Re: panic: sbuf_trim makes no sense on sbuf 0xffffff82434d8898 with drain

On Sun, Nov 25, 2012 at 2:26 PM, Niclas Zeising
<zeising+freebsd_at_daemonic.se> wrote:
> Hi!
> I consistently get this panic while trying to boot a kernel build from
> r243530.  It happens when the entropy harvesting rc.d script starts. r243380
> worked fine, I haven't tested any revisions in between. Attached is the
> backtrace from the kernel, as gotten by kgdb.  The machine uses zfs as a
> root pool, and there have been churn in this area.  To my untrained eyes,
> however, the issue seem related to hdaa.c. Please let me know if I can
> provide any more information.

r243530 added the new sysctl that is causing panic.  I'm not sure why
there's an sbuf_trim() call; there shouldn't be more than a few \n at
the end.  IMO the sbuf_trim() can be eliminated.

Alternately, the panic check can be removed and we could allow
sbuf_trim() to remove any un-emitted whitespace for an sbuf with
drain.

CC'ing mav_at_ who introduced the code.  (I introduced sbuf drains).

Thanks,
matthew