Public bug reported:
13:38:23) shorne_: Hello, I was doing some testing with migrations for my
OpenRISC SMP patch set, I noticed something that looks like a bug, wondering if
someone else wants to confirm
(13:38:47) shorne_: Basically, calling loadvm 2 times causes crash
(13:38:54) shorne_: migration/savevm.c:
qemu_event_set(&mis->main_thread_load_event)
(13:38:54) stefanha: fam: Here is my take at this change:
https://paste.debian.net/982690/
(13:38:56) shorne_: assert(ev->initialized) -
fails inside
(13:39:32) stefanha: quintela davidgiluk: ^
(13:41:23) ***davidgiluk looks
(13:41:40) shorne_: c096358e747 util/qemu-thread-posix.c (Fam Zheng
2017-07-04 20:23:25 +0800 397) assert(ev->initialized);
(13:41:51) davidgiluk: shorne_: So you're doing a loadvm to load a snapshot and
then again?
(13:42:02) shorne_: Looks like adding that assert() was done really recently
(13:42:41) shorne_: yes, just loadvm 'a' ... then wait a bit longer, loadvm 'a'
again (confirm clocks go back etc)
(13:42:50) stefanha: fam: While you're having dinner I'll work on turning my
script into a qemu-iotests test case that we can merge.
(13:44:03) gpiccoli [~gpicc...@0002093a.user.oftc.net] entered the room.
(13:44:21) davidgiluk: shorne_: Well, it looks like the c09635 assert is a
sanity check to make sure we didn't do anything stupid, and well.....
(13:44:57) pm215: migration_incoming_get_current() and
migration_incoming_state_destroy() seem a bit mismatched
(13:45:13) davidgiluk: pm215: Yep
(13:45:46) davidgiluk: pm215: Generally we've thought that an incoming
migration normally only happens once - shorne_'s case is the exception
(13:46:03) shorne_: pm215: yeah, it looked something like that I just had a few
seconds to look at today
(13:46:03) HariharanTS left the room (quit: Ping timeout: 480 seconds).
(13:46:03) shorne_ is now known as shorne
(13:48:05) shorne: davidgiluk: pm215: thanks for having a look. Unfortunately
I need to head off to bed and put kids to sleep
(13:49:11) davidgiluk: shorne: Sleep well, no nightmares about event
destroyers....
(13:49:30) davidgiluk: pm215: Yeh this is fall out from b4b076daf32
** Affects: qemu
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1712564
Title:
loadvm fails twice in sequence
Status in QEMU:
New
Bug description:
13:38:23) shorne_: Hello, I was doing some testing with migrations for my
OpenRISC SMP patch set, I noticed something that looks like a bug, wondering if
someone else wants to confirm
(13:38:47) shorne_: Basically, calling loadvm 2 times causes crash
(13:38:54) shorne_: migration/savevm.c:
qemu_event_set(&mis->main_thread_load_event)
(13:38:54) stefanha: fam: Here is my take at this change:
https://paste.debian.net/982690/
(13:38:56) shorne_: assert(ev->initialized) -
fails inside
(13:39:32) stefanha: quintela davidgiluk: ^
(13:41:23) ***davidgiluk looks
(13:41:40) shorne_: c096358e747 util/qemu-thread-posix.c (Fam Zheng
2017-07-04 20:23:25 +0800 397) assert(ev->initialized);
(13:41:51) davidgiluk: shorne_: So you're doing a loadvm to load a snapshot
and then again?
(13:42:02) shorne_: Looks like adding that assert() was done really recently
(13:42:41) shorne_: yes, just loadvm 'a' ... then wait a bit longer, loadvm
'a' again (confirm clocks go back etc)
(13:42:50) stefanha: fam: While you're having dinner I'll work on turning my
script into a qemu-iotests test case that we can merge.
(13:44:03) gpiccoli [~gpicc...@0002093a.user.oftc.net] entered the room.
(13:44:21) davidgiluk: shorne_: Well, it looks like the c09635 assert is a
sanity check to make sure we didn't do anything stupid, and well.....
(13:44:57) pm215: migration_incoming_get_current() and
migration_incoming_state_destroy() seem a bit mismatched
(13:45:13) davidgiluk: pm215: Yep
(13:45:46) davidgiluk: pm215: Generally we've thought that an incoming
migration normally only happens once - shorne_'s case is the exception
(13:46:03) shorne_: pm215: yeah, it looked something like that I just had a
few seconds to look at today
(13:46:03) HariharanTS left the room (quit: Ping timeout: 480 seconds).
(13:46:03) shorne_ is now known as shorne
(13:48:05) shorne: davidgiluk: pm215: thanks for having a look.
Unfortunately I need to head off to bed and put kids to sleep
(13:49:11) davidgiluk: shorne: Sleep well, no nightmares about event
destroyers....
(13:49:30) davidgiluk: pm215: Yeh this is fall out from b4b076daf32
To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1712564/+subscriptions
