2.5.70-mm6

2.5.70-mm6

[Andrew Morton]

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.70/2.5.70-mm6/

. Numerous little fixes and additional work against additional patches.

. Waaay too many "cleanups".  These are taking significant amounts of
  effort and it is time to start learning to live with dirty code.

. -mm kernels will be running at HZ=100 for a while.  This is because
  the anticipatory scheduler's behaviour may be altered by the lower
  resolution.  Some architectures continue to use 100Hz and we need the
  testing coverage which x86 provides.




Changes since 2.5.70-mm5:


 linus.patch

 Latest drop from Linus

-kmalloc_percpu-interface-change.patch
-kmalloc_percpu-interface-change-warning-fix.patch
-DEFINE_PERCPU-in-modules.patch
-irq-check-rate-limit.patch
-irq_desc-others.patch
-force_successful_syscall_return.patch
-proc-stat-btime-fix.patch
-console-blanking-fix.patch
-console-privacy.patch
-fix-tty-driver-mess.patch
-misc3.patch
-cs423x-fixes.patch
-remove-get_current_user.patch
-de_thread-BUG-fix.patch
-sched_best_cpu-fix-01.patch
-sched_best_cpu-fix-02.patch
-sched_best_cpu-fix-03.patch
-hweight64-warning-fix.patch
-dac960-negotiation-fix.patch
-min_free_kbytes.patch

 Merged

+HZ-100.patch

 Set HZ to 100

+kallsyms-build-fix.patch

 Fix the build for !CONFIG_KALLSYMS

+ppc64-sk_family-fix.patch

 Fix ppc64 for the socket field renaming

+x86_64-fixes.patch

 Stuff from Andi

+non-CONFIG_PROC_FS-fix.patch

 Build fix for !CONFIG_PROC_FS

-eat-keys-on-panic.patch

 Dropped, needs rethinking.

+remove_proc_entry-fix.patch

 Fix iaaa1394_core.c (at least) for !CONFIG_PROC_FS

+procfs-jffs-fix.patch

 Another !CONFIG_PROC_FS fix

+fix-numaq-apic-handling.patch

 NUMAQ fix

+cleanup-summit-subarch.patch

 Summit cleanup

+summit-bus-to-node-mapping.patch

 Summit fix

+rocket-devfs-fix.patch

 rocket.c cleanup

+alloc_bootmem_core-BUG-fix.patch

 Init ordering fix

+discontig-empty-node-fix.patch

 dicontigmem fix for nodes which have no memory

+TARGET_CPUS-cleanup-fix.patch

 More gratuitous cleanups

+show_stack-cleanup.patch

 Rationalise show_stack(), show_task(), etc, etc.

+ppc64-show_stack.patch

 Fix it for ppc64

-statfs64-fix.patch
-statfs-overflow-fix.patch
-statfs64-leftovers.patch

 Folded into statfs64.patch
 
+statfs64-x86_64-fixes.patch
+ppc64-statfs-fix.patch
+xfs-statfs-warning-fix.patch

 More statfs64() fixes

+as-autotune-write-batches.patch

 Anticipatory scheduler tuning.




All 153 patches

linus.patch

mm.patch
  add -mmN to EXTRAVERSION

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

kgdb-use-ggdb.patch

HZ-100.patch

kallsyms-build-fix.patch
  Fix build for CONFIG_KALLSYMS=n

config_spinline.patch
  uninline spinlocks for profiling accuracy.

