Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Adam Williamson]

On Sun, 2003-08-31 at 00:32, Adam Williamson wrote:
> On Sat, 2003-08-30 at 13:20, Adam Williamson wrote:
> 
> > No - they don't build unless the symlink exists, so you can assume I
> > create it myself every time this happens. Now here's something REAL odd
> > - I reinstalled 0.5.1tmb and now THAT'S got the bug too! ARGH. I'll keep
> > playing. I'm going to try some different versions of nvnet...
> 
> Who's the terribly handsome guy who wrote the above?
> 
> :)
> 
> OK, more in the Neverending Kernel Saga...I rebuilt nvnet for my new
> installation of 0.5.1tmb, using nvnet 1.0-0248 instead of 1.0-0261 this
> time, and it works correctly; i.e. I'm back where I started, non-buggy
> behaviour with the 0.5.1tmb kernel. I have no idea if it was the
> rebuild, the 0248 vs. 0261 or just the phase of the moon that made it
> work :). I'll do some more experimenting with the new tmb kernel
> tomorrow afternoon (gotta work tomorrow morning).

One more update - now using 2.4.22-3tmb again and it seems OK, using the
nvnet 1.0-0248 driver. I'm not ENTIRELY convinced it's as simple as
nvnet 1.0-0248 vs. 1.0-0261 yet, but that certainly looks like a factor.
Thomas, you got around to testing this yet?
-- 
adamw

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Charles A Edwards]

On Sun, 31 Aug 2003 00:09:32 +0100
Adam Williamson <[EMAIL PROTECTED]> wrote:

> Ah, interesting. Next time you update source, could you verify that
> before the ivh the link from /lib/modules//build to
> /usr/src/linux (or, uh, whichever way around I should write it :>)
> does not exist, and after the ivh it does? 

Well, with 2.4.22-3.tmb.1mdk there is no link created but instead the
content of /lib/modules//build is duplicated in
/usr/src/linux
This is using either -Uvh or -ivh


Charles

-- 
I'd be a poorer man if I'd never seen an eagle fly.
-- John Denver

[I saw an eagle fly once.  Fortunately, I had my eagle fly swatter
handy.  Ed.]-
Mandrake Linux 9.2 on PurpleDragon
Kernel-2.4.22-3.tmb.1mdkenterprise 
-


pgp0.pgp
Description: PGP signature

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Adam Williamson]

On Sat, 2003-08-30 at 13:20, Adam Williamson wrote:

> No - they don't build unless the symlink exists, so you can assume I
> create it myself every time this happens. Now here's something REAL odd
> - I reinstalled 0.5.1tmb and now THAT'S got the bug too! ARGH. I'll keep
> playing. I'm going to try some different versions of nvnet...

Who's the terribly handsome guy who wrote the above?

:)

OK, more in the Neverending Kernel Saga...I rebuilt nvnet for my new
installation of 0.5.1tmb, using nvnet 1.0-0248 instead of 1.0-0261 this
time, and it works correctly; i.e. I'm back where I started, non-buggy
behaviour with the 0.5.1tmb kernel. I have no idea if it was the
rebuild, the 0248 vs. 0261 or just the phase of the moon that made it
work :). I'll do some more experimenting with the new tmb kernel
tomorrow afternoon (gotta work tomorrow morning).
-- 
adamw

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Adam Williamson]

On Sat, 2003-08-30 at 13:52, Charles A Edwards wrote:
> On Sat, 30 Aug 2003 13:20:50 +0100
> Adam Williamson <[EMAIL PROTECTED]> wrote:
> 
> > No - they don't build unless the symlink exists, so you can assume I
> > create it myself every time this happens. Now here's something REAL
> > odd- I reinstalled 0.5.1tmb and now THAT'S got the bug too! ARGH. I'll
> > keep playing. I'm going to try some different versions of nvnet...
> 
> Adam I don't know if you have tried this but with the last several
> kernels, I using kernel-tmb-enterprise-2.4.22.1.tmb.1mdk-1-1mdk
> currently, I -Uvh the kernel source and then have to -ivh --force it.
> 
> I build the ATI fglrx driver with each new kernel.
> After using -Uvh for the source the build will fail with No build
> environment found if I then -ivh --force the kernel-source the driver
> nodule builds without problem.

Ah, interesting. Next time you update source, could you verify that
before the ivh the link from /lib/modules//build to
/usr/src/linux (or, uh, whichever way around I should write it :>) does
not exist, and after the ivh it does? Sounds like it's the same issue
but a different fix...
-- 
adamw

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Charles A Edwards]

On Sat, 30 Aug 2003 13:20:50 +0100
Adam Williamson <[EMAIL PROTECTED]> wrote:

> No - they don't build unless the symlink exists, so you can assume I
> create it myself every time this happens. Now here's something REAL
> odd- I reinstalled 0.5.1tmb and now THAT'S got the bug too! ARGH. I'll
> keep playing. I'm going to try some different versions of nvnet...

