[internal/hcs] Migrate package from HCS V1 to V2

cmd/containerd-shim-lcow-v2/manager.go

@@ -14,7 +14,7 @@ import (
 	"time"
 
 	runhcsopts "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/options"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/memory"
 	"github.com/Microsoft/hcsshim/internal/oc"
 	"github.com/Microsoft/hcsshim/internal/shim"

cmd/containerd-shim-lcow-v2/service/service_task_internal.go

@@ -13,8 +13,8 @@ import (
 	"github.com/Microsoft/hcsshim/internal/controller/pod"
 	"github.com/Microsoft/hcsshim/internal/controller/process"
 	"github.com/Microsoft/hcsshim/internal/controller/vm"
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/log"
 	"github.com/Microsoft/hcsshim/internal/logfields"
 	"github.com/Microsoft/hcsshim/internal/memory"

cmd/containerd-shim-runhcs-v1/delete.go

@@ -16,7 +16,7 @@ import (
 	"google.golang.org/protobuf/proto"
 	"google.golang.org/protobuf/types/known/timestamppb"
 
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/memory"
 	"github.com/Microsoft/hcsshim/internal/oc"
 	"github.com/Microsoft/hcsshim/internal/winapi"

cmd/containerd-shim-runhcs-v1/exec_hcs.go

@@ -20,7 +20,7 @@ import (
 
 	"github.com/Microsoft/hcsshim/internal/cmd"
 	"github.com/Microsoft/hcsshim/internal/cow"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/log"
 	"github.com/Microsoft/hcsshim/internal/oc"
 	"github.com/Microsoft/hcsshim/internal/protocol/guestresource"

cmd/containerd-shim-runhcs-v1/task.go

@@ -9,7 +9,7 @@ import (
 
 	"github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/options"
 	"github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/stats"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/shimdiag"
 	"github.com/Microsoft/hcsshim/pkg/ctrdtaskapi"
 	task "github.com/containerd/containerd/api/runtime/task/v2"

cmd/containerd-shim-runhcs-v1/task_hcs.go

@@ -29,10 +29,10 @@ import (
 	"github.com/Microsoft/hcsshim/internal/cmd"
 	"github.com/Microsoft/hcsshim/internal/cow"
 	"github.com/Microsoft/hcsshim/internal/guestpath"
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	"github.com/Microsoft/hcsshim/internal/hcs/resourcepaths"
 	"github.com/Microsoft/hcsshim/internal/hcs/schema1"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/hcsoci"
 	"github.com/Microsoft/hcsshim/internal/jobcontainers"
 	"github.com/Microsoft/hcsshim/internal/layers"

internal/computecore/computecore.go

@@ -8,7 +8,6 @@ import (
 	"time"
 	"unsafe"
 
-	"github.com/sirupsen/logrus"
 	"go.opencensus.io/trace"
 
 	"github.com/Microsoft/hcsshim/internal/interop"
@@ -94,39 +93,40 @@ import (
 // errVmcomputeOperationPending is an error encountered when the operation is being completed asynchronously
 const errVmcomputeOperationPending = syscall.Errno(0xC0370103)
 
+// execute runs f synchronously to completion. ctx and timeout are
+// watchdogs only: each emits a one-shot warning when exceeded and the
+// wait continues until f returns.
+//
+// Callers of the wrapping HCS API typically release the handles they
+// passed in via defer once the call returns. Abandoning f mid-flight
+// would let those defers tear down handles the syscall is still using,
+// which has been observed to crash the shim with EXCEPTION_ACCESS_VIOLATION
+// inside computecore.dll. Long-running operations should be bounded via
+// HcsCancelOperation rather than by returning early.
 func execute(ctx gcontext.Context, timeout time.Duration, f func() error) error {
-	now := time.Now()
-	if timeout > 0 {
-		var cancel gcontext.CancelFunc
-		ctx, cancel = gcontext.WithTimeout(ctx, timeout)
-		defer cancel()
-	}
+	done := make(chan error, 1)
+	go func() { done <- f() }()
 
-	deadline, ok := ctx.Deadline()
-	trueTimeout := timeout
-	if ok {
-		trueTimeout = deadline.Sub(now)
-		log.G(ctx).WithFields(logrus.Fields{
-			logfields.Timeout: trueTimeout,
-			"desiredTimeout":  timeout,
-		}).Trace("Executing syscall with deadline")
+	var watcher <-chan time.Time
+	if timeout > 0 {
+		t := time.NewTimer(timeout)
+		defer t.Stop()
+		watcher = t.C
 	}
 
