Re: [PATCH v11 4/5] migration: implementation of background snapshot thread

[Dr. David Alan Gilbert]

* Andrey Gruzdev (andrey.gruzdev@virtuozzo.com) wrote:
> On 19.01.2021 21:49, Dr. David Alan Gilbert wrote:
> > * Andrey Gruzdev (andrey.gruzdev@virtuozzo.com) wrote:
> > > Introducing implementation of 'background' snapshot thread
> > > which in overall follows the logic of precopy migration
> > > while internally utilizes completely different mechanism
> > > to 'freeze' vmstate at the start of snapshot creation.
> > > 
> > > This mechanism is based on userfault_fd with wr-protection
> > > support and is Linux-specific.
> > I noticed there weren't any bdrv_ calls in here; I guess with a snapshot
> > you still have the source running so still have it accessing the disk;
> > do you do anything to try and wire the ram snapshotting up to disk
> > snapshotting?
> 
> Block-related manipulations should be done externally, I think.
> So create backing images for RW nodes, then stop VM, switch block graph
> and start background snapshot. Something like create 'virsh 
> snapshot-create-as'
> does, but in other sequence.

If you get a chance it would be great if you could put together an
example of doing the combination RAM+block; that way we find out if there's
anything silly missing.

Dave

> //
> 
> > > Signed-off-by: Andrey Gruzdev <andrey.gruzdev@virtuozzo.com>
> > > Acked-by: Peter Xu <peterx@redhat.com>
> > > ---
> > >   migration/migration.c | 255 +++++++++++++++++++++++++++++++++++++++++-
> > >   migration/migration.h |   3 +
> > >   migration/ram.c       |   2 +
> > >   migration/savevm.c    |   1 -
> > >   migration/savevm.h    |   2 +
> > >   5 files changed, 260 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/migration/migration.c b/migration/migration.c
> > > index 2c2cb9ef01..0901a15ac5 100644
> > <snip>
> > 
> > > -    qemu_thread_create(&s->thread, "live_migration", migration_thread, s,
> > > -                       QEMU_THREAD_JOINABLE);
> > > +
> > > +    if (migrate_background_snapshot()) {
> > > +        qemu_thread_create(&s->thread, "background_snapshot",
> > Unfortunately that wont work - there's a 14 character limit on
> > the thread name length; I guess we just shorten that to bg_snapshot
> 
> Yep, missed that pthread_set_name_np() has a length limit)
> 
> > Other than that,
> > 
> > 
> > 
> > Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
> > 
> > > +                bg_migration_thread, s, QEMU_THREAD_JOINABLE);
> > > +    } else {
> > > +        qemu_thread_create(&s->thread, "live_migration",
> > > +                migration_thread, s, QEMU_THREAD_JOINABLE);
> > > +    }
> > >       s->migration_thread_running = true;
> > >   }
> > > diff --git a/migration/migration.h b/migration/migration.h
> > > index f40338cfbf..0723955cd7 100644
> > > --- a/migration/migration.h
> > > +++ b/migration/migration.h
> > > @@ -20,6 +20,7 @@
> > >   #include "qemu/thread.h"
> > >   #include "qemu/coroutine_int.h"
> > >   #include "io/channel.h"
> > > +#include "io/channel-buffer.h"
> > >   #include "net/announce.h"
> > >   #include "qom/object.h"
> > > @@ -147,8 +148,10 @@ struct MigrationState {
> > >       /*< public >*/
> > >       QemuThread thread;
> > > +    QEMUBH *vm_start_bh;
> > >       QEMUBH *cleanup_bh;
> > >       QEMUFile *to_dst_file;
> > > +    QIOChannelBuffer *bioc;
> > >       /*
> > >        * Protects to_dst_file pointer.  We need to make sure we won't
> > >        * yield or hang during the critical section, since this lock will
> > > diff --git a/migration/ram.c b/migration/ram.c
> > > index 5707382db1..05fe0c8592 100644
> > > --- a/migration/ram.c
> > > +++ b/migration/ram.c
> > > @@ -1471,6 +1471,7 @@ static RAMBlock *poll_fault_page(RAMState *rs, 
> > > ram_addr_t *offset)
> > >       page_address = (void *) uffd_msg.arg.pagefault.address;
> > >       bs = qemu_ram_block_from_host(page_address, false, offset);
> > >       assert(bs && (bs->flags & RAM_UF_WRITEPROTECT) != 0);
> > > +
> > >       return bs;
> > >   }
> > >   #endif /* CONFIG_LINUX */
> > > @@ -1836,6 +1837,7 @@ static void ram_save_host_page_post(RAMState *rs, 
> > > PageSearchStatus *pss,
> > >           /* Un-protect memory range. */
> > >           res = uffd_change_protection(rs->uffdio_fd, page_address, 
> > > run_length,
> > >                   false, false);
> > > +
> > >           /* We don't want to override existing error from 
> > > ram_save_host_page(). */
> > >           if (res < 0 && *res_override >= 0) {
> > >               *res_override = res;
> > > diff --git a/migration/savevm.c b/migration/savevm.c
> > > index 27e842812e..dd4ad0aaaf 100644
> > > --- a/migration/savevm.c
> > > +++ b/migration/savevm.c
> > > @@ -1354,7 +1354,6 @@ int 
> > > qemu_savevm_state_complete_precopy_iterable(QEMUFile *f, bool in_postcopy)
> > >       return 0;
> > >   }
> > > -static
> > >   int qemu_savevm_state_complete_precopy_non_iterable(QEMUFile *f,
> > >                                                       bool in_postcopy,
> > >                                                       bool 
> > > inactivate_disks)
> > > diff --git a/migration/savevm.h b/migration/savevm.h
> > > index ba64a7e271..aaee2528ed 100644
> > > --- a/migration/savevm.h
> > > +++ b/migration/savevm.h
> > > @@ -64,5 +64,7 @@ int qemu_loadvm_state(QEMUFile *f);
> > >   void qemu_loadvm_state_cleanup(void);
> > >   int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis);
> > >   int qemu_load_device_state(QEMUFile *f);
> > > +int qemu_savevm_state_complete_precopy_non_iterable(QEMUFile *f,
> > > +        bool in_postcopy, bool inactivate_disks);
> > >   #endif
> > > -- 
> > > 2.25.1
> > > 
> 
> -- 
> Andrey Gruzdev, Principal Engineer
> Virtuozzo GmbH  +7-903-247-6397
>                 virtuzzo.com
> 
-- 
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK