migration: multifd_send_thread always post p->sem_sync when error happen
commit2f4aefd320a8c3aeb739d59724c5c4dd6a995919
authorIvan Ren <ivanren@tencent.com>
Thu, 29 Aug 2019 02:16:36 +0000 (29 10:16 +0800)
committerDr. David Alan Gilbert <dgilbert@redhat.com>
Thu, 12 Sep 2019 09:53:33 +0000 (12 10:53 +0100)
tree91c5bc1701cf0abaeb2152dfc59dacdaadb0423d
parent89ea03a7dc83ca36b670ba7f787802791fcb04b1
migration: multifd_send_thread always post p->sem_sync when error happen

When encounter error, multifd_send_thread should always notify who pay
attention to it before exit. Otherwise it may block migration_thread
at multifd_send_sync_main forever.

Error as follow:
-------------------------------------------------------------------------------
 (gdb) bt
 #0  0x00007f4d669dfa0b in do_futex_wait.constprop.1 () from /lib64/libpthread.so.0
 #1  0x00007f4d669dfa9f in __new_sem_wait_slow.constprop.0 () from /lib64/libpthread.so.0
 #2  0x00007f4d669dfb3b in sem_wait@@GLIBC_2.2.5 () from /lib64/libpthread.so.0
 #3  0x0000562ccf0a5614 in qemu_sem_wait (sem=sem@entry=0x562cd1b698e8) at util/qemu-thread-posix.c:319
 #4  0x0000562ccecb4752 in multifd_send_sync_main (rs=<optimized out>) at /qemu/migration/ram.c:1099
 #5  0x0000562ccecb95f4 in ram_save_iterate (f=0x562cd0ecc000, opaque=<optimized out>) at /qemu/migration/ram.c:3550
 #6  0x0000562ccef43c23 in qemu_savevm_state_iterate (f=0x562cd0ecc000, postcopy=false) at migration/savevm.c:1189
 #7  0x0000562ccef3dcf3 in migration_iteration_run (s=0x562cd09fabf0) at migration/migration.c:3131
 #8  migration_thread (opaque=opaque@entry=0x562cd09fabf0) at migration/migration.c:3258
 #9  0x0000562ccf0a4c26 in qemu_thread_start (args=<optimized out>) at util/qemu-thread-posix.c:502
 #10 0x00007f4d669d9e25 in start_thread () from /lib64/libpthread.so.0
 #11 0x00007f4d6670635d in clone () from /lib64/libc.so.6
 (gdb) f 4
 #4  0x0000562ccecb4752 in multifd_send_sync_main (rs=<optimized out>) at /qemu/migration/ram.c:1099
 1099         qemu_sem_wait(&p->sem_sync);
 (gdb) list
 1094     }
 1095     for (i = 0; i < migrate_multifd_channels(); i++) {
 1096         MultiFDSendParams *p = &multifd_send_state->params[i];
 1097
 1098         trace_multifd_send_sync_main_wait(p->id);
 1099         qemu_sem_wait(&p->sem_sync);
 1100     }
 1101     trace_multifd_send_sync_main(multifd_send_state->packet_num);
 1102 }
 1103
 (gdb) p i
 $1 = 0
 (gdb)  p multifd_send_state->params[0].pending_job
 $2 = 2    //It means the job before MULTIFD_FLAG_SYNC has already fail
 (gdb)  p multifd_send_state->params[0].quit
 $3 = true

Signed-off-by: Ivan Ren <ivanren@tencent.com>
Message-Id: <1567044996-2362-1-git-send-email-ivanren@tencent.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
migration/ram.c