2.6.35 released!

falconindy · 2010-08-02 03:09:19

Omg and it's not in testing yet! *dodges incoming tomatoes*

As always, patch notes are at kernelnewbies.org. Seems like a fairly tame changelog, though the memory compaction looks like it might help out some folks on lower memory systems. Linus makes a fairly ominous mention of a real performance boost to come in 2.6.36. Might be fun to backport it to 2.6.35...

Nothing unusual on the external module front. I've got nvidia 256.35 and aufs2 compiled against my kernel with no problems; however, as usual, aufs needs a new base/standalone patch and module snapshot.

combuster · 2010-08-02 11:32:25

Well that caught me off guard That was sooner that I've expected although there werent any big changes planned for 2.6.35. Bugfix release I would say.

Gonna try it out now, but it will work because rc's behaved on my system very well during testing.

mutlu_inek · 2010-08-03 10:25:21

falconindy wrote:

Might be fun to backport it to 2.6.35...

I'm sure it wouldn't be. It's pretty invasive. But I don't think we'll stop you.

falconindy · 2010-08-03 16:49:19

mutlu_inek wrote:

I'm sure it wouldn't be. It's pretty invasive. But I don't think we'll stop you.

92 patches in the series, as pulled from Nick Piggin's repo about 10 minutes ago. They all apply cleanly to 2.6.35 source. Reading his post on LKML scares the crap out of me, but I'm thinking I'll snapshot my root and roll a kernel to see how it works.

Vi0L0 · 2010-08-03 18:54:49

falconindy wrote:

92 patches in the series, as pulled from Nick Piggin's repo about 10 minutes ago. They all apply cleanly to 2.6.35 source. Reading his post on LKML scares the crap out of me, but I'm thinking I'll snapshot my root and roll a kernel to see how it works.

Now that really looks promising! Would you be so kind to give url to those Nick's patches? Quick search return me nothing more than he's whole kernel git tree...
Edit: silly question... guess you are using git to update vanilla 2.6.35 . But then: is there any PKGBUILD i can base on to easily update vanilla kernel with git's tree?

Last edited by Vi0L0 (2010-08-03 19:04:14)

falconindy · 2010-08-03 19:04:49

I actually cloned his repo and ran git-format-patch on his working branch.

I've uploaded the patch series to omploader if anyone would like to play with it.

Working on a PKGBUILD now...

Last edited by falconindy (2010-08-03 19:14:26)

falconindy · 2010-08-03 19:46:49

$ uname -r
2.6.35-piggin

Back on topic! btrfs users might want to be aware that read performance seems to be down. I'm noticing that large tarballs (read: the kernel) takes a bit longer to decompress. The patch notes mention support for Direct I/O which I interpret to mean "not by default". This is probably a result of something else, or I might just be insane.

korpenkraxar · 2010-08-03 20:47:44

Does the 2.6.35 kernel seem stable on your machine?

Last edited by korpenkraxar (2010-08-03 20:48:00)

falconindy · 2010-08-03 23:29:48

Quite stable thus far. YMMV. Linux has never had a problem with my hardware.

Aside: Nick's patchset breaks NILFS2 and apparently Aufs2 as well. Boooo.

kernelpanic · 2010-08-04 12:48:55

Using 2.6.35 (AUR kernel26-rc package). I have problem with DHCP, it's not launching during boot. I need to start it manually. Tried to add it in rc.local but then I see only "searching for carrier" msg and nothing happens (timed out and again manually).

Edit: It's now up in testing so will check if problem occurs in it.

Last edited by kernelpanic (2010-08-04 13:22:38)

pawels64 · 2010-08-04 13:09:02

How to use power management on radeon cards with this kernel? According to this link:

http://www.overclock.net/linux-unix/731 … river.html

it's enough to: echo dynpm > /sys/class/drm/card0/device/power_method

but there's no such thing like power_method in my kernel - 2.6.35-ARCH

flamelab · 2010-08-04 13:27:13

pawels64 wrote:

How to use power management on radeon cards with this kernel? According to this link:
http://www.overclock.net/linux-unix/731 … river.html
it's enough to: echo dynpm > /sys/class/drm/card0/device/power_method
but there's no such thing like power_method in my kernel - 2.6.35-ARCH

Read the ati archwiki page

kernelpanic · 2010-08-04 13:33:03

Updated to testing kernel and problem occurs again as mention by me before.

pawels64 · 2010-08-04 13:47:44

flamelab wrote:

Read the ati archwiki page

Thanks. However radeon.dynpm=1 is not working and I'd prefer to stick to the official Arch Linux kernel.

flamelab · 2010-08-04 13:50:19

pawels64 wrote:

flamelab wrote:
Read the ati archwiki page
Thanks. However radeon.dynpm=1 is not working and I'd prefer to stick to the official Arch Linux kernel.

