitempty"`
+}
+
+// ExportLayerOptions are the set of options that are used with the `computestorage.HcsExportLayer` syscall.
+type ExportLayerOptions struct {
+	IsWritableLayer bool `json:"IsWritableLayer,omitempty"`
+}
+
+// OsLayerType is the type of layer being operated on.
+type OsLayerType string
+
+const (
+	// OsLayerTypeContainer is a container layer.
+	OsLayerTypeContainer OsLayerType = "Container"
+	// OsLayerTypeVM is a virtual machine layer.
+	OsLayerTypeVM OsLayerType = "Vm"
+)
+
+// OsLayerOptions are the set of options that are used with the `SetupBaseOSLayer` and
+// `SetupBaseOSVolume` calls.
+type OsLayerOptions struct {
+	Type                       OsLayerType `json:"Type,omitempty"`
+	DisableCiCacheOptimization bool        `json:"DisableCiCacheOptimization,omitempty"`
+	SkipUpdateBcdForBoot       bool        `json:"SkipUpdateBcdForBoot,omitempty"`
+}
@@ -0,0 +1,319 @@
+// Code generated mksyscall_windows.exe DO NOT EDIT
+
+package computestorage
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return nil
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	// TODO: add more here, after collecting data on the common
+	// error values see on Windows. (perhaps when running
+	// all.bat?)
+	return e
+}
+
+var (
+	modcomputestorage = windows.NewLazySystemDLL("computestorage.dll")
+
+	procHcsImportLayer              = modcomputestorage.NewProc("HcsImportLayer")
+	procHcsExportLayer              = modcomputestorage.NewProc("HcsExportLayer")
+	procHcsDestoryLayer             = modcomputestorage.NewProc("HcsDestoryLayer")
+	procHcsSetupBaseOSLayer         = modcomputestorage.NewProc("HcsSetupBaseOSLayer")
+	procHcsInitializeWritableLayer  = modcomputestorage.NewProc("HcsInitializeWritableLayer")
+	procHcsAttachLayerStorageFilter = modcomputestorage.NewProc("HcsAttachLayerStorageFilter")
+	procHcsDetachLayerStorageFilter = modcomputestorage.NewProc("HcsDetachLayerStorageFilter")
+	procHcsFormatWritableLayerVhd   = modcomputestorage.NewProc("HcsFormatWritableLayerVhd")
+	procHcsGetLayerVhdMountPath     = modcomputestorage.NewProc("HcsGetLayerVhdMountPath")
+	procHcsSetupBaseOSVolume        = modcomputestorage.NewProc("HcsSetupBaseOSVolume")
+)
+
+func hcsImportLayer(layerPath string, sourceFolderPath string, layerData string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(sourceFolderPath)
+	if hr != nil {
+		return
+	}
+	var _p2 *uint16
+	_p2, hr = syscall.UTF16PtrFromString(layerData)
+	if hr != nil {
+		return
+	}
+	return _hcsImportLayer(_p0, _p1, _p2)
+}
+
+func _hcsImportLayer(layerPath *uint16, sourceFolderPath *uint16, layerData *uint16) (hr error) {
+	if hr = procHcsImportLayer.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsImportLayer.Addr(), 3, uintptr(unsafe.Pointer(layerPath)), uintptr(unsafe.Pointer(sourceFolderPath)), uintptr(unsafe.Pointer(layerData)))
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsExportLayer(layerPath string, exportFolderPath string, layerData string, options string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(exportFolderPath)
+	if hr != nil {
+		return
+	}
+	var _p2 *uint16
+	_p2, hr = syscall.UTF16PtrFromString(layerData)
+	if hr != nil {
+		return
+	}
+	var _p3 *uint16
+	_p3, hr = syscall.UTF16PtrFromString(options)
+	if hr != nil {
+		return
+	}
+	return _hcsExportLayer(_p0, _p1, _p2, _p3)
+}
+
+func _hcsExportLayer(layerPath *uint16, exportFolderPath *uint16, layerData *uint16, options *uint16) (hr error) {
+	if hr = procHcsExportLayer.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall6(procHcsExportLayer.Addr(), 4, uintptr(unsafe.Pointer(layerPath)), uintptr(unsafe.Pointer(exportFolderPath)), uintptr(unsafe.Pointer(layerData)), uintptr(unsafe.Pointer(options)), 0, 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsDestroyLayer(layerPath string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	return _hcsDestroyLayer(_p0)
+}
+
+func _hcsDestroyLayer(layerPath *uint16) (hr error) {
+	if hr = procHcsDestoryLayer.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsDestoryLayer.Addr(), 1, uintptr(unsafe.Pointer(layerPath)), 0, 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsSetupBaseOSLayer(layerPath string, handle windows.Handle, options string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(options)
+	if hr != nil {
+		return
+	}
+	return _hcsSetupBaseOSLayer(_p0, handle, _p1)
+}
+
+func _hcsSetupBaseOSLayer(layerPath *uint16, handle windows.Handle, options *uint16) (hr error) {
+	if hr = procHcsSetupBaseOSLayer.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsSetupBaseOSLayer.Addr(), 3, uintptr(unsafe.Pointer(layerPath)), uintptr(handle), uintptr(unsafe.Pointer(options)))
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsInitializeWritableLayer(writableLayerPath string, layerData string, options string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(writableLayerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(layerData)
+	if hr != nil {
+		return
+	}
+	var _p2 *uint16
+	_p2, hr = syscall.UTF16PtrFromString(options)
+	if hr != nil {
+		return
+	}
+	return _hcsInitializeWritableLayer(_p0, _p1, _p2)
+}
+
+func _hcsInitializeWritableLayer(writableLayerPath *uint16, layerData *uint16, options *uint16) (hr error) {
+	if hr = procHcsInitializeWritableLayer.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsInitializeWritableLayer.Addr(), 3, uintptr(unsafe.Pointer(writableLayerPath)), uintptr(unsafe.Pointer(layerData)), uintptr(unsafe.Pointer(options)))
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsAttachLayerStorageFilter(layerPath string, layerData string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(layerData)
+	if hr != nil {
+		return
+	}
+	return _hcsAttachLayerStorageFilter(_p0, _p1)
+}
+
+func _hcsAttachLayerStorageFilter(layerPath *uint16, layerData *uint16) (hr error) {
+	if hr = procHcsAttachLayerStorageFilter.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsAttachLayerStorageFilter.Addr(), 2, uintptr(unsafe.Pointer(layerPath)), uintptr(unsafe.Pointer(layerData)), 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsDetachLayerStorageFilter(layerPath string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	return _hcsDetachLayerStorageFilter(_p0)
+}
+
+func _hcsDetachLayerStorageFilter(layerPath *uint16) (hr error) {
+	if hr = procHcsDetachLayerStorageFilter.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsDetachLayerStorageFilter.Addr(), 1, uintptr(unsafe.Pointer(layerPath)), 0, 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsFormatWritableLayerVhd(handle windows.Handle) (hr error) {
+	if hr = procHcsFormatWritableLayerVhd.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsFormatWritableLayerVhd.Addr(), 1, uintptr(handle), 0, 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsGetLayerVhdMountPath(vhdHandle windows.Handle, mountPath **uint16) (hr error) {
+	if hr = procHcsGetLayerVhdMountPath.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsGetLayerVhdMountPath.Addr(), 2, uintptr(vhdHandle), uintptr(unsafe.Pointer(mountPath)), 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
+
+func hcsSetupBaseOSVolume(layerPath string, volumePath string, options string) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(layerPath)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(volumePath)
+	if hr != nil {
+		return
+	}
+	var _p2 *uint16
+	_p2, hr = syscall.UTF16PtrFromString(options)
+	if hr != nil {
+		return
+	}
+	return _hcsSetupBaseOSVolume(_p0, _p1, _p2)
+}
+
+func _hcsSetupBaseOSVolume(layerPath *uint16, volumePath *uint16, options *uint16) (hr error) {
+	if hr = procHcsSetupBaseOSVolume.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall(procHcsSetupBaseOSVolume.Addr(), 3, uintptr(unsafe.Pointer(layerPath)), uintptr(unsafe.Pointer(volumePath)), uintptr(unsafe.Pointer(options)))
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
@@ -0,0 +1,223 @@
+package hcsshim
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"sync"
+	"time"
+
+	"github.com/Microsoft/hcsshim/internal/hcs"
+	"github.com/Microsoft/hcsshim/internal/mergemaps"
+	"github.com/Microsoft/hcsshim/internal/schema1"
+)
+
+// ContainerProperties holds the properties for a container and the processes running in that container
+type ContainerProperties = schema1.ContainerProperties
+
+// MemoryStats holds the memory statistics for a container
+type MemoryStats = schema1.MemoryStats
+
+// ProcessorStats holds the processor statistics for a container
+type ProcessorStats = schema1.ProcessorStats
+
+// StorageStats holds the storage statistics for a container
+type StorageStats = schema1.StorageStats
+
+// NetworkStats holds the network statistics for a container
+type NetworkStats = schema1.NetworkStats
+
+// Statistics is the structure returned by a statistics call on a container
+type Statistics = schema1.Statistics
+
+// ProcessList is the structure of an item returned by a ProcessList call on a container
+type ProcessListItem = schema1.ProcessListItem
+
+// MappedVirtualDiskController is the structure of an item returned by a MappedVirtualDiskList call on a container
+type MappedVirtualDiskController = schema1.MappedVirtualDiskController
+
+// Type of Request Support in ModifySystem
+type RequestType = schema1.RequestType
+
+// Type of Resource Support in ModifySystem
+type ResourceType = schema1.ResourceType
+
+// RequestType const
+const (
+	Add     = schema1.Add
+	Remove  = schema1.Remove
+	Network = schema1.Network
+)
+
+// ResourceModificationRequestResponse is the structure used to send request to the container to modify the system
+// Supported resource types are Network and Request Types are Add/Remove
+type ResourceModificationRequestResponse = schema1.ResourceModificationRequestResponse
+
+type container struct {
+	system   *hcs.System
+	waitOnce sync.Once
+	waitErr  error
+	waitCh   chan struct{}
+}
+
+// createComputeSystemAdditionalJSON is read from the environment at initialisation
+// time. It allows an environment variable to define additional JSON which
+// is merged in the CreateComputeSystem call to HCS.
+var createContainerAdditionalJSON []byte
+
+func init() {
+	createContainerAdditionalJSON = ([]byte)(os.Getenv("HCSSHIM_CREATECONTAINER_ADDITIONALJSON"))
+}
+
+// CreateContainer creates a new container with the given configuration but does not start it.
+func CreateContainer(id string, c *ContainerConfig) (Container, error) {
+	fullConfig, err := mergemaps.MergeJSON(c, createContainerAdditionalJSON)
+	if err != nil {
+		return nil, fmt.Errorf("failed to merge additional JSON '%s': %s", createContainerAdditionalJSON, err)
+	}
+
+	system, err := hcs.CreateComputeSystem(context.Background(), id, fullConfig)
+	if err != nil {
+		return nil, err
+	}
+	return &container{system: system}, err
+}
+
+// OpenContainer opens an existing container by ID.
+func OpenContainer(id string) (Container, error) {
+	system, err := hcs.OpenComputeSystem(context.Background(), id)
+	if err != nil {
+		return nil, err
+	}
+	return &container{system: system}, err
+}
+
+// GetContainers gets a list of the containers on the system that match the query
+func GetContainers(q ComputeSystemQuery) ([]ContainerProperties, error) {
+	return hcs.GetComputeSystems(context.Background(), q)
+}
+
+// Start synchronously starts the container.
+func (container *container) Start() error {
+	return convertSystemError(container.system.Start(context.Background()), container)
+}
+
+// Shutdown requests a container shutdown, but it may not actually be shutdown until Wait() succeeds.
+func (container *container) Shutdown() error {
+	err := container.system.Shutdown(context.Background())
+	if err != nil {
+		return convertSystemError(err, container)
+	}
+	return &ContainerError{Container: container, Err: ErrVmcomputeOperationPending, Operation: "hcsshim::ComputeSystem::Shutdown"}
+}
+
+// Terminate requests a container terminate, but it may not actually be terminated until Wait() succeeds.
+func (container *container) Terminate() error {
+	err := container.system.Terminate(context.Background())
+	if err != nil {
+		return convertSystemError(err, container)
+	}
+	return &ContainerError{Container: container, Err: ErrVmcomputeOperationPending, Operation: "hcsshim::ComputeSystem::Terminate"}
+}
+
+// Waits synchronously waits for the container to shutdown or terminate.
+func (container *container) Wait() error {
+	err := container.system.Wait()
+	if err == nil {
+		err = container.system.ExitError()
+	}
+	return convertSystemError(err, container)
+}
+
+// WaitTimeout synchronously waits for the container to terminate or the duration to elapse. It
+// returns false if timeout occurs.
+func (container *container) WaitTimeout(timeout time.Duration) error {
+	container.waitOnce.Do(func() {
+		container.waitCh = make(chan struct{})
+		go func() {
+			container.waitErr = container.Wait()
+			close(container.waitCh)
+		}()
+	})
+	t := time.NewTimer(timeout)
+	defer t.Stop()
+	select {
+	case <-t.C:
+		return &ContainerError{Container: container, Err: ErrTimeout, Operation: "hcsshim::ComputeSystem::Wait"}
+	case <-container.waitCh:
+		return container.waitErr
+	}
+}
+
+// Pause pauses the execution of a container.
+func (container *container) Pause() error {
+	return convertSystemError(container.system.Pause(context.Background()), container)
+}
+
+// Resume resumes the execution of a container.
+func (container *container) Resume() error {
+	return convertSystemError(container.system.Resume(context.Background()), container)
+}
+
+// HasPendingUpdates returns true if the container has updates pending to install
+func (container *container) HasPendingUpdates() (bool, error) {
+	return false, nil
+}
+
+// Statistics returns statistics for the container. This is a legacy v1 call
+func (container *container) Statistics() (Statistics, error) {
+	properties, err := container.system.Properties(context.Background(), schema1.PropertyTypeStatistics)
+	if err != nil {
+		return Statistics{}, convertSystemError(err, container)
+	}
+
+	return properties.Statistics, nil
+}
+
+// ProcessList returns an array of ProcessListItems for the container. This is a legacy v1 call
+func (container *container) ProcessList() ([]ProcessListItem, error) {
+	properties, err := container.system.Properties(context.Background(), schema1.PropertyTypeProcessList)
+	if err != nil {
+		return nil, convertSystemError(err, container)
+	}
+
+	return properties.ProcessList, nil
+}
+
+// This is a legacy v1 call
+func (container *container) MappedVirtualDisks() (map[int]MappedVirtualDiskController, error) {
+	properties, err := container.system.Properties(context.Background(), schema1.PropertyTypeMappedVirtualDisk)
+	if err != nil {
+		return nil, convertSystemError(err, container)
+	}
+
+	return properties.MappedVirtualDiskControllers, nil
+}
+
+// CreateProcess launches a new process within the container.
+func (container *container) CreateProcess(c *ProcessConfig) (Process, error) {
+	p, err := container.system.CreateProcess(context.Background(), c)
+	if err != nil {
+		return nil, convertSystemError(err, container)
+	}
+	return &process{p: p.(*hcs.Process)}, nil
+}
+
+// OpenProcess gets an interface to an existing process within the container.
+func (container *container) OpenProcess(pid int) (Process, error) {
+	p, err := container.system.OpenProcess(context.Background(), pid)
+	if err != nil {
+		return nil, convertSystemError(err, container)
+	}
+	return &process{p: p}, nil
+}
+
+// Close cleans up any state associated with the container but does not terminate or wait for it.
