Linux-pf

From ArchWiki
Revision as of 15:33, 24 March 2012 by Nous (Talk | contribs) (Installation)

Jump to: navigation, search

This template has only maintenance purposes. For linking to local translations please use interlanguage links, see Help:i18n#Interlanguage links.


Local languages: Català – Dansk – English – Español – Esperanto – Hrvatski – Indonesia – Italiano – Lietuviškai – Magyar – Nederlands – Norsk Bokmål – Polski – Português – Slovenský – Česky – Ελληνικά – Български – Русский – Српски – Українська – עברית – العربية – ไทย – 日本語 – 正體中文 – 简体中文 – 한국어


External languages (all articles in these languages should be moved to the external wiki): Deutsch – Français – Română – Suomi – Svenska – Tiếng Việt – Türkçe – فارسی

Template:Article summary start Template:Article summary text

  • Current version: 3.2.7-1
  • Kernel version: 3.2.11

Template:Article summary heading Template:Article summary wiki linux-pf mercurial repository

AUR linux-pf page

Patchset homepage

Patchset changelog Template:Article summary end

Linux-pf is a kernel package based on the stock -ARCH kernel, patched with a row of significant patches:

The reference PKGBUILD can be found at the AUR linux-pf page and at the bitbucket.org mercurial repository. Precompiled packages, generic and CPU-family optimized are uploaded at the pfkernel unofficial repository, usually within 6 hours of the AUR update for x86_64 and 2 days for i686. Append the following to /etc/pacman.conf to activate the pfkernel repo:

[pfkernel]
# Generic and optimized binaries of the ARCH kernel patched with BFS, TuxOnIce, BFQ, IMQ, Aufs3
# linux-pf, kernel26-pf, gdm-old, nvidia-pf, nvidia-96xx, xchat-greek, arora-git
Server = http://dl.dropbox.com/u/11734958/$arch

Running pacman -Syyl pfkernel will update all repos and show the available packages from pfkernel.

Compilation

There's a number of options a user is asked to choose from, should he/she select to compile from the PKGBUILD:

==> Hit <Y> to use your running kernel's config
    (needs IKCONFIG and IKCONFIG_PROC)
==> Hit <L> to run 'make localmodconfig'
==> Hit <N> (or just <ENTER>) to build an all-inclusive kernel like stock -ARCH
    (warning: it can take a looong time)

The <Y> option is for users who have already compiled and are running a custom kernel. The PKGBUILD reads the running kernel's configuration and uses it for the subsequent compilation. The <L> option tries some kind of autodetection of the user's hardware: it first tries to use the modprobed_db module database, then falls back to the linux kernel's make localmodconfig functionality. The last option is self-explanatory.

==> Kernel configuration options before build:
    <M> make menuconfig (console menu)
    <N> make nconfig (newer alternative to menuconfig)
    <G> make gconfig (needs gtk)
    <X> make xconfig (needs qt)
    <O> make oldconfig
    <ENTER> to skip configuration and start compiling

Choose one of these to use your favourite user interface for configuring the kernel. Note that the last option might still prompt with unresolved/new configuration options, if you have selected <Y> or <L> in the previous step.

==> An non-generic CPU was selected for this kernel.
==> Hit <G>     :  to create a generic package named linux-pf
==> Hit <ENTER> :  to create a package named after the selected CPU
                   (e.g. linux-pf-core2 - recommended)
==> This option affects ONLY the package name. Whether or not the
==> kernel is optimized was determined at the previous config step.

If you have selected a specific CPU optimization for your kernel in the previous step, the default action is to append the CPU to the package name. This way, a subsequent package update from the repository will pull the optimized package and not the generic one. This also will help better compatibility with 3rd party precompiled modules (e.g. nvidia-pf), which might break things if loaded on optimized linux-pf kernels.

Installation

After the compilation finishes, an additional linux-pf-headers[-cpu] package will be created. Don't forget to install it too, if you plan on using additional modules like nvidia or virtualbox. During the kernel installation, mkinitcpio will be called to create the initramfs.

If you intend to use TuxOnIce for hibernation, make sure you have added the necessary modules to the MODULES array of /etc/mkinitcpio.conf and at least the resume hook to the HOOKS array:

MODULES="... lzo tuxonice_compress tuxonice_swap tuxonice_userui ..."
HOOKS="... sata userui resume filesystems ..."

In the example above, TuxOnIce is setup to use a swap partition as the suspended image allocator. The resume hook must be placed before filesystems. Also, a progress indicator is requested with userui. Please read the TuxOnIce wiki page for more detailed information.

You need to add a boot entry in /boot/grub/menu.lst (or /boot/grub/grub.cfg) which points to linux-pf (the following example is from one of the maintainer's boxes):

title  Linux-pf 3.2
root   (hd0,4)
kernel (hd0,0)/vmlinuz-linux-pf root=/dev/disk/by-label/ROOT ro vga=0x318 lapic resume=/dev/disk/by-label/SWAP video=vesafb:ywrap,mtrr:3 fastboot quiet
initrd (hd0,0)/initramfs-linux-pf.img

Last, you must choose whether you want to suspend using pm-utils or the hibernate-script. Please, refer to the respective wiki pages for more details. TuxOnIce offers the option for a text mode or an even nicer framebuffer splash progress indicator.

Note: Due to a bug in TuxOnIce, introduced in the days of kernel26-pf 2.6.39-pf4, hibernating with LZO compression corrupts the snapshotted image and resuming fails in decompression. The bug was recently acknowledged by the TuxOnIce author, who suggested adding "ProcSetting no_multithreaded_io 1" to /etc/hibernate/tuxonice.conf until it gets fixed. This workaround has been confirmed to work, adding only 1-2 seconds of delay to hibernation.

Tweaks / Tips

  • If you notice disk-related performance problems or occational hickups, it might be an I/O scheduler issue. Try a different one than the linux-pf default (BFQ) by echoing to /sys/block/sda/queue/scheduler cfq, noop or deadline: echo noop >| /sys/block/sda/queue/scheduler. Note, the aforementioned command only sets the I/O scheduler for the 1st hard drive and additional echos will be needed if you have more. If the situation improves, then append "elevator=cfq" (or noop or deadline) to the linux-pf command line in /boot/grub/menu.lst, to make the change permanent.
  • For people who build their own tailored kernels and compilation aborts with with an error about "missing include/config/dvb/*.h files", setting <M> at DVB for Linux at Device Drivers/Multimedia support and leaving everything else out, creates just the necessary dvb.h, which allows the compilation to continue.

Forum thread for linux-pf

There's a discussion thread at the BBS for reporting errors, impressions, ideas and requests.