2.6.20-viper4: Take No Prisoner | Multiple CPU Schedulers

Message

vipernicus · Post by **vipernicus** » Fri Feb 23, 2007 2:30 pm

Viper Sources is pretty much my playground for new patches.

The main goal of this patchset is pretty much to offer the latest features, decrease latencies, decrease overhead, and improve interactivity. The only difference is experimental versions of everything. I boot test each release and run for an extended amount of time.

/* Notes */
02/23/2007
This is mostly a maintenance release. Updated to 2.6.20.1, RSDL 0.19 (and optional update to 0.21), and merged all hotfixes. While RSDL 0.19, is very stable, RSDL 0.21 is marked as (BUGGY). Also rolled up a few patches from -mm and called it 'driver-net-fixes-2.6.20.patch', which will hopefully fix some NIC IRQ issues, if not let me know.

New feature from Con Kolivas, Hz madness, which should fullfill all of your Hz needs.
Info here:
http://bhhdoa.org.au/pipermail/ck/2007- ... 06738.html
http://bhhdoa.org.au/pipermail/ck/2007- ... 06740.html
http://bhhdoa.org.au/pipermail/ck/2007- ... 06741.html

Extended changelog, downloads, and support requests have been moved to http://forums.nesl247.org/viewforum.php?f=18

RSDL Update (Optional):
http://vipernicus.nesl247.org/files/pat ... 0.21.patch
http://vipernicus.nesl247.org/files/pat ... 0.22.patch
http://vipernicus.nesl247.org/files/pat ... 0.23.patch
http://vipernicus.nesl247.org/files/pat ... 0.25.patch
http://vipernicus.nesl247.org/files/pat ... 0.26.patch
http://vipernicus.nesl247.org/files/pat ... 0.28.patch

I would like to thank Con Kolivas and William Lee Irwin III for the original concepts of the pluggable scheduler, and also Peter Williams for PlugSched. Let's also not forget, joecool, predatorfreak, cheater-conrad, JasonF, and gang with nosched.

Info on the origins of the pluggable CPU scheduler:
http://kerneltrap.org/node/4111
http://lwn.net/Articles/109049/

If you want to chat with us, we can now be found in #nesl247 on irc.freenode.org.

rmh3093 · Post by **rmh3093** » Fri Feb 23, 2007 2:41 pm

ck wrote:...but just to prove a point I've booted a
kernel at home with HZ=9861 which is the highest I've modified the code to
support, and am writing this email from that kernel which seems surprisingly
to work fine...

lol, cant wait to try this

predatorfreak · Post by **predatorfreak** » Fri Feb 23, 2007 3:12 pm

Oh you dirty Hz ricer, you!

rmh3093 · Post by **rmh3093** » Fri Feb 23, 2007 3:18 pm

predatorfreak wrote:Oh you dirty Hz ricer, you!

i've never noticed a difference switching between 100 or 1000.... im just curious what this will do

RobertDavid · Post by **RobertDavid** » Fri Feb 23, 2007 3:43 pm

I have noticed difference between 100 and 1000, on my laptop with P3 600, with 100 it is little, little bit less responsive, but dynamicly calculated CPU frequency during idle is ~2 MHz, with 1000 it is 20+ MHz, so I thing that I can save maybe few minutes of batterylife.

EDIT: It looks lile working well with the new version of RSDL..

vipernicus · Post by **vipernicus** » Fri Feb 23, 2007 3:59 pm

rmh3093 wrote:
predatorfreak wrote:Oh you dirty Hz ricer, you!
i've never noticed a difference switching between 100 or 1000.... im just curious what this will do

These values are really for buggy game servers. If you read the help topics, Con recommends 864hz for low latency desktops.

tsuehpsyde · Post by **tsuehpsyde** » Fri Feb 23, 2007 5:51 pm

rmh3093 wrote:
predatorfreak wrote:Oh you dirty Hz ricer, you!
i've never noticed a difference switching between 100 or 1000.... im just curious what this will do

Trust me, it's noticable with things like gameservers.

I host CS:S servers, and a 1000HZ kernel is all you can run, unless you want really bad shot registration.

IvanMajhen · Post by **IvanMajhen** » Fri Feb 23, 2007 5:59 pm

Compiled with RSDL,preemtible, hz=864. rcu preemtible and runs fine.
Suspend2 now works rsdl. Viper3 didn't.
Thanks.

tranquilcool · Post by **tranquilcool** » Fri Feb 23, 2007 7:01 pm

new rsdl-21 works for me but after some minutes kernel reboots itself. rsdl-19 didn't work for me at all.

vipernicus · Post by **vipernicus** » Fri Feb 23, 2007 7:18 pm

tranquilcool wrote:new rsdl-21 works for me but after some minutes kernel reboots itself. rsdl-19 didn't work for me at all.

Odd, I was able to run RSDL 0.19 on the desktop all day yesterday with no issues.

buddabrod · Post by **buddabrod** » Fri Feb 23, 2007 7:21 pm

vipernicus wrote:
tranquilcool wrote:new rsdl-21 works for me but after some minutes kernel reboots itself. rsdl-19 didn't work for me at all.
Odd, I was able to run RSDL 0.19 on the desktop all day yesterday with no issues.

Works here fine, too. Absolutely no problems.

vipernicus · Post by **vipernicus** » Sat Feb 24, 2007 2:51 am

RSDL Update (Optional):
http://vipernicus.nesl247.org/files/pat ... 0.21.patch
http://vipernicus.nesl247.org/files/pat ... 0.22.patch

mbar · Post by **mbar** » Sat Feb 24, 2007 7:30 am

yeah

darklegion · Post by **darklegion** » Sat Feb 24, 2007 8:15 am

When running with staircase, certain apps exit with a 'Killed' message.This includes:
wine 0.9.22
epsxe

This didn't occur with nicksched or RSDL 0.22.I'm not too fussed if its fixed or not, though.Nicksched still rules the cpu schedulers for me

univac^ · Post by **univac^** » Sat Feb 24, 2007 12:04 pm

rsdl is prefered for servers, or for dsktops too?

ashtophet · Post by **ashtophet** » Sat Feb 24, 2007 2:20 pm

univac^ wrote:rsdl is prefered for servers, or for dsktops too?

As stated from Con Kolivas on ck mailing list, It's mainly aimed to servers (maybe he'll be including it on cks):

Con Kolivas wrote: Staircase in -ck _will_ have better interactivity than this RSDL scheduler due
to its controlled unfairness that it meters out to interactive tasks. At the
moment I'm not sure if this will ever be a drop in replacement for staircase,
but the server edition of -ck would certainly benefit from it.

btw... great great patchset, thanks

_sys/sid · Post by **_sys/sid** » Sat Feb 24, 2007 3:05 pm

I've tried staircase and default ingo sched on my system with Reiser 4...
and i think that ingo sched & Reiser4 is much faster and responsive then staircase & Reiser4
someone have tried other scheduler with Reiser4?

santaclaws · Post by **santaclaws** » Sat Feb 24, 2007 9:11 pm

Unfortunately I get this error message:

Code: Select all

 genkernel --kernel-cc=/usr/lib/ccache/bin/gcc --xconfig all
* Gentoo Linux Genkernel; Version 3.4.6
* Running with options: --kernel-cc=/usr/lib/ccache/bin/gcc --xconfig all

* Linux Kernel 2.6.20-viper4 for x86...
* kernel: >> Running mrproper...
* config: Using config from /usr/share/genkernel/x86/kernel-config-2.6
*         Previous config backed up to .config.bak
*         >> Running oldconfig...
* kernel: >> Cleaning...
* config: >> Invoking xconfig...
*         >> Compiling 2.6.20-viper4 bzImage...
* ERROR: Failed to compile the "bzImage" target...

* -- Grepping log... --

  SHIPPED scripts/kconfig/lex.zconf.c
  SHIPPED scripts/kconfig/zconf.hash.c
  HOSTCC  scripts/kconfig/zconf.tab.o
  HOSTLD  scripts/kconfig/conf
scripts/kconfig/conf -o arch/i386/Kconfig
.config:191:warning: trying to assign nonexistent symbol REGPARM
.config:274:warning: trying to assign nonexistent symbol X86_SPEEDSTEP_CENTRINO_TABLE
.config:421:warning: trying to assign nonexistent symbol NET_DIVERT
.config:723:warning: trying to assign nonexistent symbol SCSI_SATA
.config:724:warning: trying to assign nonexistent symbol SCSI_SATA_AHCI
.config:725:warning: trying to assign nonexistent symbol SCSI_SATA_SVW
.config:726:warning: trying to assign nonexistent symbol SCSI_ATA_PIIX
.config:727:warning: trying to assign nonexistent symbol SCSI_SATA_MV
.config:728:warning: trying to assign nonexistent symbol SCSI_SATA_NV
.config:729:warning: trying to assign nonexistent symbol SCSI_PDC_ADMA
.config:730:warning: trying to assign nonexistent symbol SCSI_SATA_QSTOR
.config:731:warning: trying to assign nonexistent symbol SCSI_SATA_PROMISE
.config:732:warning: trying to assign nonexistent symbol SCSI_SATA_SX4
.config:733:warning: trying to assign nonexistent symbol SCSI_SATA_SIL
.config:734:warning: trying to assign nonexistent symbol SCSI_SATA_SIL24
.config:735:warning: trying to assign nonexistent symbol SCSI_SATA_SIS
.config:736:warning: trying to assign nonexistent symbol SCSI_SATA_ULI
.config:737:warning: trying to assign nonexistent symbol SCSI_SATA_VIA
.config:738:warning: trying to assign nonexistent symbol SCSI_SATA_VITESSE
.config:739:warning: trying to assign nonexistent symbol SCSI_SATA_INTEL_COMBINED
.config:773:warning: trying to assign nonexistent symbol SCSI_QLA2XXX_EMBEDDED_FIRMWARE
.config:812:warning: trying to assign nonexistent symbol MD_RAID5
.config:814:warning: trying to assign nonexistent symbol MD_RAID6
.config:1079:warning: trying to assign nonexistent symbol IPW_QOS
.config:1143:warning: trying to assign nonexistent symbol SYNCLINK_SYNCPPP
.config:1635:warning: trying to assign nonexistent symbol FB_FIRMWARE_EDID
.config:1693:warning: trying to assign nonexistent symbol SPEAKUP
.config:1694:warning: trying to assign nonexistent symbol SPEAKUP_ACNTSA
.config:1695:warning: trying to assign nonexistent symbol SPEAKUP_ACNTPC
.config:1696:warning: trying to assign nonexistent symbol SPEAKUP_APOLLO
.config:1697:warning: trying to assign nonexistent symbol SPEAKUP_AUDPTR
.config:1698:warning: trying to assign nonexistent symbol SPEAKUP_BNS
.config:1699:warning: trying to assign nonexistent symbol SPEAKUP_DECTLK
.config:1700:warning: trying to assign nonexistent symbol SPEAKUP_DECEXT
.config:1701:warning: trying to assign nonexistent symbol SPEAKUP_DECPC
.config:1702:warning: trying to assign nonexistent symbol SPEAKUP_DTLK
.config:1703:warning: trying to assign nonexistent symbol SPEAKUP_KEYPC
.config:1704:warning: trying to assign nonexistent symbol SPEAKUP_LTLK
.config:1705:warning: trying to assign nonexistent symbol SPEAKUP_SFTSYN
.config:1706:warning: trying to assign nonexistent symbol SPEAKUP_SPKOUT
.config:1707:warning: trying to assign nonexistent symbol SPEAKUP_TXPRT
.config:1712:warning: trying to assign nonexistent symbol SPEAKUP_DEFAULT
.config:1750:warning: trying to assign nonexistent symbol SND_AC97_BUS
.config:1926:warning: trying to assign nonexistent symbol USB_HIDINPUT
.config:1989:warning: trying to assign nonexistent symbol USB_SERIAL_ANYDATA
.config:2145:warning: trying to assign nonexistent symbol XFS_EXPORT
.config:2339:warning: trying to assign nonexistent symbol UNWIND_INFO
.config:2345:warning: trying to assign nonexistent symbol STACK_BACKTRACE_COLS
--
    2. MMConfig (PCI_GOMMCONFIG)
    3. Direct (PCI_GODIRECT)
  > 4. Any (PCI_GOANY)
  choice[1-4?]: 4
  PCI Express support (PCIEPORTBUS) [Y/n/?] y
    Root Port Advanced Error Reporting support (PCIEAER) [Y/n/?] (NEW)   Message Signaled Interrupts (MSI and MSI-X) (PCI_MSI) [N/y/?] n
--
SCSI media changer support (CHR_DEV_SCH) [M/n/y/?] m
*
* Some SCSI devices (e.g. CD jukebox) support multiple LUNs
*
Probe all LUNs on each SCSI device (SCSI_MULTI_LUN) [N/y/?] n
Verbose SCSI error reporting (kernel size +=12K) (SCSI_CONSTANTS) [N/y/?] n
--
    Verbose debugging output (INFINIBAND_MTHCA_DEBUG) [N/y/?] n
  Ammasso 1100 HCA support (INFINIBAND_AMSO1100) [N/m/?] (NEW)   IP-over-InfiniBand (INFINIBAND_IPOIB) [M/n/?] m
    IP-over-InfiniBand debugging (INFINIBAND_IPOIB_DEBUG) [N/y/?] n
  InfiniBand SCSI RDMA Protocol (INFINIBAND_SRP) [M/n/?] m
  iSCSI Extensions for RDMA (iSER) (INFINIBAND_ISER) [N/m/?] (NEW) *
* EDAC - error detection and reporting (RAS) (EXPERIMENTAL)
*
EDAC core system error reporting (EXPERIMENTAL) (EDAC) [N/m/y/?] n
--
/usr/qt/3/bin/moc -i scripts/kconfig/qconf.h -o scripts/kconfig/qconf.moc
  HOSTCC  scripts/kconfig/kconfig_load.o
  HOSTCXX scripts/kconfig/qconf.o
  HOSTLD  scripts/kconfig/qconf
scripts/kconfig/qconf arch/i386/Kconfig
Session management error: Authentication Rejected, reason : None of the authentication protocols specified are supported and host-based authentication failed
--
: undefined reference to `__mutex_fastpath_lock_retval'
make: *** [vmlinux] Fehler 1
* Gentoo Linux Genkernel; Version 3.4.6
* Running with options: --kernel-cc=/usr/lib/ccache/bin/gcc --xconfig all