ppc64-fixup.patch
  ppc64 fixup

ppc64-ioctl-pci-update.patch
  From: Anton Blanchard <anton@samba.org>
  Subject: ppc64 stuff

ppc64-reloc_hide.patch

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

ppc64-sk_family-fix.patch
  ppc64: fixup for family/sk_family rename

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

x86_64-fixes.patch
  x86_64 fixes

non-CONFIG_PROC_FS-fix.patch
  Fix the build with !CONFIG_PROC_FS

irqreturn-snd-via-fix.patch
  via sound irqreturn fix

irq_cpustat-cleanup.patch
  irq_cpustat cleanup

config-PAGE_OFFSET.patch
  Configurable kenrel/user memory split

common-ioctl32.patch
  common 32-bit ioctl code

ioctl32-cleanup-sparc64.patch
  ioctl32 cleanup: sparc64

ioctl32-cleanup-x86_64.patch
  x86_64: use common ioctl code

lru_cache_add-check.patch
  lru_cache_add debug check

remove_proc_entry-fix.patch
  remove_proc_entry() fix

procfs-jffs-fix.patch
  JFFS_PROC_FS must depend on JFFS_FS

fix-numaq-apic-handling.patch
  fix apic handling for NUMA-Q

cleanup-summit-subarch.patch
  cleanup conditionals in summit subarch

summit-bus-to-node-mapping.patch
  Subject: [PATCH] provide bus to node mapping for Summit

rocket-devfs-fix.patch
  rocket.c: devfs fix

alloc_bootmem_core-BUG-fix.patch
  add bootmem failure warning

discontig-empty-node-fix.patch
  fix discontig with 0-sized nodes

mem_driver_cleanup.patch
  driver/char/mem.c cleanup

eventpoll-use-after-free-fix.patch
  eventpoll: fix possible use-after-free

TARGET_CPUS-cleanup-fix.patch
  fix TARGET_CPUS inconsistency

fb-image-depth-fix.patch
  fbdev image depth fix

buffer-debug.patch
  buffer.c debugging

show_stack-cleanup.patch
  show_stack() portability and cleanup patch

ppc64-show_stack.patch

statfs64.patch
  Add system calls statfs64 and fstatfs64
  statfs64: handle overflows
  statfs64: remaining filesystems

statfs64-x86_64-fixes.patch
  statfs64 fixes for x86_64

ppc64-statfs-fix.patch

xfs-statfs-warning-fix.patch

VM_RESERVED-check.patch
  VM_RESERVED check

rcu-stats.patch
  RCU statistics reporting

ide_setting_sem-fix.patch

reslabify-pgds-and-pmds.patch
  re-slabify i386 pgd's and pmd's

linux-isp.patch

isp-update-1.patch

list_del-debug.patch
  list_del debug check

airo-schedule-fix.patch
  airo.c: don't sleep in atomic regions

synaptics-mouse-support.patch
  Add Synaptics touchpad tweaking to psmouse driver

resurrect-batch_requests.patch
  bring back the batch_requests function

kblockd.patch
  Create `kblockd' workqueue

cfq-infrastructure.patch

elevator-completion-api.patch
  elevator completion API

as-iosched.patch
  anticipatory I/O scheduler

as-proc-read-write.patch
  AS: pgbench improvement

as-discrete-read-fifo-batches.patch
  AS: discrete read fifo batches

as-sync-async.patch
  AS sync/async batches

as-hash-removal-fix.patch
  AS: hash removal fix

as-jumbo-patch-for-scsi.patch
  AS jumbo patch (for SCSI and TCQ)

as-stupid.patch
  AS: fix stupid thinko

as-no-batch-antic-limit.patch
  AS: no batch-antic-limit

as-autotune-write-batches.patch
  AS: autotune write batches

unplug-use-kblockd.patch
  Use kblockd for running request queues

cfq-2.patch
  CFQ scheduler, #2
  CFQ: update to rq-dyn API

cfq-hash-removal-fix.patch
  CFQ: hash removal fix

cfq-list_del-fix.patch
  CFQ: empty the queuelist

per-queue-nr_requests.patch
  per queue nr_requests

blk-invert-watermarks.patch
  blk_congestion_wait threshold cleanup

blk-fair-batches.patch
  blk-fair-batches

blk-as-hint.patch
  blk-as-hint

get_request_wait-oom-fix.patch
  handle OOM in get_request_wait().

