From 2d7b22b561f9e2dd1b86fb2c5b0d904823a189d2 Mon Sep 17 00:00:00 2001 From: Jiri Denemark Date: Tue, 6 Sep 2022 18:21:31 +0200 Subject: [PATCH] qemu: Make qemuMigrationSrcCancel optionally synchronous We have always considered "migrate_cancel" QMP command to return after successfully cancelling the migration. But this is no longer true (to be honest I'm not sure it ever was) as it just changes the migration state to "cancelling". In most cases the migration is canceled pretty quickly and we don't really notice anything, but sometimes it takes so long we even get to clearing migration capabilities before the migration is actually canceled, which fails as capabilities can only be changed when no migration is running. So to avoid this issue, we can wait for the migration to be really canceled after sending migrate_cancel. The only place where we don't need synchronous behavior is when we're cancelling migration on user's request while it is actively watched by another thread. https://bugzilla.redhat.com/show_bug.cgi?id=2114866 Signed-off-by: Jiri Denemark Reviewed-by: Peter Krempa --- src/qemu/qemu_driver.c | 2 +- src/qemu/qemu_migration.c | 53 ++++++++++++++++++++++++++++++++++----- src/qemu/qemu_migration.h | 3 ++- src/qemu/qemu_process.c | 2 +- 4 files changed, 51 insertions(+), 9 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index bb6607054e..ea7d74806c 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -12813,7 +12813,7 @@ qemuDomainAbortJobMigration(virDomainObj *vm) VIR_DEBUG("Cancelling migration job at client request"); qemuDomainObjAbortAsyncJob(vm); - return qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE); + return qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE, false); } diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 5c08e81c1b..8273fff67f 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -4611,8 +4611,36 @@ qemuMigrationSrcStart(virDomainObj *vm, } +static bool +qemuMigrationSrcIsCanceled(virDomainObj *vm) +{ + qemuDomainObjPrivate *priv = vm->privateData; + virDomainJobData *jobData = priv->job.current; + + qemuMigrationUpdateJobType(jobData); + switch (jobData->status) { + case VIR_DOMAIN_JOB_STATUS_FAILED: + case VIR_DOMAIN_JOB_STATUS_CANCELED: + case VIR_DOMAIN_JOB_STATUS_COMPLETED: + case VIR_DOMAIN_JOB_STATUS_NONE: + return true; + + case VIR_DOMAIN_JOB_STATUS_MIGRATING: + case VIR_DOMAIN_JOB_STATUS_POSTCOPY: + case VIR_DOMAIN_JOB_STATUS_PAUSED: + case VIR_DOMAIN_JOB_STATUS_HYPERVISOR_COMPLETED: + case VIR_DOMAIN_JOB_STATUS_POSTCOPY_PAUSED: + case VIR_DOMAIN_JOB_STATUS_ACTIVE: + break; + } + + return false; +} + + /** - * Requests outgoing migration to be canceled. + * Requests outgoing migration to be canceled and optionally waits for the + * cancellation to complete. * * The thread (the caller itself in most cases) which is watching the migration * will do all the cleanup once migration is canceled. If no thread is watching @@ -4620,7 +4648,8 @@ qemuMigrationSrcStart(virDomainObj *vm, */ int qemuMigrationSrcCancel(virDomainObj *vm, - virDomainAsyncJob asyncJob) + virDomainAsyncJob asyncJob, + bool wait) { qemuDomainObjPrivate *priv = vm->privateData; int rc; @@ -4633,7 +4662,19 @@ qemuMigrationSrcCancel(virDomainObj *vm, rc = qemuMonitorMigrateCancel(priv->mon); qemuDomainObjExitMonitor(vm); - return rc; + if (rc < 0) + return -1; + + if (virDomainObjIsActive(vm) && wait) { + VIR_DEBUG("Waiting for migration to be canceled"); + + while (!qemuMigrationSrcIsCanceled(vm)) { + if (qemuDomainObjWait(vm) < 0) + return -1; + } + } + + return 0; } @@ -4979,7 +5020,7 @@ qemuMigrationSrcRun(virQEMUDriver *driver, if (cancel && priv->job.current->status != VIR_DOMAIN_JOB_STATUS_HYPERVISOR_COMPLETED) - qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_MIGRATION_OUT); + qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_MIGRATION_OUT, true); /* cancel any outstanding NBD jobs */ if (mig && mig->nbd) @@ -6924,7 +6965,7 @@ qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm, virErrorPreserveLast(&orig_err); virCommandAbort(compressor); if (virDomainObjIsActive(vm)) - qemuMigrationSrcCancel(vm, asyncJob); + qemuMigrationSrcCancel(vm, asyncJob, true); } goto cleanup; } @@ -6975,7 +7016,7 @@ qemuMigrationSrcCancelUnattended(virDomainObj *vm) VIR_DEBUG("Canceling unfinished outgoing migration of domain %s", vm->def->name); - qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE); + qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE, true); for (i = 0; i < vm->def->ndisks; i++) { virDomainDiskDef *disk = vm->def->disks[i]; diff --git a/src/qemu/qemu_migration.h b/src/qemu/qemu_migration.h index 31a5547399..fbea45ad4e 100644 --- a/src/qemu/qemu_migration.h +++ b/src/qemu/qemu_migration.h @@ -245,7 +245,8 @@ qemuMigrationSrcCancelUnattended(virDomainObj *vm); int qemuMigrationSrcCancel(virDomainObj *vm, - virDomainAsyncJob asyncJob); + virDomainAsyncJob asyncJob, + bool wait); int qemuMigrationAnyFetchStats(virDomainObj *vm, diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index 4465fa89e9..08eefd0fba 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -3696,7 +3696,7 @@ qemuProcessRecoverJob(virQEMUDriver *driver, case VIR_ASYNC_JOB_SAVE: case VIR_ASYNC_JOB_DUMP: case VIR_ASYNC_JOB_SNAPSHOT: - qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE); + qemuMigrationSrcCancel(vm, VIR_ASYNC_JOB_NONE, true); /* resume the domain but only if it was paused as a result of * running a migration-to-file operation. Although we are * recovering an async job, this function is run at startup