RE: [PATCH v4 07/10] migration: split migration_incoming_co
> -Original Message- > From: Vladimir Sementsov-Ogievskiy > Sent: Thursday, May 4, 2023 6:52 AM > To: Peter Xu > Cc: qemu-devel@nongnu.org; lukasstra...@web.de; quint...@redhat.com; > Zhang, Chen ; Zhang, Hailiang > ; Leonardo Bras > Subject: Re: [PATCH v4 07/10] migration: split migration_incoming_co > > On 02.05.23 23:48, Peter Xu wrote: > > On Fri, Apr 28, 2023 at 10:49:25PM +0300, Vladimir Sementsov-Ogievskiy > wrote: > >> Originally, migration_incoming_co was introduced by > >> 25d0c16f625feb3b6 > >> "migration: Switch to COLO process after finishing loadvm" > >> to be able to enter from COLO code to one specific yield point, added > >> by 25d0c16f625feb3b6. > >> > >> Later in 923709896b1b0 > >> "migration: poll the cm event for destination qemu" > >> we reused this variable to wake the migration incoming coroutine from > >> RDMA code. > >> > >> That was doubtful idea. Entering coroutines is a very fragile thing: > >> you should be absolutely sure which yield point you are going to enter. > >> > >> I don't know how much is it safe to enter during qemu_loadvm_state() > >> which I think what RDMA want to do. But for sure RDMA shouldn't enter > >> the special COLO-related yield-point. As well, COLO code doesn't want > >> to enter during qemu_loadvm_state(), it want to enter it's own > >> specific yield-point. > >> > >> As well, when in 8e48ac95865ac97d > >> "COLO: Add block replication into colo process" we added > >> bdrv_invalidate_cache_all() call (now it's called activate_all()) it > >> became possible to enter the migration incoming coroutine during that > >> call which is wrong too. > >> > >> So, let't make these things separate and disjoint: loadvm_co for > >> RDMA, non-NULL during qemu_loadvm_state(), and colo_incoming_co for > >> COLO, non-NULL only around specific yield. > >> > >> Signed-off-by: Vladimir Sementsov-Ogievskiy > >> > >> --- > >> migration/colo.c | 4 ++-- > >> migration/migration.c | 8 ++-- > >> migration/migration.h | 9 - > >> 3 files changed, 16 insertions(+), 5 deletions(-) > > > > The idea looks right to me, but I really know mostly nothing on > > coroutines and also rdma+colo.. > > > > Is the other ref in rdma.c (rdma_cm_poll_handler()) still missing? > > > > Oops right.. I was building with rdma disabled. Will fix. > > Thanks a lot for reviewing! > Yes, I know some people and company try to enable COLO with RDMA. But in my side, I haven't tried this way yet. Thanks Chen > -- > Best regards, > Vladimir
Re: [PATCH v4 07/10] migration: split migration_incoming_co
On 02.05.23 23:48, Peter Xu wrote: On Fri, Apr 28, 2023 at 10:49:25PM +0300, Vladimir Sementsov-Ogievskiy wrote: Originally, migration_incoming_co was introduced by 25d0c16f625feb3b6 "migration: Switch to COLO process after finishing loadvm" to be able to enter from COLO code to one specific yield point, added by 25d0c16f625feb3b6. Later in 923709896b1b0 "migration: poll the cm event for destination qemu" we reused this variable to wake the migration incoming coroutine from RDMA code. That was doubtful idea. Entering coroutines is a very fragile thing: you should be absolutely sure which yield point you are going to enter. I don't know how much is it safe to enter during qemu_loadvm_state() which I think what RDMA want to do. But for sure RDMA shouldn't enter the special COLO-related yield-point. As well, COLO code doesn't want to enter during qemu_loadvm_state(), it want to enter it's own specific yield-point. As well, when in 8e48ac95865ac97d "COLO: Add block replication into colo process" we added bdrv_invalidate_cache_all() call (now it's called activate_all()) it became possible to enter the migration incoming coroutine during that call which is wrong too. So, let't make these things separate and disjoint: loadvm_co for RDMA, non-NULL during qemu_loadvm_state(), and colo_incoming_co for COLO, non-NULL only around specific yield. Signed-off-by: Vladimir Sementsov-Ogievskiy --- migration/colo.c | 4 ++-- migration/migration.c | 8 ++-- migration/migration.h | 9 - 3 files changed, 16 insertions(+), 5 deletions(-) The idea looks right to me, but I really know mostly nothing on coroutines and also rdma+colo.. Is the other ref in rdma.c (rdma_cm_poll_handler()) still missing? Oops right.. I was building with rdma disabled. Will fix. Thanks a lot for reviewing! -- Best regards, Vladimir
Re: [PATCH v4 07/10] migration: split migration_incoming_co