+func (container *container) Close() error {
+	return convertSystemError(container.system.Close(), container)
+}
+
+// Modify the System
+func (container *container) Modify(config *ResourceModificationRequestResponse) error {
+	return convertSystemError(container.system.Modify(context.Background(), config), container)
+}
@@ -0,0 +1,257 @@
+package hcsshim
+
+import (
+	"fmt"
+	"syscall"
+
+	"github.com/Microsoft/hcsshim/internal/hns"
+
+	"github.com/Microsoft/hcsshim/internal/hcs"
+	"github.com/Microsoft/hcsshim/internal/hcserror"
+)
+
+var (
+	// ErrComputeSystemDoesNotExist is an error encountered when the container being operated on no longer exists = hcs.exist
+	ErrComputeSystemDoesNotExist = hcs.ErrComputeSystemDoesNotExist
+
+	// ErrElementNotFound is an error encountered when the object being referenced does not exist
+	ErrElementNotFound = hcs.ErrElementNotFound
+
+	// ErrElementNotFound is an error encountered when the object being referenced does not exist
+	ErrNotSupported = hcs.ErrNotSupported
+
+	// ErrInvalidData is an error encountered when the request being sent to hcs is invalid/unsupported
+	// decimal -2147024883 / hex 0x8007000d
+	ErrInvalidData = hcs.ErrInvalidData
+
+	// ErrHandleClose is an error encountered when the handle generating the notification being waited on has been closed
+	ErrHandleClose = hcs.ErrHandleClose
+
+	// ErrAlreadyClosed is an error encountered when using a handle that has been closed by the Close method
+	ErrAlreadyClosed = hcs.ErrAlreadyClosed
+
+	// ErrInvalidNotificationType is an error encountered when an invalid notification type is used
+	ErrInvalidNotificationType = hcs.ErrInvalidNotificationType
+
+	// ErrInvalidProcessState is an error encountered when the process is not in a valid state for the requested operation
+	ErrInvalidProcessState = hcs.ErrInvalidProcessState
+
+	// ErrTimeout is an error encountered when waiting on a notification times out
+	ErrTimeout = hcs.ErrTimeout
+
+	// ErrUnexpectedContainerExit is the error encountered when a container exits while waiting for
+	// a different expected notification
+	ErrUnexpectedContainerExit = hcs.ErrUnexpectedContainerExit
+
+	// ErrUnexpectedProcessAbort is the error encountered when communication with the compute service
+	// is lost while waiting for a notification
+	ErrUnexpectedProcessAbort = hcs.ErrUnexpectedProcessAbort
+
+	// ErrUnexpectedValue is an error encountered when hcs returns an invalid value
+	ErrUnexpectedValue = hcs.ErrUnexpectedValue
+
+	// ErrVmcomputeAlreadyStopped is an error encountered when a shutdown or terminate request is made on a stopped container
+	ErrVmcomputeAlreadyStopped = hcs.ErrVmcomputeAlreadyStopped
+
+	// ErrVmcomputeOperationPending is an error encountered when the operation is being completed asynchronously
+	ErrVmcomputeOperationPending = hcs.ErrVmcomputeOperationPending
+
+	// ErrVmcomputeOperationInvalidState is an error encountered when the compute system is not in a valid state for the requested operation
+	ErrVmcomputeOperationInvalidState = hcs.ErrVmcomputeOperationInvalidState
+
+	// ErrProcNotFound is an error encountered when the the process cannot be found
+	ErrProcNotFound = hcs.ErrProcNotFound
+
+	// ErrVmcomputeOperationAccessIsDenied is an error which can be encountered when enumerating compute systems in RS1/RS2
+	// builds when the underlying silo might be in the process of terminating. HCS was fixed in RS3.
+	ErrVmcomputeOperationAccessIsDenied = hcs.ErrVmcomputeOperationAccessIsDenied
+
+	// ErrVmcomputeInvalidJSON is an error encountered when the compute system does not support/understand the messages sent by management
+	ErrVmcomputeInvalidJSON = hcs.ErrVmcomputeInvalidJSON
+
+	// ErrVmcomputeUnknownMessage is an error encountered guest compute system doesn't support the message
+	ErrVmcomputeUnknownMessage = hcs.ErrVmcomputeUnknownMessage
+
+	// ErrNotSupported is an error encountered when hcs doesn't support the request
+	ErrPlatformNotSupported = hcs.ErrPlatformNotSupported
+)
+
+type EndpointNotFoundError = hns.EndpointNotFoundError
+type NetworkNotFoundError = hns.NetworkNotFoundError
+
+// ProcessError is an error encountered in HCS during an operation on a Process object
+type ProcessError struct {
+	Process   *process
+	Operation string
+	ExtraInfo string
+	Err       error
+	Events    []hcs.ErrorEvent
+}
+
+// ContainerError is an error encountered in HCS during an operation on a Container object
+type ContainerError struct {
+	Container *container
+	Operation string
+	ExtraInfo string
+	Err       error
+	Events    []hcs.ErrorEvent
+}
+
+func (e *ContainerError) Error() string {
+	if e == nil {
+		return "<nil>"
+	}
+
+	if e.Container == nil {
+		return "unexpected nil container for error: " + e.Err.Error()
+	}
+
+	s := "container " + e.Container.system.ID()
+
+	if e.Operation != "" {
+		s += " encountered an error during " + e.Operation
+	}
+
+	switch e.Err.(type) {
+	case nil:
+		break
+	case syscall.Errno:
+		s += fmt.Sprintf(": failure in a Windows system call: %s (0x%x)", e.Err, hcserror.Win32FromError(e.Err))
+	default:
+		s += fmt.Sprintf(": %s", e.Err.Error())
+	}
+
+	for _, ev := range e.Events {
+		s += "\n" + ev.String()
+	}
+
+	if e.ExtraInfo != "" {
+		s += " extra info: " + e.ExtraInfo
+	}
+
+	return s
+}
+
+func makeContainerError(container *container, operation string, extraInfo string, err error) error {
+	// Don't double wrap errors
+	if _, ok := err.(*ContainerError); ok {
+		return err
+	}
+	containerError := &ContainerError{Container: container, Operation: operation, ExtraInfo: extraInfo, Err: err}
+	return containerError
+}
+
+func (e *ProcessError) Error() string {
+	if e == nil {
+		return "<nil>"
+	}
+
+	if e.Process == nil {
+		return "Unexpected nil process for error: " + e.Err.Error()
+	}
+
+	s := fmt.Sprintf("process %d in container %s", e.Process.p.Pid(), e.Process.p.SystemID())
+	if e.Operation != "" {
+		s += " encountered an error during " + e.Operation
+	}
+
+	switch e.Err.(type) {
+	case nil:
+		break
+	case syscall.Errno:
+		s += fmt.Sprintf(": failure in a Windows system call: %s (0x%x)", e.Err, hcserror.Win32FromError(e.Err))
+	default:
+		s += fmt.Sprintf(": %s", e.Err.Error())
+	}
+
+	for _, ev := range e.Events {
+		s += "\n" + ev.String()
+	}
+
+	return s
+}
+
+func makeProcessError(process *process, operation string, extraInfo string, err error) error {
+	// Don't double wrap errors
+	if _, ok := err.(*ProcessError); ok {
+		return err
+	}
+	processError := &ProcessError{Process: process, Operation: operation, ExtraInfo: extraInfo, Err: err}
+	return processError
+}
+
+// IsNotExist checks if an error is caused by the Container or Process not existing.
+// Note: Currently, ErrElementNotFound can mean that a Process has either
+// already exited, or does not exist. Both IsAlreadyStopped and IsNotExist
+// will currently return true when the error is ErrElementNotFound or ErrProcNotFound.
+func IsNotExist(err error) bool {
+	if _, ok := err.(EndpointNotFoundError); ok {
+		return true
+	}
+	if _, ok := err.(NetworkNotFoundError); ok {
+		return true
+	}
+	return hcs.IsNotExist(getInnerError(err))
+}
+
+// IsAlreadyClosed checks if an error is caused by the Container or Process having been
+// already closed by a call to the Close() method.
+func IsAlreadyClosed(err error) bool {
+	return hcs.IsAlreadyClosed(getInnerError(err))
+}
+
+// IsPending returns a boolean indicating whether the error is that
+// the requested operation is being completed in the background.
+func IsPending(err error) bool {
+	return hcs.IsPending(getInnerError(err))
+}
+
+// IsTimeout returns a boolean indicating whether the error is caused by
+// a timeout waiting for the operation to complete.
+func IsTimeout(err error) bool {
+	return hcs.IsTimeout(getInnerError(err))
+}
+
+// IsAlreadyStopped returns a boolean indicating whether the error is caused by
+// a Container or Process being already stopped.
+// Note: Currently, ErrElementNotFound can mean that a Process has either
+// already exited, or does not exist. Both IsAlreadyStopped and IsNotExist
+// will currently return true when the error is ErrElementNotFound or ErrProcNotFound.
+func IsAlreadyStopped(err error) bool {
+	return hcs.IsAlreadyStopped(getInnerError(err))
+}
+
+// IsNotSupported returns a boolean indicating whether the error is caused by
+// unsupported platform requests
+// Note: Currently Unsupported platform requests can be mean either
+// ErrVmcomputeInvalidJSON, ErrInvalidData, ErrNotSupported or ErrVmcomputeUnknownMessage
+// is thrown from the Platform
+func IsNotSupported(err error) bool {
+	return hcs.IsNotSupported(getInnerError(err))
+}
+
+func getInnerError(err error) error {
+	switch pe := err.(type) {
+	case nil:
+		return nil
+	case *ContainerError:
+		err = pe.Err
+	case *ProcessError:
+		err = pe.Err
+	}
+	return err
+}
+
+func convertSystemError(err error, c *container) error {
+	if serr, ok := err.(*hcs.SystemError); ok {
+		return &ContainerError{Container: c, Operation: serr.Op, ExtraInfo: serr.Extra, Err: serr.Err, Events: serr.Events}
+	}
+	return err
+}
+
+func convertProcessError(err error, p *process) error {
+	if perr, ok := err.(*hcs.ProcessError); ok {
+		return &ProcessError{Process: p, Operation: perr.Op, Err: perr.Err, Events: perr.Events}
+	}
+	return err
+}
@@ -0,0 +1,12 @@
+# Requirements so far:
+# dockerd running
+#  - image microsoft/nanoserver (matching host base image)  docker load -i c:\baseimages\nanoserver.tar
+#  - image alpine (linux) docker pull --platform=linux alpine
+
+
+# TODO: Add this a parameter for debugging. ie "functional-tests -debug=$true"
+#$env:HCSSHIM_FUNCTIONAL_TESTS_DEBUG="yes please"
+
+#pushd uvm
+go test -v -tags "functional uvmcreate uvmscratch uvmscsi uvmvpmem uvmvsmb uvmp9" ./...
+#popd
@@ -0,0 +1,28 @@
+// Shim for the Host Compute Service (HCS) to manage Windows Server
+// containers and Hyper-V containers.
+
+package hcsshim
+
+import (
+	"syscall"
+
+	"github.com/Microsoft/hcsshim/internal/hcserror"
+)
+
+//go:generate go run mksyscall_windows.go -output zsyscall_windows.go hcsshim.go
+
+//sys SetCurrentThreadCompartmentId(compartmentId uint32) (hr error) = iphlpapi.SetCurrentThreadCompartmentId
+
+const (
+	// Specific user-visible exit codes
+	WaitErrExecFailed = 32767
+
+	ERROR_GEN_FAILURE          = hcserror.ERROR_GEN_FAILURE
+	ERROR_SHUTDOWN_IN_PROGRESS = syscall.Errno(1115)
+	WSAEINVAL                  = syscall.Errno(10022)
+
+	// Timeout on wait calls
+	TimeoutInfinite = 0xFFFFFFFF
+)
+
+type HcsError = hcserror.HcsError
@@ -0,0 +1,104 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+// HNSEndpoint represents a network endpoint in HNS
+type HNSEndpoint = hns.HNSEndpoint
+
+// Namespace represents a Compartment.
+type Namespace = hns.Namespace
+
+//SystemType represents the type of the system on which actions are done
+type SystemType string
+
+// SystemType const
+const (
+	ContainerType      SystemType = "Container"
+	VirtualMachineType SystemType = "VirtualMachine"
+	HostType           SystemType = "Host"
+)
+
+// EndpointAttachDetachRequest is the structure used to send request to the container to modify the system
+// Supported resource types are Network and Request Types are Add/Remove
+type EndpointAttachDetachRequest = hns.EndpointAttachDetachRequest
+
+// EndpointResquestResponse is object to get the endpoint request response
+type EndpointResquestResponse = hns.EndpointResquestResponse
+
+// HNSEndpointRequest makes a HNS call to modify/query a network endpoint
+func HNSEndpointRequest(method, path, request string) (*HNSEndpoint, error) {
+	return hns.HNSEndpointRequest(method, path, request)
+}
+
+// HNSListEndpointRequest makes a HNS call to query the list of available endpoints
+func HNSListEndpointRequest() ([]HNSEndpoint, error) {
+	return hns.HNSListEndpointRequest()
+}
+
+// HotAttachEndpoint makes a HCS Call to attach the endpoint to the container
+func HotAttachEndpoint(containerID string, endpointID string) error {
+	endpoint, err := GetHNSEndpointByID(endpointID)
+	isAttached, err := endpoint.IsAttached(containerID)
+	if isAttached {
+		return err
+	}
+	return modifyNetworkEndpoint(containerID, endpointID, Add)
+}
+
+// HotDetachEndpoint makes a HCS Call to detach the endpoint from the container
+func HotDetachEndpoint(containerID string, endpointID string) error {
+	endpoint, err := GetHNSEndpointByID(endpointID)
+	isAttached, err := endpoint.IsAttached(containerID)
+	if !isAttached {
+		return err
+	}
+	return modifyNetworkEndpoint(containerID, endpointID, Remove)
+}
+
+// ModifyContainer corresponding to the container id, by sending a request
+func modifyContainer(id string, request *ResourceModificationRequestResponse) error {
+	container, err := OpenContainer(id)
+	if err != nil {
+		if IsNotExist(err) {
+			return ErrComputeSystemDoesNotExist
+		}
+		return getInnerError(err)
+	}
+	defer container.Close()
+	err = container.Modify(request)
+	if err != nil {
+		if IsNotSupported(err) {
+			return ErrPlatformNotSupported
+		}
+		return getInnerError(err)
+	}
+
+	return nil
+}
+
+func modifyNetworkEndpoint(containerID string, endpointID string, request RequestType) error {
+	requestMessage := &ResourceModificationRequestResponse{
+		Resource: Network,
+		Request:  request,
+		Data:     endpointID,
+	}
+	err := modifyContainer(containerID, requestMessage)
+
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+// GetHNSEndpointByID get the Endpoint by ID
+func GetHNSEndpointByID(endpointID string) (*HNSEndpoint, error) {
+	return hns.GetHNSEndpointByID(endpointID)
+}
+
+// GetHNSEndpointByName gets the endpoint filtered by Name
+func GetHNSEndpointByName(endpointName string) (*HNSEndpoint, error) {
+	return hns.GetHNSEndpointByName(endpointName)
+}
@@ -0,0 +1,16 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+type HNSGlobals = hns.HNSGlobals
+type HNSVersion = hns.HNSVersion
+
+var (
+	HNSVersion1803 = hns.HNSVersion1803
+)
+
+func GetHNSGlobals() (*HNSGlobals, error) {
+	return hns.GetHNSGlobals()
+}
@@ -0,0 +1,36 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+// Subnet is assoicated with a network and represents a list
+// of subnets available to the network
+type Subnet = hns.Subnet
+
+// MacPool is assoicated with a network and represents a list
+// of macaddresses available to the network
+type MacPool = hns.MacPool
+
+// HNSNetwork represents a network in HNS
+type HNSNetwork = hns.HNSNetwork
+
+// HNSNetworkRequest makes a call into HNS to update/query a single network
+func HNSNetworkRequest(method, path, request string) (*HNSNetwork, error) {
+	return hns.HNSNetworkRequest(method, path, request)
+}
+
+// HNSListNetworkRequest makes a HNS call to query the list of available networks
+func HNSListNetworkRequest(method, path, request string) ([]HNSNetwork, error) {
+	return hns.HNSListNetworkRequest(method, path, request)
+}
+
+// GetHNSNetworkByID
+func GetHNSNetworkByID(networkID string) (*HNSNetwork, error) {
+	return hns.GetHNSNetworkByID(networkID)
+}
+
+// GetHNSNetworkName filtered by Name
+func GetHNSNetworkByName(networkName string) (*HNSNetwork, error) {
+	return hns.GetHNSNetworkByName(networkName)
+}
@@ -0,0 +1,60 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+// Type of Request Support in ModifySystem
+type PolicyType = hns.PolicyType
+
+// RequestType const
+const (
+	Nat                  = hns.Nat
+	ACL                  = hns.ACL
+	PA                   = hns.PA
+	VLAN                 = hns.VLAN
+	VSID                 = hns.VSID
+	VNet                 = hns.VNet
+	L2Driver             = hns.L2Driver
+	Isolation            = hns.Isolation
+	QOS                  = hns.QOS
+	OutboundNat          = hns.OutboundNat
+	ExternalLoadBalancer = hns.ExternalLoadBalancer
+	Route                = hns.Route
+	Proxy                = hns.Proxy
+)
+
+type ProxyPolicy = hns.ProxyPolicy
+
+type NatPolicy = hns.NatPolicy
+
+type QosPolicy = hns.QosPolicy
+
+type IsolationPolicy = hns.IsolationPolicy
+
+type VlanPolicy = hns.VlanPolicy
+
+type VsidPolicy = hns.VsidPolicy
+
+type PaPolicy = hns.PaPolicy
+
+type OutboundNatPolicy = hns.OutboundNatPolicy
+
+type ActionType = hns.ActionType
+type DirectionType = hns.DirectionType
+type RuleType = hns.RuleType
+
+const (
+	Allow = hns.Allow
+	Block = hns.Block
+
+	In  = hns.In
+	Out = hns.Out
+
+	Host   = hns.Host
+	Switch = hns.Switch
+)
+
+type ACLPolicy = hns.ACLPolicy
+
+type Policy = hns.Policy
@@ -0,0 +1,47 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+// RoutePolicy is a structure defining schema for Route based Policy
+type RoutePolicy = hns.RoutePolicy
+
+// ELBPolicy is a structure defining schema for ELB LoadBalancing based Policy
+type ELBPolicy = hns.ELBPolicy
+
+// LBPolicy is a structure defining schema for LoadBalancing based Policy
+type LBPolicy = hns.LBPolicy
+
+// PolicyList is a structure defining schema for Policy list request
+type PolicyList = hns.PolicyList
+
+// HNSPolicyListRequest makes a call into HNS to update/query a single network
+func HNSPolicyListRequest(method, path, request string) (*PolicyList, error) {
+	return hns.HNSPolicyListRequest(method, path, request)
+}
+
+// HNSListPolicyListRequest gets all the policy list
+func HNSListPolicyListRequest() ([]PolicyList, error) {
+	return hns.HNSListPolicyListRequest()
+}
+
+// PolicyListRequest makes a HNS call to modify/query a network policy list
+func PolicyListRequest(method, path, request string) (*PolicyList, error) {
+	return hns.PolicyListRequest(method, path, request)
+}
+
+// GetPolicyListByID get the policy list by ID
+func GetPolicyListByID(policyListID string) (*PolicyList, error) {
+	return hns.GetPolicyListByID(policyListID)
+}
+
+// AddLoadBalancer policy list for the specified endpoints
+func AddLoadBalancer(endpoints []HNSEndpoint, isILB bool, sourceVIP, vip string, protocol uint16, internalPort uint16, externalPort uint16) (*PolicyList, error) {
+	return hns.AddLoadBalancer(endpoints, isILB, sourceVIP, vip, protocol, internalPort, externalPort)
+}
+
+// AddRoute adds route policy list for the specified endpoints
+func AddRoute(endpoints []HNSEndpoint, destinationPrefix string, nextHop string, encapEnabled bool) (*PolicyList, error) {
+	return hns.AddRoute(endpoints, destinationPrefix, nextHop, encapEnabled)
+}
@@ -0,0 +1,13 @@
+package hcsshim
+
+import (
+	"github.com/Microsoft/hcsshim/internal/hns"
+)
+
+type HNSSupportedFeatures = hns.HNSSupportedFeatures
+
+type HNSAclFeatures = hns.HNSAclFeatures
+
+func GetHNSSupportedFeatures() HNSSupportedFeatures {
+	return hns.GetHNSSupportedFeatures()
+}
@@ -0,0 +1,114 @@
+package hcsshim
+
+import (
+	"io"
+	"time"
+
+	"github.com/Microsoft/hcsshim/internal/schema1"
+)
+
+// ProcessConfig is used as both the input of Container.CreateProcess
+// and to convert the parameters to JSON for passing onto the HCS
+type ProcessConfig = schema1.ProcessConfig
+
+type Layer = schema1.Layer
+type MappedDir = schema1.MappedDir
+type MappedPipe = schema1.MappedPipe
+type HvRuntime = schema1.HvRuntime
+type MappedVirtualDisk = schema1.MappedVirtualDisk
+
+// AssignedDevice represents a device that has been directly assigned to a container
+//
+// NOTE: Support added in RS5
+type AssignedDevice = schema1.AssignedDevice
+
+// ContainerConfig is used as both the input of CreateContainer
+// and to convert the parameters to JSON for passing onto the HCS
+type ContainerConfig = schema1.ContainerConfig
+
+type ComputeSystemQuery = schema1.ComputeSystemQuery
+
+// Container represents a created (but not necessarily running) container.
