Re: [bug] stuck localhost TCP connections, v2.6.26-rc3+

["Håkon Løvdal" <hlovdal@gmail.com>]

2008/6/11 Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>:
> On Wed, 11 Jun 2008, Håkon Løvdal wrote:
>> I do not have any good explanation (bad memory?). When checking the
>> copytest003.capture.bz2 file in the http server storage directory
>> I get the same results as you. However I also have the original file
>> stored still and comparing them with cmp -l gives
>>  20558656 124 104
>> so obviously a bit has been flipped.
>
> ...Yeah, which is a bad thing because now we don't know if that same error
> causes the TCP stops too.
>
>> I have now replaced the erroneous file with a good copy, so you should
>> be able to download again.
>
> I'll have a look if there's some common denominator but this whole issue
> might well be due to hw failure.
>

Hi. I tested the old pc with memtest86 and memtest86+ for 16 and 50
hours without any errors detected, so apparently the ram is ok.

I now just did another test from a freshley booted old pc, standing
idle for around 2 hours before I started the testing (no network activity
before I plugged in the cable just before starting testing).

The copying failed after just 2.62GB, so the logs are short and
nice. Stored on same server as earlier, new 2008-06-19 directory. One
thing I noticed in the dmesg output on the old pc was the following:

BUG: unable to handle kernel paging request at virtual address ffff3cff
printing eip: ffff3cff *pde = 00793067 *pte = 00000000
Oops: 0000 [#1] PREEMPT
Modules linked in: ...
Pid: 18103, comm: sshd Not tainted (2.6.24-gafd564a8 #4)
...
Call Trace:
 [<f8c45b39>] nf_conntrack_find_get+0x36/0x3a [nf_conntrack]
 [<f8c46379>] nf_conntrack_in+0x135/0x3e8 [nf_conntrack]
 [<c0423c5d>] local_bh_disable+0xa/0xb
 [<c0626d54>] _read_lock_bh+0x12/0x1a
...

BR Håkon Løvdal