problems with soft-nfs when the server goes down

[Oliver Lehmann]

Mohan Srinivasan wrote:

> From the tcpdump, it looks like the read reply was sent by 
> the server and received by the client, but for some reason
> not delivered to the waiting process.
> 
> To say why that might have happened, it would really be 
> good to get a kernel crash dump.
> 
> To get a crash dump, you'd need to drop into KDB and type
> in "doadump". That will dump the core. The core will be
> saved upon reboot.
> 
> Then you need to point me at the coredump and the corresponding
> unstripped kernel image.
> 
> mohan

Now, compiled with all the debugging stuff in the kernel, I've seen a LOR:


pcm0: <ESS 18xx DSP> on sbc0
pcm0: [GIANT-LOCKED]
lock order reversal
 1st 0xc12a84c0 sbc0 (sound softc) @ /usr/src/sys/dev/sound/isa/sbc.c:131
 2nd 0xc13bf940 pcm0:play:0 (pcm play channel) @ /usr/src/sys/dev/sound/pcm/channel.c:525
KDB: stack backtrace:
kdb_backtrace(c06d46d5,c13bf940,c137c7d4,c06c76dc,c06c772a) at kdb_backtrace+0x2e
witness_checkorder(c13bf940,9,c06c772a,20d,c5bd2c8c) at witness_checkorder+0x6c3
_mtx_lock_flags(c13bf940,0,c06c772a,20d,c13ee580) at _mtx_lock_flags+0x8a
chn_intr(c137c780,c04f8d3a,0,c1295a58,c11a5000) at chn_intr+0x2d
ess_intr(c13ee580,c12a88c0,c11a0b00,c5bd2d00,c04eb6c2) at ess_intr+0x18d
sbc_intr(c1295a58,0,c06ce58d,220,c5bd2d00) at sbc_intr+0x24
ithread_loop(c11a0b00,c5bd2d38,c06ce384,30d,3c89ffff) at ithread_loop+0x162
fork_exit(c04eb560,c11a0b00,c5bd2d38) at fork_exit+0xc1
fork_trampoline() at fork_trampoline+0x8
--- trap 0x1, eip = 0, esp = 0xc5bd2d6c, ebp = 0 ---
nfs server 10.0.1.251:/usr: not responding



root at curry CURRY> kgdb kernel.debug /var/crash/vmcore.0
kgdb: kvm_read: invalid address (150011)
kgdb: kvm_read: invalid address (bfbfe0f8)
kgdb: kvm_read: invalid address (129)
kgdb: kvm_read: invalid address (129)

doesn't looks like it works....

-- 
 Oliver Lehmann
  http://www.pofo.de/
  http://wishlist.ans-netz.de/