+type Container interface {
+	// Start synchronously starts the container.
+	Start() error
+
+	// Shutdown requests a container shutdown, but it may not actually be shutdown until Wait() succeeds.
+	Shutdown() error
+
+	// Terminate requests a container terminate, but it may not actually be terminated until Wait() succeeds.
+	Terminate() error
+
+	// Waits synchronously waits for the container to shutdown or terminate.
+	Wait() error
+
+	// WaitTimeout synchronously waits for the container to terminate or the duration to elapse. It
+	// returns false if timeout occurs.
+	WaitTimeout(time.Duration) error
+
+	// Pause pauses the execution of a container.
+	Pause() error
+
+	// Resume resumes the execution of a container.
+	Resume() error
+
+	// HasPendingUpdates returns true if the container has updates pending to install.
+	HasPendingUpdates() (bool, error)
+
+	// Statistics returns statistics for a container.
+	Statistics() (Statistics, error)
+
+	// ProcessList returns details for the processes in a container.
+	ProcessList() ([]ProcessListItem, error)
+
+	// MappedVirtualDisks returns virtual disks mapped to a utility VM, indexed by controller
+	MappedVirtualDisks() (map[int]MappedVirtualDiskController, error)
+
+	// CreateProcess launches a new process within the container.
+	CreateProcess(c *ProcessConfig) (Process, error)
+
+	// OpenProcess gets an interface to an existing process within the container.
+	OpenProcess(pid int) (Process, error)
+
+	// Close cleans up any state associated with the container but does not terminate or wait for it.
+	Close() error
+
+	// Modify the System
+	Modify(config *ResourceModificationRequestResponse) error
+}
+
+// Process represents a running or exited process.
+type Process interface {
+	// Pid returns the process ID of the process within the container.
+	Pid() int
+
+	// Kill signals the process to terminate but does not wait for it to finish terminating.
+	Kill() error
+
+	// Wait waits for the process to exit.
+	Wait() error
+
+	// WaitTimeout waits for the process to exit or the duration to elapse. It returns
+	// false if timeout occurs.
+	WaitTimeout(time.Duration) error
+
+	// ExitCode returns the exit code of the process. The process must have
+	// already terminated.
+	ExitCode() (int, error)
+
+	// ResizeConsole resizes the console of the process.
+	ResizeConsole(width, height uint16) error
+
+	// Stdio returns the stdin, stdout, and stderr pipes, respectively. Closing
+	// these pipes does not close the underlying pipes; it should be possible to
+	// call this multiple times to get multiple interfaces.
+	Stdio() (io.WriteCloser, io.ReadCloser, io.ReadCloser, error)
+
+	// CloseStdin closes the write side of the stdin pipe so that the process is
+	// notified on the read side that there is no more data in stdin.
+	CloseStdin() error
+
+	// Close cleans up any state associated with the process but does not kill
+	// or wait on it.
+	Close() error
+}
@@ -0,0 +1,83 @@
+package cow
+
+import (
+	"context"
+	"io"
+
+	"github.com/Microsoft/hcsshim/internal/schema1"
+	hcsschema "github.com/Microsoft/hcsshim/internal/schema2"
+)
+
+// Process is the interface for an OS process running in a container or utility VM.
+type Process interface {
+	// Close releases resources associated with the process and closes the
+	// writer and readers returned by Stdio. Depending on the implementation,
+	// this may also terminate the process.
+	Close() error
+	// CloseStdin causes the process's stdin handle to receive EOF/EPIPE/whatever
+	// is appropriate to indicate that no more data is available.
+	CloseStdin(ctx context.Context) error
+	// Pid returns the process ID.
+	Pid() int
+	// Stdio returns the stdio streams for a process. These may be nil if a stream
+	// was not requested during CreateProcess.
+	Stdio() (_ io.Writer, _ io.Reader, _ io.Reader)
+	// ResizeConsole resizes the virtual terminal associated with the process.
+	ResizeConsole(ctx context.Context, width, height uint16) error
+	// Kill sends a SIGKILL or equivalent signal to the process and returns whether
+	// the signal was delivered. It does not wait for the process to terminate.
+	Kill(ctx context.Context) (bool, error)
+	// Signal sends a signal to the process and returns whether the signal was
+	// delivered. The input is OS specific (either
+	// guestrequest.SignalProcessOptionsWCOW or
+	// guestrequest.SignalProcessOptionsLCOW). It does not wait for the process
+	// to terminate.
+	Signal(ctx context.Context, options interface{}) (bool, error)
+	// Wait waits for the process to complete, or for a connection to the process to be
+	// terminated by some error condition (including calling Close).
+	Wait() error
+	// ExitCode returns the exit code of the process. Returns an error if the process is
+	// not running.
+	ExitCode() (int, error)
+}
+
+// ProcessHost is the interface for creating processes.
+type ProcessHost interface {
+	// CreateProcess creates a process. The configuration is host specific
+	// (either hcsschema.ProcessParameters or lcow.ProcessParameters).
+	CreateProcess(ctx context.Context, config interface{}) (Process, error)
+	// OS returns the host's operating system, "linux" or "windows".
+	OS() string
+	// IsOCI specifies whether this is an OCI-compliant process host. If true,
+	// then the configuration passed to CreateProcess should have an OCI process
+	// spec (or nil if this is the initial process in an OCI container).
+	// Otherwise, it should have the HCS-specific process parameters.
+	IsOCI() bool
+}
+
+// Container is the interface for container objects, either running on the host or
+// in a utility VM.
+type Container interface {
+	ProcessHost
+	// Close releases the resources associated with the container. Depending on
+	// the implementation, this may also terminate the container.
+	Close() error
+	// ID returns the container ID.
+	ID() string
+	// Properties returns the requested container properties targeting a V1 schema container.
+	Properties(ctx context.Context, types ...schema1.PropertyType) (*schema1.ContainerProperties, error)
+	// PropertiesV2 returns the requested container properties targeting a V2 schema container.
+	PropertiesV2(ctx context.Context, types ...hcsschema.PropertyType) (*hcsschema.Properties, error)
+	// Start starts a container.
+	Start(ctx context.Context) error
+	// Shutdown sends a shutdown request to the container (but does not wait for
+	// the shutdown to complete).
+	Shutdown(ctx context.Context) error
+	// Terminate sends a terminate request to the container (but does not wait
+	// for the terminate to complete).
+	Terminate(ctx context.Context) error
+	// Wait waits for the container to terminate, or for the connection to the
+	// container to be terminated by some error condition (including calling
+	// Close).
+	Wait() error
+}
@@ -0,0 +1,160 @@
+package hcs
+
+import (
+	"fmt"
+	"sync"
+	"syscall"
+
+	"github.com/Microsoft/hcsshim/internal/interop"
+	"github.com/Microsoft/hcsshim/internal/logfields"
+	"github.com/Microsoft/hcsshim/internal/vmcompute"
+	"github.com/sirupsen/logrus"
+)
+
+var (
+	nextCallback    uintptr
+	callbackMap     = map[uintptr]*notifcationWatcherContext{}
+	callbackMapLock = sync.RWMutex{}
+
+	notificationWatcherCallback = syscall.NewCallback(notificationWatcher)
+
+	// Notifications for HCS_SYSTEM handles
+	hcsNotificationSystemExited                      hcsNotification = 0x00000001
+	hcsNotificationSystemCreateCompleted             hcsNotification = 0x00000002
+	hcsNotificationSystemStartCompleted              hcsNotification = 0x00000003
+	hcsNotificationSystemPauseCompleted              hcsNotification = 0x00000004
+	hcsNotificationSystemResumeCompleted             hcsNotification = 0x00000005
+	hcsNotificationSystemCrashReport                 hcsNotification = 0x00000006
+	hcsNotificationSystemSiloJobCreated              hcsNotification = 0x00000007
+	hcsNotificationSystemSaveCompleted               hcsNotification = 0x00000008
+	hcsNotificationSystemRdpEnhancedModeStateChanged hcsNotification = 0x00000009
+	hcsNotificationSystemShutdownFailed              hcsNotification = 0x0000000A
+	hcsNotificationSystemGetPropertiesCompleted      hcsNotification = 0x0000000B
+	hcsNotificationSystemModifyCompleted             hcsNotification = 0x0000000C
+	hcsNotificationSystemCrashInitiated              hcsNotification = 0x0000000D
+	hcsNotificationSystemGuestConnectionClosed       hcsNotification = 0x0000000E
+
+	// Notifications for HCS_PROCESS handles
+	hcsNotificationProcessExited hcsNotification = 0x00010000
+
+	// Common notifications
+	hcsNotificationInvalid           hcsNotification = 0x00000000
+	hcsNotificationServiceDisconnect hcsNotification = 0x01000000
+)
+
+type hcsNotification uint32
+
+func (hn hcsNotification) String() string {
+	switch hn {
+	case hcsNotificationSystemExited:
+		return "SystemExited"
+	case hcsNotificationSystemCreateCompleted:
+		return "SystemCreateCompleted"
+	case hcsNotificationSystemStartCompleted:
+		return "SystemStartCompleted"
+	case hcsNotificationSystemPauseCompleted:
+		return "SystemPauseCompleted"
+	case hcsNotificationSystemResumeCompleted:
+		return "SystemResumeCompleted"
+	case hcsNotificationSystemCrashReport:
+		return "SystemCrashReport"
+	case hcsNotificationSystemSiloJobCreated:
+		return "SystemSiloJobCreated"
+	case hcsNotificationSystemSaveCompleted:
+		return "SystemSaveCompleted"
+	case hcsNotificationSystemRdpEnhancedModeStateChanged:
+		return "SystemRdpEnhancedModeStateChanged"
+	case hcsNotificationSystemShutdownFailed:
+		return "SystemShutdownFailed"
+	case hcsNotificationSystemGetPropertiesCompleted:
+		return "SystemGetPropertiesCompleted"
+	case hcsNotificationSystemModifyCompleted:
+		return "SystemModifyCompleted"
+	case hcsNotificationSystemCrashInitiated:
+		return "SystemCrashInitiated"
+	case hcsNotificationSystemGuestConnectionClosed:
+		return "SystemGuestConnectionClosed"
+	case hcsNotificationProcessExited:
+		return "ProcessExited"
+	case hcsNotificationInvalid:
+		return "Invalid"
+	case hcsNotificationServiceDisconnect:
+		return "ServiceDisconnect"
+	default:
+		return fmt.Sprintf("Unknown: %d", hn)
+	}
+}
+
+type notificationChannel chan error
+
+type notifcationWatcherContext struct {
+	channels notificationChannels
+	handle   vmcompute.HcsCallback
+
+	systemID  string
+	processID int
+}
+
+type notificationChannels map[hcsNotification]notificationChannel
+
+func newSystemChannels() notificationChannels {
+	channels := make(notificationChannels)
+	for _, notif := range []hcsNotification{
+		hcsNotificationServiceDisconnect,
+		hcsNotificationSystemExited,
+		hcsNotificationSystemCreateCompleted,
+		hcsNotificationSystemStartCompleted,
+		hcsNotificationSystemPauseCompleted,
+		hcsNotificationSystemResumeCompleted,
+	} {
+		channels[notif] = make(notificationChannel, 1)
+	}
+	return channels
+}
+
+func newProcessChannels() notificationChannels {
+	channels := make(notificationChannels)
+	for _, notif := range []hcsNotification{
+		hcsNotificationServiceDisconnect,
+		hcsNotificationProcessExited,
+	} {
+		channels[notif] = make(notificationChannel, 1)
+	}
+	return channels
+}
+
+func closeChannels(channels notificationChannels) {
+	for _, c := range channels {
+		close(c)
+	}
+}
+
+func notificationWatcher(notificationType hcsNotification, callbackNumber uintptr, notificationStatus uintptr, notificationData *uint16) uintptr {
+	var result error
+	if int32(notificationStatus) < 0 {
+		result = interop.Win32FromHresult(notificationStatus)
+	}
+
+	callbackMapLock.RLock()
+	context := callbackMap[callbackNumber]
+	callbackMapLock.RUnlock()
+
+	if context == nil {
+		return 0
+	}
+
+	log := logrus.WithFields(logrus.Fields{
+		"notification-type": notificationType.String(),
+		"system-id":         context.systemID,
+	})
+	if context.processID != 0 {
+		log.Data[logfields.ProcessID] = context.processID
+	}
+	log.Debug("HCS notification")
+
+	if channel, ok := context.channels[notificationType]; ok {
+		channel <- result
+	}
+
+	return 0
+}
@@ -0,0 +1,336 @@
+package hcs
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"net"
+	"syscall"
+
+	"github.com/Microsoft/hcsshim/internal/log"
+)
+
+var (
+	// ErrComputeSystemDoesNotExist is an error encountered when the container being operated on no longer exists
+	ErrComputeSystemDoesNotExist = syscall.Errno(0xc037010e)
+
+	// ErrElementNotFound is an error encountered when the object being referenced does not exist
+	ErrElementNotFound = syscall.Errno(0x490)
+
+	// ErrElementNotFound is an error encountered when the object being referenced does not exist
+	ErrNotSupported = syscall.Errno(0x32)
+
+	// ErrInvalidData is an error encountered when the request being sent to hcs is invalid/unsupported
+	// decimal -2147024883 / hex 0x8007000d
+	ErrInvalidData = syscall.Errno(0xd)
+
+	// ErrHandleClose is an error encountered when the handle generating the notification being waited on has been closed
+	ErrHandleClose = errors.New("hcsshim: the handle generating this notification has been closed")
+
+	// ErrAlreadyClosed is an error encountered when using a handle that has been closed by the Close method
+	ErrAlreadyClosed = errors.New("hcsshim: the handle has already been closed")
+
+	// ErrInvalidNotificationType is an error encountered when an invalid notification type is used
+	ErrInvalidNotificationType = errors.New("hcsshim: invalid notification type")
+
+	// ErrInvalidProcessState is an error encountered when the process is not in a valid state for the requested operation
+	ErrInvalidProcessState = errors.New("the process is in an invalid state for the attempted operation")
+
+	// ErrTimeout is an error encountered when waiting on a notification times out
+	ErrTimeout = errors.New("hcsshim: timeout waiting for notification")
+
+	// ErrUnexpectedContainerExit is the error encountered when a container exits while waiting for
+	// a different expected notification
+	ErrUnexpectedContainerExit = errors.New("unexpected container exit")
+
+	// ErrUnexpectedProcessAbort is the error encountered when communication with the compute service
+	// is lost while waiting for a notification
+	ErrUnexpectedProcessAbort = errors.New("lost communication with compute service")
+
+	// ErrUnexpectedValue is an error encountered when hcs returns an invalid value
+	ErrUnexpectedValue = errors.New("unexpected value returned from hcs")
+
+	// ErrVmcomputeAlreadyStopped is an error encountered when a shutdown or terminate request is made on a stopped container
+	ErrVmcomputeAlreadyStopped = syscall.Errno(0xc0370110)
+
+	// ErrVmcomputeOperationPending is an error encountered when the operation is being completed asynchronously
+	ErrVmcomputeOperationPending = syscall.Errno(0xC0370103)
+
+	// ErrVmcomputeOperationInvalidState is an error encountered when the compute system is not in a valid state for the requested operation
+	ErrVmcomputeOperationInvalidState = syscall.Errno(0xc0370105)
+
+	// ErrProcNotFound is an error encountered when the the process cannot be found
+	ErrProcNotFound = syscall.Errno(0x7f)
+
+	// ErrVmcomputeOperationAccessIsDenied is an error which can be encountered when enumerating compute systems in RS1/RS2
+	// builds when the underlying silo might be in the process of terminating. HCS was fixed in RS3.
