Details of 2.2.13ac1 deadlock on Alpha/SMP

[Elliot Lee <sopwith redhat com>]

I can repeat the following deadlock by installing a bunch of .rpm's from
an NFS-mounted directory (the system sometimes reboots instead of giving
me the nice message). The message looks like:

	FILE1:NNN1 spinlock stuck in klogd at ADDR1(0) owner rpm at ADDR2(1) FILE2:NNN2

(stuff in caps is not literal)

Messages:
-------------------
sched.c:30 spinlock stuck in klogd at fffffc0000321e84 owner rpm at
fffffc0000321e84

signal.c:278 spinlock stuck in klogd at fffffc000031fb00(0) owner rpm at
fffffe000002c5f0(1)
clnt.c:219

sched.c:30 spinlock stuck in klogd at fffffc0000321e84(0) owner rpm at
fffffc0000321e84(1)
sched.c:30

signal.c:278 spinlock stuck in klogd at fffffc000031fb00(0) owner rpm at
fffffe000002c5f0(1)   
clnt.c:219

(The sched.c:30 ones are actually smplock.h:30, in the
reacquire_kernel_lock() inline function. This function is called at
sched.c:828).

FILE:NNN			Source code
-------------------------------------------
clnt.c:219			spin_lock_irqsave(&current->sigmask_lock, irqflags)
signal.c:278			spin_lock_irqsave(&t->sigmask_lock, flags)
smplock.h:30			spin_lock(&kernel_flag)

Any more information I can provide to help fix this problem?

-- Elliot
Do not meddle in the affairs of dragons,
for you are crunchy and good with ketchup.