* ERROR: Failed to compile the "bzImage" target...

* -- End log... --

* Please consult /var/log/genkernel.log for more information and any
* errors that were reported above.

* Report any genkernel bugs to bugs.gentoo.org and
* assign your bug to genkernel@gentoo.org. Please include
* as much information as you can in your bug report; attaching
* /var/log/genkernel.log so that your issue can be dealt with effectively.
* 
* Please do *not* report compilation failures as genkernel bugs!
*

Do you have a suggestion, Vipernicus?

I have tried to compile it several times with different features enabled, but it always stops at the same point..

vipernicus · Post by **vipernicus** » Sat Feb 24, 2007 9:25 pm

santaclaws wrote:Do you have a suggestion, Vipernicus?

I have tried to compile it several times with different features enabled, but it always stops at the same point..

Remove debugging options.

santaclaws · Post by **santaclaws** » Sat Feb 24, 2007 9:33 pm

Ok, I will try it immediately...

santaclaws · Post by **santaclaws** » Sat Feb 24, 2007 11:01 pm

Hey Vipernicus.. thank you for your advice. It compiled successfully...

vipernicus · Post by **vipernicus** » Sun Feb 25, 2007 3:44 am

santaclaws wrote:Hey Vipernicus.. thank you for your advice. It compiled successfully...

Awesome, I hope to get a hotfix for the debugging by next release.

santaclaws · Post by **santaclaws** » Sun Feb 25, 2007 12:34 pm

:!:I found another issue but I do not know if it only concerns my laptop:

If I run "dmesg" the output at the begining is cut off. That was not the case before.

Example:

Code: Select all

localhost santaclaws # dmesg
is hidden behind transparent bridge #01 (-#01) (try 'pci=assign-busses')
Please report the result to linux-kernel to fix this permanently
PCI: Bus #06 (-#09) is hidden behind transparent bridge #01 (-#01) (try 'pci=assign-busses')
Please report the result to linux-kernel to fix this permanently
ACPI: PCI Interrupt Routing Table [\_SB_.PCI0._PRT]
 (..)

But the performance, especially the interactivity during compilation, is amazing!

vipernicus · Post by **vipernicus** » Sun Feb 25, 2007 3:49 pm

santaclaws wrote::!:I found another issue but I do not know if it only concerns my laptop:

If I run "dmesg" the output at the begining is cut off. That was not the case before.

Example:
Code: Select all
localhost santaclaws # dmesg
is hidden behind transparent bridge #01 (-#01) (try 'pci=assign-busses')
Please report the result to linux-kernel to fix this permanently
PCI: Bus #06 (-#09) is hidden behind transparent bridge #01 (-#01) (try 'pci=assign-busses')
Please report the result to linux-kernel to fix this permanently
ACPI: PCI Interrupt Routing Table [\_SB_.PCI0._PRT]
 (..)
But the performance, especially the interactivity during compilation, is amazing!

Very odd. What terminal emulator are you using?

Phlogiston · Post by **Phlogiston** » Sun Feb 25, 2007 9:27 pm

Hey viper I had some problems with viper3 and hibernate: When I resumed, I sometimes got a kernel panic, very ugly. I will try out this viper4 and I hope to never encounter that problem again