+	ErrVmcomputeOperationAccessIsDenied = syscall.Errno(0x5)
+
+	// ErrVmcomputeInvalidJSON is an error encountered when the compute system does not support/understand the messages sent by management
+	ErrVmcomputeInvalidJSON = syscall.Errno(0xc037010d)
+
+	// ErrVmcomputeUnknownMessage is an error encountered guest compute system doesn't support the message
+	ErrVmcomputeUnknownMessage = syscall.Errno(0xc037010b)
+
+	// ErrVmcomputeUnexpectedExit is an error encountered when the compute system terminates unexpectedly
+	ErrVmcomputeUnexpectedExit = syscall.Errno(0xC0370106)
+
+	// ErrNotSupported is an error encountered when hcs doesn't support the request
+	ErrPlatformNotSupported = errors.New("unsupported platform request")
+)
+
+type ErrorEvent struct {
+	Message    string `json:"Message,omitempty"`    // Fully formated error message
+	StackTrace string `json:"StackTrace,omitempty"` // Stack trace in string form
+	Provider   string `json:"Provider,omitempty"`
+	EventID    uint16 `json:"EventId,omitempty"`
+	Flags      uint32 `json:"Flags,omitempty"`
+	Source     string `json:"Source,omitempty"`
+	//Data       []EventData `json:"Data,omitempty"`  // Omit this as HCS doesn't encode this well. It's more confusing to include. It is however logged in debug mode (see processHcsResult function)
+}
+
+type hcsResult struct {
+	Error        int32
+	ErrorMessage string
+	ErrorEvents  []ErrorEvent `json:"ErrorEvents,omitempty"`
+}
+
+func (ev *ErrorEvent) String() string {
+	evs := "[Event Detail: " + ev.Message
+	if ev.StackTrace != "" {
+		evs += " Stack Trace: " + ev.StackTrace
+	}
+	if ev.Provider != "" {
+		evs += " Provider: " + ev.Provider
+	}
+	if ev.EventID != 0 {
+		evs = fmt.Sprintf("%s EventID: %d", evs, ev.EventID)
+	}
+	if ev.Flags != 0 {
+		evs = fmt.Sprintf("%s flags: %d", evs, ev.Flags)
+	}
+	if ev.Source != "" {
+		evs += " Source: " + ev.Source
+	}
+	evs += "]"
+	return evs
+}
+
+func processHcsResult(ctx context.Context, resultJSON string) []ErrorEvent {
+	if resultJSON != "" {
+		result := &hcsResult{}
+		if err := json.Unmarshal([]byte(resultJSON), result); err != nil {
+			log.G(ctx).WithError(err).Warning("Could not unmarshal HCS result")
+			return nil
+		}
+		return result.ErrorEvents
+	}
+	return nil
+}
+
+type HcsError struct {
+	Op     string
+	Err    error
+	Events []ErrorEvent
+}
+
+var _ net.Error = &HcsError{}
+
+func (e *HcsError) Error() string {
+	s := e.Op + ": " + e.Err.Error()
+	for _, ev := range e.Events {
+		s += "\n" + ev.String()
+	}
+	return s
+}
+
+func (e *HcsError) Temporary() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Temporary()
+}
+
+func (e *HcsError) Timeout() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Timeout()
+}
+
+// ProcessError is an error encountered in HCS during an operation on a Process object
+type ProcessError struct {
+	SystemID string
+	Pid      int
+	Op       string
+	Err      error
+	Events   []ErrorEvent
+}
+
+var _ net.Error = &ProcessError{}
+
+// SystemError is an error encountered in HCS during an operation on a Container object
+type SystemError struct {
+	ID     string
+	Op     string
+	Err    error
+	Extra  string
+	Events []ErrorEvent
+}
+
+var _ net.Error = &SystemError{}
+
+func (e *SystemError) Error() string {
+	s := e.Op + " " + e.ID + ": " + e.Err.Error()
+	for _, ev := range e.Events {
+		s += "\n" + ev.String()
+	}
+	if e.Extra != "" {
+		s += "\n(extra info: " + e.Extra + ")"
+	}
+	return s
+}
+
+func (e *SystemError) Temporary() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Temporary()
+}
+
+func (e *SystemError) Timeout() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Timeout()
+}
+
+func makeSystemError(system *System, op string, extra string, err error, events []ErrorEvent) error {
+	// Don't double wrap errors
+	if _, ok := err.(*SystemError); ok {
+		return err
+	}
+	return &SystemError{
+		ID:     system.ID(),
+		Op:     op,
+		Extra:  extra,
+		Err:    err,
+		Events: events,
+	}
+}
+
+func (e *ProcessError) Error() string {
+	s := fmt.Sprintf("%s %s:%d: %s", e.Op, e.SystemID, e.Pid, e.Err.Error())
+	for _, ev := range e.Events {
+		s += "\n" + ev.String()
+	}
+	return s
+}
+
+func (e *ProcessError) Temporary() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Temporary()
+}
+
+func (e *ProcessError) Timeout() bool {
+	err, ok := e.Err.(net.Error)
+	return ok && err.Timeout()
+}
+
+func makeProcessError(process *Process, op string, err error, events []ErrorEvent) error {
+	// Don't double wrap errors
+	if _, ok := err.(*ProcessError); ok {
+		return err
+	}
+	return &ProcessError{
+		Pid:      process.Pid(),
+		SystemID: process.SystemID(),
+		Op:       op,
+		Err:      err,
+		Events:   events,
+	}
+}
+
+// IsNotExist checks if an error is caused by the Container or Process not existing.
+// Note: Currently, ErrElementNotFound can mean that a Process has either
+// already exited, or does not exist. Both IsAlreadyStopped and IsNotExist
+// will currently return true when the error is ErrElementNotFound or ErrProcNotFound.
+func IsNotExist(err error) bool {
+	err = getInnerError(err)
+	return err == ErrComputeSystemDoesNotExist ||
+		err == ErrElementNotFound ||
+		err == ErrProcNotFound
+}
+
+// IsAlreadyClosed checks if an error is caused by the Container or Process having been
+// already closed by a call to the Close() method.
+func IsAlreadyClosed(err error) bool {
+	err = getInnerError(err)
+	return err == ErrAlreadyClosed
+}
+
+// IsPending returns a boolean indicating whether the error is that
+// the requested operation is being completed in the background.
+func IsPending(err error) bool {
+	err = getInnerError(err)
+	return err == ErrVmcomputeOperationPending
+}
+
+// IsTimeout returns a boolean indicating whether the error is caused by
+// a timeout waiting for the operation to complete.
+func IsTimeout(err error) bool {
+	if err, ok := err.(net.Error); ok && err.Timeout() {
+		return true
+	}
+	err = getInnerError(err)
+	return err == ErrTimeout
+}
+
+// IsAlreadyStopped returns a boolean indicating whether the error is caused by
+// a Container or Process being already stopped.
+// Note: Currently, ErrElementNotFound can mean that a Process has either
+// already exited, or does not exist. Both IsAlreadyStopped and IsNotExist
+// will currently return true when the error is ErrElementNotFound or ErrProcNotFound.
+func IsAlreadyStopped(err error) bool {
+	err = getInnerError(err)
+	return err == ErrVmcomputeAlreadyStopped ||
+		err == ErrElementNotFound ||
+		err == ErrProcNotFound
+}
+
+// IsNotSupported returns a boolean indicating whether the error is caused by
+// unsupported platform requests
+// Note: Currently Unsupported platform requests can be mean either
+// ErrVmcomputeInvalidJSON, ErrInvalidData, ErrNotSupported or ErrVmcomputeUnknownMessage
+// is thrown from the Platform
+func IsNotSupported(err error) bool {
+	err = getInnerError(err)
+	// If Platform doesn't recognize or support the request sent, below errors are seen
+	return err == ErrVmcomputeInvalidJSON ||
+		err == ErrInvalidData ||
+		err == ErrNotSupported ||
+		err == ErrVmcomputeUnknownMessage
+}
+
+// IsOperationInvalidState returns true when err is caused by
+// `ErrVmcomputeOperationInvalidState`.
+func IsOperationInvalidState(err error) bool {
+	err = getInnerError(err)
+	return err == ErrVmcomputeOperationInvalidState
+}
+
+func getInnerError(err error) error {
+	switch pe := err.(type) {
+	case nil:
+		return nil
+	case *HcsError:
+		err = pe.Err
+	case *SystemError:
+		err = pe.Err
+	case *ProcessError:
+		err = pe.Err
+	}
+	return err
+}
+
+func getOperationLogResult(err error) (string, error) {
+	switch err {
+	case nil:
+		return "Success", nil
+	default:
+		return "Error", err
+	}
+}
@@ -0,0 +1,474 @@
+package hcs
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"sync"
+	"syscall"
+	"time"
+
+	"github.com/Microsoft/hcsshim/internal/log"
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"github.com/Microsoft/hcsshim/internal/vmcompute"
+	"go.opencensus.io/trace"
+)
+
+// ContainerError is an error encountered in HCS
+type Process struct {
+	handleLock     sync.RWMutex
+	handle         vmcompute.HcsProcess
+	processID      int
+	system         *System
+	hasCachedStdio bool
+	stdioLock      sync.Mutex
+	stdin          io.WriteCloser
+	stdout         io.ReadCloser
+	stderr         io.ReadCloser
+	callbackNumber uintptr
+
+	closedWaitOnce sync.Once
+	waitBlock      chan struct{}
+	exitCode       int
+	waitError      error
+}
+
+func newProcess(process vmcompute.HcsProcess, processID int, computeSystem *System) *Process {
+	return &Process{
+		handle:    process,
+		processID: processID,
+		system:    computeSystem,
+		waitBlock: make(chan struct{}),
+	}
+}
+
+type processModifyRequest struct {
+	Operation   string
+	ConsoleSize *consoleSize `json:",omitempty"`
+	CloseHandle *closeHandle `json:",omitempty"`
+}
+
+type consoleSize struct {
+	Height uint16
+	Width  uint16
+}
+
+type closeHandle struct {
+	Handle string
+}
+
+type processStatus struct {
+	ProcessID      uint32
+	Exited         bool
+	ExitCode       uint32
+	LastWaitResult int32
+}
+
+const (
+	stdIn  string = "StdIn"
+	stdOut string = "StdOut"
+	stdErr string = "StdErr"
+)
+
+const (
+	modifyConsoleSize string = "ConsoleSize"
+	modifyCloseHandle string = "CloseHandle"
+)
+
+// Pid returns the process ID of the process within the container.
+func (process *Process) Pid() int {
+	return process.processID
+}
+
+// SystemID returns the ID of the process's compute system.
+func (process *Process) SystemID() string {
+	return process.system.ID()
+}
+
+func (process *Process) processSignalResult(ctx context.Context, err error) (bool, error) {
+	switch err {
+	case nil:
+		return true, nil
+	case ErrVmcomputeOperationInvalidState, ErrComputeSystemDoesNotExist, ErrElementNotFound:
+		select {
+		case <-process.waitBlock:
+			// The process exit notification has already arrived.
+		default:
+			// The process should be gone, but we have not received the notification.
+			// After a second, force unblock the process wait to work around a possible
+			// deadlock in the HCS.
+			go func() {
+				time.Sleep(time.Second)
+				process.closedWaitOnce.Do(func() {
+					log.G(ctx).WithError(err).Warn("force unblocking process waits")
+					process.exitCode = -1
+					process.waitError = err
+					close(process.waitBlock)
+				})
+			}()
+		}
+		return false, nil
+	default:
+		return false, err
+	}
+}
+
+// Signal signals the process with `options`.
+//
+// For LCOW `guestrequest.SignalProcessOptionsLCOW`.
+//
+// For WCOW `guestrequest.SignalProcessOptionsWCOW`.
+func (process *Process) Signal(ctx context.Context, options interface{}) (bool, error) {
+	process.handleLock.RLock()
+	defer process.handleLock.RUnlock()
+
+	operation := "hcsshim::Process::Signal"
+
+	if process.handle == 0 {
+		return false, makeProcessError(process, operation, ErrAlreadyClosed, nil)
+	}
+
+	optionsb, err := json.Marshal(options)
+	if err != nil {
+		return false, err
+	}
+
+	resultJSON, err := vmcompute.HcsSignalProcess(ctx, process.handle, string(optionsb))
+	events := processHcsResult(ctx, resultJSON)
+	delivered, err := process.processSignalResult(ctx, err)
+	if err != nil {
+		err = makeProcessError(process, operation, err, events)
+	}
+	return delivered, err
+}
+
+// Kill signals the process to terminate but does not wait for it to finish terminating.
+func (process *Process) Kill(ctx context.Context) (bool, error) {
+	process.handleLock.RLock()
+	defer process.handleLock.RUnlock()
+
+	operation := "hcsshim::Process::Kill"
+
+	if process.handle == 0 {
+		return false, makeProcessError(process, operation, ErrAlreadyClosed, nil)
+	}
+
+	resultJSON, err := vmcompute.HcsTerminateProcess(ctx, process.handle)
+	events := processHcsResult(ctx, resultJSON)
+	delivered, err := process.processSignalResult(ctx, err)
+	if err != nil {
+		err = makeProcessError(process, operation, err, events)
+	}
+	return delivered, err
+}
+
+// waitBackground waits for the process exit notification. Once received sets
+// `process.waitError` (if any) and unblocks all `Wait` calls.
+//
+// This MUST be called exactly once per `process.handle` but `Wait` is safe to
+// call multiple times.
+func (process *Process) waitBackground() {
+	operation := "hcsshim::Process::waitBackground"
+	ctx, span := trace.StartSpan(context.Background(), operation)
+	defer span.End()
+	span.AddAttributes(
+		trace.StringAttribute("cid", process.SystemID()),
+		trace.Int64Attribute("pid", int64(process.processID)))
+
+	var (
+		err      error
+		exitCode = -1
+	)
+
+	err = waitForNotification(ctx, process.callbackNumber, hcsNotificationProcessExited, nil)
+	if err != nil {
+		err = makeProcessError(process, operation, err, nil)
+		log.G(ctx).WithError(err).Error("failed wait")
+	} else {
+		process.handleLock.RLock()
+		defer process.handleLock.RUnlock()
+
+		// Make sure we didnt race with Close() here
+		if process.handle != 0 {
+			propertiesJSON, resultJSON, err := vmcompute.HcsGetProcessProperties(ctx, process.handle)
+			events := processHcsResult(ctx, resultJSON)
+			if err != nil {
+				err = makeProcessError(process, operation, err, events)
+			} else {
+				properties := &processStatus{}
+				err = json.Unmarshal([]byte(propertiesJSON), properties)
+				if err != nil {
+					err = makeProcessError(process, operation, err, nil)
+				} else {
+					if properties.LastWaitResult != 0 {
+						log.G(ctx).WithField("wait-result", properties.LastWaitResult).Warning("non-zero last wait result")
+					} else {
+						exitCode = int(properties.ExitCode)
+					}
+				}
+			}
+		}
+	}
+	log.G(ctx).WithField("exitCode", exitCode).Debug("process exited")
+
+	process.closedWaitOnce.Do(func() {
+		process.exitCode = exitCode
+		process.waitError = err
+		close(process.waitBlock)
+	})
+	oc.SetSpanStatus(span, err)
+}
+
+// Wait waits for the process to exit. If the process has already exited returns
+// the pervious error (if any).
+func (process *Process) Wait() error {
+	<-process.waitBlock
+	return process.waitError
+}
+
+// ResizeConsole resizes the console of the process.
