Re: [PATCH v2] migration: Stop migration immediately in RDMA error paths

[Peter Xu]

On Tue, Oct 24, 2023 at 06:16:27PM +0200, Juan Quintela wrote:
> Peter Xu <peterx@redhat.com> wrote:
> > In multiple places, RDMA errors are handled in a strange way, where it only
> > sets qemu_file_set_error() but not stop the migration immediately.
> >
> > It's not obvious what will happen later if there is already an error.  Make
> > all such failures stop migration immediately.
> >
> > Cc: Zhijian Li (Fujitsu) <lizhijian@fujitsu.com>
> > Cc: Markus Armbruster <armbru@redhat.com>
> > Cc: Juan Quintela <quintela@redhat.com>
> > Cc: Fabiano Rosas <farosas@suse.de>
> > Reported-by: Thomas Huth <thuth@redhat.com>
> > Signed-off-by: Peter Xu <peterx@redhat.com>
> > ---
> >
> > v2:
> > - One more line squashed into to fix the build error...  Please ignore v1,
> >   sorry for the noise.
> >
> > This patch is based on Thomas's patch:
> >
> > [PATCH v2] migration/ram: Fix compilation with -Wshadow=local
> > 20231024092220.55305-1-thuth@redhat.com">https://lore.kernel.org/r/20231024092220.55305-1-thuth@redhat.com
> >
> > Above patch should have been queued by both Markus and Juan.
> > ---
> >  migration/ram.c | 13 ++++++++-----
> >  1 file changed, 8 insertions(+), 5 deletions(-)
> >
> > diff --git a/migration/ram.c b/migration/ram.c
> > index 212add4481..1473bb593a 100644
> > --- a/migration/ram.c
> > +++ b/migration/ram.c
> > @@ -3034,11 +3034,13 @@ static int ram_save_setup(QEMUFile *f, void *opaque)
> >      ret = rdma_registration_start(f, RAM_CONTROL_SETUP);
> >      if (ret < 0) {
> >          qemu_file_set_error(f, ret);
> > +        return ret;
> 
> I agree
> 
> >      }
> >  
> >      ret = rdma_registration_stop(f, RAM_CONTROL_SETUP);
> >      if (ret < 0) {
> >          qemu_file_set_error(f, ret);
> > +        return ret;
> 
> I agree
> 
> >      }
> >  
> >      migration_ops = g_malloc0(sizeof(MigrationOps));
> > @@ -3104,6 +3106,7 @@ static int ram_save_iterate(QEMUFile *f, void *opaque)
> >          ret = rdma_registration_start(f, RAM_CONTROL_ROUND);
> >          if (ret < 0) {
> >              qemu_file_set_error(f, ret);
> > +            goto out;
> 
> Seems sensible
> 
> >          }
> >  
> >          t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
> > @@ -3208,8 +3211,6 @@ static int ram_save_complete(QEMUFile *f, void 
> > *opaque)
> >      rs->last_stage = !migration_in_colo_state();
> >  
> >      WITH_RCU_READ_LOCK_GUARD() {
> > -        int rdma_reg_ret;
> > -
> >          if (!migration_in_postcopy()) {
> >              migration_bitmap_sync_precopy(rs, true);
> >          }
> > @@ -3217,6 +3218,7 @@ static int ram_save_complete(QEMUFile *f, void 
> > *opaque)
> >          ret = rdma_registration_start(f, RAM_CONTROL_FINISH);
> >          if (ret < 0) {
> >              qemu_file_set_error(f, ret);
> > +            break;
> 
> Please
>               return ret;
> 
> 
> We can do exactly the same with pages < 0.
> 
> >          }
> >  
> >          /* try transferring iterative blocks of memory */
> > @@ -3240,9 +3242,10 @@ static int ram_save_complete(QEMUFile *f, void 
> > *opaque)
> >  
> >          ram_flush_compressed_data(rs);
> >  
> > -        rdma_reg_ret = rdma_registration_stop(f, RAM_CONTROL_FINISH);
> > -        if (rdma_reg_ret < 0) {
> > -            qemu_file_set_error(f, rdma_reg_ret);
> > +        ret = rdma_registration_stop(f, RAM_CONTROL_FINISH);
> > +        if (ret < 0) {
> > +            qemu_file_set_error(f, ret);
> > +            break;
> >          }
> >      }
> 
> And if we return here, we can just drop the:
> 
>     if (ret < 0) {
>         return ret;
>     }
> 
> 
> At the ext of the loop.

IIUC that'll be the same as this patch, but sure thing I'll prepare a v3.

Thanks,

-- 
Peter Xu