lkml.org 
[lkml]   [1998]   [Jun]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: 2.1.105 MASSIVE memory leak while stress testing
On Tue, 9 Jun 1998, Matthew Hawkins wrote:

> At 04:19PM on Mon, Jun 08, 1998, "Michael L. Galbraith" <mikeg@weiden.de> sent:
> > I was stress testing 2.1.105 with all of Bill Hawes's recent patches,
> > when the machine suddenly became ever more unresponsive.
>
> I had the same thing happen, only for me it was 2.1.104+ac1 plus the
> following patches from Bill Hawes:
>
> arch386_ldt104-patch buffer_104-patch mmap_104-patch
> nfsd_104-patch swap_vm104-patch
>
> After a while 2/3 of the processes had no resident pages.
>

Yes, in my case, the only processes which were doing _anything_ other
than swapping/paging were remote processes.

> My amazing sixth sense tells me the culprit would be the
> swap_vm104-patch (but I'm known to be wrong before:)
>

No impressions yet that mean spit. I don't _think/feel_ that this
problem is even related to Bill's patches.. I've seen this before.
The only real difference is memory which is really lost. That I
take to be a positive sign... exposed at last.

> I was trying to recompile .105 at the time, ended up rebooting and
> finishing the compile from a fresh startup as it was taking half
> an hour to compile one .c file (somewhere in the net/ hierarchy)
>
> 2.1.105+ac is running great though (and I've noticed its still a
> tad aggressive in swapping out pages)
>

I smell a race.. not a simple allocate and forget to de-allocate.
I've pushed it hard hard HARD looking for the simple stuff.

Something is wrong. Time and careful testing may uncover something.
It'll likely turn out to be something trivial, but I'll be damned
if it _looks_ trivial from my viewpoint.

Unfortunately, memleak has been virtually dead since 2.1.~85, and
I've spinlocked the whole damn thing trying to figure it out to no
effect.

-Mike


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu

\
 
 \ /
  Last update: 2005-03-22 13:43    [W:1.286 / U:0.400 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site