[OpenAFS] deadlock in OpenAFS 1.4.11 (Solaris 5.10)

[Derrick Brashear]

On Fri, Apr 9, 2010 at 1:36 PM, John Tang Boyland
<boyland@pabst.cs.uwm.edu> wrote:
> ] cmdebug or it didn't happen.
> ]
> ] On Fri, Apr 9, 2010 at 12:08 PM, John Tang Boyland
> ] <boyland@pabst.cs.uwm.edu> wrote:
> ] > We get an occasional deadlock happening on Solaris 5.10 using
> ] > OpenAFS 1.4.11. =A0After the problem starts, any attempt to use AFS
> ] > on the machine freezes: =A0For example:
>
> I foolishly thought that with every AFS access deadlocking, cmdebug
> wouldn't work. =A0But it does....

That's the best time to try it.


>
> ** Cache entry @ 0xa26da3f0 for 1.536875155.530.1418 [cs.uwm.edu]
> =A0 =A0locks: (reader_waiting, write_locked(pid:17732 at:250), 2 waiters)
> =A0 =A0 =A0 =A0 =A0 26532 bytes =A0DV =A0 =A0 =A0 =A0 =A0197 =A0refcnt =
=A0 =A0 3
> =A0 =A0callback 00000000 =A0 expires 1270781645
> =A0 =A01 opens =A0 =A0 0 writers
> =A0 =A0normal file
> =A0 =A0states (0x0)

ok, that's the rdwr vnode op, so that makes some sense.

> ** Cache entry @ 0xa27651d0 for 1.536875155.608.1458 [cs.uwm.edu]
> =A0 =A0locks: (upgrade_waiting, write_locked(pid:17679 at:66), 12 waiters=
)
> =A0 =A0 =A0 =A019094744 bytes =A0DV =A0 =A0 =A0 =A0 =A0109 =A0refcnt =A0 =
=A013
> =A0 =A0callback 00000000 =A0 expires 1270782798
> =A0 =A00 opens =A0 =A0 0 writers
> =A0 =A0normal file
> =A0 =A0states (0x0)

and 66 is GetDCache

> ** Cache entry @ 0xa2710018 for 1.536875155.610.1499 [cs.uwm.edu]
> =A0 =A0locks: (none_waiting, write_locked(pid:17889 at:250))
> =A0 =A0 =A0 =A0 =A0 23240 bytes =A0DV =A0 =A0 =A0 =A0 =A0 =A01 =A0refcnt =
=A0 =A0 1
> =A0 =A0callback 00000000 =A0 expires 1270782798
> =A0 =A01 opens =A0 =A0 0 writers
> =A0 =A0normal file
> =A0 =A0states (0x0)
> ** Cache entry @ 0xa26bf000 for 1.536873892.1.1 [cs.uwm.edu]
> =A0 =A0locks: (writer_waiting, 7 read_locks(pid:18421), 41 waiters)
> =A0 =A0 =A0 =A0 =A0 =A02048 bytes =A0DV =A0 =A0 =A0 =A0 =A0 =A08 =A0refcn=
t =A0 =A049
> =A0 =A0callback 00000000 =A0 expires 1270774475
> =A0 =A00 opens =A0 =A0 0 writers
> =A0 =A0volume root
> =A0 =A0states (0x4), read-only
> ** Cache entry @ 0xa27c99a0 for 1.536874783.4234.9047 [cs.uwm.edu]
> =A0 =A0locks: (none_waiting, write_locked(pid:17834 at:250))
> =A0 =A0 =A0 =A0 =A0 =A01641 bytes =A0DV =A0 =A0 =A0 =A0 =A0 =A01 =A0refcn=
t =A0 =A0 1
> =A0 =A0callback 00000000 =A0 expires 1270782797
> =A0 =A01 opens =A0 =A0 0 writers
> =A0 =A0normal file
> =A0 =A0states (0x0)

Can you get the fids of the  files in question?





--=20
Derrick