2.6.0-test4-mm5

2.6.0-test4-mm5

[Andrew Morton]

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/

. Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
  in evaluating the stability, efficacy and relative performance of Nick's
  work.

  We're looking for feedback on the subjective behaviour and on the usual
  server benchmarks please.


. Random other stuff.



Changes since 2.6.0-test4-mm4:


 linus.patch

 Latest from Linus

-handle-unreadable-dot-config.patch
-huge-net-update.patch
-disable-athlon-prefetch.patch
-sis900-atomicity-fix.patch
-x86_64-update-3.patch
-random-locking-fixes.patch
-random-accounting-and-sleeping-fixes.patch
-yenta-20030817-1-zv.patch
-yenta-20030817-2-override.patch
-yenta-20030817-3-sockinit.patch
-yenta-20030817-4-pm.patch
-yenta-20030817-5-pm2.patch
-yenta-20030817-6-init.patch
-yenta-20030817-7-quirks.patch
-proc-pid-setuid-ownership-fix.patch
-pid-revalidate-security-hook.patch
-dac960-GAM-IOCTLs-cleanup.patch
-kj-maintainers.patch
-ide-docs-update.patch
-v4l-use-after-free-fix.patch
-ikconfig-makefile-update.patch
-ftape-warning-fix.patch
-jffs-retval-fix.patch
-make-ACPI_SLEEP-select-SOFTWARE_SUSPEND.patch
-3GB-personality.patch
-zeromap_pmd_range-fix.patch
-no-async-write-errors-on-close.patch
-sis190-fix.patch
-remove-add_wait_queue_cond.patch
-spin_lock_irqrestore-fixes.patch
-pcmciamtd-fix.patch
-zoran-memleak-fixes.patch
-zoran-rename-debug.patch
-zoran-release-callback.patch
-zoran-pci_disable_device.patch
-zoran-cleanups.patch
-zoran-cleanups-2.patch
-zoran-naming-fix.patch
-airo-build-fix.patch
-m68k-vmlinux_lds-move.patch
-mac-ide-fix.patch
-m68k-asm-sections-fix.patch
-m68k-asm-local.patch
-amiga-z2ram-fix.patch
-amiga-floppy-fix.patch
-atari-floppy-fix.patch
-m68k-switch_to-fix.patch
-pcxx-warning-fix.patch
-pcnet32-unregister_pci-fix.patch
-hwifs-oops-unregister-fix.patch
-c99-conversions.patch
-cyc2x-fixes.patch
-noacpi-option-fix.patch
-h8300-interrupt-fix.patch
-proc-kallsyms-caching-fix.patch
-proc-kallsyms-permission-fix.patch
-cu3088-string-null-termination-fix.patch
-kobject-doc-addition.patch
-vm_enough_memory-speedup.patch
-abi-doc-update.patch
-remove-bio-boot-messages.patch
-ni5010-build-fix.patch
-sis190-build-fix.patch
-nopage-fix.patch
-parport_pc-rmmod-oops-fix.patch
-reiserfs-writepage-fix.patch
-visws-build-fix.patch
-cciss-queue-init-fix.patch
-htree-big-endian-fix.patch
-selinux-file-fcntl-fix.patch
-selinux-avtab-fix.patch
-selinux-format-specifiers-fix.patch
-selinux-binprm-hooks-rework.patch
-ext2-xattr-typo-fix.patch
-bad-inode-ops.patch
-kcore-aout-build-fix.patch
-nfs4proc-warnings-fix.patch
-bluetooth-warning-fixes.patch
-nopage-rss-accounting-fix.patch
-sonypi-update.patch
-meye-update.patch
-jbd-stfu.patch
-proc-pid-maps-32-bit-fix.patch
-acpi-pci-link-fix.patch
-rusage-context-switch-counters.patch
-rusage-context-switch-counters-fix.patch
-large-dev_t-01.patch
-large-dev_t-02.patch
-large-dev_t-03.patch
-large-dev_t-04.patch
-large-dev_t-05.patch
-large-dev_t-06.patch
-large-dev_t-07.patch
-large-dev_t-08.patch
-large-dev_t-09.patch
-large-dev_t-10.patch
-large-dev_t-11.patch
-large-dev_t-12.patch
-large-dev_t-12-fix.patch
-size_t-printk-warning-fixes.patch
-stallion-build-fix-2.patch
-evdev_ioctl-fix.patch
-as-no-initial-antic.patch
-hch-contacts-update.patch
-h8300-include-update.patch
-cyclades-isa-fix.patch
-old-module-tools-warning.patch
-old-module-tools-warning-fix.patch
-arcnet-printk-fix.patch
-floppy-cleanup.patch
-floppy-more-cleanup.patch
-v850-nommu-export-fixes.patch
-v850-RODATA-fix.patch
-dnotify-use-tgid.patch
-send_sigio-decl-fix.patch
-ipc-use-tgid.patch
-voyager-cpumask_t-fix.patch
-mtrr-attrib_to_str-consolidation.patch
-ioctl_end-fix.patch
-raw-driver-fixes.patch
-ipc_init-shuffle.patch
-zone-pressure-fixes.patch
-zone-pressure-simplification.patch
-remove-dead-sse-checks.patch
-hpet-01-late-time_init.patch
-hpet-02-boot-time-parsing.patch
-hpet-03-misc.patch
-hpet-03-misc-tweaks.patch
-hpet-04-core.patch
-hpet-05-timer-services.patch
-hpet-05-timer-services-tweaks.patch
-hpet-06-rtc-emulation.patch
-advansys-procfs-fix.patch
-usb-serial-oops-fix.patch
-handle-setup_swap_extents-error.patch
-aha1542-oops-fix.patch
-cosa-free_netdev-fix.patch
-tty_files-oops-fix.patch
-remove-kcore-aout.patch

 Merged

-deadline-requeue-workaround.patch

 Dropped: fixed properly.

-tdfx-build-fix.patch

 Folded into fbdev.patch

+misc34.patch

 Misc fixes

+cfq-3-fixes.patch

 CFQ IO scheduler fix

+ide-pm-oops-fix.patch

 IDE power management fix

+kobject-unlimited-name-lengths-use-after-free-fix.patch

 kobject oops fix

+convert-proc-stat-to-seq_file.patch

 Use seqfile for /proc/stat

+get_rtc_time-fix.patch

 RTC-vs-HPET fixes

+ricoh-mask-fix.patch

 PCMCIA fix

+visws-qla1280-needs-pio.patch

 SCSI fix

+dac960-devfs_name-fix.patch

 Fix devfs on DAC960

+ikconfig-gzipped.patch

 Move the ikconfig info to /proc/config.gz

+flush-invalidate-fixes.patch
+flush-invalidate-fixes-warning-fix.patch

 Lots of rework of the cache flushing API and code.  From DaveM

+elv-insertion-fix.patch

 IO scheduler fixes

+8250_acpi-taints-kernel.patch

 Add copyright to 8250_acpi.c

+proc_misc-build-fix.patch

 Compile fix

+slab-check-PG_slab.patch

 More slab debugging

+ide_floppy-maybe-fix.patch

 Try to mend ide_floppy

+might_sleep-improvements.patch

 Extra atomicity debug work.

