| Message ID | 20230209192043.14885-7-avihaih@nvidia.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | vfio/migration: Implement VFIO migration protocol v2 | expand |
On 2/9/23 20:20, Avihai Horon wrote: > Currently VFIO migration doesn't implement some kind of intermediate > quiescent state in which P2P DMAs are quiesced before stopping or > running the device. This can cause problems in multi-device migration > where the devices are doing P2P DMAs, since the devices are not stopped > together at the same time. > > Until such support is added, block migration of multiple devices. > > Signed-off-by: Avihai Horon <avihaih@nvidia.com> Reviewed-by: Cédric Le Goater <clg@redhat.com> Thanks, C. > --- > include/hw/vfio/vfio-common.h | 2 ++ > hw/vfio/common.c | 53 +++++++++++++++++++++++++++++++++++ > hw/vfio/migration.c | 6 ++++ > 3 files changed, 61 insertions(+) > > diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h > index e573f5a9f1..56b1683824 100644 > --- a/include/hw/vfio/vfio-common.h > +++ b/include/hw/vfio/vfio-common.h > @@ -218,6 +218,8 @@ typedef QLIST_HEAD(VFIOGroupList, VFIOGroup) VFIOGroupList; > extern VFIOGroupList vfio_group_list; > > bool vfio_mig_active(void); > +int vfio_block_multiple_devices_migration(Error **errp); > +void vfio_unblock_multiple_devices_migration(void); > int64_t vfio_mig_bytes_transferred(void); > > #ifdef CONFIG_LINUX > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > index 3a35f4afad..fe80ccf914 100644 > --- a/hw/vfio/common.c > +++ b/hw/vfio/common.c > @@ -41,6 +41,7 @@ > #include "qapi/error.h" > #include "migration/migration.h" > #include "migration/misc.h" > +#include "migration/blocker.h" > #include "sysemu/tpm.h" > > VFIOGroupList vfio_group_list = > @@ -337,6 +338,58 @@ bool vfio_mig_active(void) > return true; > } > > +static Error *multiple_devices_migration_blocker; > + > +static unsigned int vfio_migratable_device_num(void) > +{ > + VFIOGroup *group; > + VFIODevice *vbasedev; > + unsigned int device_num = 0; > + > + QLIST_FOREACH(group, &vfio_group_list, next) { > + QLIST_FOREACH(vbasedev, &group->device_list, next) { > + if (vbasedev->migration) { > + device_num++; > + } > + } > + } > + > + return device_num; > +} > + > +int vfio_block_multiple_devices_migration(Error **errp) > +{ > + int ret; > + > + if (multiple_devices_migration_blocker || > + vfio_migratable_device_num() <= 1) { > + return 0; > + } > + > + error_setg(&multiple_devices_migration_blocker, > + "Migration is currently not supported with multiple " > + "VFIO devices"); > + ret = migrate_add_blocker(multiple_devices_migration_blocker, errp); > + if (ret < 0) { > + error_free(multiple_devices_migration_blocker); > + multiple_devices_migration_blocker = NULL; > + } > + > + return ret; > +} > + > +void vfio_unblock_multiple_devices_migration(void) > +{ > + if (!multiple_devices_migration_blocker || > + vfio_migratable_device_num() > 1) { > + return; > + } > + > + migrate_del_blocker(multiple_devices_migration_blocker); > + error_free(multiple_devices_migration_blocker); > + multiple_devices_migration_blocker = NULL; > +} > + > static bool vfio_devices_all_dirty_tracking(VFIOContainer *container) > { > VFIOGroup *group; > diff --git a/hw/vfio/migration.c b/hw/vfio/migration.c > index c9f3117986..5dafa0a558 100644 > --- a/hw/vfio/migration.c > +++ b/hw/vfio/migration.c > @@ -881,6 +881,11 @@ int vfio_migration_probe(VFIODevice *vbasedev, Error **errp) > goto add_blocker; > } > > + ret = vfio_block_multiple_devices_migration(errp); > + if (ret) { > + return ret; > + } > + > trace_vfio_migration_probe(vbasedev->name, info->index); > g_free(info); > return 0; > @@ -907,6 +912,7 @@ void vfio_migration_finalize(VFIODevice *vbasedev) > qemu_del_vm_change_state_handler(migration->vm_state); > unregister_savevm(VMSTATE_IF(vbasedev->dev), "vfio", vbasedev); > vfio_migration_exit(vbasedev); > + vfio_unblock_multiple_devices_migration(); > } > > if (vbasedev->migration_blocker) {
Avihai Horon <avihaih@nvidia.com> wrote: > Currently VFIO migration doesn't implement some kind of intermediate > quiescent state in which P2P DMAs are quiesced before stopping or > running the device. This can cause problems in multi-device migration > where the devices are doing P2P DMAs, since the devices are not stopped > together at the same time. > > Until such support is added, block migration of multiple devices. > > Signed-off-by: Avihai Horon <avihaih@nvidia.com> Reviewed-by: Juan Quintela <quintela@redhat.com> With respect to how it intercat with migration.
diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h index e573f5a9f1..56b1683824 100644 --- a/include/hw/vfio/vfio-common.h +++ b/include/hw/vfio/vfio-common.h @@ -218,6 +218,8 @@ typedef QLIST_HEAD(VFIOGroupList, VFIOGroup) VFIOGroupList; extern VFIOGroupList vfio_group_list; bool vfio_mig_active(void); +int vfio_block_multiple_devices_migration(Error **errp); +void vfio_unblock_multiple_devices_migration(void); int64_t vfio_mig_bytes_transferred(void); #ifdef CONFIG_LINUX diff --git a/hw/vfio/common.c b/hw/vfio/common.c index 3a35f4afad..fe80ccf914 100644 --- a/hw/vfio/common.c +++ b/hw/vfio/common.c @@ -41,6 +41,7 @@ #include "qapi/error.h" #include "migration/migration.h" #include "migration/misc.h" +#include "migration/blocker.h" #include "sysemu/tpm.h" VFIOGroupList vfio_group_list = @@ -337,6 +338,58 @@ bool vfio_mig_active(void) return true; } +static Error *multiple_devices_migration_blocker; + +static unsigned int vfio_migratable_device_num(void) +{ + VFIOGroup *group; + VFIODevice *vbasedev; + unsigned int device_num = 0; + + QLIST_FOREACH(group, &vfio_group_list, next) { + QLIST_FOREACH(vbasedev, &group->device_list, next) { + if (vbasedev->migration) { + device_num++; + } + } + } + + return device_num; +} + +int vfio_block_multiple_devices_migration(Error **errp) +{ + int ret; + + if (multiple_devices_migration_blocker || + vfio_migratable_device_num() <= 1) { + return 0; + } + + error_setg(&multiple_devices_migration_blocker, + "Migration is currently not supported with multiple " + "VFIO devices"); + ret = migrate_add_blocker(multiple_devices_migration_blocker, errp); + if (ret < 0) { + error_free(multiple_devices_migration_blocker); + multiple_devices_migration_blocker = NULL; + } + + return ret; +} + +void vfio_unblock_multiple_devices_migration(void) +{ + if (!multiple_devices_migration_blocker || + vfio_migratable_device_num() > 1) { + return; + } + + migrate_del_blocker(multiple_devices_migration_blocker); + error_free(multiple_devices_migration_blocker); + multiple_devices_migration_blocker = NULL; +} + static bool vfio_devices_all_dirty_tracking(VFIOContainer *container) { VFIOGroup *group; diff --git a/hw/vfio/migration.c b/hw/vfio/migration.c index c9f3117986..5dafa0a558 100644 --- a/hw/vfio/migration.c +++ b/hw/vfio/migration.c @@ -881,6 +881,11 @@ int vfio_migration_probe(VFIODevice *vbasedev, Error **errp) goto add_blocker; } + ret = vfio_block_multiple_devices_migration(errp); + if (ret) { + return ret; + } + trace_vfio_migration_probe(vbasedev->name, info->index); g_free(info); return 0; @@ -907,6 +912,7 @@ void vfio_migration_finalize(VFIODevice *vbasedev) qemu_del_vm_change_state_handler(migration->vm_state); unregister_savevm(VMSTATE_IF(vbasedev->dev), "vfio", vbasedev); vfio_migration_exit(vbasedev); + vfio_unblock_multiple_devices_migration(); } if (vbasedev->migration_blocker) {
Currently VFIO migration doesn't implement some kind of intermediate quiescent state in which P2P DMAs are quiesced before stopping or running the device. This can cause problems in multi-device migration where the devices are doing P2P DMAs, since the devices are not stopped together at the same time. Until such support is added, block migration of multiple devices. Signed-off-by: Avihai Horon <avihaih@nvidia.com> --- include/hw/vfio/vfio-common.h | 2 ++ hw/vfio/common.c | 53 +++++++++++++++++++++++++++++++++++ hw/vfio/migration.c | 6 ++++ 3 files changed, 61 insertions(+)