panic: Memory modified after free

[Giorgos Keramidas]

On 2006-01-31 20:21, Steve Kargl <sgk at troutmask.apl.washington.edu> wrote:
> > Thing are looking positively horrid.  A kernel from a cvsup with
> > a date=2006.01.25.00.00.00 appears to work fine.  A kernel from with
> > date=2006.01.27.00.00.00 dies with
> >
> > dev_relthread() at dev_relthread+0x2e
> > devfs_close() at devfs_close+0x1b6
> > VOP_CLOSE_APV() at VOP_CLOSE_APV+0x74
> > vn_close() at vn_close+0x8d
> > vn_closefile() at vn_closefile+0x5a
> > fdrop_locked() at fdrop_locked+0xa1
> > closef() at closef+0x35f
> > fdfree() at fdfree+0x513
> > exit1() at exit1+0x360
> > sys_exit() 	at sys_exit+0xe
>
> a 2006.01.26.12.00.00.00 kernel dies a similar death.
> Unfortnately, this panic took out a portion of /usr/include
> and /usr/src.  I can't prove it yet, but I think the
> pts code may be the trigger.

With a kernel & userland built with DEBUG_FLAGS='-g' from last night, I
manager to kill my amd64 laptop completely.  The last kernel I had
working fine is from 25/Jan/2006, so something after this is that broke.

The bug here manifests as unusable vty's in multiuser mode, but things
seem to work relatively ok in single-user mode (i.e. I see no panics, yet).

I've reinstalled 5.4-RELEASE from a CD-ROM, so bringing this up to
CURRENT of 25-Jan will take a while though :(