blob: ac75f355dc49e14b900c7bab6cdb7866d7fa8870 (
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
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
x86: 0.795
device: 0.587
debug: 0.583
architecture: 0.564
mistranslation: 0.427
semantic: 0.372
performance: 0.369
ppc: 0.357
graphic: 0.294
boot: 0.288
virtual: 0.246
i386: 0.245
user-level: 0.240
socket: 0.229
PID: 0.221
register: 0.211
kernel: 0.205
hypervisor: 0.166
risc-v: 0.164
permissions: 0.161
vnc: 0.144
arm: 0.140
network: 0.124
files: 0.122
peripherals: 0.111
VMM: 0.090
TCG: 0.083
assembly: 0.078
KVM: 0.072
SMI trigger causes hang with multiple cores
When using qemu , SMI trigger causes hand/reboot under following conditions:
1. No KVM but there are more than 1 threads (-smp > 1)
2. When using KVM.
Info:
qemu-system-x86_64 --version
QEMU emulator version 2.11.1(Debian 1:2.11+dfsg-1ubuntu7.29)
Copyright (c) 2003-2017 Fabrice Bellard and the QEMU Project developers
SMI trigger was done by writing 0x00 in IO port 0xB2.
Does coreboot do anything to set up an SMI handler? Does it relocate SMBASE for all processors?
Misbehavior upon raising an SMI is fully expected, unless the guest (usually the guest firmware) sets up SMI handling properly.
The bug report currently includes only two bits of information about guest actions, namely "coreboot.rom" and "writing 0x00 in IO port 0xB2". Thus far a guest crash looks entirely reasonable to me.
Did you intend to attach "1.txt"?
I tried without specifying -bios parameter still hang is seen. But this time it had low memory corruption.
And built seabios with more debug logs but seabios doesn't does SMM init even when its selected in make menuconfig.
I guess fundamentally th issue is writing 0xXX in IO port 0xB2 should trigger SMI handler in all possible core but instead it triggers SMI only in Core#0.
> I guess fundamentally th issue is writing 0xXX in IO port 0xB2 should
> trigger SMI handler in all possible core but instead it triggers SMI
> only in Core#0.
For that, the guest needs to negotiate the "broadcast SMI" feature with
QEMU. See commit range 57bb40c9db40..b8bab8eb6934.
Inactive for ~two weeks, closing.
|