+MODULE_ALIAS-in-block-devices.patch
+MODULE_ALIAS-in-char-devices.patch

 Module autoloading support

+unpercpuify-in_flight-counter.patch

 microoptimise disk stats accumulation

+enable-selinux-with-boot-parameter.patch

 SELinux requires "selinux" on the kernel boot command line

+pty-devfs-fix.patch

 sort-of fix old-style pty's

+i8042-free_irq-fix.patch

 i8042 IRQ management fixlet.

+reiserfs-direct-io.patch

 direct-IO support for reiserfs

+pdflush-diag.patch

 debug stuff

+netlink-warning-fixes.patch

 Warnig fixes

+really-use-english-date-in-version-string.patch

 Locale fixes for the kernel version string

+acpi-pci-routing-fixes.patch

 ACPI fixes

+psmouse_ipms2-option.patch

 Add psmouse_ipms2 module/boot parameter to force PS/2 IPMS2 protocol

+i8042-history.patch

 Extra i8042 debug support

-sched-CAN_MIGRATE_TASK-fix.patch
-sched-balance-fix-2.6.0-test3-mm3-A0.patch
-sched-2.6.0-test2-mm2-A3.patch
-ppc-sched_clock.patch
-ppc64-sched_clock.patch
-sparc64_sched_clock.patch
-x86_64-sched_clock.patch
-sched-warning-fix.patch
-sched-balance-tuning.patch
-sched-no-tsc-on-numa.patch
-o12.2int.patch
-o12.3.patch
-o13int.patch
-o13.1int.patch
-o14int.patch
-o14int-div-fix.patch
-o14.1int.patch
-o15int.patch
-o16int.patch
-o16.1int.patch
-o16.2int.patch
-o16.3int.patch
-o18int.patch
-o18.1int.patch
-sched-cpu-migration-fix.patch

 Dropped

+np-sched-01-sched-fork-cleanup.patch
+np-sched-02-sched-migrate-fix.patch
+np-sched-03-sched-balance-tuning.patch
+np-sched-04-sched-policy-10b.patch

 Added.

-4g4g-preempt-vstack-fix.patch
-4g4g-kmap-warning-comments.patch
-4g4g-slab-__get_user-fix.patch
-4g4g-vmlinux-update-got-lost.patch
-4g4g-do_page_fault-cleanup.patch
-4g4g-cleanups.patch
-kgdb-4g4g-fix-2.patch
-4g4g-config-fix.patch
-4g4g-pmd-fix.patch
-4g4g-wli-fixes.patch
-4g4g-fpu-fix.patch
-4g4g-show_registers-fix.patch
-4g4g-pin_page-atomicity-fix.patch
-4g4g-remove-touch_all_pages.patch
-4g4g-debug-flags-fix.patch
-4g4g-TI_task-fix.patch
-cyclone-fixmap-fix.patch

 Folded into 4g-2.6.0-test2-mm2-A5.patch





All 141 patches

linus.patch

mm.patch
  add -mmN to EXTRAVERSION

kgdb-ga.patch
  kgdb stub for ia32 (George Anzinger's one)
  kgdbL warning fix

kgdb-warning-fix.patch
  kgdbL warning fix

kgdb-build-fix.patch

kgdb-spinlock-fix.patch

kgdb-fix-debug-info.patch
  kgdb: CONFIG_DEBUG_INFO fix

kgdb-cpumask_t.patch

kgdb-x86_64-fixes.patch
  x86_64 fixes

config_spinline.patch
  uninline spinlocks for profiling accuracy.

ppc64-build-fixes.patch
  Fix ppc64 breakage

ppc64-bar-0-fix.patch
  Allow PCI BARs that start at 0

ppc64-reloc_hide.patch

ppc64-semaphore-reimplementation.patch
  ppc64: use the ia32 semaphore implementation

ppc64-local.patch
  ppc64: local.h implementation

sym-do-160.patch
  make the SYM driver do 160 MB/sec

rt-tasks-special-vm-treatment.patch
  real-time enhanced page allocator and throttling

rt-tasks-special-vm-treatment-2.patch

input-use-after-free-checks.patch
  input layer debug checks

fbdev.patch
  framebbuffer driver update

cursor-flashing-fix.patch
  fbdev: fix cursor letovers

misc34.patch
  misc fixes

slab-hexdump.patch
  slab: hexdump structures when things go wrong

aic7xxx-parallel-build-fix.patch
  fix parallel builds for aic7xxx

thread-pgrp-fix-2.patch
  Fix setpgid and threads

ramdisk-cleanup.patch

delay-ksoftirqd-fallback.patch
  Try harded in IRQ context before falling back to ksoftirqd

intel8x0-cleanup.patch
  intel8x0 cleanups

claim-serio-early.patch
  Serio: claim serio early

fix-strange-code-in-bio_add_page.patch
  Fix odd code in bio_add_page

mark-devfs-obsolete.patch
  mark devfs obsolete

cfq-3.patch
  CFQ io scheduler

cfq-3-fixes.patch
  CFQ fixes

sysfs-memleak-fix.patch
  Fix sysfs memory leak

VT8231-router-detection.patch
  VT8231 IRQ router detection

block-devfs-conversions.patch
  Initialise devfs_name in various block drivers

test4-pm1.patch
  power management update

ide-pm-oops-fix.patch
  IDE power management oops fix

kobject-unlimited-name-lengths.patch
  kobject: Support unlimited name lengths.

kobject-unlimited-name-lengths-use-after-free-fix.patch
  kobject_cleanup() use-after-free-fix

convert-proc-stat-to-seq_file.patch
  convert /proc/stat to seq_file

get_rtc_time-fix.patch
  Fix rtc symbol clash and HPET config problems

ricoh-mask-fix.patch
  pcmcia: ricoh.h mask fix
  EDEC
  From: KOMURO <komujun@nifty.com>, Alan Cox <alan@lxorguk.ukuu.org.uk>
  
  RL5C4XX_16BIT_MEM_0 was wrong.

visws-qla1280-needs-pio.patch
  add config option for qla1280 SCSI MMIO/ioport selection

dac960-devfs_name-fix.patch
  dac960 devfs_name initialisation fix

ikconfig-gzipped.patch
  Move ikconfig to /proc/config.gz

flush-invalidate-fixes.patch
  memory writeback/invalidation fixes

flush-invalidate-fixes-warning-fix.patch

elv-insertion-fix.patch
  elevator insertion fixes

8250_acpi-taints-kernel.patch
  8250_acpi taints kernel

proc_misc-build-fix.patch
  proc_misc.c needs irq.h

slab-check-PG_slab.patch
  more slab page checking

ide_floppy-maybe-fix.patch
  might fix ide_floppy

might_sleep-improvements.patch
  might_sleep() improvements

MODULE_ALIAS-in-block-devices.patch
  MODULE_ALIAS() in block devices

MODULE_ALIAS-in-char-devices.patch
  MODULE_ALIAS() in char devices

unpercpuify-in_flight-counter.patch
  Remove percpufication of in_flight counter in diskstats

