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

.github/workflows/ci.yml

@@ -593,6 +593,54 @@ jobs:
           # also need to be updated and the special case for windows, removed.
           FOCUS="[^TestContainerdRestart$]" make cri-integration
 
+      - name: Stage CRI integration containerd logs into workspace
+        if: always()
+        shell: pwsh
+        run: |
+          $dest = Join-Path "${{ github.workspace }}" "ci-debug-logs"
+          New-Item -ItemType Directory -Force -Path $dest | Out-Null
+
+          # cri-integration.sh writes containerd.log to C:/Windows/Temp/test-integration
+          # and on failure tries to `mv` it to $GITHUB_WORKSPACE/report (which often fails
+          # on Windows with "Device or resource busy" because containerd still holds the
+          # file open). Look in both places. actions/upload-artifact@v7 refuses to upload
+          # files outside the workspace root, so stage everything inside the workspace.
+          $sources = @(
+            'C:/Windows/Temp/test-integration',
+            (Join-Path "${{ github.workspace }}" 'report')
+          )
+          foreach ($src in $sources) {
+            if (Test-Path $src) {
+              Write-Host "Copying logs from $src"
+              # Use robocopy for resilience against handles still being open.
+              # Exit codes 0-7 are success for robocopy; treat anything else as warning only.
+              $leaf = Split-Path -Leaf $src
+              $target = Join-Path $dest $leaf
+              New-Item -ItemType Directory -Force -Path $target | Out-Null
+              robocopy $src $target /E /R:3 /W:1 /NFL /NDL /NP | Out-Null
+              if ($LASTEXITCODE -ge 8) {
+                Write-Warning "robocopy $src returned $LASTEXITCODE (best-effort, continuing)"
+              }
+              # robocopy sets nonzero "success" codes; reset for downstream steps.
+              $global:LASTEXITCODE = 0
+            } else {
+              Write-Host "Source not found (skipping): $src"
+            }
+          }
+
+          Write-Host '::group::Staged files'
+          Get-ChildItem -Recurse -File $dest | Select-Object FullName, Length
+          Write-Host '::endgroup::'
+
+      - name: Upload CRI integration containerd.log (DEBUG-HCS / DEBUG-SHIM traces)
+        if: always()
+        uses: actions/upload-artifact@v7
+        with:
+          name: cri-integration-containerd-log
+          path: ci-debug-logs/**/*
+          if-no-files-found: warn
+          retention-days: 7
+
   #      Enable these tests once the required JobContainer images are updated.
   #
   #      - name: Install containerd service

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

@@ -4,6 +4,7 @@ package main
 
 import (
 	"context"
+	"fmt"
 	"sync"
 	"time"
 
@@ -455,10 +456,19 @@ func (he *hcsExec) waitForExit() {
 		trace.StringAttribute("tid", he.tid),
 		trace.StringAttribute("eid", he.id))
 
+	log.G(ctx).WithFields(logrus.Fields{
+		"tid": he.tid,
+		"eid": he.id,
+	}).Info("DEBUG-SHIM: hcsExec.waitForExit START")
+
 	err = he.p.Process.Wait()
 	if err != nil {
 		log.G(ctx).WithError(err).Error("failed process Wait")
 	}
+	log.G(ctx).WithFields(logrus.Fields{
+		"tid": he.tid,
+		"eid": he.id,
+	}).Info("DEBUG-SHIM: hcsExec.waitForExit Process.Wait() RETURNED")
 
 	// Issue the process cancellation to unblock the container wait as early as
 	// possible.
@@ -470,6 +480,12 @@ func (he *hcsExec) waitForExit() {
 	} else {
 		log.G(ctx).WithField("exitCode", code).Debug("exited")
 	}
+	log.G(ctx).WithFields(logrus.Fields{
+		"tid":      he.tid,
+		"eid":      he.id,
+		"exitCode": code,
+		"err":      fmt.Sprintf("%v", err),
+	}).Info("DEBUG-SHIM: hcsExec.waitForExit got exit code, transitioning to Exited")
 
 	he.sl.Lock()
 	he.state = shimExecStateExited
@@ -517,22 +533,40 @@ func (he *hcsExec) waitForContainerExit() {
 		trace.StringAttribute("tid", he.tid),
 		trace.StringAttribute("eid", he.id))
 
+	log.G(ctx).WithFields(logrus.Fields{
+		"tid": he.tid,
+		"eid": he.id,
+	}).Info("DEBUG-SHIM: hcsExec.waitForContainerExit START")
+
 	// wait for container or process to exit and ckean up resrources
 	select {
 	case <-he.c.WaitChannel():
 		// Container exited first. We need to force the process into the exited
 		// state and cleanup any resources
 		he.sl.Lock()
+		log.G(ctx).WithFields(logrus.Fields{
+			"tid":   he.tid,
+			"eid":   he.id,
+			"state": he.state,
+		}).Info("DEBUG-SHIM: hcsExec.waitForContainerExit CONTAINER WaitChannel fired (container exited first)")
 		switch he.state {
 		case shimExecStateCreated:
 			he.exitFromCreatedL(ctx, 1)
 		case shimExecStateRunning:
 			// Kill the process to unblock `he.waitForExit`.
+			log.G(ctx).WithFields(logrus.Fields{
+				"tid": he.tid,
+				"eid": he.id,
+			}).Info("DEBUG-SHIM: hcsExec.waitForContainerExit calling Process.Kill to unblock waitForExit")
 			_, _ = he.p.Process.Kill(ctx)
 		}
 		he.sl.Unlock()
 	case <-he.processDone:
 		// Process exited first. This is the normal case do nothing because
 		// `he.waitForExit` will release any waiters.
+		log.G(ctx).WithFields(logrus.Fields{
+			"tid": he.tid,
+			"eid": he.id,
+		}).Info("DEBUG-SHIM: hcsExec.waitForContainerExit processDone fired (process exited first; normal path)")
 	}
 }