Hmm, check if there is any bug report about that or make a bug report (feature request) on the bugs.archlinux.org so that the kernel devs will look upon it

okubax · 2010-08-04 13:51:10

Hope it solves the connection isuues with broadcom wireless cards. Mine stopped working after upgrading kernel to 2.6.34.2-1, had to downgrade to get it working again.

Vi0L0 · 2010-08-04 14:03:38

falconindy wrote:

$ uname -r
2.6.35-piggin
Back on topic! btrfs users might want to be aware that read performance seems to be down. I'm noticing that large tarballs (read: the kernel) takes a bit longer to decompress. The patch notes mention support for Direct I/O which I interpret to mean "not by default". This is probably a result of something else, or I might just be insane.

Thanks for piggin's patches . With them my system is kinda more responsive (but not too much) comparing to vanilla 2.6.35 and 2.6.34.2. Although performance tested with iozone on my ext4 partition is pretty the same. Anyhow i would like to see piggin's patches in company of BFS and BFQ, cuz BFS and BFQ was always making my system faster. As i can see there's BFQ for 2.6.35 already, so im gonna test how/if it's working with piggin's.

Maos · 2010-08-04 14:15:44

kernelpanic wrote:

Updated to testing kernel and problem occurs again as mention by me before.

Same problem with 2.6.35 from testing.
Dhcp times out while "waiting for carrier" but works when doing it manually it works ok

The network card in question:
00:19.0 Ethernet controller: Intel Corporation 82567LF-2 Gigabit Network Connection
Subsystem: Intel Corporation Device 5002
Kernel driver in use: e1000e
Kernel modules: e1000e

Haven't tested yet but maybe a dhcpcd problem:
http://roy.marples.name/archives/dhcpcd … /0218.html
http://forums.gentoo.org/viewtopic-p-63 … 7fe0603559

Last edited by Maos (2010-08-04 14:20:38)

Maos · 2010-08-04 14:42:29

Maos wrote:

kernelpanic wrote:
Updated to testing kernel and problem occurs again as mention by me before.
Same problem with 2.6.35 from testing.
Dhcp times out while "waiting for carrier" but works when doing it manually it works ok
The network card in question:
00:19.0 Ethernet controller: Intel Corporation 82567LF-2 Gigabit Network Connection
Subsystem: Intel Corporation Device 5002
Kernel driver in use: e1000e
Kernel modules: e1000e
Haven't tested yet but maybe a dhcpcd problem:
http://roy.marples.name/archives/dhcpcd … /0218.html
http://forums.gentoo.org/viewtopic-p-63 … 7fe0603559

Have now tested this and using dhcpcd 2.5.7 solved the issue for me

lenzy · 2010-08-04 14:48:15

my laptop (acer extensa 5235-901G16N) doesn't work anymore, since i've updated to this kernel version.
after reinstall same problem..

after grub my screen is black.
i can't do anything.

Last edited by lenzy (2010-08-04 14:56:54)

falconindy · 2010-08-04 14:52:54

lenzy wrote:

my laptop doesn't work anymore, since i've updated to this kernel version.
after reinstall same problem..

Thank you for your highly detailed and informative first post. This has been forwarded upstream to the kernel bugzilla. Rest assured that a team of skilled wombats is working quickly and decisively to resolve this issue.

pawels64 · 2010-08-04 15:28:51

flamelab wrote:

Hmm, check if there is any bug report about that or make a bug report (feature request) on the bugs.archlinux.org so that the kernel devs will look upon it

Done. The report is here:

http://bugs.archlinux.org/task/20368?st … name=&type

@Falconindy

Working on a PKGBUILD now...

Can't wait to test it.

@ViOLO

Although performance tested with iozone on my ext4 partition is pretty the same.

It should boost performance on many core systems. For tests, try using sysbench maybe.

Last edited by pawels64 (2010-08-04 15:29:26)

falconindy · 2010-08-04 15:49:40

pawels64 wrote:

@Falconindy
Working on a PKGBUILD now...
Can't wait to test it.

It's boring. I'm not interested in posting it to the AUR because it's guaranteed to be short lived, and I can't maintain it because it turns out I can't use it until Aufs works with it. I did the "hard work" of extracting the patch set. It's trivial to loop through the enclosed directory and apply the patches in order.

kernelpanic · 2010-08-04 16:09:18

Maos wrote:

Have now tested this and using dhcpcd 2.5.7 solved the issue for me

I can confirm that too. It's definitely a dhcpcd bug. Thanks for info.

