[OpenAFS-port-darwin] OpenAFS 1.4.2 crashing on Intel Macs (10.4)

Ken Hornstein kenh@cmf.nrl.navy.mil
Mon, 29 Jan 2007 14:40:08 -0500


Howdy all,

We have a series of Intel desktops here at work that are crashing in
the OpenAFS client (1.4.2).  The amount of crashing varies; on some machines
it's around once per day (it can't be reliably reproduced, though).

Here is a backtrace from one of the crashes:

Mon Jan 29 11:49:32 2007
panic(cpu 3 caller 0x001CFD88): remove_fsref: no named reference
Backtrace, Format - Frame : Return Address (4 potential args on stack) 
0x485f39d4 : 0x128d1f (0x3c9540 0x485f39f8 0x131df4 0x0) 
0x485f3a14 : 0x1cfd88 (0x3d1a28 0x7615504 0x485f3a34 0x7615504) 
0x485f3a34 : 0x757f5ef8 (0x85d2f78 0x34 0x485f3a50 0x485f3a84) 
0x485f3a90 : 0x757f2a44 (0x758720f4 0x7f5aad4 0x485f3ef0 0x0) 
0x485f3ad8 : 0x1e2df1 (0x485f3af8 0x485f3f18 0x485f3b0c 0x1) 
0x485f3b28 : 0x1cab0b (0x7f5aad4 0x485f3ddc 0x485f3ef0 0x485f3f18) 
0x485f3b98 : 0x1cb331 (0x485f3dc4 0x6f40c10 0x26 0x400) 
0x485f3c48 : 0x1d7040 (0x485f3dc4 0x0 0x0 0x0) 
0x485f3d78 : 0x1d732c (0x45c284 0x0 0x0 0x0) 
0x485f3f38 : 0x1d73c1 (0x45c284 0x0 0x0 0x0) 
0x485f3f68 : 0x378337 (0x75d11f4 0x77465f8 0x774663c 0x0) 
0x485f3fc8 : 0x19acae (0x7f805e0 0x0 0x19d0b5 0x7593624) No mapping exists for frame pointer
Backtrace terminated-invalid frame pointer 0xb011ee18
      Kernel loadable modules in backtrace (with dependencies):
	       org.openafs.filesystems.afs(1.4.2)@0x75780000

Some sleuthing reveals to me that this is happening inside of the
function afs_darwin_finalizevnode; I suspect that this happens when
that function calls vnode_removefsref() (based on the panic message).
I don't know if this is merely Intel-specific, multiprocessor-specific,
or a combination of both.  I don't see this on our single or
multi-processor PowerPC machines.

Any ideas?

--Ken