[15092] in Athena Bugs
X/keyboard + panic + watchdog lossage on portnoy
daemon@ATHENA.MIT.EDU (John Hawkinson)
Sat Apr 19 01:24:16 1997
Date: Sat, 19 Apr 1997 01:23:48 -0400
To: bugs@MIT.EDU
Cc: sipb-staff@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
portnoy (an 8.0K sparc 5) was displaying a console login when I arrived, w/
suitable dm messages. ^D caused the screen to go white and then the
console login prompt would return.
/var/adm/X0msgs contained:
Fatal server error:
failed to initialize core devices
Failed to set keyclickKIOCLAYOUT: Timer expired
Please the configuration file for Input device entries
Fatal server error:
failed to initialize core devices
Failed to set keyclickKIOCLAYOUT: Timer expired
Please the configuration file for Input device entries
we moved the keyboard and it dropped to the ROM monitor.
We plugged the keyboard back in and typed go and
the machine paniced:
BAD TRAP: type=7 rp=f03c1bb4 addr=0 mmu_fsr=0 rw=0
sched: Memory address alignment
pid=0, pc=0xffd17b80, sp=0xf03c1c00, psr=0x44000c0, context=0
g1-g7: ffd24038, ffd15a00, ffef0000, ffef0c19, ffd19fee, ffefef14, ffefebc0
Begin traceback... sp = f03c1c00
Called from f007b8d8, fp=f03c1c60, args=ffd2329c f015f0d0 1 1 ffffffff 2
Called from f00cc0b0, fp=f03c1cc0, args=ffd2329c f015f0d0 0 44000e3 2 0
Called from f00cb94c, fp=f03c1d20, args=fffffffe fc082f78 fc082f78 1 1 f015f0d0
Called from f0054118, fp=f03c1d98, args=fffffffe fc082f78 1 0 f015f0d0 f03c1df8
Called from f00842b8, fp=f03c1e00, args=f03c1ec0 f0172fa0 fc1dece0 f01878c8 f016efe4 0
Called from f0058858, fp=f03c1e60, args=f0172fa0 0 f01878c8 f01878c8 0 f0172fe0
Called from f0084264, fp=0, args=0 0 73640000 2f6f6269 6f2f6e63 72403231
End traceback...
panic: Memory address alignment
While sync-ing before writing a crash dump it paused and then
did a watchdog reset. I entered the FORTH version of ps from
/afs/sipb/user/jhawk/src/openprom/ps.forth and observed no unusual
processes in the process table.
I couldn't find a way to make the machine dump to disk.
"0 set-pc go" produced "No active program". I suspect
I could have done something with the openprom, but I failed.
I wrote /mit/watchmaker/arch/sun4m_54/bin/unsym, and fed the above
output into it, and it decodes as:
BAD TRAP: type=7 rp=f03c1bb4 addr=0 mmu_fsr=0 rw=0
sched: Memory address alignment
pid=0, pc=0xffd17b80, sp=0xf03c1c00, psr=0x44000c0, context=0
g1-g7: ffd24038, ffd15a00, ffef0000, ffef0c19, ffd19fee, ffefef14, ffefebc0
Begin traceback... sp = f03c1c00
Called from [f007b8d8=prom_getproplen+40], fp=f03c1c60, args=ffd2329c
[f015f0d0=cpr_info+1aa8] 1 1 [ffffffff=VADDR_MASK_DEBUG] 2
Called from [f00cc0b0=ddi_bus_prop_op+40], fp=f03c1cc0, args=ffd2329c
[f015f0d0=cpr_info+1aa8] 0 44000e3 2 0
Called from [f00cb94c=ddi_prop_op+2f0], fp=f03c1d20, args=[fffffffe=-2]
[fc082f78=sadminqp+11e4] [fc082f78=sadminqp+11e4] 1 1
[f015f0d0=cpr_info+1aa8]
Called from [f0054118=init_swift_idle_cpu+30], fp=f03c1d98,
args=[fffffffe=-2] [fc082f78=sadminqp+11e4] 1 0 [f015f0d0=cpr_info+1aa8]
f03c1df8
Called from [f00842b8=idle+54], fp=f03c1e00, args=f03c1ec0
[f0172fa0=cpus] [fc1dece0=ttcoinfo+dec] [f01878c8=p0]
[f016efe4=sleepq_head+190] 0
Called from [f0058858=thread_start+4], fp=f03c1e60, args=[f0172fa0=cpus] 0
[f01878c8=p0] [f01878c8=p0] 0 [f0172fe0=cpus+40]
Called from [f0084264=idle], fp=0, args=0 0 73640000 2f6f6269 6f2f6e63 72403231
End traceback...
panic: Memory address alignment
Just the stack trace itself:
Called from [f007b8d8=prom_getproplen+40]
Called from [f00cc0b0=ddi_bus_prop_op+40]
Called from [f00cb94c=ddi_prop_op+2f0]
Called from [f0054118=init_swift_idle_cpu+30]
Called from [f00842b8=idle+54]
Called from [f0058858=thread_start+4]
Called from [f0084264=idle]
That's about it, there wasn't much more data to be had, I don't
think.
I need to consider writing some code for the PROM to allow you to
send output to the message buffer (where it can be picked up by
dmesg once the machine has booted). You might want to do this
with "ps", for instance.
I don't think the above panic is very normal. It seems to have something
to do with the idle loop and also monitor properties. Ideally it should
be reported to Sun to see if they have any bright ideas.
--jhawk
