nbd-client: avoid spurious qio_channel_yield() re-entry
commit40f4a21895b5a7eae4011593837069f63460d983
authorStefan Hajnoczi <stefanha@redhat.com>
Tue, 22 Aug 2017 12:51:13 +0000 (22 13:51 +0100)
committerEric Blake <eblake@redhat.com>
Wed, 23 Aug 2017 16:22:15 +0000 (23 11:22 -0500)
treecacbb56b00763fc9e0956df21ca11d742a77aa14
parent12314f2d145c656f0a1f9aebcbe5e21be60c5ca5
nbd-client: avoid spurious qio_channel_yield() re-entry

The following scenario leads to an assertion failure in
qio_channel_yield():

1. Request coroutine calls qio_channel_yield() successfully when sending
   would block on the socket.  It is now yielded.
2. nbd_read_reply_entry() calls nbd_recv_coroutines_enter_all() because
   nbd_receive_reply() failed.
3. Request coroutine is entered and returns from qio_channel_yield().
   Note that the socket fd handler has not fired yet so
   ioc->write_coroutine is still set.
4. Request coroutine attempts to send the request body with nbd_rwv()
   but the socket would still block.  qio_channel_yield() is called
   again and assert(!ioc->write_coroutine) is hit.

The problem is that nbd_read_reply_entry() does not distinguish between
request coroutines that are waiting to receive a reply and those that
are not.

This patch adds a per-request bool receiving flag so
nbd_read_reply_entry() can avoid spurious aio_wake() calls.

Reported-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20170822125113.5025-1-stefanha@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Tested-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
block/nbd-client.c
block/nbd-client.h