VM and N-order allocation failed.
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,
David
Quoted 8 lines · David Ford
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."
Quoted 6 lines
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