overnight 2.4.11 seems to have used up all memory, slow as a dog

overnight 2.4.11 seems to have used up all memory, slow as a dog

[Kevin B. Hendricks]

Hi,

I realize this is a vague bug report but since I am new to the 2.4.X
kernel series perhaps someone could enlighten me:

I am running a benh_kernel rsync from yesterday.

I downloaded XF4.1.0 and started the build and went to bed.

In the morning I came down and everything was still functioning but
painfully slowly.  So slowly as to make every command painful to watch.

Everything seemd to be hittng the disk but "top" showed nothing unusual.
It was like a was all out of memory.

Is this a symptom of the "VM" troubles that have kept me away from the
2.4.X series?

How can I track this one down?

FWIW, this never happens with 2.2.19 at all (I can do long overnight
builds with no trouble with no slowdowns.

Ideas anyone?

I am already starting to hate the 2.4.X kernels ;-)

Kevin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: overnight 2.4.11 seems to have used up all memory, slow as a dog

[Benjamin Herrenschmidt]

>
>Is this a symptom of the "VM" troubles that have kept me away from the
>2.4.X series?
>
>How can I track this one down?
>
>FWIW, this never happens with 2.2.19 at all (I can do long overnight
>builds with no trouble with no slowdowns.

There was at least one VM bug fixed between pre4 and pre5, you
can rsync again (or pull from bk if you are using bk _devel)
to check if it's better.

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: overnight 2.4.11 seems to have used up all memory, slow as a dog

[Dan Malek]

Benjamin Herrenschmidt wrote:

> There was at least one VM bug fixed between pre4 and pre5,

I just saw David Miller post another one on the Sparc Linux vger
tree.  We will have to track it down and see if it affects us.


	-- Dan

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: overnight 2.4.11 seems to have used up all memory, slow as a dog

[Kevin Hendricks]

Hi,

Once that fix makes it into Ben's tree, please let me know and I would
be happy to repeat that experiment to see if the problem is fixed.

Thanks,

Kevin


ps:

A download of official 2.4.10 with Alan's ac10 patch applied (cleanly)
seems to be missing a lot of agp pieces (drivers/char/agp/agpgart_be.c)
and has a bug (parse error) in drivers/char/pc_keyb.c in line 400 when
compared to Ben's tree.  Is this to be expected?

Thanks,

Kevin



On Tuesday, October 9, 2001, at 02:04 PM, Dan Malek wrote:

> Benjamin Herrenschmidt wrote:
>
>> There was at least one VM bug fixed between pre4 and pre5,
>
> I just saw David Miller post another one on the Sparc Linux vger
> tree.  We will have to track it down and see if it affects us.
>
>
> 	-- Dan
>

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/