Drivers: hv: hv_fcopy: fix a race condition for SMP guest
authorDexuan Cui <decui@microsoft.com>
Wed, 16 Jul 2014 07:00:45 +0000 (00:00 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 18 Jul 2014 01:43:10 +0000 (18:43 -0700)
We should schedule the 5s "timer work" before starting the data transfer,
otherwise, the data transfer code may finish so fast on another
virtual cpu that when the code(fcopy_write()) trying to cancel the 5s
"timer work" can occasionally fail because the "timer work" may haven't
been scheduled yet and as a result the fcopy process will be aborted
wrongly by fcopy_work_func() in 5s.

Thank Liz Zhang <lizzha@microsoft.com> for the initial investigation
on the bug.

This addresses https://bugzilla.redhat.com/show_bug.cgi?id=1118123

Tested-by: Liz Zhang <lizzha@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/hv/hv_fcopy.c

index eaaa3d843b8052c4c0d5b513567da720c6cfbe7a..23b2ce294c4ca78ea3f961ee102550e4aeb02632 100644 (file)
@@ -246,8 +246,8 @@ void hv_fcopy_onchannelcallback(void *context)
                /*
                 * Send the information to the user-level daemon.
                 */
-               fcopy_send_data();
                schedule_delayed_work(&fcopy_work, 5*HZ);
+               fcopy_send_data();
                return;
        }
        icmsghdr->icflags = ICMSGHDRFLAG_TRANSACTION | ICMSGHDRFLAG_RESPONSE;