unmap-page-debugging-2.patch
  debug patch: unmap unused kernel pages

unmap-page-debugging-2-fix.patch

slab-poisoning-fix.patch
  slab poisoning fix

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

mmap-prefault.patch
  prefault of executable mmaps

bio-debug-trap.patch
  BIO debugging patch

sound-irq-hack.patch

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

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

lockmeter.patch

ext3-no-bkl.patch

journal_get_write_access-speedup.patch
  JBD: journal_get_write_access() speedup

ext3-concurrent-block-inode-allocation.patch
  ext3: concurrent block/inode allocation
  Fix orlov allocator boundary case

ext3-concurrent-block-allocation-hashed.patch
  ext3: scalable counters and locks
  fix ext3 inode allocator race

jbd-010-b_committed_data-race-fix.patch
  JBD: fix race over access to b_committed_data

jbd-020-locking-schema.patch
  JBD: plan JBD locking schema

jbd-030-remove-splice_lock.patch
  JBD: remove jh_splice_lock

jbd-040-journal_add_journal_head-locking.patch
  JBD: fine-grain journal_add_journal_head locking

jbd-045-rename-journal_unlock_journal_head.patch
  JBD: rename journal_unlock_journal_head to journal_put_journal_head

jbd-050-b_frozen_data-locking.patch
  JBD: Finish protection of journal_head.b_frozen_data

jbd-060-b_committed_data-locking.patch
  JBD: implement b_committed_data locking

jbd-070-b_transaction-locking.patch
  JBD: implement b_transaction locking rules

jbd-080-b_next_transaction-locking.patch
  JBD: Implement b_next_transaction locking rules

jbd-090-b_tnext-locking.patch
  JBD: b_tnext locking

jbd-100-remove-journal_datalist_lock.patch
  JBD: remove journal_datalist_lock

jbd-110-t_nr_buffers-locking.patch
  JBD: t_nr_buffers locking

jbd-120-t_updates-locking.patch
  JBD: t_updates locking

jbd-130-t_outstanding_credits-locking.patch
  JBD: implement t_outstanding_credits locking

jbd-140-t_jcb-locking.patch
  JBD: implement t_jcb locking

jbd-150-j_barrier_count-locking.patch
  JBD: implement j_barrier_count locking

jbd-160-j_running_transaction-locking.patch
  JBD: implement j_running_transaction locking

jbd-170-j_committing_transaction-locking.patch
  JBD: implement j_committing_transaction locking

jbd-180-j_checkpoint_transactions.patch
  JBD: implement j_checkpoint_transactions locking

jbd-190-j_head-locking.patch
  JBD: implement journal->j_head locking

jbd-200-j_tail-locking.patch
  JBD: implement journal->j_tail locking

jbd-210-j_free-locking.patch
  JBD: implement journal->j_free locking

jbd-220-j_commit_sequence-locking.patch
  JBD: implement journal->j_commit_sequence locking

jbd-230-j_commit_request-locking.patch
  JBD: implement j_commit_request locking

jbd-240-dual-revoke-tables.patch
  JBD: implement dual revoke tables.

jbd-250-remove-sleep_on.patch
  JBD: remove remaining sleep_on()s

jbd-300-remove-lock_kernel.patch
  JBD: remove lock_kernel()

jbd-400-remove-lock_journal.patch
  JBD: remove lock_journal()

jbd-510-h_credits-fix.patch
  JBD: journal_release_buffer: handle credits fix

jbd-520-journal_unmap_buffer-race.patch
  JBD: journal_unmap_buffer race fix

jbd-530-walk_page_buffers-race-fix.patch
  ext3: ext3_writepage race fix

jbd-540-journal_try_to_free_buffers-race-fix.patch
  JBD: buffer freeing non-race comment

jbd-550-locking-checks.patch
  JBD: add some locking assertions

jbd-570-transaction-state-locking.patch
  JBD: additional transaction shutdown locking

jbd-580-log_start_commit-race-fix.patch
  JBD: fix log_start_commit race

jbd-590-do_get_write_access-speedup.patch
  JBD: do_get_write_access() speedup

ext3-010-fix-journalled-data.patch
  ext3: fix data=journal mode

ext3-035-journal_try_to_free_buffers-race-fix.patch

ext3-040-recursive-ext3_write_inode-check.patch
  ext3: add a dump_stack()

