Zen-Kernel Discussion/Support Thread - Part 4

[kernelOfTruth]

git guys are reading the logs, right ?
+EXTRAVERSION = -zen0
+NAME = Untested Master

v2.6.35-zen0
- This master branch has not been tested or even compiled yet. Please hold bug reports until 2.6.35-zen1 is tagged in the master.
- Feel free to test as is, use at your own risk

definitely i´m log browsing, just wanted to tell how i managed to compile the kernel for me in its early 35´s stage

the swap-problem should be fixed now

thanks cheater !

[rahulthewall]

git guys are reading the logs, right ?
+EXTRAVERSION = -zen0
+NAME = Untested Master

v2.6.35-zen0
- This master branch has not been tested or even compiled yet. Please hold bug reports until 2.6.35-zen1 is tagged in the master.
- Feel free to test as is, use at your own risk

definitely i´m log browsing, just wanted to tell how i managed to compile the kernel for me in its early 35´s stage

the swap-problem should be fixed now

thanks cheater !

Yes, yes, I know that the current branch is completely untested, pre-alpha material. That is why I am testing. Anyways, latest is kernel panic while booting with the message (produced as well as I can remember)

EDIT: Added more panic info.

Code: Select all

TuxOnIce can't translate /dev/sda5 into a device
....
Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)
sda6 is not a valid device (or something similar)
VFS: Root filesystem not found
....

I am going to put an empty string for default resume partition and then try again.

[Anon-E-moose]

Compiled and running unstable snapshot zen-1350978e18a133ec938e90d87f9946e296f9c1ea.tar.gz from 8/11

VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 35
EXTRAVERSION = .1
NAME = Sheep on Meth

so far things seem to be working fine.

System seems to be just a touch snappier than the 2.6.34 series I was running.

[wrc1944]

FWIW, I think that mm/swapfile problem might be 2.6.35- ck1. I just tried to patch 2.6.35.2 with ck1, and got this:

Code: Select all

patching file mm/swapfile.c
Hunk #1 FAILED at 319.
Hunk #2 succeeded at 410 (offset 2 lines).
Hunk #3 succeeded at 425 (offset 2 lines).
Hunk #4 succeeded at 721 (offset 2 lines).
1 out of 4 hunks FAILED -- saving rejects to file mm/swapfile.c.rej

If anyone likes, I an post the little relevant excepts of mm/swapfilec.orig, .rej, and the ck1 patched "new" version.

[cheater1034]

Working on it guys, appreciate the bug reports

[rahulthewall]

Disregard the bug reports about the kernel panic - I forgot to select the driver for my HD. :oops

[mantoo]

[..]I forgot to select the driver for my HD. :oops

me didnt, but got kind of the same (t)error

[rahulthewall]

Just to make sure, are these patches included in the current zen-stable. When I checked, they seemed to be there, but then I could have easily have made a mistake.

http://bugs.freedesktop.org/attachment.cgi?id=36464

These patches are needed for OpenGL compositing in KDE 4.5 to work on Intel 945GM.

[kernelOfTruth]

this should improve fsync performance on high-end disks (e.g. raid arrays, SSDs) and possibly also firefox interactivity

please add if you find it worth it

[Anon-E-moose]

Those changes seem to be in the version I'm using (I only looked at the 1st two files from the patch) so YMMV.

As I said in an earlier post, the whole system seems snappier than the 2.6.34 version I was using, including firefox.

[kernelOfTruth]

this should improve fsync performance on high-end disks (e.g. raid arrays, SSDs) and possibly also firefox interactivity

please add if you find it worth it

there seems to be a lot going on in relation to speeding / improving fsync behavior, combining & reducing barriers, etc.:
Re: [RFC] ext4: Don't send extra barrier during fsync if there are no dirty pages.

[cheater1034]

this should improve fsync performance on high-end disks (e.g. raid arrays, SSDs) and possibly also firefox interactivity

please add if you find it worth it

there seems to be a lot going on in relation to speeding / improving fsync behavior, combining & reducing barriers, etc.:
Re: [RFC] ext4: Don't send extra barrier during fsync if there are no dirty pages.

