summary refs log tree commit diff stats
path: root/results/classifier/108/none/1888971
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/108/none/1888971')
-rw-r--r--results/classifier/108/none/188897153
1 files changed, 53 insertions, 0 deletions
diff --git a/results/classifier/108/none/1888971 b/results/classifier/108/none/1888971
new file mode 100644
index 000000000..ef4f97b8f
--- /dev/null
+++ b/results/classifier/108/none/1888971
@@ -0,0 +1,53 @@
+device: 0.587
+debug: 0.583
+other: 0.419
+semantic: 0.372
+performance: 0.369
+graphic: 0.294
+boot: 0.288
+socket: 0.229
+PID: 0.221
+permissions: 0.161
+vnc: 0.144
+network: 0.124
+files: 0.122
+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.
+