enable-selinux-with-boot-parameter.patch
  Enable SELinux via boot parameter

pty-devfs-fix.patch
  devfs pty fix

i8042-free_irq-fix.patch
  i8042 free_irq() aliasing fix

reiserfs-direct-io.patch
  resierfs direct-IO support

pdflush-diag.patch

netlink-warning-fixes.patch

really-use-english-date-in-version-string.patch
  really use english date in version string

acpi-pci-routing-fixes.patch
  Fixing USB interrupt problems with ACPI enabled

p00001_synaptics-restore-on-close.patch

p00002_psmouse-reset-timeout.patch

p00003_synaptics-multi-button.patch

p00004_synaptics-optional.patch

p00005_synaptics-pass-through.patch

p00006_psmouse-suspend-resume.patch

p00007_synaptics-old-proto.patch

synaptics-mode-set.patch
  Synaptics mode setting

syn-multi-btn-fix.patch
  synaptics multibutton fix

keyboard-resend-fix.patch
  keyboard resend fix

psmouse_ipms2-option.patch
  Force mouse detection as imps/2 (and fix my KVM switch)

i8042-history.patch
  debug: i8042 history dumping

linux-isp-2.patch

linux-isp-2-fix-again.patch
  lost feral fix

feral-bounce-fix.patch
  Feral driver - highmem issues

feral-bounce-fix-2.patch
  Feral driver bouncing fix

list_del-debug.patch
  list_del debug check

print-build-options-on-oops.patch
  print a few config options on oops

show_task-free-stack-fix.patch
  show_task() fix and cleanup

put_task_struct-debug.patch

ia32-mknod64.patch
  mknod64 for ia32

ext2-64-bit-special-inodes.patch
  ext2: support for 64-bit device nodes

ext3-64-bit-special-inodes.patch
  ext3: support for 64-bit device nodes

64-bit-dev_t-kdev_t.patch
  64-bit dev_t and kdev_t

64-bit-dev_t-other-archs.patch
  enable 64-bit dev_t for other archs

mknod64-64-bit-fix.patch
  dev_t: fix mknod for 64-bit archs

ustat64.patch
  ustat64

ppc-64-bit-stat.patch
  fix ppc stat.h for 64-bit dev_t

64-bit-dev_t-init_rd-fixes.patch
  initrd fixes for 64-bit dev_t

arch-dev_t-stat-fixes.patch
  Fix all asm-*/stat.h dev_t instances

oops-dump-preceding-code.patch
  i386 oops output: dump preceding code

lockmeter.patch

sparc64-lockmeter-fix.patch

sparc64-lockmeter-fix-2.patch
  Fix lockmeter on sparc64

printk-oops-mangle-fix.patch
  disentangle printk's whilst oopsing on SMP

20-odirect_enable.patch

21-odirect_cruft.patch

22-read_proc.patch

23-write_proc.patch

24-commit_proc.patch

25-odirect.patch

nfs-O_DIRECT-always-enabled.patch
  Force CONFIG_NFS_DIRECTIO

np-sched-01-sched-fork-cleanup.patch
  move fork-related code to fork.c

np-sched-02-sched-migrate-fix.patch
  scheduler migration fix

np-sched-03-sched-balance-tuning.patch
  scheduler balancing tuning

np-sched-04-sched-policy-10b.patch
  scheduler policy changes

4g-2.6.0-test2-mm2-A5.patch
  4G/4G split patch
  4G/4G: remove debug code
  4g4g: pmd fix
  4g/4g: fixes from Bill
  4g4g: fpu emulation fix
  4g/4g usercopy atomicity fix
  4G/4G: remove debug code
  4g4g: pmd fix
  4g/4g: fixes from Bill
  4g4g: fpu emulation fix
  4g/4g usercopy atomicity fix
  4G/4G preempt on vstack
  4G/4G: even number of kmap types
  4g4g: fix __get_user in slab
  4g4g: Remove extra .data.idt section definition
  4g/4g linker error (overlapping sections)
  4G/4G: remove debug code
  4g4g: pmd fix
  4g/4g: fixes from Bill
  4g4g: fpu emulation fix
  4g4g: show_registers() fix
  4g/4g usercopy atomicity fix
  4g4g: debug flags fix
  4g4g: Fix wrong asm-offsets entry
  cyclone time fixmap fix
  4G/4G preempt on vstack
  4G/4G: even number of kmap types
  4g4g: fix __get_user in slab
  4g4g: Remove extra .data.idt section definition
  4g/4g linker error (overlapping sections)
  4G/4G: remove debug code
  4g4g: pmd fix
  4g/4g: fixes from Bill
  4g4g: fpu emulation fix
  4g4g: show_registers() fix
  4g/4g usercopy atomicity fix
  4g4g: debug flags fix
  4g4g: Fix wrong asm-offsets entry
  cyclone time fixmap fix

ppc-fixes.patch
  make mm4 compile on ppc

aic7xxx_old-oops-fix.patch

aio-01-retry.patch
  AIO: Core retry infrastructure

io_submit_one-EINVAL-fix.patch
  Fix aio process hang on EINVAL

aio-02-lockpage_wq.patch
  AIO: Async page wait

aio-03-fs_read.patch
  AIO: Filesystem aio read

aio-04-buffer_wq.patch
  AIO: Async buffer wait

aio-05-fs_write.patch
  AIO: Filesystem aio write

aio-05-fs_write-fix.patch

aio-06-bread_wq.patch
  AIO: Async block read

aio-06-bread_wq-fix.patch

aio-07-ext2getblk_wq.patch
  AIO: Async get block for ext2

O_SYNC-speedup-2.patch
  speed up O_SYNC writes

aio-09-o_sync.patch
  aio O_SYNC

aio-10-BUG-fix.patch
  AIO: fix a BUG

aio-11-workqueue-flush.patch
  AIO: flush workqueues before destroying ioctx'es

aio-12-readahead.patch
  AIO: readahead fixes

aio-dio-no-readahead.patch
  aio O_DIRECT no readahead

lock_buffer_wq-fix.patch
  lock_buffer_wq fix

unuse_mm-locked.patch
  AIO: hold the context lock across unuse_mm

aio-take-task_lock.patch
  From: Suparna Bhattacharya <suparna@in.ibm.com>
  Subject: Re: 2.5.72-mm1 - Under heavy testing with AIO,.. vmstat seems to blow the kernel

aio-O_SYNC-fix.patch
  Unify o_sync changes for aio and regular writes

aio-O_SYNC-fix-missing-bit.patch
  aio-O_SYNC-fix bits got lost

O_SYNC-speedup-nolock-fix.patch

aio-writev-nsegs-fix.patch
  aio: writev nr_segs fix

aio-remove-lseek-triggerable-BUG_ONs.patch

aio-readahead-rework.patch
  Unified page range readahead for aio and regular reads

aio-readahead-speedup.patch
  Readahead issues and AIO read speedup

aio-osync-fix-2.patch
  More AIO O_SYNC related fixes

Re: 2.6.0-test4-mm5

[Christian Axelsson]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

How is the work on CPU scheduler selection coming along? It would be a
Good Thing (TM) to have in this one imho.

