FS#23937 - comp becames very slow with kernel26-2.6.38.4-1_x86_64

Attached to Project: Arch Linux
Opened by Marat Kagarmanov (corvinusz) - Monday, 25 April 2011, 21:14 GMT
Last edited by Jan de Groot (JGC) - Wednesday, 11 May 2011, 09:41 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To Tobias Powalowski (tpowa)
Architecture x86_64
Severity Critical
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 7
Private No

Details

Description:
After update packages kernel26 and kernel26-headers from 2.6.38.3.-1 to 2.6.38.4-1 my old notebook "Roverbook Explorer E15" (AMD64 Turion MT-30, 800MHz, RAM 512Mb) became freezy. Boot process goes about 5 minutes (instead 20sec on 2.6.38.3-1). Ctrl-Alt-F1 takes tty1 after 1 minute.

Additional info:
* package version - kernel26-2.6.38.4-1-x86_64
* /usr/bin/top says "all ok", cpu load ~1%. Of course, /usr/bin/top bring main screen after 30sec lock.
* i have no special partition for /boot. /boot is hosted on /
* grub-string "kernel /boot/vmlinuz26 root=/dev/disk/by-uuid/xxxxxxx ro quiet"
adding rootfstype=ext4 does not anything
* i did not verify kernel26-2.6.38.4-1-i686

Steps to reproduce: just power on.
Steps to restore good state: downgrade to kernel26 2.38.3-1
This task depends upon

Closed by  Jan de Groot (JGC)
Wednesday, 11 May 2011, 09:41 GMT
Reason for closing:  Fixed
Additional comments about closing:  Fixed in 2.6.38.6 which is in testing at this moment.
Comment by Balazs Heder (hbalazs1980) - Tuesday, 26 April 2011, 15:52 GMT
I experience the same behaviour on my i686 laptop with kernel26 (2.6.38.4) and also with kernel26-ck (2.6.38.4) kernels.
Comment by Edwin Orrico (edorr) - Tuesday, 26 April 2011, 20:04 GMT
same thing happens to me. After update kernel26 and kernel26-headers from 2.6.38.3-1 to 2.6.38.4-1, my laptop became freezy.

boot takes about 8 to 10 min.

I have an AMD64 Turion 1.6 Ghz 2GB RAM running x86_64.

Problem disappear when booting whit kernel26-lts
Comment by synflag (synflag) - Wednesday, 27 April 2011, 06:30 GMT
Not a problem kernel 2.6.38.4, I use the same kernel but compiled for me, downloading the code from kernel.org and using my .config own, and do not use ABS, but, make, make modules_install make install and mkinitcpio -k 2.6.38.4 -c /etc/mkinitcpio.conf -g /boot/vmlinuz-2.6.38.4.img.
Rule out that it is a kernel problem, I do not think that is a problem .config used, and would think that is a problem in the compilation of any module as the gcc 4.6 is more restrictive, but I use it without the parameter -fpermissive and all is well, use all ext4 least the / boot partition (ext2).
CPU AMD Phenom II and chipset ATI, no problems here.
Try compile your kernel.
Comment by synflag (synflag) - Wednesday, 27 April 2011, 06:35 GMT
Try rebuilding your init ramfs.

mkinitcpio -k 2.6.38.4-1 -c /etc/mkinicpio.conf -g /boot/kernel26.img
Comment by Rares Aioanei (lmov) - Wednesday, 27 April 2011, 15:22 GMT
Works here too just fine (HP Pavilion, AMD Turion TL-2, 2GB RAM, NVidia GPU) running x86_64.
Comment by Edwin Orrico (edorr) - Friday, 29 April 2011, 14:50 GMT
I tried rebuilding my init ramfs with mkinitcpio -k 2.6.38.4-1 -c /etc/mkinicpio.conf -g /boot/kernel26.img but i'm still having the same problem, the system is very slow.

i'm using the default /etc/mkinicpio.conf

booting whit kernel26-lts have no problems.

also, I tried to compile the kernel 2.6.39-rc5 from kernel.org using the whit makepkg and pacman using the make oldconfig, just like the https://wiki.archlinux.org/index.php/Kernel_Compilation_from_Source recomend, but the system doesn't boot with that kernel.

Comment by Edwin Orrico (edorr) - Saturday, 30 April 2011, 01:27 GMT
Well, I downgrade mi kernel to version 2.6.38.3-1 and my laptop boots fast again. I will not update my kernel until the next new version. I hope next version fix this bug
Comment by Ionut Biru (wonder) - Saturday, 30 April 2011, 06:40 GMT
@Edwin, it won't be fixed unless is reported.

https://bugzilla.kernel.org/
Comment by Balazs Heder (hbalazs1980) - Tuesday, 03 May 2011, 12:42 GMT
The problem occurs also with kernel26-lts-ck-2.6.35.13-2-i686 from the following repo (https://wiki.archlinux.org/index.php/Kernel26-ck):

[kernel26-ck]
Server = http://home.comcast.net/~repo-ck/i686

The previous lts-ck kernel (kernel26-lts-ck-headers-2.6.35.12-5-i686) worked fine.

Moreover, I noticed a very strange thing:
As far as we know, booting with the affected kernels (kernel26-2.6.38.4, kernel26-ck-2.6.38.4 or kernel26-lts-ck-2.6.35.13-2) is very slow.
But accidentally I noticed the following: when the booting process seems to be frozen for some seconds and I press the SPACE button, the booting process goes forward to the next step and seems to be frozen again. If I press SPACE again, the process goes one more step forward and so on.

