Re: 2.2.17 --- extreme format string weirdness in /proc

From: Jesse Pollard (pollard@cats-chateau.net)
Date: Sat Sep 30 2000 - 16:00:07 EST


On Sat, 30 Sep 2000, Nix wrote:
>Andries Brouwer <aeb@veritas.com> writes:
>
>> On Sat, Sep 30, 2000 at 03:09:10PM +0100, Nix wrote:
>>
>> > Yesterday, I noticed that netstat had stopped working on my 2.2.17 box
>> > The reason is fairly self-evident:
>> >
>> > : loki:/# cat /proc/net/dev
>> ...
>> > : lo:%lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu
>> > : eth0:%lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu
>>
>> This could be explained by a single-bit corruption of your kernel,
>> namely if the place where printk tests a format character against '%'
>> is broken. Maybe the '%' is bad, or maybe the comparison instruction.
>
>Agreed. /proc/{pid}/stat backs you up there too. %u is fine, only %lu is
>mangled by something.
>
>It's certainly an amusing failure; I'm surprised by how well everything
>runs with the system in this state; netstat and top fall over, uptime
>calculations go wrong, and ps and inndwatch complain. Everything else is
>happy as could be.
>
>
>Whatever it is it probably isn't memory error; I do lots of gcc
>compilations on this machine, and don't get nonrepeatable bootstrap
>comparison failures...
>
>I'll try pulling out some of the patches and modules (lm-sensors can go
>easily; reiserfs not so easily) and see what happens.

Check back a few weeks in the archives - I think this has to do with some
compiler issues where %lu isn't done because printk couldn't do the conversion.
I seem to remember that a patch was provided too...
                               

-- 
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@cats-chateau.net

Any opinions expressed are solely my own. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sat Sep 30 2000 - 21:00:27 EST