- --
Christan Axelsson
smiler@lanil.mine.nu

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQE/VZV5yqbmAWw8VdkRAvr1AJ4mEWvJJg2S8jIMk5OOBx+/kHmgtACg4CG2
YLVDbf13vBtm6uFNoQUZG3M=
=81tT
-----END PGP SIGNATURE-----

Re: 2.6.0-test4-mm5

[Adrian Bunk]

I got the following compile error (using gcc 2.95) in 2.6.0-test4-mm5
(but it seems to come from Linus' tree):

<--  snip  -->

...
  CC [M]  drivers/net/sungem.o
drivers/net/sungem.c:2444: duplicate initializer
drivers/net/sungem.c:2444: (near initialization for `gem_ethtool_ops.get_link')
make[2]: *** [drivers/net/sungem.o] Error 1

<--  snip  -->


It seems gcc is right, there are two .get_link members in this struct:


<--  snip  -->

...
static struct ethtool_ops gem_ethtool_ops = {
        .get_drvinfo            = gem_get_drvinfo,
        .get_link               = ethtool_op_get_link,
        .get_settings           = gem_get_settings,
        .set_settings           = gem_set_settings,
        .nway_reset             = gem_nway_reset,
        .get_link               = gem_get_link,
        .get_msglevel           = gem_get_msglevel,
        .set_msglevel           = gem_set_msglevel,
};
...

<--  snip  -->


cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

[PATCH] Re: 2.6.0-test4-mm5

[Jeff Garzik]

[-- Attachment #1: Type: text/plain, Size: 581 bytes --]

Adrian Bunk wrote:
> It seems gcc is right, there are two .get_link members in this struct:
> 
> 
> <--  snip  -->
> 
> ...
> static struct ethtool_ops gem_ethtool_ops = {


David, would you look over this patch and apply/modify?

I would prefer to use the generic ethtool_op_get_link, because (a) 
sungem is already using netif_carrier_xxx, and (b) if ->get_link ever 
returns an incorrect value, that signals a netif_carrier_xxx bug exists.

As a tangent, gem_pcs_interrupt appears to call netif_carrier_xxx but 
not set gem->lstate.  David/Ben, is that a bug?

Thanks,

	Jeff



[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 645 bytes --]

===== drivers/net/sungem.c 1.44 vs edited =====
--- 1.44/drivers/net/sungem.c	Sun Aug 24 08:58:18 2003
+++ edited/drivers/net/sungem.c	Wed Sep  3 12:28:30 2003
@@ -2416,13 +2416,6 @@
 	return 0;
 }
 
-static u32 gem_get_link(struct net_device *dev)
-{
-	struct gem *gp = dev->priv;
-
-	return (gp->lstate == link_up);
-}
-
 static u32 gem_get_msglevel(struct net_device *dev)
 {
 	struct gem *gp = dev->priv;
@@ -2441,7 +2434,6 @@
 	.get_settings		= gem_get_settings,
 	.set_settings		= gem_set_settings,
 	.nway_reset		= gem_nway_reset,
-	.get_link		= gem_get_link,
 	.get_msglevel		= gem_get_msglevel,
 	.set_msglevel		= gem_set_msglevel,
 };

Re: [PATCH] Re: 2.6.0-test4-mm5

[Benjamin Herrenschmidt]

> As a tangent, gem_pcs_interrupt appears to call netif_carrier_xxx but 
> not set gem->lstate.  David/Ben, is that a bug?

Looks like it is, David, you are the one who knows the PCS stuff ...

BTW. David: Any reason why you wouldn't let me change all occurences
of spin_{lock,unlock}_irq into the ...{save,restore} versions ?

I don't like force re-enabling interrupts the way we do it now with
spin_unlock_irq() the way we do it now. Among other things, that
breaks SA_INTERRUPT semantics and that breaks some pet project of
mine which is to run the IRQ handler with interrupts off when the
kernel stack is below a given threshold (to limit interrupt depth)

Ben.

2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Adrian Bunk]

The following compile error (tested with gcc 2.95) seems to come from 
Linus' tree:

<--  snip  -->

...
  CC [M]  drivers/scsi/imm.o
In file included from drivers/scsi/imm.c:55:
drivers/scsi/imm.h:105: duplicate array index in initializer
drivers/scsi/imm.h:105: (near initialization for `IMM_MODE_STRING')
make[2]: *** [drivers/scsi/imm.o] Error 1

<--  snip  -->

The problem is the following code in imm.h (with 
CONFIG_SCSI_IZIP_EPP16 enabled):

<--  snip  -->

...
static char *IMM_MODE_STRING[] =
{
        [IMM_AUTODETECT] = "Autodetect",
        [IMM_NIBBLE]     = "SPP",
        [IMM_PS2]        = "PS/2",
        [IMM_EPP_8]      = "EPP 8 bit",
        [IMM_EPP_16]     = "EPP 16 bit",
#ifdef CONFIG_SCSI_IZIP_EPP16
        [IMM_EPP_16]     = "EPP 16 bit",
#else
        [IMM_EPP_32]     = "EPP 32 bit",
#endif
        [IMM_UNKNOWN]    = "Unknown",
};
...

<--  snip  -->

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.0-test4-mm5

[William Lee Irwin III]

On Wed, Sep 03, 2003 at 09:17:13AM +0200, Christian Axelsson wrote:
> How is the work on CPU scheduler selection coming along? It would be a
> Good Thing (TM) to have in this one imho.

No idea if anyone's doing it, but it should be relatively easy to do.


-- wli

Re: 2.6.0-test4-mm5

[Diego Calleja García]

El Tue, 2 Sep 2003 23:18:12 -0700 Andrew Morton <akpm@osdl.org> escribió:

> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
> 
> . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>   in evaluating the stability, efficacy and relative performance of Nick's
>   work.
> 
>   We're looking for feedback on the subjective behaviour and on the usual
>   server benchmarks please.


I must say that this one doesn't feel nice under heavy gcc load. Huge mp3
skips that didn't happened before, big pauses in X...gcc starves anything else.
-mm4 was better there.

Re: [PATCH] Re: 2.6.0-test4-mm5

[Jeff Garzik]

Benjamin Herrenschmidt wrote:
> BTW. David: Any reason why you wouldn't let me change all occurences
> of spin_{lock,unlock}_irq into the ...{save,restore} versions ?


IMO... even though you do lose a tiny bit of performance, I definitely
prefer the save/restore versions.

It allows the arch a lot more flexibility, so I even have a [weak]
argument that {save,restore} variants increase portability.  And it's
safer, as I like to avoid code which winds up doing (as it passes
through layers) spin_unlock_irq() followed by spin_unlock_irqrestore().

	Jeff

Re: 2.6.0-test4-mm5

[Bill Huey]

On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/

make[1]: `arch/i386/kernel/asm-offsets.s' is up to date.
  CHK     include/linux/compile.h
  CC      drivers/acpi/pci_link.o
drivers/acpi/pci_link.c: In function `acpi_pci_link_try_get_current':
drivers/acpi/pci_link.c:290: error: `_dbg' undeclared (first use in this function)
drivers/acpi/pci_link.c:290: error: (Each undeclared identifier is reported only once
drivers/acpi/pci_link.c:290: error: for each function it appears in.)
make[2]: *** [drivers/acpi/pci_link.o] Error 1
make[1]: *** [drivers/acpi] Error 2
make: *** [drivers] Error 2

------------------------------------------------------------------------------------------

bill

Re: 2.6.0-test4-mm5

[Nick Piggin]

Diego Calleja García wrote:

>El Tue, 2 Sep 2003 23:18:12 -0700 Andrew Morton <akpm@osdl.org> escribió:
>
>  
>
>>ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
>>
>>. Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>>  in evaluating the stability, efficacy and relative performance of Nick's
>>  work.
>>
>>  We're looking for feedback on the subjective behaviour and on the usual
>>  server benchmarks please.
>>    
>>
>
>
>I must say that this one doesn't feel nice under heavy gcc load. Huge mp3
>skips that didn't happened before, big pauses in X...gcc starves anything else.
>-mm4 was better there.
>  
>

Hmm... what's heavy gcc load?

