[Crash-utility] [PATCH] x86_64: do not process zero size exception stack

HAGIO KAZUHITO(萩尾 一仁) k-hagio-ab at nec.com
Wed Nov 11 08:05:07 UTC 2020


-----Original Message-----
> Hi Kazu,
> >
> > Apparently the following patch causes this and it's in 5.10-rc ?
> >
> > commit 02772fb9b68e6a72a5e17f994048df832fe2b15e
> > Author: Joerg Roedel <jroedel at suse.de>
> > Date:   Mon Sep 7 15:15:43 2020 +0200
> >
> >    x86/sev-es: Allocate and map an IST stack for #VC handler
> >
> > If so, it seems we should support this properly.
> >
> Not this one. I use 5.9.6 kernel for testing. It does not use VC handlers ( no SEV-ES support yet).
> First look in the kernel - didn’t show anything that sets 5th stack. Address seems valid. Will dig
> deeper. Might need to take a look at crash code, around ebase[] fetching.
> Will send update on my findings.

ok, thanks.

I thought the above because 5.9.4 vmcore on hand doesn't show the 5th address
and 5.10.0-rc1 one does:

5.9.4:
crash> help -m
...
   [0]: fffffe0000009000 fffffe000000b000 fffffe000000d000 fffffe000000f000 
        0000000000000000 0000000000000000 0000000000000000

5.10.0-rc1:
   [0]: fffffe0000009000 fffffe000000b000 fffffe000000d000 fffffe000000f000
        fffffe0000012000 0000000000000000 0000000000000000

> 
> > btw, hopefully we would like to release the new version (7.2.9) of crash
> > next week and are preparing, so some patches might be deferred to the next
> > (7.3.0 ?), especially patches for Linux 5.10.
> Got it. Thanks for update.
> My bigger patch that I've sent Today should definitely go post 7.2.9.

Yes, that will.  Thanks for understanding.

Kazu





More information about the Crash-utility mailing list