[K42-discussion] DirLinuxFSVolatile deadlock

[Patrick Bozeman]

I'm triggering it with /proc using code you don't have yet.  I'll  
send instructions on how to trigger it with my next patch.

The thing that makes it unclear about the locks for me is that they  
are on the nameholder and not the file itself.  Currently, I am under  
the assumption that the dirlock is required to grab a namholder lock,  
and then the dirlock can be released.  In this case, the nameholder  
points to a stale object.  It seems that they would both trigger the  
invalidation logic prior to releasing the dir lock..  Although, I  
haven't studied and reasoned through those paths enough, and  
definitely not across all the fs ops that can trigger the code below.

On Sep 21, 2006, at 10:00 AM, Dilma DaSilva wrote:

>
> Patrick,
>
> I'll look at the deadlock you observed, and also try to document why
> the locks are necessary.
>
> I would like to understand how you're triggering invalidation of
> directories. This can be triggered if the files/directories in the
> remote NFS server are changed as k42 is running and caching these
> entities. Is this the scenario you're exercising?
>
> Dilma
> _______________________________________________
> K42-discussion mailing list
> K42-discussion at ozlabs.org
> https://ozlabs.org/mailman/listinfo/k42-discussion