Blame SOURCES/kvm-util-async-use-qemu_aio_coroutine_enter-in-co_schedu.patch

26ba25
From f29b1e17713739baf416b64eeee9549f07717ea8 Mon Sep 17 00:00:00 2001
26ba25
From: Kevin Wolf <kwolf@redhat.com>
26ba25
Date: Wed, 10 Oct 2018 20:21:53 +0100
26ba25
Subject: [PATCH 27/49] util/async: use qemu_aio_coroutine_enter in
26ba25
 co_schedule_bh_cb
26ba25
26ba25
RH-Author: Kevin Wolf <kwolf@redhat.com>
26ba25
Message-id: <20181010202213.7372-15-kwolf@redhat.com>
26ba25
Patchwork-id: 82604
26ba25
O-Subject: [RHEL-8 qemu-kvm PATCH 24/44] util/async: use qemu_aio_coroutine_enter in co_schedule_bh_cb
26ba25
Bugzilla: 1637976
26ba25
RH-Acked-by: Max Reitz <mreitz@redhat.com>
26ba25
RH-Acked-by: John Snow <jsnow@redhat.com>
26ba25
RH-Acked-by: Thomas Huth <thuth@redhat.com>
26ba25
26ba25
From: Sergio Lopez <slp@redhat.com>
26ba25
26ba25
AIO Coroutines shouldn't by managed by an AioContext different than the
26ba25
one assigned when they are created. aio_co_enter avoids entering a
26ba25
coroutine from a different AioContext, calling aio_co_schedule instead.
26ba25
26ba25
Scheduled coroutines are then entered by co_schedule_bh_cb using
26ba25
qemu_coroutine_enter, which just calls qemu_aio_coroutine_enter with the
26ba25
current AioContext obtained with qemu_get_current_aio_context.
26ba25
Eventually, co->ctx will be set to the AioContext passed as an argument
26ba25
to qemu_aio_coroutine_enter.
26ba25
26ba25
This means that, if an IO Thread's AioConext is being processed by the
26ba25
Main Thread (due to aio_poll being called with a BDS AioContext, as it
26ba25
happens in AIO_WAIT_WHILE among other places), the AioContext from some
26ba25
coroutines may be wrongly replaced with the one from the Main Thread.
26ba25
26ba25
This is the root cause behind some crashes, mainly triggered by the
26ba25
drain code at block/io.c. The most common are these abort and failed
26ba25
assertion:
26ba25
26ba25
util/async.c:aio_co_schedule
26ba25
456     if (scheduled) {
26ba25
457         fprintf(stderr,
26ba25
458                 "%s: Co-routine was already scheduled in '%s'\n",
26ba25
459                 __func__, scheduled);
26ba25
460         abort();
26ba25
461     }
26ba25
26ba25
util/qemu-coroutine-lock.c:
26ba25
286     assert(mutex->holder == self);
26ba25
26ba25
But it's also known to cause random errors at different locations, and
26ba25
even SIGSEGV with broken coroutine backtraces.
26ba25
26ba25
By using qemu_aio_coroutine_enter directly in co_schedule_bh_cb, we can
26ba25
pass the correct AioContext as an argument, making sure co->ctx is not
26ba25
wrongly altered.
26ba25
26ba25
Signed-off-by: Sergio Lopez <slp@redhat.com>
26ba25
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
26ba25
(cherry picked from commit 6808ae0417131f8dbe7b051256dff7a16634dc1d)
26ba25
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
26ba25
Signed-off-by: Danilo C. L. de Paula <ddepaula@redhat.com>
26ba25
---
26ba25
 util/async.c | 2 +-
26ba25
 1 file changed, 1 insertion(+), 1 deletion(-)
26ba25
26ba25
diff --git a/util/async.c b/util/async.c
26ba25
index 4dd9d95..5693191 100644
26ba25
--- a/util/async.c
26ba25
+++ b/util/async.c
26ba25
@@ -391,7 +391,7 @@ static void co_schedule_bh_cb(void *opaque)
26ba25
 
26ba25
         /* Protected by write barrier in qemu_aio_coroutine_enter */
26ba25
         atomic_set(&co->scheduled, NULL);
26ba25
-        qemu_coroutine_enter(co);
26ba25
+        qemu_aio_coroutine_enter(ctx, co);
26ba25
         aio_context_release(ctx);
26ba25
     }
26ba25
 }
26ba25
-- 
26ba25
1.8.3.1
26ba25