Re: kern/68442: panic - acquiring duplicate lock of same type: "sleepq chain"

On Tuesday 29 June 2004 02:35 pm, Daniel Lang wrote:
> Hi Brian,
>
> Brian Fundakowski Feldman wrote on Tue, Jun 29, 2004 at 01:00:14PM -0400:
> [..]
>
> > You can find it out without using gdb, too.  This will work for
> > only main kernel symbols, but you can do something similar for
> > KLDs.  Say I want to find a symbol that's in the main kernel object:
> >
> > $ objdump -t /boot/kernel/kernel | ruby -ne 'fields = $_.split; if
> > fields[3] == ".text" and fields[2] == "F" and
> > 0xc048a800.between?(fields[0].hex, fields[0].hex + fields[4].hex) then
> > puts $_ end' c048a7ac l     F .text  0000006b cbb_removal
>
> [..]
>
> Thanks for that hint, but Colin suggested to use "addr2line", which
> produced some result. Maybe addr2line does a similar thing
> as your objdump/ruby script:
>
> So I have a line of code for the failing address:
>
> # addr2line -e kernel.debug 0xc053932b
> /usr/src/sys/kern/subr_witness.c:898
>
> which is (in my kernel):
>
> [..]
>         lock1 = &(*lock_list)->ll_children[(*lock_list)->ll_count - 1];
> [..]
>
> Well, I'm not sure if this is a big help. I doubt there is a bug
> in witness code.
>
> I'll cross-check with gdb to see if there is the same
> result. Too bad I couldn't get a crashdump.
>
> (What hurts most, is, that in one occasion I had a ddb prompt
> and could call doadump() successfully. But after reboot, damn
> /var was full, so savecore could not write it to disk, argl!).

Well, it may be a hint sadly enough.  The fact that it thinks Giant is a spin 
lock means that witness is confused, and this panic may be further such 
confusion.  One possibility is that somehow the sleep queue chain mutexes 
have been corrupted.

-- 
John Baldwin <jhb_at_FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org