Re: 2.6.0-test4-mm5

[Mike Fedyk]

On Thu, Sep 04, 2003 at 01:08:52AM +0200, Diego Calleja Garc?a wrote:
> El Tue, 2 Sep 2003 23:18:12 -0700 Andrew Morton <akpm@osdl.org> escribi?:
> 
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
> > 
> > . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
> >   in evaluating the stability, efficacy and relative performance of Nick's
> >   work.
> > 
> >   We're looking for feedback on the subjective behaviour and on the usual
> >   server benchmarks please.
> 
> 
> I must say that this one doesn't feel nice under heavy gcc load. Huge mp3
> skips that didn't happened before, big pauses in X...gcc starves anything else.
> -mm4 was better there.

Can you put your Xserver back to nice -10, and try again?

Re: 2.6.0-test4-mm5

[Nick Piggin]

Mike Fedyk wrote:

>On Thu, Sep 04, 2003 at 01:08:52AM +0200, Diego Calleja Garc?a wrote:
>
>>El Tue, 2 Sep 2003 23:18:12 -0700 Andrew Morton <akpm@osdl.org> escribi?:
>>
>>
>>>ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
>>>
>>>. Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>>>  in evaluating the stability, efficacy and relative performance of Nick's
>>>  work.
>>>
>>>  We're looking for feedback on the subjective behaviour and on the usual
>>>  server benchmarks please.
>>>
>>
>>I must say that this one doesn't feel nice under heavy gcc load. Huge mp3
>>skips that didn't happened before, big pauses in X...gcc starves anything else.
>>-mm4 was better there.
>>
>
>Can you put your Xserver back to nice -10, and try again?
>

This would help X, but regardless mp3 playing should not skip. I think its
giving newly forked children much too high a priority. Basically new 
children
can't get less than 50% sleep time, which is stupid on my behalf because a
program which continually forks children that do a little bit of work then
exit would basically be at 50% sleep time when it should be at or close 
to 0.

Re: 2.6.0-test4-mm5

[Chris Wright]

* Bill Huey (billh@gnuppy.monkey.org) wrote:
> On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
> 
> make[1]: `arch/i386/kernel/asm-offsets.s' is up to date.
>   CHK     include/linux/compile.h
>   CC      drivers/acpi/pci_link.o
> drivers/acpi/pci_link.c: In function `acpi_pci_link_try_get_current':
> drivers/acpi/pci_link.c:290: error: `_dbg' undeclared (first use in this function)
> drivers/acpi/pci_link.c:290: error: (Each undeclared identifier is reported only once
> drivers/acpi/pci_link.c:290: error: for each function it appears in.)
> make[2]: *** [drivers/acpi/pci_link.o] Error 1
> make[1]: *** [drivers/acpi] Error 2
> make: *** [drivers] Error 2

try the patch below.
thanks,
-chris
-- 
Linux Security Modules     http://lsm.immunix.org     http://lsm.bkbits.net

--- drivers/acpi/pci_link.c.try	Wed Sep  3 15:06:33 2003
+++ drivers/acpi/pci_link.c	Wed Sep  3 15:07:22 2003
@@ -285,6 +285,8 @@ acpi_pci_link_try_get_current (
 {
 	int result;
 
+	ACPI_FUNCTION_TRACE("acpi_pci_link_try_get_current");
+
 	result = acpi_pci_link_get_current(link);
 	if (result && link->irq.active) {
  		return_VALUE(result);

Re: [PATCH] Re: 2.6.0-test4-mm5

[David S. Miller]

On Wed, 03 Sep 2003 12:32:41 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> David, would you look over this patch and apply/modify?

Applied, thanks Jeff.

> I would prefer to use the generic ethtool_op_get_link, because (a) 
> sungem is already using netif_carrier_xxx, and (b) if ->get_link ever 
> returns an incorrect value, that signals a netif_carrier_xxx bug exists.

Agreed.

> As a tangent, gem_pcs_interrupt appears to call netif_carrier_xxx but 
> not set gem->lstate.  David/Ben, is that a bug?

Doesn't really matter, I don't think the rest of the PCS code even
cares about the setting of gem-lstate.

Re: [PATCH] Re: 2.6.0-test4-mm5

[Benjamin Herrenschmidt]

On Thu, 2003-09-04 at 01:31, Jeff Garzik wrote:
> Benjamin Herrenschmidt wrote:
> > BTW. David: Any reason why you wouldn't let me change all occurences
> > of spin_{lock,unlock}_irq into the ...{save,restore} versions ?
> 
> 
> IMO... even though you do lose a tiny bit of performance, I definitely
> prefer the save/restore versions.

I'm not even sure you actually lose perfs... at least on ppc ;)

Ben.

Re: 2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Arnaldo Carvalho de Melo]

Em Wed, Sep 03, 2003 at 07:02:56PM +0200, Adrian Bunk escreveu:
> The following compile error (tested with gcc 2.95) seems to come from 
> Linus' tree:
> 
> <--  snip  -->
> 
> ...
>   CC [M]  drivers/scsi/imm.o
> In file included from drivers/scsi/imm.c:55:
> drivers/scsi/imm.h:105: duplicate array index in initializer
> drivers/scsi/imm.h:105: (near initialization for `IMM_MODE_STRING')
> make[2]: *** [drivers/scsi/imm.o] Error 1
> 
> <--  snip  -->
> 
> The problem is the following code in imm.h (with 
> CONFIG_SCSI_IZIP_EPP16 enabled):
> 
> <--  snip  -->
> 
> ...
> static char *IMM_MODE_STRING[] =
> {
>         [IMM_AUTODETECT] = "Autodetect",
>         [IMM_NIBBLE]     = "SPP",
>         [IMM_PS2]        = "PS/2",
>         [IMM_EPP_8]      = "EPP 8 bit",
>         [IMM_EPP_16]     = "EPP 16 bit",
> #ifdef CONFIG_SCSI_IZIP_EPP16
>         [IMM_EPP_16]     = "EPP 16 bit",
> #else
>         [IMM_EPP_32]     = "EPP 32 bit",
> #endif
>         [IMM_UNKNOWN]    = "Unknown",
> };
> ...
> 
> <--  snip  -->

Original code was:

 static char *IMM_MODE_STRING[] =
 {
    "Autodetect",
    "SPP",
    "PS/2",
    "EPP 8 bit",
    "EPP 16 bit",
 #ifdef CONFIG_SCSI_IZIP_EPP16
    "EPP 16 bit",
 #else
    "EPP 32 bit",
 #endif
    "Unknown"};

I just converted it to the more safe c99 init style, but haven't noticed
the original bug, that is "EPP 16 bit" was duplicated... But this is already
fixed by Andrew Morton on current Linus bk tree.

