thaloco/banger
1
Fork 0
Code Pull requests Releases Packages Activity Actions

Rollback partial dm snapshot startup

Browse source 
Prevent partial VM startup failures from leaking loop devices and dm state on the host.

Move root snapshot setup into a rollback-safe helper that records loop and mapper handles incrementally, tears them down in reverse order on failure, and reuses the same dm/loop cleanup path during normal runtime teardown. Also switch the daemon runner field to a small command-runner interface so the snapshot path can be tested with injected failures.

Add failure-injection coverage for losetup, blockdev, dmsetup, partial teardown, and joined rollback errors. Validated with go test ./... and make build.
This commit is contained in:
Thales Maciel 2026-03-16 14:06:17 -03:00
parent 171009b30b
commit 375900cf65
No known key found for this signature in database
GPG key ID: 33112E6833C34679
5 changed files with 401 additions and 47 deletions
Download patch file Download diff file Expand all files Collapse all files

58
internal/daemon/vm.go
View file

@ -153,14 +153,14 @@ func (d *Daemon) startVMLocked(ctx context.Context, vm model.VMRecord, image mod
return model.VMRecord{}, err
}
baseLoop, cowLoop, dmDev, err := d.createDMSnapshot(ctx, image.RootfsPath, vm.Runtime.SystemOverlay, dmName)
handles, err := d.createDMSnapshot(ctx, image.RootfsPath, vm.Runtime.SystemOverlay, dmName)
if err != nil {
return model.VMRecord{}, err
}
vm.Runtime.BaseLoop = baseLoop
vm.Runtime.COWLoop = cowLoop
vm.Runtime.DMName = dmName
vm.Runtime.DMDev = dmDev
vm.Runtime.BaseLoop = handles.BaseLoop
vm.Runtime.COWLoop = handles.COWLoop
vm.Runtime.DMName = handles.DMName
vm.Runtime.DMDev = handles.DMDev
vm.Runtime.APISockPath = apiSock
vm.Runtime.TapDevice = tap
vm.Runtime.State = model.VMStateRunning
@ -171,7 +171,9 @@ func (d *Daemon) startVMLocked(ctx context.Context, vm model.VMRecord, image mod
vm.State = model.VMStateError
vm.Runtime.State = model.VMStateError
vm.Runtime.LastError = err.Error()
_ = d.cleanupRuntime(context.Background(), vm, true)
if cleanupErr := d.cleanupRuntime(context.Background(), vm, true); cleanupErr != nil {
err = errors.Join(err, cleanupErr)
}
clearRuntimeHandles(&vm)
_ = d.store.UpsertVM(context.Background(), vm)
return model.VMRecord{}, err
@ -273,6 +275,7 @@ func (d *Daemon) StopVM(ctx context.Context, idOrName string) (model.VMRecord, e
return vm, nil
}
func (d *Daemon) RestartVM(ctx context.Context, idOrName string) (model.VMRecord, error) {
vm, err := d.StopVM(ctx, idOrName)
if err != nil {
@ -506,28 +509,6 @@ func (d *Daemon) ensureWorkDisk(ctx context.Context, vm *model.VMRecord) error {
return nil
}
func (d *Daemon) createDMSnapshot(ctx context.Context, rootfsPath, cowPath, dmName string) (baseLoop, cowLoop, dmDev string, err error) {
baseBytes, err := d.runner.RunSudo(ctx, "losetup", "-f", "--show", "--read-only", rootfsPath)
if err != nil {
return "", "", "", err
}
baseLoop = strings.TrimSpace(string(baseBytes))
cowBytes, err := d.runner.RunSudo(ctx, "losetup", "-f", "--show", cowPath)
if err != nil {
return "", "", "", err
}
cowLoop = strings.TrimSpace(string(cowBytes))
sectorsBytes, err := d.runner.RunSudo(ctx, "blockdev", "--getsz", baseLoop)
if err != nil {
return "", "", "", err
}
sectors := strings.TrimSpace(string(sectorsBytes))
if _, err := d.runner.RunSudo(ctx, "dmsetup", "create", dmName, "--table", fmt.Sprintf("0 %s snapshot %s %s P 8", sectors, baseLoop, cowLoop)); err != nil {
return "", "", "", err
}
return baseLoop, cowLoop, "/dev/mapper/" + dmName, nil
}
func (d *Daemon) ensureBridge(ctx context.Context) error {
if _, err := d.runner.Run(ctx, "ip", "link", "show", d.config.BridgeName); err == nil {
_, err = d.runner.RunSudo(ctx, "ip", "link", "set", d.config.BridgeName, "up")
@ -638,25 +619,20 @@ func (d *Daemon) cleanupRuntime(ctx context.Context, vm model.VMRecord, preserve
if vm.Runtime.APISockPath != "" {
_ = os.Remove(vm.Runtime.APISockPath)
}
if vm.Runtime.DMName != "" {
_, _ = d.runner.RunSudo(ctx, "dmsetup", "remove", vm.Runtime.DMName)
} else if vm.Runtime.DMDev != "" {
_, _ = d.runner.RunSudo(ctx, "dmsetup", "remove", vm.Runtime.DMDev)
}
if vm.Runtime.COWLoop != "" {
_, _ = d.runner.RunSudo(ctx, "losetup", "-d", vm.Runtime.COWLoop)
}
if vm.Runtime.BaseLoop != "" {
_, _ = d.runner.RunSudo(ctx, "losetup", "-d", vm.Runtime.BaseLoop)
}
snapshotErr := d.cleanupDMSnapshot(ctx, dmSnapshotHandles{
BaseLoop: vm.Runtime.BaseLoop,
COWLoop: vm.Runtime.COWLoop,
DMName: vm.Runtime.DMName,
DMDev: vm.Runtime.DMDev,
})
if vm.Spec.NATEnabled {
_ = d.ensureNAT(ctx, vm, false)
}
_ = d.removeDNS(ctx, vm.Runtime.DNSName)
if !preserveDisks && vm.Runtime.VMDir != "" {
return os.RemoveAll(vm.Runtime.VMDir)
return errors.Join(snapshotErr, os.RemoveAll(vm.Runtime.VMDir))
}
return nil
return snapshotErr
}
func clearRuntimeHandles(vm *model.VMRecord) {