| |||||||||||||
The 2.6.13 kernel is out
Linus has announced the availability of the
2.6.13 kernel. Only a small number of relatively important fixes went in
since -rc7. For those just tuning in, 2.6.13 includes inotify, support for the Xtensa architecture,
kexec and kdump, execute-in-place support, a configuration-time selectable clock interrupt
frequency (the default for i386 changes to 250 Hz), a much-improved CFQ I/O scheduler
with I/O priority support, the voluntary preemption patches, and more. The
long-format changelog contains the details
for the patches merged since 2.6.13-rc7.
(Log in to post comments)
Looks like devfs is really gone this time Posted Aug 29, 2005 3:52 UTC (Mon) by donio (subscriber, #94) [Link] The files are still there in fs/devfs but the CONFIG_DEVFS_FS optionis not in make config anymore.
I am still using devfs in a somewhat special environment, I guess I
Looks like devfs is really gone this time Posted Aug 29, 2005 4:49 UTC (Mon) by set (guest, #4788) [Link] Yeah, I noticed that when testing an rc release not too long ago,and after fighting off my innate laziness, typed 'emerge udev', and was pleasantly suprised that nothing else was required of me;) This is under gentoo, obviously, one of the last devfs holdouts; what are all the 'major' distributions using now?
(I think devfs is still there, just the config option is missing, just
Looks like devfs is really gone this time Posted Aug 29, 2005 5:05 UTC (Mon) by rahulsundaram (subscriber, #21946) [Link]
Pretty much all the distributions have switched over to udev now.
Rahul
Looks like devfs is really gone this time Posted Aug 29, 2005 6:24 UTC (Mon) by gavino (guest, #16214) [Link] OpenWRT, which is an embedded distro I've installed on my Linksys WRT54G wireless router, is running DevFS. The point is moot though as OpenWRT uses the 2.4 kernel. It is a 'current distro' though, albeit an esoteric one.
Looks like devfs is really gone this time Posted Aug 29, 2005 17:59 UTC (Mon) by ncm (subscriber, #165) [Link] The WRT distros still use 2.4 because they depend on a binary-only driver for the Broadcom 802.11x chips, right?
Looks like devfs is really gone this time Posted Aug 29, 2005 20:11 UTC (Mon) by nchip (guest, #13292) [Link] That, and the fact that 2.6 kernels are bigger (atleast without serioustweaking), and thus slow on on systems with little memory and slow caches. There is broadcom driver reverse engineering effort at http://linux-bcom4301.sourceforge.net/go/progress
Looks like devfs is really gone this time Posted Aug 29, 2005 21:27 UTC (Mon) by kundor (guest, #14621) [Link] well, gentoo isn't really a devfs holdout, considering it's actually theplatform that Greg KH developed udev on and the first distro to distribute and support it. Gentoo is in fact the udev 'flagship.'
Looks like devfs is really gone this time Posted Aug 29, 2005 20:17 UTC (Mon) by job (guest, #670) [Link] I used devfs but switched my work computer to udev when Gentoo moved to it. It has caused many annoyances for me. No big problems or anything, but not as nice from the user perspective as devfs was. Examples include:1. The kernel still needs the console (perhaps also the root device?) inside /dev. I have no choice but to manually create those nodes. Not a problem, but not a completely dynamic system. 2. The nice namespace like /dev/discs doesn't work anymore. This is probably just a bug with the scripts, one I could fix, but is an annoyance. 3. ptys and ttys are not dynamically created, there's an insane amount of them. 4. Old devices are left under /dev. I had six partitions on sda for a while, they are removed now, but all the devices nodes of all six are still there.
The 2.6.13 kernel is out Posted Aug 29, 2005 8:29 UTC (Mon) by lacostej (guest, #2760) [Link] Sounds like a lot of nice features went into this release. The new kernel development model seems to work very well.
The 2.6.13 kernel is out Posted Aug 29, 2005 8:58 UTC (Mon) by dwalters (subscriber, #4207) [Link] Well said!
I'd hate to have had to wait for months to get my hands on some of these features in a stable kernel. On the flip side, it remains to be seen just how stable 2.6.13 is now that it has been released into the wild.
Although not really a stability issue per se, I imagine that changing the default value of HZ from 1000 to 250 will "break" some multimedia applications (but at least tech-savvy users will be able to reconfigure it back to 1000 and recompile the kernel).
CFQ priorities Posted Aug 29, 2005 11:36 UTC (Mon) by thames (guest, #238) [Link] Is there any command (similar to nice) that changes a process' io-priority?It seems that CFQ priorities are inherited from the normal (cpu) priority, so it would be nice with a simple way of changeing them.
CFQ priorities Posted Aug 29, 2005 13:12 UTC (Mon) by joib (guest, #8541) [Link] Were would it actually be needed?
My guess is that processes which are deemed important enough to get more cpu time also should get more I/O time and vice versa.
Or to put it another way, since I/O is orders of magnitude slower than just about anything else, if I/O is the bottleneck then the I/O heavy process whould also get a high cpu priority, so that when it needs cpu it gets it so it can go back to doing I/O ASAP.
Certainly one could design any number of synthetic benchmarks were different cpu and I/O priorities would be beneficial, but do these situations really occur in real life?
CFQ priorities Posted Aug 29, 2005 16:49 UTC (Mon) by larryr (guest, #4030) [Link] I may want an IO bound process to have a lower IO priority than other processes, but that doesnt mean I necessarily want it to have to wait for CPU time when it is ready for it. By definition an IO bound process is not going to be competing much with other IO bound processes for CPU time, so it doesnt seem to make much sense to force them to use CPU priority for their IO contention, when that will put them at a disadvantage relative to CPU bound processes; it seems reasonable for all IO bound processes to have a high CPU priority, as has traditionally been the case. Larry
CFQ priorities Posted Aug 29, 2005 13:36 UTC (Mon) by LogicG8 (guest, #11076) [Link] Robert Love's ionice is in linux-utils as of version 2.13-pre1IIRC it was previously contained in a package called sched-utils
CFQ priorities Posted Aug 29, 2005 18:12 UTC (Mon) by ncm (subscriber, #165) [Link] If I can't apt-get it, it doesn't exist. :-)
CFQ priorities Posted Aug 29, 2005 18:40 UTC (Mon) by jwb (guest, #15467) [Link] Package: schedutilsPriority: optional Section: universe/utils Installed-Size: 84 Maintainer: Guus Sliepen <guus@debian.org> Architecture: amd64 Version: 1.4.0-1 Depends: libc6 (>= 2.3.4-1) Filename: pool/universe/s/schedutils/schedutils_1.4.0-1_amd64.deb Size: 16560 MD5sum: e2c7d859db4ad078a1dbe029933800ea
CFQ priorities Posted Aug 29, 2005 20:28 UTC (Mon) by micampe (guest, #4384) [Link] The programs that are included in this package are taskset and chrt. Together with nice and renice (not included), they allow full control of process scheduling parameters. Suggestions for related utilities are welcome, although it is believed (barring new interfaces) that all scheduling interfaces are covered.... ;)
CFQ priorities Posted Aug 30, 2005 1:40 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link] in that case you can ignore this entire topic as 2.6.13 doesn't exist for you
CFQ priorities Posted Aug 30, 2005 0:08 UTC (Tue) by darthmdh (guest, #8032) [Link] Jens posted an ionice utility for cfq to the ck mailing list a while ago.Check the archives off http://ck.kolivas.org/
I only tested it once, it did successfully cripple an annoying cron job just as I asked it to ;)
|
|||||||||||||