On 27/1/23 21:37, Joerg Roedel wrote:
On Fri, Jan 27, 2023 at 10:08:14AM +0100, Peter Zijlstra wrote:
Welcome to the wonderful shit show that is x86 exceptions :/
I thought sev_es_*() is supposed to fix this. Joerg, any clue?
Hmm, no, not yet, the stack-trace doesn't make much sense to me. The
sev_es_* function calls in the NMI path are for re-enabling NMI and
adjusting the #VC IST stack to allow nested VCs.
Alexey, can you try to get a more stable backtrace? For example by
building the kernel with frame pointers?
Do you mean these guys (below)?
aik@aik-Standard-PC-i440FX-PIIX-1996:~$ grep FRAME_POINTER
/boot/config-$(uname -r)
CONFIG_SCHED_OMIT_FRAME_POINTER=y
CONFIG_FRAME_POINTER=y
CONFIG_UNWINDER_FRAME_POINTER=y
Here is the complete output of that VM (200k so not in the email):
https://github.com/aik/linux/commit/d0d6bbb58fcd927ddd1f8e9d42ab121920c7eafc
Note that the original long backtrace appears more than once, I just did
not copy all of that in the first email.
--
Alexey