Thanks Andrew for fixing, Adrian for noticing.

- Arnaldo

Re: 2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 807 bytes --]

On Thu, 2003-09-04 10:30:56 -0300, Arnaldo Carvalho de Melo <acme@conectiva.com.br>
wrote in message <20030904133056.GA2411@conectiva.com.br>:
> Em Wed, Sep 03, 2003 at 07:02:56PM +0200, Adrian Bunk escreveu:
> > The following compile error (tested with gcc 2.95) seems to come from 
> > Linus' tree:
> 
> I just converted it to the more safe c99 init style, but haven't noticed

C99 style is

	.element = initializer,

not
	[element] = initializer,

which is a GNU/GCCism.

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
      ret = do_actions((curr | FREE_SPEECH) & ~(IRAQ_WAR_2 | DRM | TCPA));

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.0-test4-mm5

[Bruno T. Moura]

[-- Attachment #1: Type: text/plain, Size: 1140 bytes --]

On Tue, 2 Sep 2003 23:18:12 -0700
Andrew Morton <akpm@osdl.org> wrote:

> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
> 
> . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>   in evaluating the stability, efficacy and relative performance of Nick's
>   work.
> 
 [ snip. ]

Getting this one when setting dma on a barracuda SATA ST380013AS, i875 chipset.

Sep  4 09:53:26 pyro kernel: hde: set_drive_speed_status: status=0xd0 { Busy }
Sep  4 09:53:26 pyro kernel: hde: channel busy
Sep  4 09:53:26 pyro kernel: hde: Speed warnings UDMA 3/4/5 is not functional.
Sep  4 09:53:26 pyro kernel: hde: dma_timer_expiry: dma status == 0x20
Sep  4 09:53:26 pyro kernel: hde: DMA timeout retry
Sep  4 09:53:26 pyro kernel: hde: timeout waiting for DMA
Sep  4 09:53:26 pyro kernel: hde: status timeout: status=0xd0 { Busy }
Sep  4 09:53:26 pyro kernel:
Sep  4 09:53:26 pyro kernel: hde: drive not ready for command
Sep  4 09:53:26 pyro kernel: hde: status timeout: status=0xd0 { Busy }
Sep  4 09:53:26 pyro kernel:
Sep  4 09:53:26 pyro kernel: hde: drive not ready for command

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Christoph Hellwig]

On Thu, Sep 04, 2003 at 03:52:56PM +0200, Jan-Benedict Glaw wrote:
> C99 style is
> 
> 	.element = initializer,
> 
> not
> 	[element] = initializer,
> 
> which is a GNU/GCCism.

We're talking about arrays here..

.element obviously never works for arrays and [constant] never
works for structs..

Re: 2.6.0-test4-mm5

[Bruno T. Moura]

[-- Attachment #1: Type: text/plain, Size: 1629 bytes --]

On Thu, 4 Sep 2003 10:29:18 -0400
"Bruno T. Moura" <bruno.tavares@terra.com.br> wrote:

> On Tue, 2 Sep 2003 23:18:12 -0700
> Andrew Morton <akpm@osdl.org> wrote:
> 
> > 
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test4/2.6.0-test4-mm5/
> > 
> > . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
> >   in evaluating the stability, efficacy and relative performance of Nick's
> >   work.
> > 
>  [ snip. ]
> 
> Getting this one when setting dma on a barracuda SATA ST380013AS, i875 chipset.
> 
> Sep  4 09:53:26 pyro kernel: hde: set_drive_speed_status: status=0xd0 { Busy }
> Sep  4 09:53:26 pyro kernel: hde: channel busy
> Sep  4 09:53:26 pyro kernel: hde: Speed warnings UDMA 3/4/5 is not functional.
> Sep  4 09:53:26 pyro kernel: hde: dma_timer_expiry: dma status == 0x20
> Sep  4 09:53:26 pyro kernel: hde: DMA timeout retry
> Sep  4 09:53:26 pyro kernel: hde: timeout waiting for DMA
> Sep  4 09:53:26 pyro kernel: hde: status timeout: status=0xd0 { Busy }
> Sep  4 09:53:26 pyro kernel:
> Sep  4 09:53:26 pyro kernel: hde: drive not ready for command
> Sep  4 09:53:26 pyro kernel: hde: status timeout: status=0xd0 { Busy }
> Sep  4 09:53:26 pyro kernel:
> Sep  4 09:53:26 pyro kernel: hde: drive not ready for command
> 

forgot this log part :

Sep  4 09:53:26 pyro kernel: ide2: reset: success
Sep  4 09:53:26 pyro kernel: hde: task_no_data_intr: status=0x51 { DriveReady SeekComplete Error }
Sep  4 09:53:26 pyro kernel: hde: task_no_data_intr: error=0x04 { DriveStatusError }
Sep  4 09:53:26 pyro kernel: blk: queue 41d72600, I/O limit 4095Mb (mask 0xffffffff)



[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 738 bytes --]

On Thu, 2003-09-04 15:30:23 +0100, Christoph Hellwig <hch@infradead.org>
wrote in message <20030904153023.A32549@infradead.org>:
> On Thu, Sep 04, 2003 at 03:52:56PM +0200, Jan-Benedict Glaw wrote:
> > C99 style is
> > 
> > 	.element = initializer,
> > 
> > not
> > 	[element] = initializer,
> > 
> > which is a GNU/GCCism.
> 
> We're talking about arrays here..

*plonk* I stand corrected... I'm sorry...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
      ret = do_actions((curr | FREE_SPEECH) & ~(IRAQ_WAR_2 | DRM | TCPA));

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

[new patch] Re: 2.6.0-test4-mm5: SCSI imm driver doesn't compile

[Adrian Bunk]

On Thu, Sep 04, 2003 at 10:30:56AM -0300, Arnaldo Carvalho de Melo wrote:
>...
> I just converted it to the more safe c99 init style, but haven't noticed
> the original bug, that is "EPP 16 bit" was duplicated... But this is already
> fixed by Andrew Morton on current Linus bk tree.
> 
> Thanks Andrew for fixing, Adrian for noticing.

Andrews patch removed the first IMM_EPP_16 line in the array.

This isn't correct especially in the !CONFIG_SCSI_IZIP_EPP16 case,
reading all uses of this array (IMM_MODE_STRING is used to print the
corresponding string in printks).

If I'm not misunderstanding it, CONFIG_SCSI_IZIP_EPP16 means "use 16bit 
even when 32bit is requested".

It seems the right solution is


static char *IMM_MODE_STRING[] =
{
        [IMM_AUTODETECT] = "Autodetect",
        [IMM_NIBBLE]     = "SPP",
        [IMM_PS2]        = "PS/2",
        [IMM_EPP_8]      = "EPP 8 bit",
        [IMM_EPP_16]     = "EPP 16 bit",
#ifdef CONFIG_SCSI_IZIP_EPP16
        [IMM_EPP_32]     = "EPP 16 bit",
#else
        [IMM_EPP_32]     = "EPP 32 bit",
#endif
        [IMM_UNKNOWN]    = "Unknown",
};


A patch against the current BK tree is below.


