By popular demand, here is release -v22 of the CFS scheduler. It is a

full backport of the latest & greatest sched-devel.git code to

v2.6.23-rc8, v2.6.22.8, v2.6.21.7 and v2.6.20.20. The patches can be

downloaded from the usual place:
    http://people.redhat.com/mingo/cfs-scheduler/
This is the first time the development version of the scheduler has been

fed back into the stable backport series, so there's many changes since

v20.5:
 15 files changed, 1103 insertions(+), 840 deletions(-)
Even if CFS v20.5 worked well for you, please try this release too, with

a good focus on interactivity testing - because, unless some major

showstopper is found, this codebase is intended for a v2.6.24 upstream

merge.
( Even a quick, subjective report of: "checked this patch, it didnt

  crash and it feels like v20.5" or "laggier than v20.5" or "feels

  better than v20.5" is useful to us and enables us to judge the general

  direction of interactivity. )
The changes in v22 consist of lots of mostly small enhancements,

speedups, interactivity improvements, debug enhancements and tidy-ups -

many of which can be user-visible. (These enhancements have been

contributed by many people - see the changelog below and the git tree

for detailed credits.)
The biggest individual new feature is per UID group scheduling, written

by Srivatsa Vaddagiri, which can be enabled via the

CONFIG_FAIR_USER_SCHED=y .config option. With this feature enabled, each

user gets a fair share of the CPU time, regardless of how many tasks

each user is running.
For example, it took me 0.1 seconds to log in over ssh as root on a

testbox that was running a kernel with per UID group scheduling enabled:
  $ time ssh root@testbox /bin/true
  real    0m0.125s

  user    0m0.013s

  sys     0m0.011s
Which testbox had a system load of 1000.17 at this time, due to a rogue

runaway workload of one thousand (!) non-reniced infinite loops:
  top - 14:34:05 up 30 min,  3 users,  load average: 1000.17...
[ message continues ]

Better late than never...
Hi Ingo & cie.,
I tested CFS v22 on a 2.6.20.10 and it is working like a charm.  I used to have a few lags once in a while with v20.5 on a 2.6.20.6 kernel but they simply vanished with the v22 (although I'm currenctly mostly testing on etch with xorg 7.1 instead of sarge with xfree 4.3 so this might also have an effect...)
It seems stable enough to put it in operation by next week on all our workstations.
This is great work, thx to all CFS contributors.
- vin
-

El Wed, 26 Sep 2007 13:13:57 +0200
 I feel it better than 20.5 but the later is more stable. Let me explain.
 I patched a 2.6.22.9 kernel with both versions 22 and 20.5[1]. With the 22

 version if i lock the screen (run screensaver) or i try to run a wine program

 i experience a hard lock up no keyboard or mouse and i have to reboot the

 machine (i can not try to access it via ssh because i do not have a second

 machine).

 I run with the nvidia kernel module so I know my report is useless but just

 for the record...

 Here it is my config:
 [1] Hand edited the Makefile hunk no other rejects
#

# Automatically generated make config: don't edit

# Linux kernel version: 2.6.22.9-cfs-v20.5

# Fri Sep 28 16:12:07 2007

#

CONFIG_X86_32=y

CONFIG_GENERIC_TIME=y

CONFIG_CLOCKSOURCE_WATCHDOG=y

CONFIG_GENERIC_CLOCKEVENTS=y

CONFIG_GENERIC_CLOCKEVENTS_BROADCAST=y

CONFIG_LOCKDEP_SUPPORT=y

CONFIG_STACKTRACE_SUPPORT=y

CONFIG_SEMAPHORE_SLEEPERS=y

CONFIG_X86=y

CONFIG_MMU=y

CONFIG_ZONE_DMA=y

CONFIG_QUICKLIST=y

CONFIG_GENERIC_ISA_DMA=y

CONFIG_GENERIC_IOMAP=y

CONFIG_GENERIC_BUG=y

CONFIG_GENERIC_HWEIGHT=y

CONFIG_ARCH_MAY_HAVE_PC_FDC=y

CONFIG_DMI=y

CONFIG_DEFCONFIG_LIST="/lib/modules/$UNAME_RELEASE/.config"
#

# Code maturity level options

#

CONFIG_EXPERIMENTAL=y

CONFIG_LOCK_KERNEL=y

CONFIG_INIT_ENV_ARG_LIMIT=32
#

# General setup

#

CONFIG_LOCALVERSION=""