ext3-050-ioctl-transaction-leak.patch
  ext3: fix error-path handle leak

ext3-070-xattr-clone-leak-fix.patch
  Fix leak in ext3_acl_chmod()

ext3-080-remove-block-inode-count-message.patch
  ext3: remove mount-time diagnostic messages

jbd-600-journal_dirty_metadata-speedup.patch
  JBD: journal_dirty_metadata() speedup

jbd-610-journal_dirty_metadata-diags.patch
  JBD: journal_dirty_metadata diagnostics

jbd-620-commit-vs-start-race-fix.patch
  JBD: fix race between journal_commit_transaction and start_this_handle

invalidate_mmap_range.patch
  Interface to invalidate regions of mmaps

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

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

aio-poll.patch
  aio_poll
  aio-poll: don't put extern decls in .c!

O_SYNC-speedup-2.patch
  speed up O_SYNC writes

aio-09-o_sync.patch
  aio O_SYNC

vfsmount_lock.patch
  From: Maneesh Soni <maneesh@in.ibm.com>
  Subject: [patch 1/2] vfsmount_lock

syncppp-locking-fix.patch
  syncppp locking fix

s390-dirty-bit-cleaning.patch
  dirty bit clearing on s390.

sched-hot-balancing-fix.patch
  fix for CPU scheduler load distribution

Re: 2.5.70-mm6

[Alexander Hoogerhuis]

Andrew Morton <akpm@digeo.com> writes:
>
> [SNIP]
>

It builds nicely here and runs nicely so far, but my USB-drive still
blows up after a few gigs and I have this one when plugging it in:

Attached scsi generic sg0 at scsi0, channel 0, id 0, lun 0,  type 0
spurious 8259A interrupt: IRQ7.
SCSI device sda: 490232832 512-byte hdwr sectors (250999 MB)
sda: cache data unavailable
sda: assuming drive cache: write through
 /dev/scsi/host0/bus0/target0/lun0: p1
devfs_mk_dir(scsi/host0/bus0/target0/lun0): could not append to dir: ea549820 "target0"
Attached scsi disk sda at scsi0, channel 0, id 0, lun 0
kjournald starting.  Commit interval 5 seconds
EXT3 FS 2.4-0.9.16, 02 Dec 2001 on sda1, internal journal
EXT3-fs: recovery complete.
EXT3-fs: mounted filesystem with ordered data mode.

mvh,
A
-- 
Alexander Hoogerhuis                               | alexh@ihatent.com
CCNP - CCDP - MCNE - CCSE                          | +47 908 21 485
"You have zero privacy anyway. Get over it."  --Scott McNealy

Re: 2.5.70-mm6

[Andrew Morton]

Alexander Hoogerhuis <alexh@ihatent.com> wrote:
>
> Andrew Morton <akpm@digeo.com> writes:
> >
> > [SNIP]
> >
> 
> It builds nicely here and runs nicely so far, but my USB-drive still
> blows up after a few gigs

Is that usb-storage?  There seem to have been a few reports of
erratic behaviour lately.

> and I have this one when plugging it in:
> 
> Attached scsi generic sg0 at scsi0, channel 0, id 0, lun 0,  type 0
> spurious 8259A interrupt: IRQ7.
> SCSI device sda: 490232832 512-byte hdwr sectors (250999 MB)
> sda: cache data unavailable
> sda: assuming drive cache: write through
>  /dev/scsi/host0/bus0/target0/lun0: p1
> devfs_mk_dir(scsi/host0/bus0/target0/lun0): could not append to dir: ea549820 "target0"

Maybe Christph can decode this one for us.

