[OpenAFS-devel] Re: ptserver coredumps (1.5.78 on FreeBSD 8.1)

[Marcus Watts]

Various wrote:
> Date:    Wed, 08 Dec 2010 10:12:20 CST
> To:      openafs-devel@openafs.org
> From:    Andrew Deason <adeason@sinenomine.net>
> Subject: [OpenAFS-devel] Re: ptserver coredumps (1.5.78 on FreeBSD 8.1)
> 
> On Wed, 8 Dec 2010 14:22:42 +0100
> u-openafsdev-njsf@aetey.se wrote:
> 
> > Ok, I wanted to give FreeBSD an extra chance.
> 
> Well, if you're running 1.5, my first guess would be it's the 1.5 branch
> that's broken, not FreeBSD. But the rest of the email suggests we're
> just broken on FreeBSD :)
> 
> > Compiled 1.4.12.1 which needed some tweaking to go through
> > but in the end I see:
> 
> Can you share what that tweaking was? I assume you mean you needed to do
> some extra steps to get it to compile?
> 
> > (gdb) bt
> > #0  0x0000000000415570 in free_map ()
> > #1  0x00000000004156a3 in bic_bitmap ()
> > #2  0x0000000000407f13 in GetListSG2 ()
> > #3  0x00000000004082f2 in GetList ()
> > #4  0x000000000040b9ba in getCPS ()
> > #5  0x000000000040bded in SPR_GetCPS ()
> > #6  0x0000000000412a8a in PR_ExecuteRequest ()
> > #7  0x0000000000439e0a in rxi_ServerProc ()
> > #8  0x000000000043187d in rx_ServerProc ()
> 
> Can you try running without supergroups enabled, and see if it works?
> I'm not saying that's the real solution, but it would help to know if
> it's just supergroups that's broken. It wouldn't surprise me too greatly
> if some platform-specific quirk makes it break.
> 
> And this might be a similar issue as in 1.5, as it's hard to tell what
> the deal was in the 1.5 stack trace, as it didn't make much sense.
> 
> > I get a feeling that FreeBSD is hardly well supported, not even as
> > a server platform? Fime with me but I will miss zfs if I have
> > to switch to Linux.
> 
> I'm not aware of many people that run OpenAFS servers on FreeBSD, but
> we'll still listen to and fix bug reports.
> 
> -- 
> Andrew Deason
> adeason@sinenomine.net
> 
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel

ptserver for 1.4.12x should be using lwp.  Stack thread size matters.
Try increasing
RX_DEFAULT_STACK_SIZE

After that, try building/running with -g - or failing that, besides the stack
backtrace try to include register values and enough context surrounding
the failing instruction to understand what died.  For this case, free_map
isn't very big, so just disassemblying the whole thing should suffice.
Wouldn't hurt to do that anyways.

I take it you're not just doing freebsd, you're doing freebsd/amd64?
If you're doing something more exotic like ppc64 there may be add'l issues.
You might want to try other versions of freebsd in any case.

			-Marcus Watts