Re: Oops in 2.6.21-rc4, 2.6.23

[Jarek Poplawski <jarkao2@o2.pl>]

On 15-10-2007 17:39, Darko K. wrote:
> Hello,
> 
> after recent upgrade to kernel 2.6.23 (from 2.6.20) I have started
> seeing kernel oops-es in networking code. The problem is 100%
> reproducible in my environment. I've seen two slightly different
> backtraces but both seem to be caused by the same commit. Please
> put me on Cc: on replies.
> 
> I've performed the git bisect and tracked down the problem to the
> commit:
> 
> 53cdcc04c1e85d4e423b2822b66149b6f2e52c2c [TCP]: Fix tcp_mem[]
> initialization
> 
> Once I reverse this commit in 2.6.23 the problem goes away (this
> is true also for the kernel version generated by git bisect,
> 2.6.21-rc4).
> 
> Description of my machine: Athlon XP1700, VIA KT-333 chipset, 512MB
> RAM, VIA rhine on-board NIC. The problem appears when approximately
> 70% of RAM is in use and I start a large file transfer (via FTP)
> from another machine on the LAN. Sample output from free command:
>        total   used free   shared buffers cached
> Mem:  516508 510600 5908        0    1004 154084
> -/+ buffers/cache:     355512     160996
> Swap: 899600  25240 874360
> 
> Backtrace #1:
> page allocation failure. order:1, mode:0x20
>  [<c0131581>] __alloc_pages+0x2e1/0x300   
>  [<c0144bee>] cache_alloc_refill+0x29e/0x4b0
>  [<c0144e6e>] __kmalloc+0x6e/0x80
>  [<c0227103>] __alloc_skb+0x53/0x110
>  [<c024de5c>] tcp_collapse+0x1ac/0x370

Hi,

I hope you've found this by yourself by now, but:

1. These are warnings only - not oopses.
2. It seems this patch you've found to be responsible for this all
slightly changes some limits, which is not necessarily wrong; the
lucky thing is you can change these limits with sysctl or some
echos to /proc/sys/net/ipv4/tcp_mem, I presume. (Or more safely by
adding some memory.)

Regards,
Jarek P.