Adam I don't know if you have tried this but with the last several
kernels, I using kernel-tmb-enterprise-2.4.22.1.tmb.1mdk-1-1mdk
currently, I -Uvh the kernel source and then have to -ivh --force it.

I build the ATI fglrx driver with each new kernel.
After using -Uvh for the source the build will fail with No build
environment found if I then -ivh --force the kernel-source the driver
nodule builds without problem.


Charles

-- 
There's really not much that can go wrong with your computer untill you
turn it on. 
-- Murphy's Law of Passive Dynamics
-
Mandrake Linux 9.2 on PurpleDragon
Kernel-2.4.22-1.tmb.1mdkenterprise 
-


pgp0.pgp
Description: PGP signature

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Adam Williamson]

On Sat, 2003-08-30 at 13:08, Thomas Backlund wrote:
> From: "Adam Williamson" <[EMAIL PROTECTED]>
> > On Sat, 2003-08-30 at 01:28, Adam Williamson wrote:
> > > See topic. I just tried out the latest Cooker kernel,
> > > 2.4.22-2mdkenterprise, and it's STILL missing the
> > > /lib/modules/2.4.22-2mdkenterprise/build -> /usr/src/linux-2.4.22-2mdk
> > > symlink and STILL exhibits the bug I reported a couple of weeks back
> > > where network activity results in extreme system sluggishness. I'm going
> > > to test Thomas Backlund's most recent kernel now and see how that one
> > > behaves; will report back. Until I tried 2.4.22-2 I've been using
> > > 2.4.22-0.5.1tmb_mdkenterprise, which works great.
> >
> > 2.4.22-1.tmb.1mdkenterprise , built from the .src.rpm , is missing the
> > equivalent symlink and displays the sluggish behaviour :(.
> > 2.4.22-0.5.1tmb_mdkenterprise is my only known-good kernel. Are there
> > any other nforce2 users out there using the nvnet module who can help
> > find out if this is isolated or not?
> > -- 
> > adamw
> >
> 
> I have a nForce2...

well, that's handy :)

> I just stopped compiling nVidia drivers since I do
> so many testboots and testruns on different kernels...
> so I got bored with rebuilding the nvidia drivers for every testboot...

yeah, it can be a pain, but it'd be great if you could test using the
onboard network with a few different kernels (yours, mandrake's) and see
what happens for you.

> Seems I'll have to check the specfile scripts ...
> and retry building the nvidia drivers...
> 
> as for the symlinks:
> I just recreated my kernel-tmb.spec from the kernel-2.4.spec in
> 2.4.22-1mdk
> 
> If you create the symlink yourself, and rebuild the nVidia
> drivers, do they work as they should then?

No - they don't build unless the symlink exists, so you can assume I
create it myself every time this happens. Now here's something REAL odd
- I reinstalled 0.5.1tmb and now THAT'S got the bug too! ARGH. I'll keep
playing. I'm going to try some different versions of nvnet...

> BTW, any reason you compiled the enterprise ? it is in the contribs
> with the rest... unless you are recompiling for athlon...

Because your page is faster than contrib :). It's not in contrib on my
mirror yet, or wasn't last night when I built it.
-- 
adamw

Re: [Cooker] current cooker kernel STILL exhibiting thatnetwork-activity related slowdown bug

[Adam Williamson]

On Sat, 2003-08-30 at 01:28, Adam Williamson wrote:
> See topic. I just tried out the latest Cooker kernel,
> 2.4.22-2mdkenterprise, and it's STILL missing the
> /lib/modules/2.4.22-2mdkenterprise/build -> /usr/src/linux-2.4.22-2mdk
> symlink and STILL exhibits the bug I reported a couple of weeks back
> where network activity results in extreme system sluggishness. I'm going
> to test Thomas Backlund's most recent kernel now and see how that one
> behaves; will report back. Until I tried 2.4.22-2 I've been using
> 2.4.22-0.5.1tmb_mdkenterprise, which works great.

2.4.22-1.tmb.1mdkenterprise , built from the .src.rpm , is missing the
equivalent symlink and displays the sluggish behaviour :(.
2.4.22-0.5.1tmb_mdkenterprise is my only known-good kernel. Are there
any other nforce2 users out there using the nvnet module who can help
find out if this is isolated or not?
-- 
adamw