Re: 2.5.70-mm6

[Alexander Hoogerhuis]

Andrew Morton <akpm@digeo.com> writes:

> Alexander Hoogerhuis <alexh@ihatent.com> wrote:
> >
> > Andrew Morton <akpm@digeo.com> writes:
> > >
> > > [SNIP]
> > >
> > 
> > It builds nicely here and runs nicely so far, but my USB-drive still
> > blows up after a few gigs
> 
> Is that usb-storage?  There seem to have been a few reports of
> erratic behaviour lately.
> 

Indeed. Lots of the noise is from me :) You CC'ed Matthew Dharm on a
mail about it, so I tucked him in the CC here too.

mvh,
A
-- 
Alexander Hoogerhuis                               | alexh@ihatent.com
CCNP - CCDP - MCNE - CCSE                          | +47 908 21 485
"You have zero privacy anyway. Get over it."  --Scott McNealy

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Sun, 2003-06-08 at 00:14, Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.70/2.5.70-mm6/
> 
> . Numerous little fixes and additional work against additional patches.
> 
> . Waaay too many "cleanups".  These are taking significant amounts of
>   effort and it is time to start learning to live with dirty code.
> 
> . -mm kernels will be running at HZ=100 for a while.  This is because
>   the anticipatory scheduler's behaviour may be altered by the lower
>   resolution.  Some architectures continue to use 100Hz and we need the
>   testing coverage which x86 provides.

Testing it right now... It compiles nicely with gcc 3.3 (remember the
problems I had with snd-ymfpci when using gcc 3.2), boots and seems
functional.

Re: 2.5.70-mm6

[Christoph Hellwig]

On Sat, Jun 07, 2003 at 05:56:49PM -0700, Andrew Morton wrote:
> > Attached scsi generic sg0 at scsi0, channel 0, id 0, lun 0,  type 0
> > spurious 8259A interrupt: IRQ7.
> > SCSI device sda: 490232832 512-byte hdwr sectors (250999 MB)
> > sda: cache data unavailable
> > sda: assuming drive cache: write through
> >  /dev/scsi/host0/bus0/target0/lun0: p1
> > devfs_mk_dir(scsi/host0/bus0/target0/lun0): could not append to dir: ea549820 "target0"
> 
> Maybe Christph can decode this one for us.

There's multiple gendisks with the same .devfs_name in scsi and we call
devfs_mk_dir on each of them.  I think the right fix is to let
devfs_mk_dir succeed on an existing directory.


--- 1.91/fs/devfs/base.c	Wed Jun  4 09:50:29 2003
+++ edited/fs/devfs/base.c	Sat Jun  7 16:09:26 2003
@@ -1684,7 +1684,10 @@
 	}
 
 	error = _devfs_append_entry(dir, de, &old);
-	if (error) {
+	if (error == -EEXIST) {
+		error = 0;
+		devfs_put(old);
+	} else if (error) {
 		PRINTK("(%s): could not append to dir: %p \"%s\"\n",
 				buf, dir, dir->name);
 		devfs_put(old);

[patch] 2.5.70-mm6: ethertap.c doesn't compile

[Adrian Bunk]

I got the following compile error in 2.5.70-mm6:

<--  snip  -->

...
  CC      drivers/net/ethertap.o
drivers/net/ethertap.c: In function `ethertap_rx':
drivers/net/ethertap.c:295: structure has no member named `protocol'
drivers/net/ethertap.c:300: structure has no member named `receive_queue'
drivers/net/ethertap.c:307: structure has no member named `receive_queue'
drivers/net/ethertap.c: In function `ethertap_close':
drivers/net/ethertap.c:323: structure has no member named `socket'
make[2]: *** [drivers/net/ethertap.o] Error 1

<--   snip  -->

The patch below fixes the compilation.



cu
Adrian

--- linux-2.5.70-mm6/drivers/net/ethertap.c.old	2003-06-08 17:48:57.000000000 +0200
+++ linux-2.5.70-mm6/drivers/net/ethertap.c	2003-06-08 17:49:53.000000000 +0200
@@ -292,19 +292,19 @@
 
 static void ethertap_rx(struct sock *sk, int len)
 {
-	struct net_device *dev = tap_map[sk->protocol];
+	struct net_device *dev = tap_map[sk->sk_protocol];
 	struct sk_buff *skb;
 
 	if (dev==NULL) {
 		printk(KERN_CRIT "ethertap: bad unit!\n");
-		skb_queue_purge(&sk->receive_queue);
+		skb_queue_purge(&sk->sk_receive_queue);
 		return;
 	}
 
 	if (ethertap_debug > 3)
 		printk("%s: ethertap_rx()\n", dev->name);
 
