Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 02 Jan 2007 16:39:51 +0000
From:      Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>
To:        Jeremy Chadwick <koitsu@freebsd.org>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: Interrupt (SCSI?) hang on 4.x
Message-ID:  <1167755991.84652.6.camel@buffy.york.ac.uk>
In-Reply-To: <20070102153608.GA78405@icarus.home.lan>
References:  <20070102153608.GA78405@icarus.home.lan>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2007-01-02 at 07:36 -0800, Jeremy Chadwick wrote:
> Yes, I know 4.11 is EOL'd at the end of this month, but hopefully
> someone can shed some light on this problem anyways.  I simply don't
> have the knowledge of what's going on on a low-level to determine
> the cause.
> 
> I do have serial console on this box, and after enabling some
> debugging for the ahc(4) driver a few months back, was able to
> get something intelligent out of the system regarding SCBs this
> morning.  This may not be useful (or the cause), though.  I also
> cannot enable drop-to-DDB-on-serial-break because our Portmaster 2
> has been known to send a serial break on rare occasion.  :-(
> 
> Every so often (sometimes hours, sometimes months -- usually months),
> the 4.11 box we have "locks up" in the sense that both NICs on the
> box stop working, and the SCSI controller also appears hung.  This
> problem has existed for a couple years; it's not specific to 4.11
> (versus 4.10 or 4.9).

> 
> # vmstat -i
> ata0 irq14                      6          0
> fxp0 irq10                  14874         28
> mux irq11                   65028        125
> fdc0 irq6                       1          0
> sio0 irq4                     948          1
> clk irq0                   516187        998
> rtc irq8                    66071        127
> Total                      663115       1282

Do any of these numbers continue to increase after the hang?  You may
find that if you are already logged in over the serial port before the
hang and have run vmstat recently, it'll still be runnable due to it
being cached.

If the serial port is dead, you will probably still find you can get
output from the serial port, so start "date; vmstat -i" in a loop over
the serial port before it hangs, and watch the output once it wedges.

Gavin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1167755991.84652.6.camel>