[libvirt] [PATCH 0/3] qemu: Fix two block-job related regressions (blockdev-add saga)
Peter Krempa (3): qemu: domain: Fix logic bug in qemuDomainStorageSourceAccessAllow qemu: domain: Allow overriding images to read-write in qemuDomainStorageSourceAccessAllow qemu: Use proper block job name when reconnecting to VM src/qemu/qemu_domain.c | 21 ++++++++++++++------- src/qemu/qemu_migration.c | 2 +- src/qemu/qemu_monitor.c | 7 ++++--- src/qemu/qemu_monitor.h | 3 ++- src/qemu/qemu_monitor_json.c | 12 ++++++++---- src/qemu/qemu_monitor_json.h | 3 ++- src/qemu/qemu_process.c | 2 +- 7 files changed, 32 insertions(+), 18 deletions(-) -- 2.21.0
In commit 76b9aba2ba6 I tried to refactor qemuDomainStorageSourceAccessAllow but used wrong operators for adding bitwise flags. This way the flags would result in 0 if any of them would be applied. Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_domain.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index 4d3a8868b2..e521bd3982 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -9491,10 +9491,10 @@ qemuDomainStorageSourceAccessAllow(virQEMUDriverPtr driver, qemuDomainStorageSourceAccessFlags flags = QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE; if (readonly) - flags &= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY; + flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY; if (!newSource) - flags &= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS; + flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS; return qemuDomainStorageSourceAccessModify(driver, vm, elem, flags); } -- 2.21.0
On Tue, Jun 11, 2019 at 04:56:23PM +0200, Peter Krempa wrote:
In commit 76b9aba2ba6 I tried to refactor qemuDomainStorageSourceAccessAllow but used wrong operators for adding bitwise flags.
How did that get past review? O:-)
This way the flags would result in 0 if any of them would be applied.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_domain.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Reviewed-by: Ján Tomko <jtomko@redhat.com> Jano
In commit 76b9aba2ba6 I refactored how the function treats the readonly flag which introduced a bug when we'd not allow to force read-write state for an image. This created problems with blockjobs where we need to temporarily override images to read-write. Rename QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY to QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY and also introduce a complement QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE which will allow to force wire access. https://bugzilla.redhat.com/show_bug.cgi?id=1717768 Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_domain.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index e521bd3982..595150708b 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -9305,12 +9305,14 @@ typedef enum { QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_REVOKE = 1 << 0, /* operate on full backing chain rather than single image */ QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_CHAIN = 1 << 1, - /* force permissions to read-only when allowing */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY = 1 << 2, + /* force permissions to read-only/read-write when allowing */ + /* currently does not properly wrok with QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_CHAIN */ + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY = 1 << 2, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE = 1 << 3, /* don't revoke permissions when modification has failed */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE = 1 << 3, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE = 1 << 4, /* VM already has access to the source and we are just modifying it */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS = 1 << 4, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS = 1 << 5, } qemuDomainStorageSourceAccessFlags; @@ -9344,9 +9346,12 @@ qemuDomainStorageSourceAccessModify(virQEMUDriverPtr driver, bool revoke_namespace = false; bool revoke_lockspace = false; - if (flags & QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY) + if (flags & QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY) src->readonly = true; + if (flags & QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE) + src->readonly = false; + /* just tear down the disk access */ if (flags & QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_REVOKE) { virErrorPreserveLast(&orig_err); @@ -9491,7 +9496,9 @@ qemuDomainStorageSourceAccessAllow(virQEMUDriverPtr driver, qemuDomainStorageSourceAccessFlags flags = QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE; if (readonly) - flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY; + flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY; + else + flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE; if (!newSource) flags |= QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS; -- 2.21.0
On Tue, Jun 11, 2019 at 04:56:24PM +0200, Peter Krempa wrote:
In commit 76b9aba2ba6 I refactored how the function treats the readonly flag which introduced a bug when we'd not allow to force read-write state for an image.
This created problems with blockjobs where we need to temporarily override images to read-write.