> - Arnaldo

cu
Adrian

--- linux-2.6.0-test4/drivers/scsi/imm.h.old	2003-09-04 19:47:23.000000000 +0200
+++ linux-2.6.0-test4/drivers/scsi/imm.h	2003-09-04 19:48:05.000000000 +0200
@@ -100,8 +100,9 @@
 	[IMM_NIBBLE]	 = "SPP",
 	[IMM_PS2]	 = "PS/2",
 	[IMM_EPP_8]	 = "EPP 8 bit",
-#ifdef CONFIG_SCSI_IZIP_EPP16
 	[IMM_EPP_16]	 = "EPP 16 bit",
+#ifdef CONFIG_SCSI_IZIP_EPP16
+	[IMM_EPP_32]	 = "EPP 16 bit",
 #else
 	[IMM_EPP_32]	 = "EPP 32 bit",
 #endif

Re: 2.6.0-test4-mm5

[Diego Calleja García]

El Thu, 04 Sep 2003 11:32:49 +1000 Nick Piggin <piggin@cyberone.com.au> escribió:

> Hmm... what's heavy gcc load?

make -j25 with 256 MB RAM.

My X server is reniced at -1; but reniced X to -10 and it didn't helped;
-j15 was better (less swapping) but still I saw various mp3 & mouse skips.

Re: 2.6.0-test4-mm5

[Mike Fedyk]

On Thu, Sep 04, 2003 at 08:23:19PM +0200, Diego Calleja Garc?a wrote:
> El Thu, 04 Sep 2003 11:32:49 +1000 Nick Piggin <piggin@cyberone.com.au> escribi?:
> 
> > Hmm... what's heavy gcc load?
> 
> make -j25 with 256 MB RAM.
> 
> My X server is reniced at -1; but reniced X to -10 and it didn't helped;
> -j15 was better (less swapping) but still I saw various mp3 & mouse skips.

And this worked good with Con's scheduler?

Try both schedulers on the same base (test4), and see if you see similair
differences.

I doubt it's the scheduler that's causing this problem.  Once you get into
swap like that, the scheduler shouldn't affect it too much...

Re: 2.6.0-test4-mm5

[Alan Cox]

On Iau, 2003-09-04 at 15:29, Bruno T. Moura wrote:
> Sep  4 09:53:26 pyro kernel: hde: Speed warnings UDMA 3/4/5 is not functional.
> Sep  4 09:53:26 pyro kernel: hde: dma_timer_expiry: dma status == 0x20
> Sep  4 09:53:26 pyro kernel: hde: DMA timeout retry

SATA mode on the ICH5 really wants to be driven by Jeff Garziks libata
SATA drivers not the ide layer code. The IDE code doesnt currently know
about cable detect not being valid on the SATA ports.

Re: 2.6.0-test4-mm5

[Martin Schlemmer]

On Thu, 2003-09-04 at 20:23, Diego Calleja García wrote:
> El Thu, 04 Sep 2003 11:32:49 +1000 Nick Piggin <piggin@cyberone.com.au> escribió:
> 
> > Hmm... what's heavy gcc load?
> 
> make -j25 with 256 MB RAM.
> 
> My X server is reniced at -1; but reniced X to -10 and it didn't helped;
> -j15 was better (less swapping) but still I saw various mp3 & mouse skips.
> -

Without trying to be insulting, don't you think that you might
be expecting too much ?  I have a P4-2.4C (HT) on a i785 board
with 1GB DDR400 memory running dual channel, and if I run two
or three compile jobs at -j12 (more for testing Nick/Con's stuff,
usually use -j[46] and never really more than 2 of 3 of them), I
do not expect everything to be blazing fast.  Yes, while it is
not swapping (rare if ever do swap), I do expect xmms not to
skip, I do not expect the mouse to jerk, and yes I do expect
changing desktops to be fairly smooth and responsive.  I do not
however expect apps to still start with the same speed, or doing
the "window wiggle thing" to still be 100% smooth.

Nick's stuff with X reniced to -10 do fix the "window wiggle"
issue as far as I am concerned, as it is relative smooth (not
croaking like in vanilla), although not 100% as great as Con's,
but then Nick's finish the two make jobs at -j12 faster than
Con's stuff, and fairly close to vanilla (yes not 100% scientific),
and that is what I expect, even though it is a fairly ok machine.

Point I want to make, is that yes, for desktop you want you xmms
to not skip, you window switching to be ok while compiling a new
kernel at make -j[246], but come on, this is like expecting a
mini to out drag a F1 car because you fitted it with a turbo.
Especially when you start to swap heavily, you cannot really
expect the system to be responsive, especially when it is something
that needs to access memory that is swapped, or need to come out
of free swap, or that need to access the disk.

Con/Nick do need feedback, but really, expecting the scheduler to
be nice while your disk/memory is thrashing and whatever also need
a slice of that pie.  Can we try to keep it real ?


Thanks,

-- 
Martin Schlemmer

Re: 2.6.0-test4-mm5

[Nick Piggin]

Martin Schlemmer wrote:

>On Thu, 2003-09-04 at 20:23, Diego Calleja García wrote:
>
>>El Thu, 04 Sep 2003 11:32:49 +1000 Nick Piggin <piggin@cyberone.com.au> escribió:
>>
>>
>>>Hmm... what's heavy gcc load?
>>>
>>make -j25 with 256 MB RAM.
>>
>>My X server is reniced at -1; but reniced X to -10 and it didn't helped;
>>-j15 was better (less swapping) but still I saw various mp3 & mouse skips.
>>-
>>
>
>Without trying to be insulting, don't you think that you might
>be expecting too much ?  I have a P4-2.4C (HT) on a i785 board
>with 1GB DDR400 memory running dual channel, and if I run two
>or three compile jobs at -j12 (more for testing Nick/Con's stuff,
>usually use -j[46] and never really more than 2 of 3 of them), I
>

I think Martin is right here. I don't know what would be a good reason
for wanting X to work nicely with a make -j25 running. X typically
needs at least 75% CPU on my box to be nicely interactive when moving
a window or scrolling something complex. This gives only 1% to each
cc1.

I am still working on my scheduler. I've removed backboost. It is
hypocritical of me to worry about complexity or difficult traceability
of say Con's implementation when backboost is probably "worse" than
anything he has ;)

So I've found I'm getting more consistent behaviour, but it is now
very dependant on nice to get X running well under load. I'm
concentrating on getting it working well with make -j <= 6.

Re: 2.6.0-test4-mm5

[Diego Calleja García]

El Sat, 06 Sep 2003 02:17:22 +1000 Nick Piggin <piggin@cyberone.com.au> escribió:

[chaging email to avoid smart spam detectors]

> I think Martin is right here. I don't know what would be a good reason
> for wanting X to work nicely with a make -j25 running. X typically
> needs at least 75% CPU on my box to be nicely interactive when moving
> a window or scrolling something complex. This gives only 1% to each
> cc1.

I know; obiously it has not sense to ask that people makes -j25
smooth here; there's no way of doing that (with gcc 3.x :)

I just pointed out that -mm4 was a bit better under that load, specially
mp3 skips.


