[PVFS2-developers]

david.s.metheny at conwaycorp.net david.s.metheny at conwaycorp.net
Tue Jun 14 14:38:46 EDT 2005


Quoting robl at mcs.anl.gov (Robert Latham):

> On Tue, Jun 14, 2005 at 09:01:22AM -0500, david.s.metheny at conwaycorp.net
> wrote:
> > The s_op->scheduled_id = 0 statement would prevent the operation
> > from being released from the request scheduler, and thus it would
> > block other operations on the directory. I triggered it by
> > accidentally trying to create the "/" entry, which basically blocked
> > all other requests to the server at that point.
> 
> Ok, so removing the call to isprint() lets more characters through
> (any VFS-legal character). got it.  
> 
> I'm still a bit puzzled as to why removing the s_op_>scheduled_id is
> the correct thing to do.  Clearly it *works*, but I just want to
> understand it better.  final_response_release() is supposed to catch
> the places wehre s_op->scheduled_id is zero, right?  

I think what's happening here is the final_response_release checks to see if the
job has been scheduled. This checks the s_op->scheduled_id field. If it's zero,
it consideres that this is a job that has never been scheduled, and skips the
call to job_req_sched_release. This would leave the job in the request
scheduler. 

> 
> ==rob
> 
> -- 
> Rob Latham
> Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
> Argonne National Labs, IL USA                B29D F333 664A 4280 315B
> _______________________________________________
> PVFS2-developers mailing list
> PVFS2-developers at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
> 




-----------------------------------------------------------------------------
This message was sent using Conway Corporation WebMail -- www.conwaycorp.net


More information about the PVFS2-developers mailing list