+func (process *Process) ResizeConsole(ctx context.Context, width, height uint16) error {
+	process.handleLock.RLock()
+	defer process.handleLock.RUnlock()
+
+	operation := "hcsshim::Process::ResizeConsole"
+
+	if process.handle == 0 {
+		return makeProcessError(process, operation, ErrAlreadyClosed, nil)
+	}
+
+	modifyRequest := processModifyRequest{
+		Operation: modifyConsoleSize,
+		ConsoleSize: &consoleSize{
+			Height: height,
+			Width:  width,
+		},
+	}
+
+	modifyRequestb, err := json.Marshal(modifyRequest)
+	if err != nil {
+		return err
+	}
+
+	resultJSON, err := vmcompute.HcsModifyProcess(ctx, process.handle, string(modifyRequestb))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return makeProcessError(process, operation, err, events)
+	}
+
+	return nil
+}
+
+// ExitCode returns the exit code of the process. The process must have
+// already terminated.
+func (process *Process) ExitCode() (int, error) {
+	select {
+	case <-process.waitBlock:
+		if process.waitError != nil {
+			return -1, process.waitError
+		}
+		return process.exitCode, nil
+	default:
+		return -1, makeProcessError(process, "hcsshim::Process::ExitCode", ErrInvalidProcessState, nil)
+	}
+}
+
+// StdioLegacy returns the stdin, stdout, and stderr pipes, respectively. Closing
+// these pipes does not close the underlying pipes. Once returned, these pipes
+// are the responsibility of the caller to close.
+func (process *Process) StdioLegacy() (_ io.WriteCloser, _ io.ReadCloser, _ io.ReadCloser, err error) {
+	operation := "hcsshim::Process::StdioLegacy"
+	ctx, span := trace.StartSpan(context.Background(), operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("cid", process.SystemID()),
+		trace.Int64Attribute("pid", int64(process.processID)))
+
+	process.handleLock.RLock()
+	defer process.handleLock.RUnlock()
+
+	if process.handle == 0 {
+		return nil, nil, nil, makeProcessError(process, operation, ErrAlreadyClosed, nil)
+	}
+
+	process.stdioLock.Lock()
+	defer process.stdioLock.Unlock()
+	if process.hasCachedStdio {
+		stdin, stdout, stderr := process.stdin, process.stdout, process.stderr
+		process.stdin, process.stdout, process.stderr = nil, nil, nil
+		process.hasCachedStdio = false
+		return stdin, stdout, stderr, nil
+	}
+
+	processInfo, resultJSON, err := vmcompute.HcsGetProcessInfo(ctx, process.handle)
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, nil, nil, makeProcessError(process, operation, err, events)
+	}
+
+	pipes, err := makeOpenFiles([]syscall.Handle{processInfo.StdInput, processInfo.StdOutput, processInfo.StdError})
+	if err != nil {
+		return nil, nil, nil, makeProcessError(process, operation, err, nil)
+	}
+
+	return pipes[0], pipes[1], pipes[2], nil
+}
+
+// Stdio returns the stdin, stdout, and stderr pipes, respectively.
+// To close them, close the process handle.
+func (process *Process) Stdio() (stdin io.Writer, stdout, stderr io.Reader) {
+	process.stdioLock.Lock()
+	defer process.stdioLock.Unlock()
+	return process.stdin, process.stdout, process.stderr
+}
+
+// CloseStdin closes the write side of the stdin pipe so that the process is
+// notified on the read side that there is no more data in stdin.
+func (process *Process) CloseStdin(ctx context.Context) error {
+	process.handleLock.RLock()
+	defer process.handleLock.RUnlock()
+
+	operation := "hcsshim::Process::CloseStdin"
+
+	if process.handle == 0 {
+		return makeProcessError(process, operation, ErrAlreadyClosed, nil)
+	}
+
+	modifyRequest := processModifyRequest{
+		Operation: modifyCloseHandle,
+		CloseHandle: &closeHandle{
+			Handle: stdIn,
+		},
+	}
+
+	modifyRequestb, err := json.Marshal(modifyRequest)
+	if err != nil {
+		return err
+	}
+
+	resultJSON, err := vmcompute.HcsModifyProcess(ctx, process.handle, string(modifyRequestb))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return makeProcessError(process, operation, err, events)
+	}
+
+	process.stdioLock.Lock()
+	if process.stdin != nil {
+		process.stdin.Close()
+		process.stdin = nil
+	}
+	process.stdioLock.Unlock()
+
+	return nil
+}
+
+// Close cleans up any state associated with the process but does not kill
+// or wait on it.
+func (process *Process) Close() (err error) {
+	operation := "hcsshim::Process::Close"
+	ctx, span := trace.StartSpan(context.Background(), operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(
+		trace.StringAttribute("cid", process.SystemID()),
+		trace.Int64Attribute("pid", int64(process.processID)))
+
+	process.handleLock.Lock()
+	defer process.handleLock.Unlock()
+
+	// Don't double free this
+	if process.handle == 0 {
+		return nil
+	}
+
+	process.stdioLock.Lock()
+	if process.stdin != nil {
+		process.stdin.Close()
+		process.stdin = nil
+	}
+	if process.stdout != nil {
+		process.stdout.Close()
+		process.stdout = nil
+	}
+	if process.stderr != nil {
+		process.stderr.Close()
+		process.stderr = nil
+	}
+	process.stdioLock.Unlock()
+
+	if err = process.unregisterCallback(ctx); err != nil {
+		return makeProcessError(process, operation, err, nil)
+	}
+
+	if err = vmcompute.HcsCloseProcess(ctx, process.handle); err != nil {
+		return makeProcessError(process, operation, err, nil)
+	}
+
+	process.handle = 0
+	process.closedWaitOnce.Do(func() {
+		process.exitCode = -1
+		process.waitError = ErrAlreadyClosed
+		close(process.waitBlock)
+	})
+
+	return nil
+}
+
+func (process *Process) registerCallback(ctx context.Context) error {
+	callbackContext := &notifcationWatcherContext{
+		channels:  newProcessChannels(),
+		systemID:  process.SystemID(),
+		processID: process.processID,
+	}
+
+	callbackMapLock.Lock()
+	callbackNumber := nextCallback
+	nextCallback++
+	callbackMap[callbackNumber] = callbackContext
+	callbackMapLock.Unlock()
+
+	callbackHandle, err := vmcompute.HcsRegisterProcessCallback(ctx, process.handle, notificationWatcherCallback, callbackNumber)
+	if err != nil {
+		return err
+	}
+	callbackContext.handle = callbackHandle
+	process.callbackNumber = callbackNumber
+
+	return nil
+}
+
+func (process *Process) unregisterCallback(ctx context.Context) error {
+	callbackNumber := process.callbackNumber
+
+	callbackMapLock.RLock()
+	callbackContext := callbackMap[callbackNumber]
+	callbackMapLock.RUnlock()
+
+	if callbackContext == nil {
+		return nil
+	}
+
+	handle := callbackContext.handle
+
+	if handle == 0 {
+		return nil
+	}
+
+	// vmcompute.HcsUnregisterProcessCallback has its own synchronization to
+	// wait for all callbacks to complete. We must NOT hold the callbackMapLock.
+	err := vmcompute.HcsUnregisterProcessCallback(ctx, handle)
+	if err != nil {
+		return err
+	}
+
+	closeChannels(callbackContext.channels)
+
+	callbackMapLock.Lock()
+	delete(callbackMap, callbackNumber)
+	callbackMapLock.Unlock()
+
+	handle = 0
+
+	return nil
+}
@@ -0,0 +1,49 @@
+package hcs
+
+import (
+	"context"
+	"encoding/json"
+
+	hcsschema "github.com/Microsoft/hcsshim/internal/schema2"
+	"github.com/Microsoft/hcsshim/internal/vmcompute"
+)
+
+// GetServiceProperties returns properties of the host compute service.
+func GetServiceProperties(ctx context.Context, q hcsschema.PropertyQuery) (*hcsschema.ServiceProperties, error) {
+	operation := "hcsshim::GetServiceProperties"
+
+	queryb, err := json.Marshal(q)
+	if err != nil {
+		return nil, err
+	}
+	propertiesJSON, resultJSON, err := vmcompute.HcsGetServiceProperties(ctx, string(queryb))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, &HcsError{Op: operation, Err: err, Events: events}
+	}
+
+	if propertiesJSON == "" {
+		return nil, ErrUnexpectedValue
+	}
+	properties := &hcsschema.ServiceProperties{}
+	if err := json.Unmarshal([]byte(propertiesJSON), properties); err != nil {
+		return nil, err
+	}
+	return properties, nil
+}
+
+// ModifyServiceSettings modifies settings of the host compute service.
+func ModifyServiceSettings(ctx context.Context, settings hcsschema.ModificationRequest) error {
+	operation := "hcsshim::ModifyServiceSettings"
+
+	settingsJSON, err := json.Marshal(settings)
+	if err != nil {
+		return err
+	}
+	resultJSON, err := vmcompute.HcsModifyServiceSettings(ctx, string(settingsJSON))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return &HcsError{Op: operation, Err: err, Events: events}
+	}
+	return nil
+}
@@ -0,0 +1,605 @@
+package hcs
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"strings"
+	"sync"
+	"syscall"
+
+	"github.com/Microsoft/hcsshim/internal/cow"
+	"github.com/Microsoft/hcsshim/internal/log"
+	"github.com/Microsoft/hcsshim/internal/oc"
+	"github.com/Microsoft/hcsshim/internal/schema1"
+	hcsschema "github.com/Microsoft/hcsshim/internal/schema2"
+	"github.com/Microsoft/hcsshim/internal/timeout"
+	"github.com/Microsoft/hcsshim/internal/vmcompute"
+	"go.opencensus.io/trace"
+)
+
+type System struct {
+	handleLock     sync.RWMutex
+	handle         vmcompute.HcsSystem
+	id             string
+	callbackNumber uintptr
+
+	closedWaitOnce sync.Once
+	waitBlock      chan struct{}
+	waitError      error
+	exitError      error
+	os, typ        string
+}
+
+func newSystem(id string) *System {
+	return &System{
+		id:        id,
+		waitBlock: make(chan struct{}),
+	}
+}
+
+// CreateComputeSystem creates a new compute system with the given configuration but does not start it.
+func CreateComputeSystem(ctx context.Context, id string, hcsDocumentInterface interface{}) (_ *System, err error) {
+	operation := "hcsshim::CreateComputeSystem"
+
+	// hcsCreateComputeSystemContext is an async operation. Start the outer span
+	// here to measure the full create time.
+	ctx, span := trace.StartSpan(ctx, operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("cid", id))
+
+	computeSystem := newSystem(id)
+
+	hcsDocumentB, err := json.Marshal(hcsDocumentInterface)
+	if err != nil {
+		return nil, err
+	}
+
+	hcsDocument := string(hcsDocumentB)
+
+	var (
+		identity    syscall.Handle
+		resultJSON  string
+		createError error
+	)
+	computeSystem.handle, resultJSON, createError = vmcompute.HcsCreateComputeSystem(ctx, id, hcsDocument, identity)
+	if createError == nil || IsPending(createError) {
+		defer func() {
+			if err != nil {
+				computeSystem.Close()
+			}
+		}()
+		if err = computeSystem.registerCallback(ctx); err != nil {
+			// Terminate the compute system if it still exists. We're okay to
+			// ignore a failure here.
+			computeSystem.Terminate(ctx)
+			return nil, makeSystemError(computeSystem, operation, "", err, nil)
+		}
+	}
+
+	events, err := processAsyncHcsResult(ctx, createError, resultJSON, computeSystem.callbackNumber, hcsNotificationSystemCreateCompleted, &timeout.SystemCreate)
+	if err != nil {
+		if err == ErrTimeout {
+			// Terminate the compute system if it still exists. We're okay to
+			// ignore a failure here.
+			computeSystem.Terminate(ctx)
+		}
+		return nil, makeSystemError(computeSystem, operation, hcsDocument, err, events)
+	}
+	go computeSystem.waitBackground()
+	if err = computeSystem.getCachedProperties(ctx); err != nil {
+		return nil, err
+	}
+	return computeSystem, nil
+}
+
+// OpenComputeSystem opens an existing compute system by ID.
+func OpenComputeSystem(ctx context.Context, id string) (*System, error) {
+	operation := "hcsshim::OpenComputeSystem"
+
+	computeSystem := newSystem(id)
+	handle, resultJSON, err := vmcompute.HcsOpenComputeSystem(ctx, id)
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, events)
+	}
+	computeSystem.handle = handle
+	defer func() {
+		if err != nil {
+			computeSystem.Close()
+		}
+	}()
+	if err = computeSystem.registerCallback(ctx); err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+	go computeSystem.waitBackground()
+	if err = computeSystem.getCachedProperties(ctx); err != nil {
+		return nil, err
+	}
+	return computeSystem, nil
+}
+
+func (computeSystem *System) getCachedProperties(ctx context.Context) error {
+	props, err := computeSystem.Properties(ctx)
+	if err != nil {
+		return err
+	}
+	computeSystem.typ = strings.ToLower(props.SystemType)
+	computeSystem.os = strings.ToLower(props.RuntimeOSType)
+	if computeSystem.os == "" && computeSystem.typ == "container" {
+		// Pre-RS5 HCS did not return the OS, but it only supported containers
+		// that ran Windows.
+		computeSystem.os = "windows"
+	}
+	return nil
+}
+
+// OS returns the operating system of the compute system, "linux" or "windows".
+func (computeSystem *System) OS() string {
+	return computeSystem.os
+}
+
+// IsOCI returns whether processes in the compute system should be created via
+// OCI.
+func (computeSystem *System) IsOCI() bool {
+	return computeSystem.os == "linux" && computeSystem.typ == "container"
+}
+
+// GetComputeSystems gets a list of the compute systems on the system that match the query
+func GetComputeSystems(ctx context.Context, q schema1.ComputeSystemQuery) ([]schema1.ContainerProperties, error) {
+	operation := "hcsshim::GetComputeSystems"
+
+	queryb, err := json.Marshal(q)
+	if err != nil {
+		return nil, err
+	}
+
+	computeSystemsJSON, resultJSON, err := vmcompute.HcsEnumerateComputeSystems(ctx, string(queryb))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, &HcsError{Op: operation, Err: err, Events: events}
+	}
+
+	if computeSystemsJSON == "" {
+		return nil, ErrUnexpectedValue
+	}
+	computeSystems := []schema1.ContainerProperties{}
+	if err = json.Unmarshal([]byte(computeSystemsJSON), &computeSystems); err != nil {
+		return nil, err
+	}
+
+	return computeSystems, nil
+}
+
+// Start synchronously starts the computeSystem.
+func (computeSystem *System) Start(ctx context.Context) (err error) {
+	operation := "hcsshim::System::Start"
+
+	// hcsStartComputeSystemContext is an async operation. Start the outer span
+	// here to measure the full start time.
+	ctx, span := trace.StartSpan(ctx, operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("cid", computeSystem.id))
+
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	if computeSystem.handle == 0 {
+		return makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	resultJSON, err := vmcompute.HcsStartComputeSystem(ctx, computeSystem.handle, "")
+	events, err := processAsyncHcsResult(ctx, err, resultJSON, computeSystem.callbackNumber, hcsNotificationSystemStartCompleted, &timeout.SystemStart)
+	if err != nil {
+		return makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	return nil
+}
+
+// ID returns the compute system's identifier.
+func (computeSystem *System) ID() string {
+	return computeSystem.id
+}
+
+// Shutdown requests a compute system shutdown.
+func (computeSystem *System) Shutdown(ctx context.Context) error {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::Shutdown"
+
+	if computeSystem.handle == 0 {
+		return nil
+	}
+
+	resultJSON, err := vmcompute.HcsShutdownComputeSystem(ctx, computeSystem.handle, "")
+	events := processHcsResult(ctx, resultJSON)
+	switch err {
+	case nil, ErrVmcomputeAlreadyStopped, ErrComputeSystemDoesNotExist, ErrVmcomputeOperationPending:
+	default:
+		return makeSystemError(computeSystem, operation, "", err, events)
+	}
+	return nil
+}
+
+// Terminate requests a compute system terminate.
+func (computeSystem *System) Terminate(ctx context.Context) error {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::Terminate"
+
+	if computeSystem.handle == 0 {
+		return nil
+	}
+
+	resultJSON, err := vmcompute.HcsTerminateComputeSystem(ctx, computeSystem.handle, "")
+	events := processHcsResult(ctx, resultJSON)
+	switch err {
+	case nil, ErrVmcomputeAlreadyStopped, ErrComputeSystemDoesNotExist, ErrVmcomputeOperationPending:
+	default:
+		return makeSystemError(computeSystem, operation, "", err, events)
+	}
+	return nil
+}
+
+// waitBackground waits for the compute system exit notification. Once received
+// sets `computeSystem.waitError` (if any) and unblocks all `Wait` calls.
+//
+// This MUST be called exactly once per `computeSystem.handle` but `Wait` is
+// safe to call multiple times.