> So I've found I'm getting more consistent behaviour, but it is now
> very dependant on nice to get X running well under load. I'm
> concentrating on getting it working well with make -j <= 6.

I just hope 2.6 is shipped with a "decent" scheduler. Mp3 players are
probably the most important apps nowadays :P

2.6.0-test4-mm5 and below: Wine and XMMS problems

[Adrian Bunk]

On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
>...
> . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>   in evaluating the stability, efficacy and relative performance of Nick's
>   work.
> 
>   We're looking for feedback on the subjective behaviour and on the usual
>   server benchmarks please.
>...

Short story:

I'm still using 2.5.72, all of the 2.6.0-test?{,-mm?} kernels have 
problems


Long story:

System:
K6-2 @ 500 MHz
128 MB RAM
1 GB swap
Debian unstable

Workload:
XFree86
FVWM
XMMS
Wine running "Master of Orion 2" (a round based space strategy game)

With 2.4 kernels and 2.5.72 everything works fine.

With 2.6.0-test? and 2.6.0-test?-mm? kernels up to 2.6.0-test4-mm4 the
XMMS sound sometimes skips or sounds slow (like when wou manually retard
a record). That's much more awful than skips.

RAM usage is low, even after a "swapoff -a" at about half of my RAM
would be enough.

The problems might be related to the fact that after I start Wine three
wine.bin processes run and each of them tries to get as much CPU time as
possible.

It might be part of the problem that although Wine is the interactive 
task a working XMMS is subjectively more important.

With 2.6.0-test4-mm5 these problems don't occur. Instead, Wine feels 
slow. I couldn;t test it much since after the first fast mouse movement 
the X mouse cursor has lost the mouse cursor of the game (this might be 
a bug in Wine, but it doesnt occur with other kernels).

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.0-test4-mm5 and below: Wine and XMMS problems

[Nick Piggin]

Adrian Bunk wrote:

>On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
>
>>...
>>. Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
>>  in evaluating the stability, efficacy and relative performance of Nick's
>>  work.
>>
>>  We're looking for feedback on the subjective behaviour and on the usual
>>  server benchmarks please.
>>...
>>
>
>Short story:
>
>I'm still using 2.5.72, all of the 2.6.0-test?{,-mm?} kernels have 
>problems
>
>
>Long story:
>
>System:
>K6-2 @ 500 MHz
>128 MB RAM
>1 GB swap
>Debian unstable
>
>Workload:
>XFree86
>FVWM
>XMMS
>Wine running "Master of Orion 2" (a round based space strategy game)
>
>With 2.4 kernels and 2.5.72 everything works fine.
>
>With 2.6.0-test? and 2.6.0-test?-mm? kernels up to 2.6.0-test4-mm4 the
>XMMS sound sometimes skips or sounds slow (like when wou manually retard
>a record). That's much more awful than skips.
>
>RAM usage is low, even after a "swapoff -a" at about half of my RAM
>would be enough.
>
>The problems might be related to the fact that after I start Wine three
>wine.bin processes run and each of them tries to get as much CPU time as
>possible.
>
>It might be part of the problem that although Wine is the interactive 
>task a working XMMS is subjectively more important.
>
>With 2.6.0-test4-mm5 these problems don't occur. Instead, Wine feels 
>slow. I couldn;t test it much since after the first fast mouse movement 
>the X mouse cursor has lost the mouse cursor of the game (this might be 
>a bug in Wine, but it doesnt occur with other kernels).
>
>cu
>Adrian
>

Hi Adrian,
It would be great if you could test the latest mm kernel (mm6 as of now
I think), which has Con's latest stuff in it. You could also test my
newest scheduler patch. Thanks for the feedback.

Re: 2.6.0-test4-mm5 and below: Wine and XMMS problems

[Con Kolivas]

On Sun, 7 Sep 2003 20:08, Adrian Bunk wrote:
> On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
> >...
> > . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
> >   in evaluating the stability, efficacy and relative performance of
> > Nick's work.
> >
> >   We're looking for feedback on the subjective behaviour and on the usual
> >   server benchmarks please.
> >...
>
> Short story:
>
> I'm still using 2.5.72, all of the 2.6.0-test?{,-mm?} kernels have
> problems

What's your X and xmms nice values? Many X servers are reniced to -10 and some 
shells spawn new apps at nice 5. After that the most common thing I find in 
reports are upgrades to newer kernels losing hard disk dma at some stage (due 
to config changes/movements) and it not being noticed.

Con

Re: 2.6.0-test4-mm5 and below: Wine and XMMS problems

[Adrian Bunk]

On Sun, Sep 07, 2003 at 10:18:27PM +1000, Con Kolivas wrote:
> On Sun, 7 Sep 2003 20:08, Adrian Bunk wrote:
> > On Tue, Sep 02, 2003 at 11:18:12PM -0700, Andrew Morton wrote:
> > >...
> > > . Dropped out Con's CPU scheduler work, added Nick's.  This is to help us
> > >   in evaluating the stability, efficacy and relative performance of
> > > Nick's work.
> > >
> > >   We're looking for feedback on the subjective behaviour and on the usual
> > >   server benchmarks please.
> > >...
> >
> > Short story:
> >
> > I'm still using 2.5.72, all of the 2.6.0-test?{,-mm?} kernels have
> > problems
> 
> What's your X and xmms nice values? Many X servers are reniced to -10 and some 

They have both nice value 0.

> shells spawn new apps at nice 5. After that the most common thing I find in 
> reports are upgrades to newer kernels losing hard disk dma at some stage (due 
> to config changes/movements) and it not being noticed.

DMA is wrking without problems.

> Con

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.0-test4-mm5 and below: Wine and XMMS problems

[Adrian Bunk]

On Sun, Sep 07, 2003 at 08:39:14PM +1000, Nick Piggin wrote:
> 
> Hi Adrian,

Hi Nick,

> It would be great if you could test the latest mm kernel (mm6 as of now
> I think), which has Con's latest stuff in it. You could also test my
> newest scheduler patch. Thanks for the feedback.

I didn't check -mm6 (I had a different problem with -mm6 and not that 
much time).

I tried plain test4 with your sched-rollup-v14 and I got these awful
slower sound like when wou manually retard a record.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.0-test4-mm5 and below: Wine and XMMS problems

[Adrian Bunk]

On Tue, Sep 09, 2003 at 01:08:20AM +0200, Adrian Bunk wrote:
> On Sun, Sep 07, 2003 at 08:39:14PM +1000, Nick Piggin wrote:
> > 
> > Hi Adrian,
> 
> Hi Nick,
> 
> > It would be great if you could test the latest mm kernel (mm6 as of now
> > I think), which has Con's latest stuff in it. You could also test my
> > newest scheduler patch. Thanks for the feedback.
> 
> I didn't check -mm6 (I had a different problem with -mm6 and not that 
> much time).
> 
> I tried plain test4 with your sched-rollup-v14 and I got these awful
> slower sound like when wou manually retard a record.

More data:
I tried test5 and test5-mm1.

Both produced this awful slower sound like when wou manually retard a 
record, although my subjective impression was that it happens fewer than 
with test4 with your sched-rollup-v14.

2.5.72 is still better than all recent kernels I've tested...  :-(

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed