Which bits of s10 are you running? What is the system type?
Could you boot the system in interactive mode using kmdb and after
main is entered could you check the value of 'mtc_off' and in case
it is set, reset it to '0' as:
kmdb[0]> mtc_off/W 0
I believe system should come up fine then.
cheers.
C, Kiran (STSD) wrote:
> Hi,
> Yeah, the whole is system is hung..
> When you do a reboot.. It is hung at a point and it is not coming up
> after that..
>
> When I do ::filndlocks it Is showing a thread which is holding the
> lock.. When I check the stack trace of this thread
> there are various functions which is blocked and the thread is waiting
> for cv to be released.
>
> ::findlocks
> 1852a38 (struct di_cache.cache_lock) is owned by 300025d4640
> findlocks: nota bene: other locks may be held
>
>>300025d4640::findstack
>
> stack pointer for thread 300025d4640: 2a1005a0c41
> [ 000002a1005a0c41 cv_wait+0x38() ]
> 000002a1005a0cf1 mt_config_fini+0x20()
> 000002a1005a0da1 devi_config_common+0x118()
> 000002a1005a0e51 di_copytree+0x50()
> 000002a1005a0f01 di_snapshot+0xf0()
> 000002a1005a1011 di_cache_update+0x28()
> 000002a1005a10c1 di_ioctl+0x564()
> 000002a1005a1191 ioctl+0x184()
> 000002a1005a12e1 syscall_trap32+0xcc()
>
>
> I am stuck at this point. Now my doubt is who is holding the lock and
> how to release it..
>
> Alexander, "You need to look at stack traces for threads and see whether
> something looks suspicious. "
>
> I am really sorry I did not get what you meant by this statement. Could
> be more specific
>
> Cheers
> Kiran
>
>
> -----Original Message-----
> From: Alexander Kolbasov [mailto:akolb at eng.sun.com]
> Sent: Wednesday, September 06, 2006 3:33 AM
> To: C, Kiran (STSD)
> Cc: mdb-discuss at opensolaris.org
> Subject: Re: [mdb-discuss] How to debug a Kernel Hang using mdb???
>
>
>>Hi,
>>
>>
>>I have a Solaris 10 system.. and it hanged I initiated a forced panic
>
> using "sync" I have a core dump when i do $c it shows the stack
> backtrace. and i can see it is hanging in a solaris system call..
>
>>Now my question is how to debug the kernel hang... I am suspecting a
>
> locking issue in one of the functions.. for the system hang..
>
>>Is there any command to see the locking issues and not being unlocked
>>in any functions (doing a mutex_enter and not exiting)
>>
>>is there any commands in mdb where in i can find the pssible reason
>
> for the system hang..
>
> When investigating system hangs you need to figure out
>
> - Is it the whole system that is hang or some processes?
> - Is there any forward progress?
>
> Before you force the core dump you may try running ::cpuinfo -v command
>
> It will show you what threads are on CPUs and on run queues. You may use
> ::findstack to get stack information for these threads.
>
> The $c will not show you anything useful if the core dump was produced
> by sync.
> You need to look at stack traces for threads and see whether something
> looks suspicious.
>
> To check for possible locking issues you can do
>
> ::typegraph
>
> followed by
>
> ::findlocks
>
> This will show kernel locks hold, you may check stack traces of lock
> owners.
>
> - Alexander Kolbasov
>
> _______________________________________________
> mdb-discuss mailing list
> mdb-discuss at opensolaris.org
--
http://blogs.sun.com/sprakki
