[OpenAFS] openafs panic after chroot

[emoy@apple.com]

On Tuesday, February 11, 2003, at 09:14  AM, Chaskiel M Grundman wrote:

> --On Tuesday, February 11, 2003 01:28:53 -0800 emoy@apple.com wrote:
>
>>  replaced the panic() call with a printf(), and going in and out of
>> chroot a couple of times resulted in the following from the system log
>> file:
> Firstly, if you replace the panic with a printf, you should avoid  
> calling
> osi_VM_TryReclaim if the if is triggered. TryReclaim might successfully
> dissociate the vnode from the VM object and drop the refcount to 0. if  
> it's
> really open, that would be bad.
>
> Assuming you have the time to try it (I do not), could you see if the
> following makes the "ui_refcount panic averted" messages go away:
> - comment out all calls to osi_VM_TryReclaim. leave the 'flushactive
> open...' printf in.
> - run afsd with a large -stat parameter (at least 2000. you may want  
> more)
> - Don't run this configuration for extended periods of time, since it  
> will
> eventually panic in NewVCache.
>
> (Basically, I'd like to know if the problem is the dodgy code in
> TryReclaim, or something more fundamental)

OK, I comment out all osi_VM_TryReclaim and set -stat to 3000.  In  
running chroot or dynamic libraries for AFS, no "ui_refcount panic  
averted" messages were seen.  However, the "flushactive open" message  
produced by chroot, continued to be logged every >1 minute.

I then tried commenting out just the osi_VM_TryReclaim associated with  
the "flushactive open" problem.  Again, the "flushactive open" message  
repeated after doing chroot.
------------------------------------------------------------------------ 
--
Edward Moy
Apple Computer, Inc.
emoy@apple.com

(This message is from me as a reader of this list, and not a statement
from Apple.)