Rename QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY to QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY and also introduce a complement QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE which
Rather than OVERRIDE_ I'd use FORCE_. overriding read-only access might imply that you want to force it to read-write.
will allow to force wire access.
s/wire/write/
https://bugzilla.redhat.com/show_bug.cgi?id=1717768
Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_domain.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index e521bd3982..595150708b 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -9305,12 +9305,14 @@ typedef enum { QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_REVOKE = 1 << 0, /* operate on full backing chain rather than single image */ QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_CHAIN = 1 << 1, - /* force permissions to read-only when allowing */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_READ_ONLY = 1 << 2, + /* force permissions to read-only/read-write when allowing */ + /* currently does not properly wrok with QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_CHAIN */
s/wrok/work/
+ QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_ONLY = 1 << 2, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_OVERRIDE_READ_WRITE = 1 << 3, /* don't revoke permissions when modification has failed */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE = 1 << 3, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_SKIP_REVOKE = 1 << 4, /* VM already has access to the source and we are just modifying it */ - QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS = 1 << 4, + QEMU_DOMAIN_STORAGE_SOURCE_ACCESS_MODIFY_ACCESS = 1 << 5, } qemuDomainStorageSourceAccessFlags;
@@ -9344,9 +9346,12 @@ qemuDomainStorageSourceAccessModify(virQEMUDriverPtr driver,
Reviewed-by: Ján Tomko <jtomko@redhat.com> Jano
The hash table returned by qemuMonitorGetAllBlockJobInfo is organized by the frontend name (which skipps the 'drive-' prefix). While our code properly matches the jobs to the disk, qemu needs the full job name including the 'drive-' prefix to be able to identify jobs. Fix this by adding an argument to qemuMonitorGetAllBlockJobInfo which does not modify the job name before filling the hash. This fixes a regression where users would not be able to cancel/pivot block jobs after restarting libvirtd while a blockjob is running. Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_migration.c | 2 +- src/qemu/qemu_monitor.c | 7 ++++--- src/qemu/qemu_monitor.h | 3 ++- src/qemu/qemu_monitor_json.c | 12 ++++++++---- src/qemu/qemu_monitor_json.h | 3 ++- src/qemu/qemu_process.c | 2 +- 6 files changed, 18 insertions(+), 11 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 32b3040473..1ea817004d 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -5545,7 +5545,7 @@ qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver, if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0) return -1; - blockinfo = qemuMonitorGetAllBlockJobInfo(priv->mon); + blockinfo = qemuMonitorGetAllBlockJobInfo(priv->mon, false); if (qemuDomainObjExitMonitor(driver, vm) < 0 || !blockinfo) return -1; diff --git a/src/qemu/qemu_monitor.c b/src/qemu/qemu_monitor.c index 6b731cd91a..9826426b29 100644 --- a/src/qemu/qemu_monitor.c +++ b/src/qemu/qemu_monitor.c @@ -3442,10 +3442,11 @@ qemuMonitorBlockJobSetSpeed(qemuMonitorPtr mon, virHashTablePtr -qemuMonitorGetAllBlockJobInfo(qemuMonitorPtr mon) +qemuMonitorGetAllBlockJobInfo(qemuMonitorPtr mon, + bool rawjobname) { QEMU_CHECK_MONITOR_NULL(mon); - return qemuMonitorJSONGetAllBlockJobInfo(mon); + return qemuMonitorJSONGetAllBlockJobInfo(mon, rawjobname); } @@ -3465,7 +3466,7 @@ qemuMonitorGetBlockJobInfo(qemuMonitorPtr mon, VIR_DEBUG("alias=%s, info=%p", alias, info); - if (!(all = qemuMonitorGetAllBlockJobInfo(mon))) + if (!