Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 14 Aug 2004 22:04:53 -0700 (PDT)
From:      Doug White <dwhite@gumbysoft.com>
To:        current@freebsd.org
Subject:   traceback from hung system
Message-ID:  <20040814214821.S6429@carver.gumbysoft.com>

next in thread | raw e-mail | index | archive | help
Hey folks,

Hoping that Peter & other hardware people see this.  This is on a dual
Xeon 2.4GHz Dell PE1750, for reference.

It looks like the hangs that Robert and I have been experiencing recently
with buildworld are related to Scott & sandvine's problems with IPI
delivery.  I was able, with a hack from rwatson to avoid stop_cpus() when
going into ddb, to get into ddb and get a traceback:

stopped at      smp_tlb_shootdown+0x45: jb      smp_tlb_shootdown+0x3c
db> tr
smp_tlb_shootdown(f6,db5f7000,db5f8000) at smp_tlb_shootdown+0x45
smp_invlpg_range(db5f7000,db5f8000) at smp_invlpg_range+0x1c
pmap_invalidate_range(c0775de0,db5f7000,db5f8000,c227a000,c22774a4) at pmap_invalidate_range+0xb5
pmap_qenter(db5f7000,c227a010,1) at pmap_qenter+0x50
sf_buf_alloc(c1526388,0,0,0,0) at sf_buf_alloc+0x1a9
uiomove_fromphys(c3f05d58,0,27a5,dfe8cc88,0) at uiomove_fromphys+0x92
pipe_read(c287fdd0,dfe8cc88,c2c11080,0,c2c679a0) at pipe_read+0x238
dofileread(c2c679a0,c287fdd0,0,812a000,4000) at dofileread+0x95
read(c2c679a0,dfe8cd14,3,0,296) at read+0x3b
syscall(2f,2f,2f,80da500,80f7034) at syscall+0x287
Xint0x80_syscall() at Xint0x80_syscall+0x1f
--- syscall (3, FreeBSD ELF32, read), eip = 0x80a5783, esp = 0xbfbfe72c,
ebp = 0xbfbfe748 ---

Additional details and data collection at
http://www.gumbysoft.com/debug-20040814
(includes 'ps' and 'tr' output for all processes on CPUs.)

I can get the hang to come up easily enough, although I don't know how
reliable the jump to DDB is. I tried to get into gdb via firewire but it
wasn't working for me.

The variable smp_tlb_wait was set to 0 according to 'x smp_tlb_wait' in
ddb. ddb isn't one of my strong points so any hints on things to inspect
would be appreciated. I doubt a crashdump will work in this context.

I'm also going to try the same trick to get into my dual 600MHz P3, which
also hangs.

-- 
Doug White                    |  FreeBSD: The Power to Serve
dwhite@gumbysoft.com          |  www.FreeBSD.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040814214821.S6429>