-	done := make(chan error, 1)
-	go func() {
-		done <- f()
-	}()
-	select {
-	case <-ctx.Done():
-		if ctx.Err() == gcontext.DeadlineExceeded {
-			log.G(ctx).WithField(logfields.Timeout, trueTimeout).
-				Warning("Syscall did not complete within operation timeout. This may indicate a platform issue. " +
-					"If it appears to be making no forward progress, obtain the stacks and see if there is a syscall " +
-					"stuck in the platform API for a significant length of time.")
+	for {
+		select {
+		case err := <-done:
+			return err
+		case <-watcher:
+			log.G(ctx).WithField(logfields.Timeout, timeout).
+				Warning("HCS syscall exceeded timeout; still waiting to avoid use-after-free in computecore.dll")
+			watcher = nil
+		case <-ctx.Done():
+			log.G(ctx).WithError(ctx.Err()).
+				Warning("HCS syscall context canceled; still waiting to avoid use-after-free in computecore.dll")
+			ctx = gcontext.WithoutCancel(ctx)
 		}
-		return ctx.Err()
-	case err := <-done:
-		return err
 	}
 }
 

internal/controller/device/plan9/controller.go

@@ -11,7 +11,7 @@ import (
 	"github.com/Microsoft/go-winio/pkg/guid"
 	"github.com/Microsoft/hcsshim/internal/controller/device/plan9/mount"
 	"github.com/Microsoft/hcsshim/internal/controller/device/plan9/share"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/log"
 	"github.com/Microsoft/hcsshim/internal/logfields"
 	"github.com/sirupsen/logrus"

internal/controller/linuxcontainer/container_test.go

@@ -14,8 +14,8 @@ import (
 	"github.com/Microsoft/hcsshim/internal/controller/linuxcontainer/mocks"
 	"github.com/Microsoft/hcsshim/internal/controller/process"
 	"github.com/Microsoft/hcsshim/internal/gcs"
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/protocol/guestresource"
 	"github.com/Microsoft/hcsshim/internal/signals"
 

internal/controller/network/network_lcow_test.go

@@ -14,8 +14,8 @@ import (
 	"github.com/Microsoft/hcsshim/internal/controller/network/mocks"
 	"github.com/Microsoft/hcsshim/internal/gcs"
 	"github.com/Microsoft/hcsshim/internal/guest/prot"
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/protocol/guestresource"
 )
 

internal/controller/network/network_wcow_test.go

@@ -13,9 +13,9 @@ import (
 	"github.com/Microsoft/hcsshim/hcn"
 	"github.com/Microsoft/hcsshim/internal/controller/network/mocks"
 	"github.com/Microsoft/hcsshim/internal/gcs"
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	"github.com/Microsoft/hcsshim/internal/hcs/schema1"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/protocol/guestrequest"
 )
 

internal/controller/process/process.go

@@ -10,7 +10,7 @@ import (
 
 	"github.com/Microsoft/hcsshim/internal/cmd"
 	"github.com/Microsoft/hcsshim/internal/cow"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/log"
 	"github.com/Microsoft/hcsshim/internal/logfields"
 	eventstypes "github.com/containerd/containerd/api/events"

internal/controller/process/process_test.go

@@ -14,7 +14,7 @@ import (
 	"go.uber.org/mock/gomock"
 
 	"github.com/Microsoft/hcsshim/internal/controller/process/mocks"
-	"github.com/Microsoft/hcsshim/internal/hcs"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 )
 
 const (

internal/cpugroup/cpugroup.go

@@ -8,8 +8,8 @@ import (
 	"fmt"
 	"strings"
 
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/pkg/errors"
 )
 

internal/credentials/credentials.go

@@ -9,8 +9,8 @@ import (
 	"errors"
 	"fmt"
 
-	"github.com/Microsoft/hcsshim/internal/hcs"
 	hcsschema "github.com/Microsoft/hcsshim/internal/hcs/schema2"
+	hcs "github.com/Microsoft/hcsshim/internal/hcs/v2"
 	"github.com/Microsoft/hcsshim/internal/log"
 )
 

internal/gcs/process.go

@@ -187,7 +187,10 @@ func (p *Process) ExitCode() (_ int, err error) {
 		return -1, errors.New("process not exited")
 	}
 	if err := p.waitCall.Err(); err != nil {
-		return -1, err
+		var rerr *rpcError
+		if !errors.As(err, &rerr) || uint32(rerr.result) != hrNotFound {
+			return -1, err
+		}
 	}
 	return int(p.waitResp.ExitCode), nil
 }
@@ -274,7 +277,8 @@ func (p *Process) Stdio() (stdin io.Writer, stdout, stderr io.Reader) {
 // Wait waits for the process (or guest connection) to terminate.
 func (p *Process) Wait() error {
 	p.waitCall.Wait()
-	return p.waitCall.Err()
+	_, err := p.ExitCode()
+	return err
 }
 
 func (p *Process) waitBackground() {