It is strange, isn't it?


Comment by Edwin Orrico (edorr) - Tuesday, 03 May 2011, 12:52 GMT
@Balazs I noticed that too, I press the ENTER button and the process continue but seems frozen again one or two steps forward. it's very strange this.

have anybody been trying to boot another distro with those kernel version? just to verify if it's not a problem with Arch package.
Comment by synflag (synflag) - Tuesday, 03 May 2011, 22:51 GMT
@edwin @balazs try this:

1) wget http://www.kernel.org/pub/linux/kernel/v2.6/linux-2.6.38.4.tar.bz2 in /usr/src/
2) tar jxvf linux-2.6.38.4.tar.bz2
3) cd linux-2.6.38.4
4) create .config with .config file attach
5) make menuconfig
6) modify the kernel for you hardware (video sound chipset etc)
7) save and exit
8) make -jX (X == number of cores)
9) make modules_install
10) make install
11) mkinitcpio -k 2.6.38.4 -c /etc/mkinitcpio.conf -g /boot/vmlinuz-2.6.38.4.img
12) mv /boot/vmlinuz /boot/vmlinuz-2.6.38.4
13) add a line in /boot/grub/menu.lst with:

title Arch Linux 2.6.38.4
root (hd0,0)
kernel /vmlinuz-2.6.38.4 root=(your route here, copie of last line of grub uuid of device) ro
initrd /vmlinuz-2.6.38.4.img

14) save and exit
15) boot and press Esc in grub and type "e" and "e" new over kernel line, add number 3 at the end of the line (boot in runlevel 3) and put enter and "b" over kernel line to boot
16) Login as root, compile the nvidia driver or your driver (use links to download from ftp://download.nvidia.com)
17) chmod +x NVIDIA-xxx and ./NVIxxx.run
18) reboot and test

please tell me how did it work for you and if you need further assistance
Comment by Frank Phillips (fphillips) - Wednesday, 04 May 2011, 17:37 GMT
@synflag: this bug seems to be specific to certain Turion laptops (i686/x86_64). Unless you can say that you have the bug on your hardware with the Arch kernel, and don't have it when you use this config, then it is probably a waste of time for them to try.
Comment by Balazs Heder (hbalazs1980) - Wednesday, 04 May 2011, 18:02 GMT
Yes, I can confirm that this is a Turion specific bug: On my laptop with Turion processor the problem occurs, but on my desktop with Athlon X2 there isn't any problem...
Comment by Edwin Orrico (edorr) - Wednesday, 04 May 2011, 19:05 GMT
Yes, I tried on a AMD Phenom II X4 and there's no problem. So it's a Turion problem.

i'm gonna look on https://bugzilla.kernel.org/ if it's reported. Maybe you can do the same to make sure they will fix the bug on the next release
Comment by Edwin Orrico (edorr) - Wednesday, 04 May 2011, 19:44 GMT
I reported the bug in https://bugzilla.kernel.org/show_bug.cgi?id=34502 so, if you want, can add your comments there
Comment by Frank Phillips (fphillips) - Wednesday, 04 May 2011, 19:48 GMT
Found it:

commit "x86, AMD: Set ARAT feature on AMD processors" in both 2.6.38.4 and 2.6.35.13
http://www.gossamer-threads.com/lists/linux/kernel/1369696#1369696

fix "x86, AMD: Fix APIC timer erratum 400 affecting K8 Rev.A-E processors"
http://www.gossamer-threads.com/lists/linux/kernel/1373217#1373217

It has been sent to -stable, so it should be in 2.6.38.6. You can wait for it to show up in the queue here (when a new ".38 patches" is there):
http://git.kernel.org/?p=linux/kernel/git/stable/stable-queue.git;a=summary
Comment by Edwin Orrico (edorr) - Wednesday, 04 May 2011, 20:00 GMT
thank you so much for the info @fphillips, i'll be waiting for the patch
Comment by Frank Phillips (fphillips) - Thursday, 05 May 2011, 06:24 GMT
@edwin, don't forget to close your upstream report, or at least point them at the fix.
Comment by F. Bogo the III. (fogobogo) - Friday, 06 May 2011, 10:00 GMT
I had the same trouble here. box is a
Linux voodoo 2.6.38-ARCH #1 SMP PREEMPT Tue May 3 06:04:40 UTC 2011 i686 Mobile AMD Sempron(tm) Processor 2600+ AuthenticAMD GNU/Linux

The way I fixed it here was adding nohz=off notsc to the grub kernel line. I honestly think the notsc is superfluous since the clocksource is still acpi_pm and the only available one is acpi_pm too so it seems to be a AMD bug with the tickless kernel. Anyway, that might help whomever is affected.
Comment by Romain Forlot (RFOR) - Sunday, 08 May 2011, 01:13 GMT
Same here, with another AMD chip :

Linux percevale 2.6.38-ARCH #1 SMP PREEMPT Tue May 3 07:40:34 CEST 2011 x86_64 AMD Athlon(tm) 64 Processor 3000+ AuthenticAMD GNU/Linux

The trick of fogobogo works fine, thank for that. No "notsc" needed indeed.

Loading...