I added all 3 of the patches you posted

[kernelOfTruth]

this should improve fsync performance on high-end disks (e.g. raid arrays, SSDs) and possibly also firefox interactivity

please add if you find it worth it

there seems to be a lot going on in relation to speeding / improving fsync behavior, combining & reducing barriers, etc.:
Re: [RFC] ext4: Don't send extra barrier during fsync if there are no dirty pages.

I added all 3 of the patches you posted

thanks !

we however should be careful with the last one:

We could actually improve the scheme to work for data as well. I wrote
a proof-of-concept patches (attached) and they nicely avoid second barrier
when doing:
echo "aaa" >file1; echo "aaa" >file2; fsync file2; fsync file1

I'm confident that Jan Kara has experience in this area but nevertheless - bugs could occur so (it's proof-of-concept after all): YMMV and happy testing

oh - and btw: could you also please add the vmscan:
http://git.zen-kernel.org/?p=kernel/zen ... ebcc2a62f1

and the fix to replace:

Code: Select all

if (should_reclaim_stall(nr_taken, nr_reclaimed, priority, sc)) {

with

Code: Select all

if (should_reclaim_stall(nr_taken, nr_freed, priority, sc)) {

many thanks in advance ! (this one's crucial for desktop linux)

[ponciarello]

http://theinvisiblethings.blogspot.com/ ... loset.html
http://www.invisiblethingslab.com/resou ... ttacks.pdf
https://bugzilla.redhat.com/show_bug.cgi?id=606611





but the fix (320b2b8de12698082609ebbc1a17165727f4c893) it's already in 2.6.35.2

[kernelOfTruth]

http://theinvisiblethings.blogspot.com/ ... loset.html
http://www.invisiblethingslab.com/resou ... ttacks.pdf
https://bugzilla.redhat.com/show_bug.cgi?id=606611





but the fix (320b2b8de12698082609ebbc1a17165727f4c893) it's already in 2.6.35.2

well - there's a glitch:

https://bugzilla.kernel.org/show_bug.cgi?id=16588
potential fix (probably will also be included in 2.6.35.3)
http://www.kernel.org/pub/linux/kernel/ ... le-review/

[Ant P.]

http://theinvisiblethings.blogspot.com/ ... loset.html

This link is just corporate propaganda to promote a product.

[cheater1034]

zen1 is tagged and ready for release, but unfortunately due to technical difficulties it won't be uploaded or released on the website yet.

[Anon-E-moose]

"Omgwtf Finallayz"


Like the name

[cheater1034]

idk if you guys have seen either, but my zendroid development may take off soon too

zen on android ftw

[kernelOfTruth]

idk if you guys have seen either, but my zendroid development may take off soon too

zen on android ftw

then I might have a useful patch for it (if zendroid isn't based on cyanogenmod):
Re: android-kernel memory reclaim x20 boost?

dunno if that's the new one with the non-requirement for profiling (seems rather like the original with the requirement for profiling):
staging: android: lowmemkiller: Substantially reduce overhead during reclaim

anyways: if you can find the new & improved version of that patch could you please also apply it to zen-stable or at least post a reference here so that I can apply it manually (for testing) here ?

many thanks in advance

edit:

It seems to be not a issue of mainline but only one of android lowmemkiller.

?

[Ant P.]

With latest zen-stable, my laptop won't boot (kernel panic with something about agp/drm, but it scrolls off the top of the screen). The same config with the unstable tree works fine.

[Kollin]

A tar snapshot of latest stable *2.6.35-r* for the portage please

[DigitalCorpus]

My kernel panicked upon boot from the latest stable git. I't couldn't mount my reiser4 root. I triple checked the config and redid make oldconfig.

[albright]

A tar snapshot of latest stable *2.6.35-r* for the portage

yeah to that ... we're already up to 35.3 I think

[devl547]

Considering interactive governor - 5 days of non-stop notebook usage with no problems.

Linux localhost 2.6.35.1+ #9 ZEN SMP PREEMPT Tue Aug 16 22:44:45 GMT 2010 x86_64 Intel(R) Core(TM) i5 CPU M 430 @ 2.27GHz GenuineIntel GNU/Linux

gcc version 4.6.0-alpha20100731 20100731 (experimental) (Gentoo 4.6.0_alpha20100731)