On Sun, Oct 18, 2009 at 3:56 PM, Ray Rashif <[email protected]> wrote: > 2009/10/18 P.V.Anthony <[email protected]> >> >> Hi, >> >> Was going to compile the latest 2.6 kernel and noticed that I do not know >> what all the options mean. Especially the "General Setup" and "Processor >> Type and Features". >> >> Is there some detail guide explaining the options, about the meaning and >> implications of these options for an Intel Xeon multi-core cpu? >> >> Any web links or books? >> >> P.V.Anthony > > You could press "?" on each option and get a rough summary with menuconfig. > The Linux Kernel in a Nutshell is a good book, and there's an online version > at: > http://www.linuxtopia.org/online_books/linux_kernel/kernel_configuration/ > But that one is mostly a technical guide, not necessarily what you're > looking for. You could give up on finding up-to-date "tutorials", especially > with the massive progress there has been over the years. I remember when .19 > rolled out, when pata deprecated ide, I had to dig through and read the > changelogs to see where the options had moved to. > If you're building for the desktop or workstation, there aren't many > "tunables". Selecting the processor type can yield a slightly better > performance from the optimisation, but not so much as you go down the scale. > Removing options and modules can reduce the kernel size, but that doesn't > make much difference today as it did in 1995. Coupled with SMP, Dynticks > (the defaults basically), you're good to go. Take a look at the staging > drivers for new and testing development stuff; there may just be salvation > for hardware that didn't work before. > As for the building, you can take a look at buildscripts like > http://aur.archlinux.org/packages/kernel26rt/kernel26rt/PKGBUILD and adapt > it to your needs (look inside the build function). For example, this one > patches the kernel for realtime (hard) preemption which basically is almost > anti-scheduler, meant for embedded and industrial applications as well as > audio. Using the Deadline I/O sched can improve the performance under > certain workloads against CFQ. > You may also want to check out BFS: http://ck.kolivas.org/patches/bfs/ > Though some claim it to be "bloat", I personally don't care because the > response I'm getting is simply awesome to say the least. It's not something > hypothetical, you can actually notice the difference especially for > multimedia (playing videos, fullscreen flash HD, running a multitude of > processes yet not facing any kind of stutter anywhere, etc.). CFS fails in > comparison. > _______________________________________________ > LUGS Mailing list - [email protected] > List FAQ: http://wiki.lugs.org.sg/LugsMailingListFaq > Info page: http://www.lugs.org.sg/mailman/listinfo/slugnet > To unsubscribe send an empty email to: [email protected] > >
May be you already have a running kernel that can be used as a base for your config. That's often what I do when I recompile, I start from some config for a particular kernel -- with corresponding kernel sources --, then I fine tune. For exemple, say I have the following in /boot: $ ls /boot/*2.6.30* /boot/config-2.6.29.6-217.2.16.fc11.i586 /boot/config-2.6.30.8-64.fc11.i586 /boot/config-2.6.30.5-43.fc11.i586 /boot/config-2.6.30.8-64.homebuild I could use the config of my current running kernel as a base, and fine tune it. To do so (assuming you have a /usr/src/linux pointing to your kernel sources): $ cd /usr/src/linux $ make mrproper #only required if your tree is not clean $ cp /boot/config-2.6.30.8-64.homebuild .config $ make oldconfig $ make prepare #only required if you just cleaned your tree with mrproper After this step, you can run your favourite configuration interface: make xconfig, make menuconfig, etc... Be careful, the kernel source you're using should correspond to the precedent configuration. Cheers, _______________________________________________ LUGS Mailing list - [email protected] List FAQ: http://wiki.lugs.org.sg/LugsMailingListFaq Info page: http://www.lugs.org.sg/mailman/listinfo/slugnet To unsubscribe send an empty email to: [email protected]
