migration: fix multifd_send_pages() next channel
commit7e89a1401a9674c9882948f05f4d17ea7be1c4eb
authorLaurent Vivier <lvivier@redhat.com>
Wed, 17 Jun 2020 11:31:54 +0000 (17 13:31 +0200)
committerDr. David Alan Gilbert <dgilbert@redhat.com>
Wed, 17 Jun 2020 16:48:39 +0000 (17 17:48 +0100)
tree1090373a6623a9362ba18b80f0c8ff349ff39339
parent6bcd361a52e73889d2123033ce48450289a1933e
migration: fix multifd_send_pages() next channel

multifd_send_pages() loops around the available channels,
the next channel to use between two calls to multifd_send_pages() is stored
inside a local static variable, next_channel.

It works well, except if the number of channels decreases between two calls
to multifd_send_pages(). In this case, the loop can try to access the
data of a channel that doesn't exist anymore.

The problem can be triggered if we start a migration with a given number of
channels and then we cancel the migration to restart it with a lower number.
This ends generally with an error like:
qemu-system-ppc64: .../util/qemu-thread-posix.c:77: qemu_mutex_lock_impl: Assertion `mutex->initialized' failed.

This patch fixes the error by capping next_channel with the current number
of channels before using it.

Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Message-Id: <20200617113154.593233-1-lvivier@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
migration/multifd.c