+func (computeSystem *System) waitBackground() {
+	operation := "hcsshim::System::waitBackground"
+	ctx, span := trace.StartSpan(context.Background(), operation)
+	defer span.End()
+	span.AddAttributes(trace.StringAttribute("cid", computeSystem.id))
+
+	err := waitForNotification(ctx, computeSystem.callbackNumber, hcsNotificationSystemExited, nil)
+	switch err {
+	case nil:
+		log.G(ctx).Debug("system exited")
+	case ErrVmcomputeUnexpectedExit:
+		log.G(ctx).Debug("unexpected system exit")
+		computeSystem.exitError = makeSystemError(computeSystem, operation, "", err, nil)
+		err = nil
+	default:
+		err = makeSystemError(computeSystem, operation, "", err, nil)
+	}
+	computeSystem.closedWaitOnce.Do(func() {
+		computeSystem.waitError = err
+		close(computeSystem.waitBlock)
+	})
+	oc.SetSpanStatus(span, err)
+}
+
+// Wait synchronously waits for the compute system to shutdown or terminate. If
+// the compute system has already exited returns the previous error (if any).
+func (computeSystem *System) Wait() error {
+	<-computeSystem.waitBlock
+	return computeSystem.waitError
+}
+
+// ExitError returns an error describing the reason the compute system terminated.
+func (computeSystem *System) ExitError() error {
+	select {
+	case <-computeSystem.waitBlock:
+		if computeSystem.waitError != nil {
+			return computeSystem.waitError
+		}
+		return computeSystem.exitError
+	default:
+		return errors.New("container not exited")
+	}
+}
+
+// Properties returns the requested container properties targeting a V1 schema container.
+func (computeSystem *System) Properties(ctx context.Context, types ...schema1.PropertyType) (*schema1.ContainerProperties, error) {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::Properties"
+
+	queryBytes, err := json.Marshal(schema1.PropertyQuery{PropertyTypes: types})
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	propertiesJSON, resultJSON, err := vmcompute.HcsGetComputeSystemProperties(ctx, computeSystem.handle, string(queryBytes))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	if propertiesJSON == "" {
+		return nil, ErrUnexpectedValue
+	}
+	properties := &schema1.ContainerProperties{}
+	if err := json.Unmarshal([]byte(propertiesJSON), properties); err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	return properties, nil
+}
+
+// PropertiesV2 returns the requested container properties targeting a V2 schema container.
+func (computeSystem *System) PropertiesV2(ctx context.Context, types ...hcsschema.PropertyType) (*hcsschema.Properties, error) {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::PropertiesV2"
+
+	queryBytes, err := json.Marshal(hcsschema.PropertyQuery{PropertyTypes: types})
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	propertiesJSON, resultJSON, err := vmcompute.HcsGetComputeSystemProperties(ctx, computeSystem.handle, string(queryBytes))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	if propertiesJSON == "" {
+		return nil, ErrUnexpectedValue
+	}
+	properties := &hcsschema.Properties{}
+	if err := json.Unmarshal([]byte(propertiesJSON), properties); err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	return properties, nil
+}
+
+// Pause pauses the execution of the computeSystem. This feature is not enabled in TP5.
+func (computeSystem *System) Pause(ctx context.Context) (err error) {
+	operation := "hcsshim::System::Pause"
+
+	// hcsPauseComputeSystemContext is an async peration. Start the outer span
+	// here to measure the full pause time.
+	ctx, span := trace.StartSpan(ctx, operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("cid", computeSystem.id))
+
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	if computeSystem.handle == 0 {
+		return makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	resultJSON, err := vmcompute.HcsPauseComputeSystem(ctx, computeSystem.handle, "")
+	events, err := processAsyncHcsResult(ctx, err, resultJSON, computeSystem.callbackNumber, hcsNotificationSystemPauseCompleted, &timeout.SystemPause)
+	if err != nil {
+		return makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	return nil
+}
+
+// Resume resumes the execution of the computeSystem. This feature is not enabled in TP5.
+func (computeSystem *System) Resume(ctx context.Context) (err error) {
+	operation := "hcsshim::System::Resume"
+
+	// hcsResumeComputeSystemContext is an async operation. Start the outer span
+	// here to measure the full restore time.
+	ctx, span := trace.StartSpan(ctx, operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("cid", computeSystem.id))
+
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	if computeSystem.handle == 0 {
+		return makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	resultJSON, err := vmcompute.HcsResumeComputeSystem(ctx, computeSystem.handle, "")
+	events, err := processAsyncHcsResult(ctx, err, resultJSON, computeSystem.callbackNumber, hcsNotificationSystemResumeCompleted, &timeout.SystemResume)
+	if err != nil {
+		return makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	return nil
+}
+
+func (computeSystem *System) createProcess(ctx context.Context, operation string, c interface{}) (*Process, *vmcompute.HcsProcessInformation, error) {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	if computeSystem.handle == 0 {
+		return nil, nil, makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	configurationb, err := json.Marshal(c)
+	if err != nil {
+		return nil, nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	configuration := string(configurationb)
+	processInfo, processHandle, resultJSON, err := vmcompute.HcsCreateProcess(ctx, computeSystem.handle, configuration)
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, nil, makeSystemError(computeSystem, operation, configuration, err, events)
+	}
+
+	log.G(ctx).WithField("pid", processInfo.ProcessId).Debug("created process pid")
+	return newProcess(processHandle, int(processInfo.ProcessId), computeSystem), &processInfo, nil
+}
+
+// CreateProcess launches a new process within the computeSystem.
+func (computeSystem *System) CreateProcess(ctx context.Context, c interface{}) (cow.Process, error) {
+	operation := "hcsshim::System::CreateProcess"
+	process, processInfo, err := computeSystem.createProcess(ctx, operation, c)
+	if err != nil {
+		return nil, err
+	}
+	defer func() {
+		if err != nil {
+			process.Close()
+		}
+	}()
+
+	pipes, err := makeOpenFiles([]syscall.Handle{processInfo.StdInput, processInfo.StdOutput, processInfo.StdError})
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+	process.stdin = pipes[0]
+	process.stdout = pipes[1]
+	process.stderr = pipes[2]
+	process.hasCachedStdio = true
+
+	if err = process.registerCallback(ctx); err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+	go process.waitBackground()
+
+	return process, nil
+}
+
+// OpenProcess gets an interface to an existing process within the computeSystem.
+func (computeSystem *System) OpenProcess(ctx context.Context, pid int) (*Process, error) {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::OpenProcess"
+
+	if computeSystem.handle == 0 {
+		return nil, makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	processHandle, resultJSON, err := vmcompute.HcsOpenProcess(ctx, computeSystem.handle, uint32(pid))
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, events)
+	}
+
+	process := newProcess(processHandle, pid, computeSystem)
+	if err = process.registerCallback(ctx); err != nil {
+		return nil, makeSystemError(computeSystem, operation, "", err, nil)
+	}
+	go process.waitBackground()
+
+	return process, nil
+}
+
+// Close cleans up any state associated with the compute system but does not terminate or wait for it.
+func (computeSystem *System) Close() (err error) {
+	operation := "hcsshim::System::Close"
+	ctx, span := trace.StartSpan(context.Background(), operation)
+	defer span.End()
+	defer func() { oc.SetSpanStatus(span, err) }()
+	span.AddAttributes(trace.StringAttribute("cid", computeSystem.id))
+
+	computeSystem.handleLock.Lock()
+	defer computeSystem.handleLock.Unlock()
+
+	// Don't double free this
+	if computeSystem.handle == 0 {
+		return nil
+	}
+
+	if err = computeSystem.unregisterCallback(ctx); err != nil {
+		return makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	err = vmcompute.HcsCloseComputeSystem(ctx, computeSystem.handle)
+	if err != nil {
+		return makeSystemError(computeSystem, operation, "", err, nil)
+	}
+
+	computeSystem.handle = 0
+	computeSystem.closedWaitOnce.Do(func() {
+		computeSystem.waitError = ErrAlreadyClosed
+		close(computeSystem.waitBlock)
+	})
+
+	return nil
+}
+
+func (computeSystem *System) registerCallback(ctx context.Context) error {
+	callbackContext := &notifcationWatcherContext{
+		channels: newSystemChannels(),
+		systemID: computeSystem.id,
+	}
+
+	callbackMapLock.Lock()
+	callbackNumber := nextCallback
+	nextCallback++
+	callbackMap[callbackNumber] = callbackContext
+	callbackMapLock.Unlock()
+
+	callbackHandle, err := vmcompute.HcsRegisterComputeSystemCallback(ctx, computeSystem.handle, notificationWatcherCallback, callbackNumber)
+	if err != nil {
+		return err
+	}
+	callbackContext.handle = callbackHandle
+	computeSystem.callbackNumber = callbackNumber
+
+	return nil
+}
+
+func (computeSystem *System) unregisterCallback(ctx context.Context) error {
+	callbackNumber := computeSystem.callbackNumber
+
+	callbackMapLock.RLock()
+	callbackContext := callbackMap[callbackNumber]
+	callbackMapLock.RUnlock()
+
+	if callbackContext == nil {
+		return nil
+	}
+
+	handle := callbackContext.handle
+
+	if handle == 0 {
+		return nil
+	}
+
+	// hcsUnregisterComputeSystemCallback has its own syncronization
+	// to wait for all callbacks to complete. We must NOT hold the callbackMapLock.
+	err := vmcompute.HcsUnregisterComputeSystemCallback(ctx, handle)
+	if err != nil {
+		return err
+	}
+
+	closeChannels(callbackContext.channels)
+
+	callbackMapLock.Lock()
+	delete(callbackMap, callbackNumber)
+	callbackMapLock.Unlock()
+
+	handle = 0
+
+	return nil
+}
+
+// Modify the System by sending a request to HCS
+func (computeSystem *System) Modify(ctx context.Context, config interface{}) error {
+	computeSystem.handleLock.RLock()
+	defer computeSystem.handleLock.RUnlock()
+
+	operation := "hcsshim::System::Modify"
+
+	if computeSystem.handle == 0 {
+		return makeSystemError(computeSystem, operation, "", ErrAlreadyClosed, nil)
+	}
+
+	requestBytes, err := json.Marshal(config)
+	if err != nil {
+		return err
+	}
+
+	requestJSON := string(requestBytes)
+	resultJSON, err := vmcompute.HcsModifyComputeSystem(ctx, computeSystem.handle, requestJSON)
+	events := processHcsResult(ctx, resultJSON)
+	if err != nil {
+		return makeSystemError(computeSystem, operation, requestJSON, err, events)
+	}
+
+	return nil
+}
@@ -0,0 +1,62 @@
+package hcs
+
+import (
+	"context"
+	"io"
+	"syscall"
+
+	"github.com/Microsoft/go-winio"
+	diskutil "github.com/Microsoft/go-winio/vhd"
+	"github.com/Microsoft/hcsshim/computestorage"
+	"github.com/pkg/errors"
+	"golang.org/x/sys/windows"
+)
+
+// makeOpenFiles calls winio.MakeOpenFile for each handle in a slice but closes all the handles
+// if there is an error.
+func makeOpenFiles(hs []syscall.Handle) (_ []io.ReadWriteCloser, err error) {
+	fs := make([]io.ReadWriteCloser, len(hs))
+	for i, h := range hs {
+		if h != syscall.Handle(0) {
+			if err == nil {
+				fs[i], err = winio.MakeOpenFile(h)
+			}
+			if err != nil {
+				syscall.Close(h)
+			}
+		}
+	}
+	if err != nil {
+		for _, f := range fs {
+			if f != nil {
+				f.Close()
+			}
+		}
+		return nil, err
+	}
+	return fs, nil
+}
+
+// CreateNTFSVHD creates a VHD formatted with NTFS of size `sizeGB` at the given `vhdPath`.
+func CreateNTFSVHD(ctx context.Context, vhdPath string, sizeGB uint32) (err error) {
+	if err := diskutil.CreateVhdx(vhdPath, sizeGB, 1); err != nil {
+		return errors.Wrap(err, "failed to create VHD")
+	}
+
+	vhd, err := diskutil.OpenVirtualDisk(vhdPath, diskutil.VirtualDiskAccessNone, diskutil.OpenVirtualDiskFlagNone)
+	if err != nil {
+		return errors.Wrap(err, "failed to open VHD")
+	}
+	defer func() {
+		err2 := windows.CloseHandle(windows.Handle(vhd))
+		if err == nil {
+			err = errors.Wrap(err2, "failed to close VHD")
+		}
+	}()
+
+	if err := computestorage.FormatWritableLayerVhd(ctx, windows.Handle(vhd)); err != nil {
+		return errors.Wrap(err, "failed to format VHD")
+	}
+
+	return nil
+}
@@ -0,0 +1,69 @@
+package hcs
+
+import (
+	"context"
+	"time"
+
+	"github.com/Microsoft/hcsshim/internal/log"
+)
+
+func processAsyncHcsResult(ctx context.Context, err error, resultJSON string, callbackNumber uintptr, expectedNotification hcsNotification, timeout *time.Duration) ([]ErrorEvent, error) {
+	events := processHcsResult(ctx, resultJSON)
+	if IsPending(err) {
+		return nil, waitForNotification(ctx, callbackNumber, expectedNotification, timeout)
+	}
+
+	return events, err
+}
+
+func waitForNotification(ctx context.Context, callbackNumber uintptr, expectedNotification hcsNotification, timeout *time.Duration) error {
+	callbackMapLock.RLock()
+	if _, ok := callbackMap[callbackNumber]; !ok {
+		callbackMapLock.RUnlock()
+		log.G(ctx).WithField("callbackNumber", callbackNumber).Error("failed to waitForNotification: callbackNumber does not exist in callbackMap")
+		return ErrHandleClose
+	}
+	channels := callbackMap[callbackNumber].channels
+	callbackMapLock.RUnlock()
+
+	expectedChannel := channels[expectedNotification]
+	if expectedChannel == nil {
+		log.G(ctx).WithField("type", expectedNotification).Error("unknown notification type in waitForNotification")
+		return ErrInvalidNotificationType
+	}
+
+	var c <-chan time.Time
+	if timeout != nil {
+		timer := time.NewTimer(*timeout)
+		c = timer.C
+		defer timer.Stop()
+	}
+
+	select {
+	case err, ok := <-expectedChannel:
+		if !ok {
+			return ErrHandleClose
+		}
+		return err
+	case err, ok := <-channels[hcsNotificationSystemExited]:
+		if !ok {
+			return ErrHandleClose
+		}
+		// If the expected notification is hcsNotificationSystemExited which of the two selects
+		// chosen is random. Return the raw error if hcsNotificationSystemExited is expected
+		if channels[hcsNotificationSystemExited] == expectedChannel {
+			return err
+		}
+		return ErrUnexpectedContainerExit
+	case _, ok := <-channels[hcsNotificationServiceDisconnect]:
+		if !ok {
+			return ErrHandleClose
+		}
+		// hcsNotificationServiceDisconnect should never be an expected notification
+		// it does not need the same handling as hcsNotificationSystemExited
+		return ErrUnexpectedProcessAbort
+	case <-c:
+		return ErrTimeout
+	}
+	return nil
+}
@@ -0,0 +1,47 @@
+package hcserror
+
+import (
+	"fmt"
+	"syscall"
+)
+
+const ERROR_GEN_FAILURE = syscall.Errno(31)
+
+type HcsError struct {
+	title string
+	rest  string
+	Err   error
+}
+
+func (e *HcsError) Error() string {
+	s := e.title
+	if len(s) > 0 && s[len(s)-1] != ' ' {
+		s += " "
+	}
+	s += fmt.Sprintf("failed in Win32: %s (0x%x)", e.Err, Win32FromError(e.Err))
+	if e.rest != "" {
+		if e.rest[0] != ' ' {
+			s += " "
+		}
+		s += e.rest
+	}
+	return s
+}
+
+func New(err error, title, rest string) error {
+	// Pass through DLL errors directly since they do not originate from HCS.
+	if _, ok := err.(*syscall.DLLError); ok {
+		return err
+	}
+	return &HcsError{title, rest, err}
+}
+
+func Win32FromError(err error) uint32 {
+	if herr, ok := err.(*HcsError); ok {
+		return Win32FromError(herr.Err)
+	}
+	if code, ok := err.(syscall.Errno); ok {
+		return uint32(code)
+	}
+	return uint32(ERROR_GEN_FAILURE)
+}
@@ -0,0 +1,23 @@
+package hns
+
+import "fmt"
+
+//go:generate go run ../../mksyscall_windows.go -output zsyscall_windows.go hns.go
+
+//sys _hnsCall(method string, path string, object string, response **uint16) (hr error) = vmcompute.HNSCall?
+
+type EndpointNotFoundError struct {
+	EndpointName string
+}
+
+func (e EndpointNotFoundError) Error() string {
+	return fmt.Sprintf("Endpoint %s not found", e.EndpointName)
+}
+
+type NetworkNotFoundError struct {
+	NetworkName string
+}
+
+func (e NetworkNotFoundError) Error() string {
+	return fmt.Sprintf("Network %s not found", e.NetworkName)
+}
@@ -0,0 +1,308 @@
+package hns
+
+import (
+	"encoding/json"
+	"net"
+	"strings"
+
+	"github.com/sirupsen/logrus"
+)
+
+// HNSEndpoint represents a network endpoint in HNS
+type HNSEndpoint struct {
+	Id                 string            `json:"ID,omitempty"`
+	Name               string            `json:",omitempty"`
+	VirtualNetwork     string            `json:",omitempty"`
+	VirtualNetworkName string            `json:",omitempty"`
+	Policies           []json.RawMessage `json:",omitempty"`
+	MacAddress         string            `json:",omitempty"`
+	IPAddress          net.IP            `json:",omitempty"`
+	IPv6Address        net.IP            `json:",omitempty"`
+	DNSSuffix          string            `json:",omitempty"`
+	DNSServerList      string            `json:",omitempty"`
+	GatewayAddress     string            `json:",omitempty"`
+	GatewayAddressV6   string            `json:",omitempty"`
+	EnableInternalDNS  bool              `json:",omitempty"`
+	DisableICC         bool              `json:",omitempty"`
+	PrefixLength       uint8             `json:",omitempty"`
+	IPv6PrefixLength   uint8             `json:",omitempty"`
+	IsRemoteEndpoint   bool              `json:",omitempty"`
+	EnableLowMetric    bool              `json:",omitempty"`
+	Namespace          *Namespace        `json:",omitempty"`
+	EncapOverhead      uint16            `json:",omitempty"`
+}
+
+//SystemType represents the type of the system on which actions are done
+type SystemType string
+
+// SystemType const
+const (
+	ContainerType      SystemType = "Container"
+	VirtualMachineType SystemType = "VirtualMachine"
+	HostType           SystemType = "Host"
+)
+
+// EndpointAttachDetachRequest is the structure used to send request to the container to modify the system
+// Supported resource types are Network and Request Types are Add/Remove
+type EndpointAttachDetachRequest struct {
+	ContainerID    string     `json:"ContainerId,omitempty"`
+	SystemType     SystemType `json:"SystemType"`
+	CompartmentID  uint16     `json:"CompartmentId,omitempty"`
+	VirtualNICName string     `json:"VirtualNicName,omitempty"`
+}
+
+// EndpointResquestResponse is object to get the endpoint request response
+type EndpointResquestResponse struct {
+	Success bool
+	Error   string
+}
+
+// HNSEndpointRequest makes a HNS call to modify/query a network endpoint
+func HNSEndpointRequest(method, path, request string) (*HNSEndpoint, error) {
+	endpoint := &HNSEndpoint{}
+	err := hnsCall(method, "/endpoints/"+path, request, &endpoint)
+	if err != nil {
+		return nil, err
+	}
+
+	return endpoint, nil
+}
+
+// HNSListEndpointRequest makes a HNS call to query the list of available endpoints
+func HNSListEndpointRequest() ([]HNSEndpoint, error) {
+	var endpoint []HNSEndpoint
+	err := hnsCall("GET", "/endpoints/", "", &endpoint)
+	if err != nil {
+		return nil, err
+	}
+
+	return endpoint, nil
+}
+
+// GetHNSEndpointByID get the Endpoint by ID
+func GetHNSEndpointByID(endpointID string) (*HNSEndpoint, error) {
+	return HNSEndpointRequest("GET", endpointID, "")
+}
+
+// GetHNSEndpointByName gets the endpoint filtered by Name
+func GetHNSEndpointByName(endpointName string) (*HNSEndpoint, error) {
+	hnsResponse, err := HNSListEndpointRequest()
+	if err != nil {
+		return nil, err
+	}
+	for _, hnsEndpoint := range hnsResponse {
+		if hnsEndpoint.Name == endpointName {
+			return &hnsEndpoint, nil
+		}
+	}
+	return nil, EndpointNotFoundError{EndpointName: endpointName}
+}
+
+type endpointAttachInfo struct {
+	SharedContainers json.RawMessage `json:",omitempty"`
+}
+
+func (endpoint *HNSEndpoint) IsAttached(vID string) (bool, error) {
+	attachInfo := endpointAttachInfo{}
+	err := hnsCall("GET", "/endpoints/"+endpoint.Id, "", &attachInfo)
+
+	// Return false allows us to just return the err
+	if err != nil {
+		return false, err
+	}
+
+	if strings.Contains(strings.ToLower(string(attachInfo.SharedContainers)), strings.ToLower(vID)) {
+		return true, nil
+	}
+
+	return false, nil
+
+}
+
+// Create Endpoint by sending EndpointRequest to HNS. TODO: Create a separate HNS interface to place all these methods
+func (endpoint *HNSEndpoint) Create() (*HNSEndpoint, error) {
+	operation := "Create"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	jsonString, err := json.Marshal(endpoint)
+	if err != nil {
+		return nil, err
+	}
+	return HNSEndpointRequest("POST", "", string(jsonString))
+}
+
+// Delete Endpoint by sending EndpointRequest to HNS
+func (endpoint *HNSEndpoint) Delete() (*HNSEndpoint, error) {
+	operation := "Delete"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	return HNSEndpointRequest("DELETE", endpoint.Id, "")
+}
+
+// Update Endpoint
+func (endpoint *HNSEndpoint) Update() (*HNSEndpoint, error) {
+	operation := "Update"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+	jsonString, err := json.Marshal(endpoint)
+	if err != nil {
+		return nil, err
+	}
+	err = hnsCall("POST", "/endpoints/"+endpoint.Id, string(jsonString), &endpoint)
+
+	return endpoint, err
+}
+
+// ApplyACLPolicy applies a set of ACL Policies on the Endpoint
+func (endpoint *HNSEndpoint) ApplyACLPolicy(policies ...*ACLPolicy) error {
+	operation := "ApplyACLPolicy"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	for _, policy := range policies {
+		if policy == nil {
+			continue
+		}
+		jsonString, err := json.Marshal(policy)
+		if err != nil {
+			return err
+		}
+		endpoint.Policies = append(endpoint.Policies, jsonString)
+	}
+
+	_, err := endpoint.Update()
+	return err
+}
+
+// ApplyProxyPolicy applies a set of Proxy Policies on the Endpoint
+func (endpoint *HNSEndpoint) ApplyProxyPolicy(policies ...*ProxyPolicy) error {
+	operation := "ApplyProxyPolicy"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	for _, policy := range policies {
+		if policy == nil {
+			continue
+		}
+		jsonString, err := json.Marshal(policy)
+		if err != nil {
+			return err
+		}
+		endpoint.Policies = append(endpoint.Policies, jsonString)
+	}
+
+	_, err := endpoint.Update()
+	return err
+}
+
+// ContainerAttach attaches an endpoint to container
+func (endpoint *HNSEndpoint) ContainerAttach(containerID string, compartmentID uint16) error {
+	operation := "ContainerAttach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	requestMessage := &EndpointAttachDetachRequest{
+		ContainerID:   containerID,
+		CompartmentID: compartmentID,
+		SystemType:    ContainerType,
+	}
+	response := &EndpointResquestResponse{}
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/attach", string(jsonString), &response)
+}
+
+// ContainerDetach detaches an endpoint from container
+func (endpoint *HNSEndpoint) ContainerDetach(containerID string) error {
+	operation := "ContainerDetach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	requestMessage := &EndpointAttachDetachRequest{
+		ContainerID: containerID,
+		SystemType:  ContainerType,
+	}
+	response := &EndpointResquestResponse{}
+
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/detach", string(jsonString), &response)
+}
+
+// HostAttach attaches a nic on the host
+func (endpoint *HNSEndpoint) HostAttach(compartmentID uint16) error {
+	operation := "HostAttach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+	requestMessage := &EndpointAttachDetachRequest{
+		CompartmentID: compartmentID,
+		SystemType:    HostType,
+	}
+	response := &EndpointResquestResponse{}
+
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/attach", string(jsonString), &response)
+
+}
+
+// HostDetach detaches a nic on the host
+func (endpoint *HNSEndpoint) HostDetach() error {
+	operation := "HostDetach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+	requestMessage := &EndpointAttachDetachRequest{
+		SystemType: HostType,
+	}
+	response := &EndpointResquestResponse{}
+
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/detach", string(jsonString), &response)
+}
+
+// VirtualMachineNICAttach attaches a endpoint to a virtual machine
+func (endpoint *HNSEndpoint) VirtualMachineNICAttach(virtualMachineNICName string) error {
+	operation := "VirtualMachineNicAttach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+	requestMessage := &EndpointAttachDetachRequest{
+		VirtualNICName: virtualMachineNICName,
+		SystemType:     VirtualMachineType,
+	}
+	response := &EndpointResquestResponse{}
+
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/attach", string(jsonString), &response)
+}
+
+// VirtualMachineNICDetach detaches a endpoint  from a virtual machine
+func (endpoint *HNSEndpoint) VirtualMachineNICDetach() error {
+	operation := "VirtualMachineNicDetach"
+	title := "hcsshim::HNSEndpoint::" + operation
+	logrus.Debugf(title+" id=%s", endpoint.Id)
+
+	requestMessage := &EndpointAttachDetachRequest{
+		SystemType: VirtualMachineType,
+	}
+	response := &EndpointResquestResponse{}
+
+	jsonString, err := json.Marshal(requestMessage)
+	if err != nil {
+		return err
+	}
+	return hnsCall("POST", "/endpoints/"+endpoint.Id+"/detach", string(jsonString), &response)
+}
@@ -0,0 +1,49 @@
+package hns
+
+import (
+	"encoding/json"
+	"fmt"
+
+	"github.com/Microsoft/hcsshim/internal/hcserror"
+	"github.com/Microsoft/hcsshim/internal/interop"
+	"github.com/sirupsen/logrus"
+)
+
+func hnsCallRawResponse(method, path, request string) (*hnsResponse, error) {
+	var responseBuffer *uint16
+	logrus.Debugf("[%s]=>[%s] Request : %s", method, path, request)
+
+	err := _hnsCall(method, path, request, &responseBuffer)
+	if err != nil {
+		return nil, hcserror.New(err, "hnsCall ", "")
+	}
+	response := interop.ConvertAndFreeCoTaskMemString(responseBuffer)
+
+	hnsresponse := &hnsResponse{}
+	if err = json.Unmarshal([]byte(response), &hnsresponse); err != nil {
+		return nil, err
+	}
+	return hnsresponse, nil
+}
+
+func hnsCall(method, path, request string, returnResponse interface{}) error {
+	hnsresponse, err := hnsCallRawResponse(method, path, request)
+	if err != nil {
+		return fmt.Errorf("failed during hnsCallRawResponse: %v", err)
+	}
+	if !hnsresponse.Success {
+		return fmt.Errorf("hns failed with error : %s", hnsresponse.Error)
+	}
+
+	if len(hnsresponse.Output) == 0 {
+		return nil
+	}
+
+	logrus.Debugf("Network Response : %s", hnsresponse.Output)
+	err = json.Unmarshal(hnsresponse.Output, returnResponse)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
@@ -0,0 +1,28 @@
+package hns
+
+type HNSGlobals struct {
+	Version HNSVersion `json:"Version"`
+}
+
+type HNSVersion struct {
+	Major int `json:"Major"`
+	Minor int `json:"Minor"`
+}
+
+var (
+	HNSVersion1803 = HNSVersion{Major: 7, Minor: 2}
+)
+
+func GetHNSGlobals() (*HNSGlobals, error) {
+	var version HNSVersion
+	err := hnsCall("GET", "/globals/version", "", &version)
+	if err != nil {
+		return nil, err
+	}
+
+	globals := &HNSGlobals{
+		Version: version,
+	}
+
+	return globals, nil
+}
@@ -0,0 +1,147 @@
+package hns
+
+import (
+	"encoding/json"
+	"errors"
+	"github.com/sirupsen/logrus"
+	"net"
+)
+
+// Subnet is assoicated with a network and represents a list
+// of subnets available to the network
+type Subnet struct {
+	AddressPrefix  string            `json:",omitempty"`
+	GatewayAddress string            `json:",omitempty"`
+	Policies       []json.RawMessage `json:",omitempty"`
+}
+
+// MacPool is assoicated with a network and represents a list
+// of macaddresses available to the network
+type MacPool struct {
+	StartMacAddress string `json:",omitempty"`
+	EndMacAddress   string `json:",omitempty"`
+}
+
+// HNSNetwork represents a network in HNS
+type HNSNetwork struct {
+	Id                   string            `json:"ID,omitempty"`
+	Name                 string            `json:",omitempty"`
+	Type                 string            `json:",omitempty"`
+	NetworkAdapterName   string            `json:",omitempty"`
+	SourceMac            string            `json:",omitempty"`
+	Policies             []json.RawMessage `json:",omitempty"`
+	MacPools             []MacPool         `json:",omitempty"`
+	Subnets              []Subnet          `json:",omitempty"`
+	DNSSuffix            string            `json:",omitempty"`
+	DNSServerList        string            `json:",omitempty"`
+	DNSServerCompartment uint32            `json:",omitempty"`
+	ManagementIP         string            `json:",omitempty"`
+	AutomaticDNS         bool              `json:",omitempty"`
+}
+
+type hnsNetworkResponse struct {
+	Success bool
+	Error   string
+	Output  HNSNetwork
+}
+
+type hnsResponse struct {
+	Success bool
+	Error   string
+	Output  json.RawMessage
+}
+
+// HNSNetworkRequest makes a call into HNS to update/query a single network
+func HNSNetworkRequest(method, path, request string) (*HNSNetwork, error) {
+	var network HNSNetwork
+	err := hnsCall(method, "/networks/"+path, request, &network)
+	if err != nil {
+		return nil, err
+	}
+
+	return &network, nil
+}
+
+// HNSListNetworkRequest makes a HNS call to query the list of available networks
+func HNSListNetworkRequest(method, path, request string) ([]HNSNetwork, error) {
+	var network []HNSNetwork
+	err := hnsCall(method, "/networks/"+path, request, &network)
+	if err != nil {
+		return nil, err
+	}
+
+	return network, nil
+}
+
+// GetHNSNetworkByID
+func GetHNSNetworkByID(networkID string) (*HNSNetwork, error) {
+	return HNSNetworkRequest("GET", networkID, "")
+}
+
+// GetHNSNetworkName filtered by Name
+func GetHNSNetworkByName(networkName string) (*HNSNetwork, error) {
+	hsnnetworks, err := HNSListNetworkRequest("GET", "", "")
+	if err != nil {
+		return nil, err
+	}
+	for _, hnsnetwork := range hsnnetworks {
+		if hnsnetwork.Name == networkName {
+			return &hnsnetwork, nil
+		}
+	}
+	return nil, NetworkNotFoundError{NetworkName: networkName}
+}
+
+// Create Network by sending NetworkRequest to HNS.
+func (network *HNSNetwork) Create() (*HNSNetwork, error) {
+	operation := "Create"
+	title := "hcsshim::HNSNetwork::" + operation
+	logrus.Debugf(title+" id=%s", network.Id)
+
+	for _, subnet := range network.Subnets {
+		if (subnet.AddressPrefix != "") && (subnet.GatewayAddress == "") {
+			return nil, errors.New("network create error, subnet has address prefix but no gateway specified")
+		}
+	}
+
+	jsonString, err := json.Marshal(network)
+	if err != nil {
+		return nil, err
+	}
+	return HNSNetworkRequest("POST", "", string(jsonString))
+}
+
+// Delete Network by sending NetworkRequest to HNS
+func (network *HNSNetwork) Delete() (*HNSNetwork, error) {
+	operation := "Delete"
+	title := "hcsshim::HNSNetwork::" + operation
+	logrus.Debugf(title+" id=%s", network.Id)
+
+	return HNSNetworkRequest("DELETE", network.Id, "")
+}
+
+// Creates an endpoint on the Network.
