Re: wedged machine in disklabel run from /etc/daily

[mlelstv%serpens.de@localhost (Michael van Elst)]

brook%nmsu.edu@localhost (Brook Milligan) writes:


>> On May 13, 2019, at 3:43 AM, Michael van Elst <mlelstv%serpens.de@localhost> wrote:
>> 
>> brook%nmsu.edu@localhost (Brook Milligan) writes:
>> 
>>> # ps -lp 4904
>>> UID  PID PPID  CPU PRI NI  VSZ  RSS WCHAN  STAT TTY    TIME COMMAND
>>> 0 4904    1 1968 127  0 9924 1160 tstile D    ?   0:00.00 disklabel sd0 
>> 
>>> What exactly does the 'tstile' mean?
>> 
>> It means, that the process is waiting for a mutex. There is no information on
>> what mutex it is waiting on, probably something related to a disk driver.
>> 
>> You can use the 'crash' utility or the DDB kernel debugger to get a
>> backtrace. This shows where the process is waiting, which is a good
>> hint on what it is waiting for.

>What is your suggestion for how to use 'crash'?  I have never done that so I am not sure what commands are available once I run it.  Running ps in crash does not seem to reveal much, but I'm not really sure how to interpret all the hex information.

There is a backtrace command, like (the 0t means 'decimal number'):

bt/t 0t$PID

or

bt/a $LWPADDRESS

the LWPADDRESS can be found in the ps output ("STRUCT LWP *").

You only need the latter to select between different threads of a process
or to select a kernel thread. That's not necessary for disklabel.



>This sounds like the only option is to get someone to power cycle it.  Is that correct?  It would still be nice to know what happened so I can understand what to do in the future; hence my questions about troubleshooting.

You might try to reboot, but this may then hang completely and then you still
need a hard reset. A 'reboot -n' that doesn't try to sync disks probably works.



-- 
-- 
                                Michael van Elst
Internet: mlelstv%serpens.de@localhost
                                "A potential Snark may lurk in every tree."