On Sun, 26 Jul 98 20:56 BST, alan@lxorguk.ukuu.org.uk (Alan Cox) said:
> 4Mb 8KNFS kernel builds generally fail
> 4KNFS Except when being attacked or very heavily loaded
> kernel builds work fine. Under extreme attack
> I can about force I/O errors and NFS deadlocks
> 1K NFS Builds fine
With the patch to fail free_memory_available if a recent high order
allocation failed, 4MB 4k and 8k nfs works here and recovers just fine
after attack.
Is this important enough to address? A strict time limit on the error
condition will eliminate the risk of a single allocation failure sending
us into swap storm, but if we have persistant errors from a component
which retries (eg. NFS) then we either accept a lockup of that component
or we keep trying until something unsticks. Without such a mechanism,
large-packet ping attacks can really screw up a box which has /usr or /
NFS-mounted, since stalling NFS on such a box can freeze the background
VM activity which might otherwise eventually free us a higher order
page.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html