blob: 95298e7cf5f2b28976ba2de45ef8d80fc4ddbe1a (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
|
graphic: 0.831
instruction: 0.733
device: 0.723
socket: 0.684
vnc: 0.682
assembly: 0.651
other: 0.642
network: 0.640
mistranslation: 0.592
semantic: 0.553
boot: 0.466
KVM: 0.410
loadvm fails twice in sequence
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 [~gpiccoli@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
Posted:
migration: Reset rather than destroy main_thread_load_event
snapshot/tests: Try loadvm twice
Commit 5089e1862fe80b6f23ba4c494e2902cbe3d9d48e
|