yeahuh / rpms / qemu-kvm

Forked from rpms/qemu-kvm 2 years ago
Clone

Blame SOURCES/kvm-migration-multifd-fix-nullptr-access-in-multifd_send.patch

902636
From 517a99c5fba163bf684978fe3d9476b619481391 Mon Sep 17 00:00:00 2001
902636
From: Juan Quintela <quintela@redhat.com>
902636
Date: Tue, 3 Mar 2020 14:51:42 +0000
902636
Subject: [PATCH 10/18] migration/multifd: fix nullptr access in
902636
 multifd_send_terminate_threads
902636
902636
RH-Author: Juan Quintela <quintela@redhat.com>
902636
Message-id: <20200303145143.149290-10-quintela@redhat.com>
902636
Patchwork-id: 94117
902636
O-Subject: [RHEL-AV-8.2.0 qemu-kvm PATCH v2 09/10] migration/multifd: fix nullptr access in multifd_send_terminate_threads
902636
Bugzilla: 1738451
902636
RH-Acked-by: Laurent Vivier <lvivier@redhat.com>
902636
RH-Acked-by: Peter Xu <peterx@redhat.com>
902636
RH-Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
902636
902636
From: Zhimin Feng <fengzhimin1@huawei.com>
902636
902636
If the multifd_send_threads is not created when migration is failed,
902636
multifd_save_cleanup would be called twice. In this senario, the
902636
multifd_send_state is accessed after it has been released, the result
902636
is that the source VM is crashing down.
902636
902636
Here is the coredump stack:
902636
    Program received signal SIGSEGV, Segmentation fault.
902636
    0x00005629333a78ef in multifd_send_terminate_threads (err=err@entry=0x0) at migration/ram.c:1012
902636
    1012            MultiFDSendParams *p = &multifd_send_state->params[i];
902636
    #0  0x00005629333a78ef in multifd_send_terminate_threads (err=err@entry=0x0) at migration/ram.c:1012
902636
    #1  0x00005629333ab8a9 in multifd_save_cleanup () at migration/ram.c:1028
902636
    #2  0x00005629333abaea in multifd_new_send_channel_async (task=0x562935450e70, opaque=<optimized out>) at migration/ram.c:1202
902636
    #3  0x000056293373a562 in qio_task_complete (task=task@entry=0x562935450e70) at io/task.c:196
902636
    #4  0x000056293373a6e0 in qio_task_thread_result (opaque=0x562935450e70) at io/task.c:111
902636
    #5  0x00007f475d4d75a7 in g_idle_dispatch () from /usr/lib64/libglib-2.0.so.0
902636
    #6  0x00007f475d4da9a9 in g_main_context_dispatch () from /usr/lib64/libglib-2.0.so.0
902636
    #7  0x0000562933785b33 in glib_pollfds_poll () at util/main-loop.c:219
902636
    #8  os_host_main_loop_wait (timeout=<optimized out>) at util/main-loop.c:242
902636
    #9  main_loop_wait (nonblocking=nonblocking@entry=0) at util/main-loop.c:518
902636
    #10 0x00005629334c5acf in main_loop () at vl.c:1810
902636
    #11 0x000056293334d7bb in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4471
902636
902636
If the multifd_send_threads is not created when migration is failed.
902636
In this senario, we don't call multifd_save_cleanup in multifd_new_send_channel_async.
902636
902636
Signed-off-by: Zhimin Feng <fengzhimin1@huawei.com>
902636
Reviewed-by: Juan Quintela <quintela@redhat.com>
902636
Signed-off-by: Juan Quintela <quintela@redhat.com>
902636
(cherry picked from commit 9c4d333c092e9c26d38f740ff3616deb42f21681)
902636
Signed-off-by: Danilo C. L. de Paula <ddepaula@redhat.com>
902636
---
902636
 migration/ram.c | 10 +++++++++-
902636
 1 file changed, 9 insertions(+), 1 deletion(-)
902636
902636
diff --git a/migration/ram.c b/migration/ram.c
902636
index 902c56c..3891eff 100644
902636
--- a/migration/ram.c
902636
+++ b/migration/ram.c
902636
@@ -1229,7 +1229,15 @@ static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
902636
     trace_multifd_new_send_channel_async(p->id);
902636
     if (qio_task_propagate_error(task, &local_err)) {
902636
         migrate_set_error(migrate_get_current(), local_err);
902636
-        multifd_save_cleanup();
902636
+        /* Error happen, we need to tell who pay attention to me */
902636
+        qemu_sem_post(&multifd_send_state->channels_ready);
902636
+        qemu_sem_post(&p->sem_sync);
902636
+        /*
902636
+         * Although multifd_send_thread is not created, but main migration
902636
+         * thread neet to judge whether it is running, so we need to mark
902636
+         * its status.
902636
+         */
902636
+        p->quit = true;
902636
     } else {
902636
         p->c = QIO_CHANNEL(sioc);
902636
         qio_channel_set_delay(p->c, false);
902636
-- 
902636
1.8.3.1
902636