On Fri, Apr 28, 2023 at 10:49:25PM +0300, Vladimir Sementsov-Ogievskiy wrote: > Originally, migration_incoming_co was introduced by > 25d0c16f625feb3b6 >"migration: Switch to COLO process after finishing loadvm" > to be able to enter from COLO code to one specific yield point, added > by 25d0c16f625feb3b6. > > Later in 923709896b1b0 > "migration: poll the cm event for destination qemu" > we reused this variable to wake the migration incoming coroutine from > RDMA code. > > That was doubtful idea. Entering coroutines is a very fragile thing: > you should be absolutely sure which yield point you are going to enter. > > I don't know how much is it safe to enter during qemu_loadvm_state() > which I think what RDMA want to do. But for sure RDMA shouldn't enter > the special COLO-related yield-point. As well, COLO code doesn't want > to enter during qemu_loadvm_state(), it want to enter it's own specific > yield-point. > > As well, when in 8e48ac95865ac97d > "COLO: Add block replication into colo process" we added > bdrv_invalidate_cache_all() call (now it's called activate_all()) > it became possible to enter the migration incoming coroutine during > that call which is wrong too. > > So, let't make these things separate and disjoint: loadvm_co for RDMA, > non-NULL during qemu_loadvm_state(), and colo_incoming_co for COLO, > non-NULL only around specific yield. > > Signed-off-by: Vladimir Sementsov-Ogievskiy > --- > migration/colo.c | 4 ++-- > migration/migration.c | 8 ++-- > migration/migration.h | 9 - > 3 files changed, 16 insertions(+), 5 deletions(-) The idea looks right to me, but I really know mostly nothing on coroutines and also rdma+colo.. Is the other ref in rdma.c (rdma_cm_poll_handler()) still missing? -- Peter Xu
[PATCH v4 07/10] migration: split migration_incoming_co
Originally, migration_incoming_co was introduced by 25d0c16f625feb3b6 "migration: Switch to COLO process after finishing loadvm" to be able to enter from COLO code to one specific yield point, added by 25d0c16f625feb3b6. Later in 923709896b1b0 "migration: poll the cm event for destination qemu" we reused this variable to wake the migration incoming coroutine from RDMA code. That was doubtful idea. Entering coroutines is a very fragile thing: you should be absolutely sure which yield point you are going to enter. I don't know how much is it safe to enter during qemu_loadvm_state() which I think what RDMA want to do. But for sure RDMA shouldn't enter the special COLO-related yield-point. As well, COLO code doesn't want to enter during qemu_loadvm_state(), it want to enter it's own specific yield-point. As well, when in 8e48ac95865ac97d "COLO: Add block replication into colo process" we added bdrv_invalidate_cache_all() call (now it's called activate_all()) it became possible to enter the migration incoming coroutine during that call which is wrong too. So, let't make these things separate and disjoint: loadvm_co for RDMA, non-NULL during qemu_loadvm_state(), and colo_incoming_co for COLO, non-NULL only around specific yield. Signed-off-by: Vladimir Sementsov-Ogievskiy --- migration/colo.c | 4 ++-- migration/migration.c | 8 ++-- migration/migration.h | 9 - 3 files changed, 16 insertions(+), 5 deletions(-) diff --git a/migration/colo.c b/migration/colo.c index 6c7c313956..a688ac553a 100644 --- a/migration/colo.c +++ b/migration/colo.c @@ -145,8 +145,8 @@ static void secondary_vm_do_failover(void) qemu_sem_post(>colo_incoming_sem); /* For Secondary VM, jump to incoming co */ -if (mis->migration_incoming_co) { -qemu_coroutine_enter(mis->migration_incoming_co); +if (mis->colo_incoming_co) { +qemu_coroutine_enter(mis->colo_incoming_co); } } diff --git a/migration/migration.c b/migration/migration.c index 8db0892317..23b2d187de 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -505,12 +505,14 @@ process_incoming_migration_co(void *opaque) Error *local_err = NULL; assert(mis->from_src_file); -mis->migration_incoming_co = qemu_coroutine_self(); mis->largest_page_size = qemu_ram_pagesize_largest(); postcopy_state_set(POSTCOPY_INCOMING_NONE); migrate_set_state(>state, MIGRATION_STATUS_NONE, MIGRATION_STATUS_ACTIVE); + +mis->loadvm_co = qemu_coroutine_self(); ret = qemu_loadvm_state(mis->from_src_file); +mis->loadvm_co = NULL; ps = postcopy_state_get(); trace_process_incoming_migration_co_end(ret, ps); @@ -551,7 +553,10 @@ process_incoming_migration_co(void *opaque) qemu_thread_create(_incoming_thread, "COLO incoming", colo_process_incoming_thread, mis, QEMU_THREAD_JOINABLE); + +mis->colo_incoming_co = qemu_coroutine_self(); qemu_coroutine_yield(); +mis->colo_incoming_co = NULL; qemu_mutex_unlock_iothread(); /* Wait checkpoint incoming thread exit before free resource */ @@ -563,7 +568,6 @@ process_incoming_migration_co(void *opaque) mis->bh = qemu_bh_new(process_incoming_migration_bh, mis); qemu_bh_schedule(mis->bh); -mis->migration_incoming_co = NULL; return; fail: local_err = NULL; diff --git a/migration/migration.h b/migration/migration.h index 7721c7658b..48a46123a0 100644 --- a/migration/migration.h +++ b/migration/migration.h @@ -162,8 +162,15 @@ struct MigrationIncomingState { int state; +/* + * The incoming migration coroutine, non-NULL during qemu_loadvm_state(). + * Used to wake the migration incoming coroutine from rdma code. How much is + * it safe - it's a question. + */ +Coroutine *loadvm_co; + /* The coroutine we should enter (back) after failover */ -Coroutine *migration_incoming_co; +Coroutine *colo_incoming_co; QemuSemaphore colo_incoming_sem; /* -- 2.34.1