Re: [PATCH 00/11] New Caching mechanism for ib_core

[Goldwyn Rodrigues <rgoldwyn-l3A5Bk7waGM@public.gmane.org>]

Hi Sean,

On Thu, Nov 10, 2011 at 07:03:40AM +0000, Hefty, Sean wrote:
> > > Yes, but the use of the cache is hidden from the user.
> > 
> > user who?
> 
> ib_cm, rdma_cm, ib_mad, ib_ipoib, etc.  You would need to trace the use up to all ULPs.
> 
> > Since the drivers would be users of cache calls, the behavior would be
> > as assumed for the query calls for the driver. Refer mthca
> > implementation (PATCH 11/11): If the entry is found in cache, it
> > will return immediately, and will not sleep.
> 
> > If the entry is not found, it will query the device and may sleep.
> 
> This is the issue.  You changed calls from ib_find_cached_pkey() and ib_find_cached_gid() to ib_query_pkey() and ib_query_gid().  The latter calls may sleep, but I don't believe that we can sleep in all places where the new functions are used.  Changing from ib_find_cached_gid to ib_query_gid changes the locking restrictions up through the stack.  It doesn't matter if the cache doesn't block, the exposed calls may.

Thanks for explaining. I see your point now. This had been in the
orignal discussion, but I missed on this point while working
on this. I will redo the patches after reviewing the function usage.

-- 
Goldwyn
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html