Vi0L0 · 2010-08-04 20:12:07

pawels64 wrote:

It should boost performance on many core systems. For tests, try using sysbench maybe.

And i was hoping for that boost since my system got 4 cores...
So i run sysbench

$ sysbench --num-threads=16 --test=fileio --file-total-size=3G --file-test-mode=rndrw prepare
$ sysbench --num-threads=16 --test=fileio --file-total-size=3G --file-test-mode=rndrw run

128 files, 24576Kb each, 3072Mb total
Number of threads: 16

Results:

2.6.34.2-ARCH wrote:

Operations performed: 6007 Read, 3999 Write, 12804 Other = 22810 Total
Read 93.859Mb Written 62.484Mb Total transferred 156.34Mb (5.5756Mb/sec)
356.84 Requests/sec executed
Test execution summary:
total time: 28.0406s
total number of events: 10006
total time taken by event execution: 253.5010
per-request statistics:
min: 0.01ms
avg: 25.33ms
max: 1638.84ms
approx. 95 percentile: 211.99ms
Threads fairness:
events (avg/stddev): 625.3750/83.92
execution time (avg/stddev): 15.8438/0.99

2.6.35-piggin wrote:

Operations performed: 6001 Read, 4001 Write, 12801 Other = 22803 Total
Read 93.766Mb Written 62.516Mb Total transferred 156.28Mb (6.7751Mb/sec)
433.61 Requests/sec executed
Test execution summary:
total time: 23.0669s
total number of events: 10002
total time taken by event execution: 114.6167
per-request statistics:
min: 0.01ms
avg: 11.46ms
max: 1510.54ms
approx. 95 percentile: 77.08ms
Threads fairness:
events (avg/stddev): 625.1250/138.08
execution time (avg/stddev): 7.1635/0.81

So there is a boost! (~22% faster with piggin's patches)

Although i am still surprised about poor result of piggin and BFQ company:

2.6.35-piggin+bfq wrote:

Operations performed: 6007 Read, 3999 Write, 12673 Other = 22679 Total
Read 93.859Mb Written 62.484Mb Total transferred 156.34Mb (4.3039Mb/sec)
275.45 Requests/sec executed
Test execution summary:
total time: 36.3257s
total number of events: 10006
total time taken by event execution: 243.5208
per-request statistics:
min: 0.01ms
avg: 24.34ms
max: 993.32ms
approx. 95 percentile: 194.19ms
Threads fairness:
events (avg/stddev): 625.3750/106.26
execution time (avg/stddev): 15.2200/2.30

Last edited by Vi0L0 (2010-08-04 21:11:34)

Arch Linux

#1 2010-08-02 03:09:19

2.6.35 released!

#2 2010-08-02 11:32:25

Re: 2.6.35 released!

#3 2010-08-03 10:25:21

Re: 2.6.35 released!

#4 2010-08-03 16:49:19

Re: 2.6.35 released!

#5 2010-08-03 18:54:49

Re: 2.6.35 released!

#6 2010-08-03 19:04:49

Re: 2.6.35 released!

#7 2010-08-03 19:46:49

Re: 2.6.35 released!

#8 2010-08-03 20:47:44

Re: 2.6.35 released!

#9 2010-08-03 23:29:48

Re: 2.6.35 released!

#10 2010-08-04 12:48:55

Re: 2.6.35 released!

#11 2010-08-04 13:09:02

Re: 2.6.35 released!

#12 2010-08-04 13:27:13

Re: 2.6.35 released!

#13 2010-08-04 13:33:03

Re: 2.6.35 released!

#14 2010-08-04 13:47:44

Re: 2.6.35 released!

#15 2010-08-04 13:50:19

Re: 2.6.35 released!

#16 2010-08-04 13:51:10

Re: 2.6.35 released!

#17 2010-08-04 14:03:38

Re: 2.6.35 released!

#18 2010-08-04 14:15:44

Re: 2.6.35 released!

#19 2010-08-04 14:42:29

Re: 2.6.35 released!

#20 2010-08-04 14:48:15

Re: 2.6.35 released!

#21 2010-08-04 14:52:54

Re: 2.6.35 released!

#22 2010-08-04 15:28:51

Re: 2.6.35 released!

#23 2010-08-04 15:49:40

Re: 2.6.35 released!

#24 2010-08-04 16:09:18

Re: 2.6.35 released!

#25 2010-08-04 20:12:07

Re: 2.6.35 released!

Board footer