CONFIG_LOCALVERSION_AUTO=y

CONFIG_SWAP=y

CONFIG_SYSVIPC=y

# CONFIG_IPC_NS is not set

CONFIG_SYSVIPC_SYSCTL=y

CONFIG_POSIX_MQUEUE=y

CONFIG_BSD_PROCESS_ACCT=y

# CONFIG_BSD_PROCESS_ACCT_V3 is not set

# CONFIG_TASKSTATS is not set

# CONFIG_UTS_NS is not set

# CONFIG_AUDIT is not set

CONFIG_IKCONFIG=y

CONFIG_IKCONFIG_PROC=y

CONFIG_LOG_BUF_SHIFT=15

# CONFIG_CPUSETS is not set

CONFIG_SYSFS_DEPRECATED=y

CONFIG_RELAY=y

CONFIG_BLK_DEV_INITRD=y

CONFIG_INITRAMFS_SOURCE=""

# CONFIG_CC_OPTIMIZE_FOR_SIZE is not set

CONFIG_SYSCTL=y

CONFIG_EMBEDDED=y

# CONFIG_UID16 is not set

# CONFIG_SYSCTL_SYSCALL is not set

# CONFIG_KALLS...
[ message continues ]

hm. There are no similar reports so far. Could you try to run the wine

program from a VGA text console (first do 'xhost +' in an xterm under X

and then 'export DISPLAY=:0' in the text console) and see whether

anything gets output to that console? Also enable nmi_watchdog=2 - which

could print a backtrace of the point of the hard lockup.
but in general there's nothing scheduler-specific about screensavers or

wine - but it could be related to 3D mode of the chip (hence related to

the nvidia module) - do you get a similar lockup if you try to run

'glxgears'?
	Ingo

-

El Sun, 30 Sep 2007 17:26:22 +0200
 Ignore the last mail there was no need to use pictures this no an oops on

boot ;)
This is the back trace

  Oct  2 17:37:13 localhost kernel: [  886.216868] BUG: unable to handle kernel NULL pointer dereference at virtual address 0000012c

Oct  2 17:37:13 localhost kernel: [  886.216874]  printing eip:

Oct  2 17:37:13 localhost kernel: [  886.216876] f9374d0c

Oct  2 17:37:13 localhost kernel: [  886.216877] *pde = 00000000

Oct  2 17:37:13 localhost kernel: [  886.216880] Oops: 0002 [#1]

Oct  2 17:37:13 localhost kernel: [  886.216881] PREEMPT SMP

Oct  2 17:37:13 localhost kernel: [  886.216884] Modules linked in: nls_cp850 isofs udf nls_utf8 ntfs binfmt_misc rfcomm l2cap bluetooth ipt_ULOG x_tables jfs ipv6 dm_crypt af_packet kvm_amd kvm fuse w83627ehf hwmon_vid i2c_dev lp snd_mpu401 snd_mpu401_uart snd_hda_intel snd_pcm_oss snd_mixer_oss snd_seq_dummy snd_pcm snd_seq_oss snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq snd_timer nvidia(P) snd_seq_device usbhid i2c_ali15x3 i2c_ali1535 8250_pnp usblp ff_memless uli526x snd i2c_core 8250 serial_core button ali_agp parport_pc parport k8temp hwmon pcspkr evdev rtc snd_page_alloc rt2500 soundcore sg floppy sr_mod raid10 cdrom sd_mod ehci_hcd ohci_hcd ata_generic usbcore raid456 async_xor async_memcpy async_tx xor raid1 raid0 multipath linear md_mod thermal processor fan unix dm_mod