+func (network *HNSNetwork) NewEndpoint(ipAddress net.IP, macAddress net.HardwareAddr) *HNSEndpoint {
+	return &HNSEndpoint{
+		VirtualNetwork: network.Id,
+		IPAddress:      ipAddress,
+		MacAddress:     string(macAddress),
+	}
+}
+
+func (network *HNSNetwork) CreateEndpoint(endpoint *HNSEndpoint) (*HNSEndpoint, error) {
+	operation := "CreateEndpoint"
+	title := "hcsshim::HNSNetwork::" + operation
+	logrus.Debugf(title+" id=%s, endpointId=%s", network.Id, endpoint.Id)
+
+	endpoint.VirtualNetwork = network.Id
+	return endpoint.Create()
+}
+
+func (network *HNSNetwork) CreateRemoteEndpoint(endpoint *HNSEndpoint) (*HNSEndpoint, error) {
+	operation := "CreateRemoteEndpoint"
+	title := "hcsshim::HNSNetwork::" + operation
+	logrus.Debugf(title+" id=%s", network.Id)
+	endpoint.IsRemoteEndpoint = true
+	return network.CreateEndpoint(endpoint)
+}
@@ -0,0 +1,109 @@
+package hns
+
+// Type of Request Support in ModifySystem
+type PolicyType string
+
+// RequestType const
+const (
+	Nat                  PolicyType = "NAT"
+	ACL                  PolicyType = "ACL"
+	PA                   PolicyType = "PA"
+	VLAN                 PolicyType = "VLAN"
+	VSID                 PolicyType = "VSID"
+	VNet                 PolicyType = "VNET"
+	L2Driver             PolicyType = "L2Driver"
+	Isolation            PolicyType = "Isolation"
+	QOS                  PolicyType = "QOS"
+	OutboundNat          PolicyType = "OutBoundNAT"
+	ExternalLoadBalancer PolicyType = "ELB"
+	Route                PolicyType = "ROUTE"
+	Proxy                PolicyType = "PROXY"
+)
+
+type NatPolicy struct {
+	Type         PolicyType `json:"Type"`
+	Protocol     string
+	InternalPort uint16
+	ExternalPort uint16
+}
+
+type QosPolicy struct {
+	Type                            PolicyType `json:"Type"`
+	MaximumOutgoingBandwidthInBytes uint64
+}
+
+type IsolationPolicy struct {
+	Type               PolicyType `json:"Type"`
+	VLAN               uint
+	VSID               uint
+	InDefaultIsolation bool
+}
+
+type VlanPolicy struct {
+	Type PolicyType `json:"Type"`
+	VLAN uint
+}
+
+type VsidPolicy struct {
+	Type PolicyType `json:"Type"`
+	VSID uint
+}
+
+type PaPolicy struct {
+	Type PolicyType `json:"Type"`
+	PA   string     `json:"PA"`
+}
+
+type OutboundNatPolicy struct {
+	Policy
+	VIP          string   `json:"VIP,omitempty"`
+	Exceptions   []string `json:"ExceptionList,omitempty"`
+	Destinations []string `json:",omitempty"`
+}
+
+type ProxyPolicy struct {
+	Type          PolicyType `json:"Type"`
+	IP            string     `json:",omitempty"`
+	Port          string     `json:",omitempty"`
+	ExceptionList []string   `json:",omitempty"`
+	Destination   string     `json:",omitempty"`
+	OutboundNat   bool       `json:",omitempty"`
+}
+
+type ActionType string
+type DirectionType string
+type RuleType string
+
+const (
+	Allow ActionType = "Allow"
+	Block ActionType = "Block"
+
+	In  DirectionType = "In"
+	Out DirectionType = "Out"
+
+	Host   RuleType = "Host"
+	Switch RuleType = "Switch"
+)
+
+type ACLPolicy struct {
+	Type            PolicyType `json:"Type"`
+	Id              string     `json:"Id,omitempty"`
+	Protocol        uint16
+	Protocols       string `json:"Protocols,omitempty"`
+	InternalPort    uint16
+	Action          ActionType
+	Direction       DirectionType
+	LocalAddresses  string
+	RemoteAddresses string
+	LocalPorts      string `json:"LocalPorts,omitempty"`
+	LocalPort       uint16
+	RemotePorts     string `json:"RemotePorts,omitempty"`
+	RemotePort      uint16
+	RuleType        RuleType `json:"RuleType,omitempty"`
+	Priority        uint16
+	ServiceName     string
+}
+
+type Policy struct {
+	Type PolicyType `json:"Type"`
+}
@@ -0,0 +1,201 @@
+package hns
+
+import (
+	"encoding/json"
+
+	"github.com/sirupsen/logrus"
+)
+
+// RoutePolicy is a structure defining schema for Route based Policy
+type RoutePolicy struct {
+	Policy
+	DestinationPrefix string `json:"DestinationPrefix,omitempty"`
+	NextHop           string `json:"NextHop,omitempty"`
+	EncapEnabled      bool   `json:"NeedEncap,omitempty"`
+}
+
+// ELBPolicy is a structure defining schema for ELB LoadBalancing based Policy
+type ELBPolicy struct {
+	LBPolicy
+	SourceVIP string   `json:"SourceVIP,omitempty"`
+	VIPs      []string `json:"VIPs,omitempty"`
+	ILB       bool     `json:"ILB,omitempty"`
+	DSR       bool     `json:"IsDSR,omitempty"`
+}
+
+// LBPolicy is a structure defining schema for LoadBalancing based Policy
+type LBPolicy struct {
+	Policy
+	Protocol     uint16 `json:"Protocol,omitempty"`
+	InternalPort uint16
+	ExternalPort uint16
+}
+
+// PolicyList is a structure defining schema for Policy list request
+type PolicyList struct {
+	ID                 string            `json:"ID,omitempty"`
+	EndpointReferences []string          `json:"References,omitempty"`
+	Policies           []json.RawMessage `json:"Policies,omitempty"`
+}
+
+// HNSPolicyListRequest makes a call into HNS to update/query a single network
+func HNSPolicyListRequest(method, path, request string) (*PolicyList, error) {
+	var policy PolicyList
+	err := hnsCall(method, "/policylists/"+path, request, &policy)
+	if err != nil {
+		return nil, err
+	}
+
+	return &policy, nil
+}
+
+// HNSListPolicyListRequest gets all the policy list
+func HNSListPolicyListRequest() ([]PolicyList, error) {
+	var plist []PolicyList
+	err := hnsCall("GET", "/policylists/", "", &plist)
+	if err != nil {
+		return nil, err
+	}
+
+	return plist, nil
+}
+
+// PolicyListRequest makes a HNS call to modify/query a network policy list
+func PolicyListRequest(method, path, request string) (*PolicyList, error) {
+	policylist := &PolicyList{}
+	err := hnsCall(method, "/policylists/"+path, request, &policylist)
+	if err != nil {
+		return nil, err
+	}
+
+	return policylist, nil
+}
+
+// GetPolicyListByID get the policy list by ID
+func GetPolicyListByID(policyListID string) (*PolicyList, error) {
+	return PolicyListRequest("GET", policyListID, "")
+}
+
+// Create PolicyList by sending PolicyListRequest to HNS.
+func (policylist *PolicyList) Create() (*PolicyList, error) {
+	operation := "Create"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" id=%s", policylist.ID)
+	jsonString, err := json.Marshal(policylist)
+	if err != nil {
+		return nil, err
+	}
+	return PolicyListRequest("POST", "", string(jsonString))
+}
+
+// Delete deletes PolicyList
+func (policylist *PolicyList) Delete() (*PolicyList, error) {
+	operation := "Delete"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" id=%s", policylist.ID)
+
+	return PolicyListRequest("DELETE", policylist.ID, "")
+}
+
+// AddEndpoint add an endpoint to a Policy List
+func (policylist *PolicyList) AddEndpoint(endpoint *HNSEndpoint) (*PolicyList, error) {
+	operation := "AddEndpoint"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" id=%s, endpointId:%s", policylist.ID, endpoint.Id)
+
+	_, err := policylist.Delete()
+	if err != nil {
+		return nil, err
+	}
+
+	// Add Endpoint to the Existing List
+	policylist.EndpointReferences = append(policylist.EndpointReferences, "/endpoints/"+endpoint.Id)
+
+	return policylist.Create()
+}
+
+// RemoveEndpoint removes an endpoint from the Policy List
+func (policylist *PolicyList) RemoveEndpoint(endpoint *HNSEndpoint) (*PolicyList, error) {
+	operation := "RemoveEndpoint"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" id=%s, endpointId:%s", policylist.ID, endpoint.Id)
+
+	_, err := policylist.Delete()
+	if err != nil {
+		return nil, err
+	}
+
+	elementToRemove := "/endpoints/" + endpoint.Id
+
+	var references []string
+
+	for _, endpointReference := range policylist.EndpointReferences {
+		if endpointReference == elementToRemove {
+			continue
+		}
+		references = append(references, endpointReference)
+	}
+	policylist.EndpointReferences = references
+	return policylist.Create()
+}
+
+// AddLoadBalancer policy list for the specified endpoints
+func AddLoadBalancer(endpoints []HNSEndpoint, isILB bool, sourceVIP, vip string, protocol uint16, internalPort uint16, externalPort uint16) (*PolicyList, error) {
+	operation := "AddLoadBalancer"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" endpointId=%v, isILB=%v, sourceVIP=%s, vip=%s, protocol=%v, internalPort=%v, externalPort=%v", endpoints, isILB, sourceVIP, vip, protocol, internalPort, externalPort)
+
+	policylist := &PolicyList{}
+
+	elbPolicy := &ELBPolicy{
+		SourceVIP: sourceVIP,
+		ILB:       isILB,
+	}
+
+	if len(vip) > 0 {
+		elbPolicy.VIPs = []string{vip}
+	}
+	elbPolicy.Type = ExternalLoadBalancer
+	elbPolicy.Protocol = protocol
+	elbPolicy.InternalPort = internalPort
+	elbPolicy.ExternalPort = externalPort
+
+	for _, endpoint := range endpoints {
+		policylist.EndpointReferences = append(policylist.EndpointReferences, "/endpoints/"+endpoint.Id)
+	}
+
+	jsonString, err := json.Marshal(elbPolicy)
+	if err != nil {
+		return nil, err
+	}
+	policylist.Policies = append(policylist.Policies, jsonString)
+	return policylist.Create()
+}
+
+// AddRoute adds route policy list for the specified endpoints
+func AddRoute(endpoints []HNSEndpoint, destinationPrefix string, nextHop string, encapEnabled bool) (*PolicyList, error) {
+	operation := "AddRoute"
+	title := "hcsshim::PolicyList::" + operation
+	logrus.Debugf(title+" destinationPrefix:%s", destinationPrefix)
+
+	policylist := &PolicyList{}
+
+	rPolicy := &RoutePolicy{
+		DestinationPrefix: destinationPrefix,
+		NextHop:           nextHop,
+		EncapEnabled:      encapEnabled,
+	}
+	rPolicy.Type = Route
+
+	for _, endpoint := range endpoints {
+		policylist.EndpointReferences = append(policylist.EndpointReferences, "/endpoints/"+endpoint.Id)
+	}
+
+	jsonString, err := json.Marshal(rPolicy)
+	if err != nil {
+		return nil, err
+	}
+
+	policylist.Policies = append(policylist.Policies, jsonString)
+	return policylist.Create()
+}
@@ -0,0 +1,49 @@
+package hns
+
+import (
+	"github.com/sirupsen/logrus"
+)
+
+type HNSSupportedFeatures struct {
+	Acl HNSAclFeatures `json:"ACL"`
+}
+
+type HNSAclFeatures struct {
+	AclAddressLists       bool `json:"AclAddressLists"`
+	AclNoHostRulePriority bool `json:"AclHostRulePriority"`
+	AclPortRanges         bool `json:"AclPortRanges"`
+	AclRuleId             bool `json:"AclRuleId"`
+}
+
+func GetHNSSupportedFeatures() HNSSupportedFeatures {
+	var hnsFeatures HNSSupportedFeatures
+
+	globals, err := GetHNSGlobals()
+	if err != nil {
+		// Expected on pre-1803 builds, all features will be false/unsupported
+		logrus.Debugf("Unable to obtain HNS globals: %s", err)
+		return hnsFeatures
+	}
+
+	hnsFeatures.Acl = HNSAclFeatures{
+		AclAddressLists:       isHNSFeatureSupported(globals.Version, HNSVersion1803),
+		AclNoHostRulePriority: isHNSFeatureSupported(globals.Version, HNSVersion1803),
+		AclPortRanges:         isHNSFeatureSupported(globals.Version, HNSVersion1803),
+		AclRuleId:             isHNSFeatureSupported(globals.Version, HNSVersion1803),
+	}
+
+	return hnsFeatures
+}
+
+func isHNSFeatureSupported(currentVersion HNSVersion, minVersionSupported HNSVersion) bool {
+	if currentVersion.Major < minVersionSupported.Major {
+		return false
+	}
+	if currentVersion.Major > minVersionSupported.Major {
+		return true
+	}
+	if currentVersion.Minor < minVersionSupported.Minor {
+		return false
+	}
+	return true
+}
@@ -0,0 +1,111 @@
+package hns
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+	"path"
+	"strings"
+)
+
+type namespaceRequest struct {
+	IsDefault bool `json:",omitempty"`
+}
+
+type namespaceEndpointRequest struct {
+	ID string `json:"Id"`
+}
+
+type NamespaceResource struct {
+	Type string
+	Data json.RawMessage
+}
+
+type namespaceResourceRequest struct {
+	Type string
+	Data interface{}
+}
+
+type Namespace struct {
+	ID            string
+	IsDefault     bool                `json:",omitempty"`
+	ResourceList  []NamespaceResource `json:",omitempty"`
+	CompartmentId uint32              `json:",omitempty"`
+}
+
+func issueNamespaceRequest(id *string, method, subpath string, request interface{}) (*Namespace, error) {
+	var err error
+	hnspath := "/namespaces/"
+	if id != nil {
+		hnspath = path.Join(hnspath, *id)
+	}
+	if subpath != "" {
+		hnspath = path.Join(hnspath, subpath)
+	}
+	var reqJSON []byte
+	if request != nil {
+		if reqJSON, err = json.Marshal(request); err != nil {
+			return nil, err
+		}
+	}
+	var ns Namespace
+	err = hnsCall(method, hnspath, string(reqJSON), &ns)
+	if err != nil {
+		if strings.Contains(err.Error(), "Element not found.") {
+			return nil, os.ErrNotExist
+		}
+		return nil, fmt.Errorf("%s %s: %s", method, hnspath, err)
+	}
+	return &ns, err
+}
+
+func CreateNamespace() (string, error) {
+	req := namespaceRequest{}
+	ns, err := issueNamespaceRequest(nil, "POST", "", &req)
+	if err != nil {
+		return "", err
+	}
+	return ns.ID, nil
+}
+
+func RemoveNamespace(id string) error {
+	_, err := issueNamespaceRequest(&id, "DELETE", "", nil)
+	return err
+}
+
+func GetNamespaceEndpoints(id string) ([]string, error) {
+	ns, err := issueNamespaceRequest(&id, "GET", "", nil)
+	if err != nil {
+		return nil, err
+	}
+	var endpoints []string
+	for _, rsrc := range ns.ResourceList {
+		if rsrc.Type == "Endpoint" {
+			var endpoint namespaceEndpointRequest
+			err = json.Unmarshal(rsrc.Data, &endpoint)
+			if err != nil {
+				return nil, fmt.Errorf("unmarshal endpoint: %s", err)
+			}
+			endpoints = append(endpoints, endpoint.ID)
+		}
+	}
+	return endpoints, nil
+}
+
+func AddNamespaceEndpoint(id string, endpointID string) error {
+	resource := namespaceResourceRequest{
+		Type: "Endpoint",
+		Data: namespaceEndpointRequest{endpointID},
+	}
+	_, err := issueNamespaceRequest(&id, "POST", "addresource", &resource)
+	return err
+}
+
+func RemoveNamespaceEndpoint(id string, endpointID string) error {
+	resource := namespaceResourceRequest{
+		Type: "Endpoint",
+		Data: namespaceEndpointRequest{endpointID},
+	}
+	_, err := issueNamespaceRequest(&id, "POST", "removeresource", &resource)
+	return err
+}
@@ -0,0 +1,76 @@
+// Code generated mksyscall_windows.exe DO NOT EDIT
+
+package hns
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return nil
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	// TODO: add more here, after collecting data on the common
+	// error values see on Windows. (perhaps when running
+	// all.bat?)
+	return e
+}
+
+var (
+	modvmcompute = windows.NewLazySystemDLL("vmcompute.dll")
+
+	procHNSCall = modvmcompute.NewProc("HNSCall")
+)
+
+func _hnsCall(method string, path string, object string, response **uint16) (hr error) {
+	var _p0 *uint16
+	_p0, hr = syscall.UTF16PtrFromString(method)
+	if hr != nil {
+		return
+	}
+	var _p1 *uint16
+	_p1, hr = syscall.UTF16PtrFromString(path)
+	if hr != nil {
+		return
+	}
+	var _p2 *uint16
+	_p2, hr = syscall.UTF16PtrFromString(object)
+	if hr != nil {
+		return
+	}
+	return __hnsCall(_p0, _p1, _p2, response)
+}
+
+func __hnsCall(method *uint16, path *uint16, object *uint16, response **uint16) (hr error) {
+	if hr = procHNSCall.Find(); hr != nil {
+		return
+	}
+	r0, _, _ := syscall.Syscall6(procHNSCall.Addr(), 4, uintptr(unsafe.Pointer(method)), uintptr(unsafe.Pointer(path)), uintptr(unsafe.Pointer(object)), uintptr(unsafe.Pointer(response)), 0, 0)
+	if int32(r0) < 0 {
+		if r0&0x1fff0000 == 0x00070000 {
+			r0 &= 0xffff
+		}
+		hr = syscall.Errno(r0)
+	}
+	return
+}
@@ -0,0 +1,23 @@
+package interop
+
+import (
+	"syscall"
+	"unsafe"
+)
+
+//go:generate go run ../../mksyscall_windows.go -output zsyscall_windows.go interop.go
+
+//sys coTaskMemFree(buffer unsafe.Pointer) = api_ms_win_core_com_l1_1_0.CoTaskMemFree
+
+func ConvertAndFreeCoTaskMemString(buffer *uint16) string {
+	str := syscall.UTF16ToString((*[1 << 29]uint16)(unsafe.Pointer(buffer))[:])
+	coTaskMemFree(unsafe.Pointer(buffer))
+	return str
+}
+
+func Win32FromHresult(hr uintptr) syscall.Errno {
+	if hr&0x1fff0000 == 0x00070000 {
+		return syscall.Errno(hr & 0xffff)
+	}
+	return syscall.Errno(hr)
+}
--- a/Show More
+++ b/Show More
				`@@ -0,0 +1 @@`
				`package snapshot_service // import "yunion.io/x/onecloud/pkg/hostman/container/snapshot_service"`