Serial console - how to reboot?

Eirik Oeverby ltning at
Tue Jun 15 06:19:22 GMT 2004

Doug White wrote:
> On Sat, 12 Jun 2004, Eirik Oeverby wrote:
>>I have recently been getting the following messages on my console:
>>swap_pager: indefinite wait buffer: device: #amrd/0x20001, blkno: 1072,
>>size: 4096
>>swap_pager: indefinite wait buffer: device: #amrd/0x20001, blkno: 832,
>>size: 4096
>>swap_pager: indefinite wait buffer: device: #amrd/0x20001, blkno: 512,
>>size: 4096
>>swap_pager: indefinite wait buffer: device: #amrd/0x20001, blkno: 576,
>>size: 4096
>>swap_pager: indefinite wait buffer: device: #amrd/0x20001, blkno: 584,
>>size: 4096
>>I am aware that this usually indicates a bad drive or something, but
>>given that it's a RAID volume hanging on an AMI MegaRAID controller
>>(SCSI), and none of the drives in the array are degraded, that can't
>>really be the case. I'm suspecting a bug in the driver (though freebsd
>>4.10 should be stable, right?) or a faulty cable or such.
> Yes, usually it means the device is not responding to requests.  What kind
> of container is your swap in? Mirror? Stripe?

Mirror. Two 10k SCSI drives, 36gb.
Tried with each of the drives connected in single, no difference. A
reboot solves the problem for another couple of days. What would be
interesting to know is if anything actually triggers this (somehting
that can be easily reproduced) or if it's truly 'random' in terms of
weak HW (cables etc.).

Any ideas?

>>What would mitigate the problem for now though would be to have a way to
>>reboot the system. I'm about 1000 miles away from the box, so I can't go
>>there, but I do have access via serial console (which is where I'm
>>getting these messages).
> [...]
>>Entering the kernel debugger might help, though I'm actually not sure if
>>it's compiled into the running kernel. But if it is, how can I get to it
>>via serial?
> If you built the kernel with 'options BREAK_TO_DEBUGGER', then a serial
> break will drop to ddb. From there you could do 'call boot(0)' to attempt
> and orderly shutdown, or 'reset' to, well, reset :) There's also an
> 'ALT_BREAK_TO_DEBUGGER' option that emulates the sun alt-break -- [cr] ~
> ^B.

Okay cool. Is this going to work on -STABLE (4.10) aswell? That would
make my life somewhat easier..

> Without either of those, though, maybe you should call someone nearby and
> entice them to press some buttons for you :)

Done, and done. Kudos to that person ;)

Thanks again,

More information about the freebsd-current mailing list