Re: VM and N-order allocation failed.

[Daniel Phillips]

On August 15, 2001 11:21 pm, David Ford wrote:
> I've a friend who has a small box doing routing/firewalling/nat for a 
> /27 and the bandwidth is pretty tiny.  The machine has 32megs of ram and 
> 100 in swap.
> 
> It dies anywhere from an hour after reboot to 12 hours later, on console 
> are messages like so "__alloc_pages: 0-order allocation failed." and 
> "ip_conntrack: table full, dropping packet."
> 
> His prior kernel was 2.4.7, nothing special, he's using 3com 
> 3cSOHO100-TX and Cpq Neteligent 10/100.  I just upped him to 2.4.8 as a 
> starter and we'll see how it's going.  I don't have direct access to the 
> box so I can't give complete information.
> 
> Comments and suggestions welcomed,

We should add the GFP_* allocation flags to that allocation failed
message so we know whether it's atomic, noio, nofs, highmem or what.

Please define 'it dies' a little more precisely - oops, deadlock,
livelock, app terminated or what.  Then please supply the /proc/meminfo
if possible or shift-ScrLk dump to console (roughly the same thing).

He needs to upgrade 2.4.9-preX because that is where the VM work is
being done.  Finally, it would be worth merging this thread with the
following thread on linux-mm:

  0-order allocation problem

--
Daniel