[gentoo-user] too many kernel problems!

[gentoo-user] too many kernel problems!

[Iain Buchanan]

Hi list!

you may remember my post recently[1] on 2.6.13 problems (it was very
slow to run).  This turned out to be a hyperthreaded issue[2], which was
fixed by a couple of patches [3], [4].

However, I can't find a kernel that will run both a) hyperthreading and
b) suspend / resume in a stable manner!

Here's a little table to show what I mean.  The c-state patch is the one
required to get HT working.  With it, suspend (to disk) doesn't work,
without it, usability is crap!

    kernel  c-state patch?   SMP speed  suspend    general
2.6.12-gr7       no            fast     untested    stable
----------
2.6.13-gr2       no            slow      stable     stable
2.6.13-gr2      yes            fast     unstable
----------
2.6.13-gr3       no
2.6.13-gr3      yes            fast     unstable   unstable

I can't yet post any of this on bugzilla.kernel.org because I'm not
using vanilla sources.  It would take more time that I have right now to
try it all again with vanilla.

So, my question is: Do I have to give up suspend? (I hope not, as when
it does work it works great and only takes a few seconds to suspend /
start up).  Is it worth a bugzilla entry anywhere?

[I'm using only the gentoo-sources for suspend, no extra patches or
utilities.  To suspend, I just type `echo shutdown > /sys/power/disk;
echo disk > /sys/power/state` as root]

I'm hoping there are some kernel guru's out there somewhere that can
give me some pointers.

Many thanks,

[1] http://thread.gmane.org/gmane.linux.gentoo.user/141911
[2] http://bugzilla.kernel.org/show_bug.cgi?id=5165
[3] http://bugzilla.kernel.org/show_bug.cgi?id=5165#c16
[4] http://bugzilla.kernel.org/show_bug.cgi?id=5165#c17
[5] 5? There is no 5! Why do you want 5?!

-- 
Iain Buchanan <iaindb@netspace.net.au>

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] too many kernel problems!

[Richard Fish]

Iain Buchanan wrote:

>Hi list!
>
>you may remember my post recently[1] on 2.6.13 problems (it was very
>slow to run).  This turned out to be a hyperthreaded issue[2], which was
>fixed by a couple of patches [3], [4].
>
>However, I can't find a kernel that will run both a) hyperthreading and
>b) suspend / resume in a stable manner!
>
>Here's a little table to show what I mean.  The c-state patch is the one
>required to get HT working.  With it, suspend (to disk) doesn't work,
>without it, usability is crap!
>
>    kernel  c-state patch?   SMP speed  suspend    general
>2.6.12-gr7       no            fast     untested    stable
>----------
>2.6.13-gr2       no            slow      stable     stable
>2.6.13-gr2      yes            fast     unstable
>----------
>2.6.13-gr3       no
>2.6.13-gr3      yes            fast     unstable   unstable
>
>I can't yet post any of this on bugzilla.kernel.org because I'm not
>using vanilla sources.  It would take more time that I have right now to
>try it all again with vanilla.
>
>So, my question is: Do I have to give up suspend? (I hope not, as when
>it does work it works great and only takes a few seconds to suspend /
>start up).  Is it worth a bugzilla entry anywhere?
>
>[I'm using only the gentoo-sources for suspend, no extra patches or
>utilities.  To suspend, I just type `echo shutdown > /sys/power/disk;
>echo disk > /sys/power/state` as root]
>
>I'm hoping there are some kernel guru's out there somewhere that can
>give me some pointers.
>  
>

You might trying playing with some kernel boot options 
(kernel-parameters.txt).  I have, at various times, needed one of:

acpi=ht
pci=routeirq

in order for suspend/resume to become stable.

HTH,
-Richard


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] too many kernel problems!

[Mark Knecht]

On 10/9/05, Iain Buchanan <iaindb@netspace.net.au> wrote:

> So, my question is: Do I have to give up suspend? (I hope not, as when
> it does work it works great and only takes a few seconds to suspend /
> start up).  Is it worth a bugzilla entry anywhere?

In my opinion, most definitely it's worth filing a report in Gentoo's
bugzilla. My thought is that ANYTHING that doesn't work the way you
want it to work is a candidate for a bug report.

I think that before you file a bug report you owe it to the developers
to check for other reports, both in bugzilla and the forums. It sounds
like you did that. Now you file and see what happens. I've filed a
number of bug reports. Very few of them have ever been blown off. A
few have come back 'won't fix', for various reasons. Most (I think)
have been fixed. Some have been duplicates that I couldn't find using
the search system. You may find that your issue is already in there
and you just haven't found it. A few bug reports have been pushed up
to the main application's developers.

