Re: panic: System call old.recv returning with 1 locks held

On Sun, Sep 10, 2006 at 11:00:03PM +0200, Alexander Leidinger wrote:
> Quoting Kris Kennaway <kris_at_obsecurity.org> (Sun, 10 Sep 2006 14:43:59 -0400):
> 
> > Enable DEBUG_LOCKS and DEBUG_VFS_LOCKS then run 'show lockedvnods'
> > from DDB to find out where the lockmgr lock was acquired.
> 
> I've enabled them (and WITNESS, INVARIANTS and DIAGNOSTICS) and rebuild
> the kernel. Not much files where rebuild.

That's a warning sign; almost every file in the kernel should have
been rebuilt when you turned on WITNESS and INVARIANTS.  Perhaps you
didn't run 'make depend'?

> show lockedvnods did not show any locks.

Odd, I thought your initial panic was from a held lockmgr lock, so it
should have showed up here.

> Other show XXX commands did
> show nothing too. show lockchain showed a process. It's send01 (a LTP
> testcase). The ddb bt command showed not much more. I got "syscall
> -90" (dup2?) and the registers eip=0x28120706, esp=0xbfbfe9fc and
> ebp=0xbfbfea28.
> 
> While rebooting I got another panic:

This panic also looks like a symptom of an inconsistent kernel:

> at ../../../i386/i386/trap.c:461 #6  0xc057fefa in calltrap ()
> at ../../../i386/i386/exception.s:138 #7  0xc04e3035 in vfs_filteropt
> (opts=0xc05b95d6, legal=0xc2960ae4) at ../../../kern/vfs_mount.c:1648
> #8  0xc2949dd9 in ?? () #9  0xc05b95d6 in ?? () #10 0xc2960ae4 in ?? ()
> #11 0x000003a1 in ?? () #12 0x00000008 in ?? () #13 0xc0607758 in
> w_lock_list_free () #14 0xc05b46a4 in ?? () #15 0xc05aff94 in ?? () #16
> 0xc2460010 in ?? () #17 0xd49498d8 in ?? () #18 0xc04ba3fb in
> fixup_filename (file=0xc05ef780 " ???^???")
> at ../../../kern/subr_witness.c:764 #19 0xc04e2244 in vfs_donmount
> (td=0xc26c66c0, fsflags=0xc292d000, fsoptions=0xd4949bb4)
> at ../../../kern/vfs_mount.c:947 #20 0xc04e3962 in kernel_mount
> (ma=0xc26c3610, flags=0x10000000) at pcpu.h:163 #21 0xc2949671 in ?? ()

Kris