i386/121148: Repeatable sysctl crash (Fatal Trap 12) with ACPI enabled

gavin at FreeBSD.org gavin at FreeBSD.org
Thu Feb 28 17:53:26 UTC 2008


Synopsis: Repeatable sysctl crash (Fatal Trap 12) with ACPI enabled

State-Changed-From-To: open->feedback
State-Changed-By: gavin
State-Changed-When: Thu Feb 28 15:34:11 UTC 2008
State-Changed-Why: 
To submitter:  Firstly, it looks to me like the commit that you narrowed
the panic down to is not actually responsible for the problem you are
seeing - my suspicion is that it actually just moves the layout of memory
around enough to avoid seeing the problem.

>From single user mode, can you determine which of the following panic:

sysctl dev.pcib.3
sysctl dev.pcib.4

(I'm guessing it's the latter, but it's worth checking).

Secondly, I wonder if you could test setting debug.acpi.disabled="ec" from
the loader, and see if that makes any difference?  I notice that the
"fault virtual address" is 0x2043455c, or " CE\", but this may be a
coincidence...

Lastly, are you able to recompile the kernel with debugging support
(options KDB and DDB), and also add printf's to
/usr/src/sys/kern/kern_sysctl.c at lines 618 and 630 (between the setting
of lsp and calling sysctl_sysctl_next_ls()) to show the value of the various
variables?  Something like this line should work:

printf("lsp=%p, oidp=%p, oidpp=%p\n", lsp, oidp, oidpp);

If you can still recreate the panic with these printf's and the debugger
compiled in, hopefully we can get more information out of your system as to
exactly what is happening.



Responsible-Changed-From-To: freebsd-i386->gavin
Responsible-Changed-By: gavin
Responsible-Changed-When: Thu Feb 28 15:34:11 UTC 2008
Responsible-Changed-Why: 
Track

http://www.freebsd.org/cgi/query-pr.cgi?pr=121148


More information about the freebsd-i386 mailing list