(all = qemuMonitorGetAllBlockJobInfo(mon, false))) return -1; if ((data = virHashLookup(all, alias))) { diff --git a/src/qemu/qemu_monitor.h b/src/qemu/qemu_monitor.h index dee594fa66..cf4d567667 100644 --- a/src/qemu/qemu_monitor.h +++ b/src/qemu/qemu_monitor.h @@ -956,7 +956,8 @@ struct _qemuMonitorBlockJobInfo { int ready; /* -1 if unknown, 0 if not ready, 1 if ready */ }; -virHashTablePtr qemuMonitorGetAllBlockJobInfo(qemuMonitorPtr mon); +virHashTablePtr qemuMonitorGetAllBlockJobInfo(qemuMonitorPtr mon, + bool rawjobname); int qemuMonitorGetBlockJobInfo(qemuMonitorPtr mon, const char *device, qemuMonitorBlockJobInfoPtr info) diff --git a/src/qemu/qemu_monitor_json.c b/src/qemu/qemu_monitor_json.c index 53a7de8b77..2873399dc3 100644 --- a/src/qemu/qemu_monitor_json.c +++ b/src/qemu/qemu_monitor_json.c @@ -4662,7 +4662,8 @@ int qemuMonitorJSONScreendump(qemuMonitorPtr mon, static int qemuMonitorJSONParseBlockJobInfo(virHashTablePtr blockJobs, - virJSONValuePtr entry) + virJSONValuePtr entry, + bool rawjobname) { qemuMonitorBlockJobInfoPtr info = NULL; const char *device; @@ -4674,7 +4675,9 @@ qemuMonitorJSONParseBlockJobInfo(virHashTablePtr blockJobs, _("entry was missing 'device'")); return -1; } - device = qemuAliasDiskDriveSkipPrefix(device); + + if (!rawjobname) + device = qemuAliasDiskDriveSkipPrefix(device); if (VIR_ALLOC(info) < 0 || virHashAddEntry(blockJobs, device, info) < 0) { @@ -4724,7 +4727,8 @@ qemuMonitorJSONParseBlockJobInfo(virHashTablePtr blockJobs, } virHashTablePtr -qemuMonitorJSONGetAllBlockJobInfo(qemuMonitorPtr mon) +qemuMonitorJSONGetAllBlockJobInfo(qemuMonitorPtr mon, + bool rawjobname) { virJSONValuePtr cmd = NULL; virJSONValuePtr reply = NULL; @@ -4756,7 +4760,7 @@ qemuMonitorJSONGetAllBlockJobInfo(qemuMonitorPtr mon) _("missing array element")); goto error; } - if (qemuMonitorJSONParseBlockJobInfo(blockJobs, entry) < 0) + if (qemuMonitorJSONParseBlockJobInfo(blockJobs, entry, rawjobname) < 0) goto error; } diff --git a/src/qemu/qemu_monitor_json.h b/src/qemu/qemu_monitor_json.h index acef1a0a79..c8fde1d1b6 100644 --- a/src/qemu/qemu_monitor_json.h +++ b/src/qemu/qemu_monitor_json.h @@ -317,7 +317,8 @@ int qemuMonitorJSONBlockJobSetSpeed(qemuMonitorPtr mon, unsigned long long speed) ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2); -virHashTablePtr qemuMonitorJSONGetAllBlockJobInfo(qemuMonitorPtr mon) +virHashTablePtr qemuMonitorJSONGetAllBlockJobInfo(qemuMonitorPtr mon, + bool rawjobname) ATTRIBUTE_NONNULL(1); int qemuMonitorJSONSetLink(qemuMonitorPtr mon, diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index 50a76aa0ed..f114611e02 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -8011,7 +8011,7 @@ qemuProcessRefreshLegacyBlockjobs(virQEMUDriverPtr driver, int ret = -1; qemuDomainObjEnterMonitor(driver, vm); - blockJobs = qemuMonitorGetAllBlockJobInfo(qemuDomainGetMonitor(vm)); + blockJobs = qemuMonitorGetAllBlockJobInfo(qemuDomainGetMonitor(vm), true); if (qemuDomainObjExitMonitor(driver, vm) < 0 || !blockJobs) goto cleanup; -- 2.21.0
On Tue, Jun 11, 2019 at 04:56:25PM +0200, Peter Krempa wrote:
The hash table returned by qemuMonitorGetAllBlockJobInfo is organized by the frontend name (which skipps the 'drive-' prefix). While our code properly matches the jobs to the disk, qemu needs the full job name including the 'drive-' prefix to be able to identify jobs.
Fix this by adding an argument to qemuMonitorGetAllBlockJobInfo which does not modify the job name before filling the hash.
This fixes a regression where users would not be able to cancel/pivot block jobs after restarting libvirtd while a blockjob is running.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> --- src/qemu/qemu_migration.c | 2 +- src/qemu/qemu_monitor.c | 7 ++++--- src/qemu/qemu_monitor.h | 3 ++- src/qemu/qemu_monitor_json.c | 12 ++++++++---- src/qemu/qemu_monitor_json.h | 3 ++- src/qemu/qemu_process.c | 2 +- 6 files changed, 18 insertions(+), 11 deletions(-)
Reviewed-by: Ján Tomko <jtomko@redhat.com> Jano
participants (2)
-
Ján Tomko -
Peter Krempa