Oct  2 17:37:13 localhost kernel: [  886.216933] CPU:    0

Oct  2 17:37:13 localhost kernel: [  886.216934] EIP:    0060:[<f9374d0c>]    Tainted: P        VLI

Oct  2 17:37:13 localhost kernel: [  886.216935] EFLAGS: 00010246   (2.6.23-rc9-cfs-v22-dirty #82)

Oct  2 17:37:13 localhost kernel: [  886.217129] EIP is at os_set_mlock_capability+0xc/0x30 [nvidia]

Oct  2 17:37:13 localhost kernel: [  886.217131] eax: 00000000   ebx: 00000000   ecx: 00000050   edx: 00000202

Oct  2 17:37:13 localhost kernel: [  886.217134] esi: f74c3000   edi: f749cc00   ebp: df4b3ff0   esp: d319bed4

Oct  2 17:37:13 localhost kernel: [  886.217136] ds: ...
[ message continues ]

I have seen that twice with v20.5 under 2.6.22.7/.8. Didn't see it yet with

v22.  xterm would get pissed if I tried to "secure keyboard", so something

stole the keyboard focus and wouldn't give it back to xorg.  Maybe a locking

bug in xorg somewhere?
Anyway, my kernel is NOT tainted, so rest assured that at least this one is

not nVidia's fault.
BTW, v20.5 in 2.6.21.7 would sometimes cause dmcrypt to oops when first

mounting a LUKS volume if there was some non-extra-light disk activity going

on.  Probably something missing in the backport to 2.6.21...
Sorry, days down here have been very busy, so I didn't have time to send a

proper bug report.
--

  "One disk to rule them all, One disk to find them. One disk to bring

  them all and in the darkness grind them. In the Land of Redmond

  where the shadows lie." -- The Silicon Valley Tarot

  Henrique Holschuh

-

El Fri, 28 Sep 2007 23:20:13 -0300
 I can add another data point 2.6.23-rc8-cfs-v22-g1bef7dc0-dirty does not have

 any problem with wine or xscreensaver is working fine. Again i use nvidia and

 a 3 party gpl driver for my wifi rt2500 which have compiled fine with the new

 kernel (surprise surprise!)

 If you need any more info just ask.
 Thanks
 P.S: I use ubuntu feisty fwiw
--

Nunca discutas con un idiota. Al final te hacen rebajarte a su nivel y entonces

te acaban ganando debido a su mayor experiencia.
-

Compilation [against 2.6.20.20] fails with
buildfarm linux-2.6.20 # make

  CHK     include/linux/version.h

  CHK     include/linux/utsrelease.h

  CHK     include/linux/compile.h

  CC      kernel/sched.o

In file included from kernel/sched.c:850:

kernel/sched_fair.c:49: error: sysctl_sched_nr_latency causes a section type

conflict

make[1]: *** [kernel/sched.o] Error 1

make: *** [kernel] Error 2
@kernel/sched_fair.c

-const_debug unsigned int sysctl_sched_nr_latency __read_mostly = 20;

+const_debug unsigned int sysctl_sched_nr_latency = 20;
seems solve that issue but i'm not sure this is the right thing to do or

not :)
Cheers

--

S.Çağlar Onur <caglar@pardus.org.tr>

http://cekirdek.pardus.org.tr/~caglar/
Linux is like living in a teepee. No Windows, no Gates and an Apache in house!

-

oops, indeed! This happens with certain .config combinations only and it 
your fix is the right one - and i've updated the patches with this small

fix. Anyone who hits this issue should re-download the same patch once

more, or should apply the small patch below.
	Ingo
Index: linux/kernel/sched_fair.c

===================================================================

--- linux.orig/kernel/sched_fair.c

+++ linux/kernel/sched_fair.c

@@ -46,7 +46,7 @@ const_debug unsigned int sysctl_sched_ch

  * Minimal preemption granularity for CPU-bound tasks:

  * (default: 2 msec, units: nanoseconds)

  */

-const_debug unsigned int sysctl_sched_nr_latency __read_mostly = 20;

+const_debug unsigned int sysctl_sched_nr_latency = 20;
 /*

  * sys_sched_yield() compat mode

-