>In article <199902101407.OAA22565@mauve.demon.co.uk>,
> Ian Stirling <root@mauve.demon.co.uk> writes:
>> 2.2.1, on 16Mb ram.
>> After an unfortunate incident, that seems fortunately not to have harmed
>> my system, I wanted to check my hard drives.
>> However, when I did badblocks -w /dev/xxx count>40000, the system became very
>> unresponsive, (0.2-0.5 seconds for a char to be echoed on a text vc), and
>> then started killing stuff, beginning with my webcache, and continuing with
>> inetd, update, and after a few seconds init.
>> Fortunately, a shell stayed live, so I was able to shutdown gracefully.
>> Badblocks in write mode
>>does: write 1024 bytes of data, lseek 1024 bytes into the file, write 1024,
>> lseek 2048, .....
>>
>> Happens with dma on, or off, on two seperate (quantum bigfoot) drives, X
>> running or not, kernel compile running or not.
>
>`badblocks' program effectively kills *any* linux kernel I tried it on,
>when run in write mode. Not necessarily hangs, but it becomes so slow
>that you would rather reboot than wait a few days till it finishes...
It's because kflushd don't wait for I/O completation if the number of
dirty buffers is high. Grep for too_many in the kernel list and you'll
find some patch that will fix the bad behavior. I guess Linus will include
one of the patches in 2.2.2.
Andrea Arcangeli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/