Everything's fair if you do your part.

Cheers,
Mark

-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: too many kernel problems!

[James]

Iain Buchanan <iaindb <at> netspace.net.au> writes:


> However, I can't find a kernel that will run both a) hyperthreading and
> b) suspend / resume in a stable manner!

Well, I'm no genius here, but, I can help you by using
a portable with a 3GHz HT intel processor to build
equivalent kernels (differing only in hardware on the mobo)
to test/verify your troubles, if you like.

>From cpuinfo:
model name      : Intel(R) Pentium(R) 4 CPU 3.06GHz
stepping        : 9
cpu MHz         : 3067.472
cache size      : 512 KB

flags           : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat
pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe cid xtpr



> Here's a little table to show what I mean.  The c-state patch is the one
> required to get HT working.  With it, suspend (to disk) doesn't work,
> without it, usability is crap!
> 
>     kernel  c-state patch?   SMP speed  suspend    general
> 2.6.12-gr7       no            fast     untested    stable
> ----------
> 2.6.13-gr2       no            slow      stable     stable
> 2.6.13-gr2      yes            fast     unstable
> ----------
> 2.6.13-gr3       no
> 2.6.13-gr3      yes            fast     unstable   unstable

Hm, this table looks dubious, how was the results generated, i.e.
which variables are quantitative(objective) and which are qualitative
(subjective) in nature?

> So, my question is: Do I have to give up suspend? (I hope not, as when
> it does work it works great and only takes a few seconds to suspend /
> start up).  Is it worth a bugzilla entry anywhere?

Lets approach this systematically, and yes, you can/should open up
a trouble ticket with bugzilla. You can lead the discussion/direction
 here on this list, as this is YOUR thread.

> [I'm using only the gentoo-sources for suspend, no extra patches or
> utilities.  To suspend, I just type `echo shutdown > /sys/power/disk;
> echo disk > /sys/power/state` as root.

I use these gentoo-sources too. In fact I keep lots of different kernel
versions around on each machine, as I'm always noodling with kernels
options, patches, hacks, and sporadic assembler code....



> I'm hoping there are some kernel guru's out there somewhere that can
> give me some pointers.

Folks with the sort of expertise you seek, rarely have time for
us commoners, in my experience....YMMV.

So if you are interested in a systematic approach, let's agree on 
a common set of USE flags. 
Here are mine from make.conf:
CFLAGS="-O3 -march=pentium4 -fomit-frame-pointer -pipe -g"
CHOST="i686-pc-linux-gnu"

I do not deviate from these for any code/packages I build
on this Intel ht portable.


The kernel versions we can use
(for example both a gentoo-sources 2.6.12-r9 ( I use
on a variety of systems.) and 2.6.13-r3). Build both and let's 
use a common set of tools for testing and evaluation.

Send me your .config for each, privately, unless the list
wants those files or relevant portions thereof.

Let's also agree on how what we use as testing tools, I use
KDE and it has many gui related front-end apps. that can
be used to interface to ACPI.

Interested in this approach?

James





-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: too many kernel problems!

[Petteri Räty]

[-- Attachment #1: Type: text/plain, Size: 606 bytes --]

James wrote:
> Iain Buchanan <iaindb <at> netspace.net.au> writes:
> 
> So if you are interested in a systematic approach, let's agree on 
> a common set of USE flags. 
> Here are mine from make.conf:
> CFLAGS="-O3 -march=pentium4 -fomit-frame-pointer -pipe -g"
> CHOST="i686-pc-linux-gnu"
> 

Let me just point out that your CFLAGS are kind of stupid. You have -g
for producing debugging information but then you have
fomit-frame-pointer, which makes debugging impossible on some machines.

See http://www.gnu.org/software/gcc/onlinedocs/ for more information.

Regards,
Petteri Räty


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 256 bytes --]

[gentoo-user] Re: too many kernel problems!

[James]

Petteri Räty <petteri.raty <at> saunalahti.fi> writes:


> Let me just point out that your CFLAGS are kind of stupid. You have -g
> for producing debugging information but then you have
> fomit-frame-pointer, which makes debugging impossible on some machines.

Yes, a vestige, on a system that begs for attention....


James        



-- 
gentoo-user@gentoo.org mailing list