On Fri, May 11, 2012 at 01:10:34AM -0400, David Miller wrote: > From: Mel Gorman <mgorman@xxxxxxx> > Date: Thu, 10 May 2012 14:54:14 +0100 > > > It could happen that all !SOCK_MEMALLOC sockets have buffered so > > much data that we're over the global rmem limit. This will prevent > > SOCK_MEMALLOC buffers from receiving data, which will prevent userspace > > from running, which is needed to reduce the buffered data. > > > > Fix this by exempting the SOCK_MEMALLOC sockets from the rmem limit. > > > > Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> > > Signed-off-by: Mel Gorman <mgorman@xxxxxxx> > > This introduces an invariant which I am not so sure is enforced. > > With this change it is absolutely required that once a socket > becomes SOCK_MEMALLOC it must never _ever_ lose that attribute. > This is effectively true. In the NFS case, the flag is cleared on swapoff after all the entries have been paged in. In the NBD case, SOCK_MEMALLOC is left set until the socket is destroyed. I'll update the changelog. -- Mel Gorman SUSE Labs -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