-	while ((skb = skb_dequeue(&sk->receive_queue)) != NULL)
+	while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL)
 		ethertap_rx_skb(skb, dev);
 }
 
@@ -320,7 +320,7 @@
 
 	if (sk) {
 		lp->nl = NULL;
-		sock_release(sk->socket);
+		sock_release(sk->sk_socket);
 	}
 
 	return 0;

Re: [patch] 2.5.70-mm6: ethertap.c doesn't compile

[Arnaldo Carvalho de Melo]

Thanks for the patch, but it was already submitted by Martin@gentoo and I
pushed to Linus that already has this in his tree.

- Arnaldo

Em Sun, Jun 08, 2003 at 05:56:23PM +0200, Adrian Bunk escreveu:
> 
> I got the following compile error in 2.5.70-mm6:
> 
> <--  snip  -->
> 

Re: [patch] 2.5.70-mm6: ethertap.c doesn't compile

[Adrian Bunk]

On Sun, Jun 08, 2003 at 01:00:55PM -0300, Arnaldo Carvalho de Melo wrote:

> Thanks for the patch, but it was already submitted by Martin@gentoo and I
> pushed to Linus that already has this in his tree.

Ups, yes, I missed this patch...

> - Arnaldo

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: 2.5.70-mm6

[Pasi Savolainen]

* Andrew Morton <akpm@digeo.com>:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.70/2.5.70-mm6/

Xfree86 4.3.0 won't start on this one. -mm4 started fine.
X will stop (and seemingly hang) on PCI initialization and iteration stage.

> linus.patch

I'd say this is the source of this. Some cleanup along pci_for_each_dev
removal. All the 'fixes' (into while) don't even compile warningless.


-- 
   Psi -- <http://www.iki.fi/pasi.savolainen>

[patch] 2.5.70-mm6: isp driver cleanups

[Adrian Bunk]

On Sat, Jun 07, 2003 at 03:14:40PM -0700, Andrew Morton wrote:
>...
> linux-isp.patch
> 
> isp-update-1.patch
>...

isp_linux.h already states kernels < 2.4 are not supported. The patch 
below removes #ifdef'd code for kernels < 2.4.

diffstat output:

 isp_linux.c |   42 ----------
 isp_linux.h |  100 +-----------------------
 isp_pci.c   |  248 ------------------------------------------------------------
 3 files changed, 6 insertions(+), 384 deletions(-)

I've tested the compilation with 2.5.70-mm6.

cu
Adrian

--- linux-2.5.70-mm6/drivers/scsi/isp/isp_linux.h.old	2003-06-09 01:07:27.000000000 +0200
+++ linux-2.5.70-mm6/drivers/scsi/isp/isp_linux.h	2003-06-09 00:58:53.000000000 +0200
@@ -39,35 +39,13 @@
 #ifndef _ISP_LINUX_H
 #define _ISP_LINUX_H
 
-#ifndef	ISP_MODULE
-#define	__NO_VERSION__
-#endif
-#ifdef	LINUX_ISP_TARGET_MODE
-#define	EXPORT_SYMTAB
-#endif
-
 #include <linux/version.h>
-#ifndef	KERNEL_VERSION
-#define KERNEL_VERSION(v,p,s)		(((v)<<16)+(p<<8)+s)
-#endif
-#define	_KVC	KERNEL_VERSION
-
-#if LINUX_VERSION_CODE <= _KVC(2,2,0)
-#error	"Linux 2.0 and 2.1 kernels are not supported anymore"
-#endif
-#if LINUX_VERSION_CODE >= _KVC(2,3,0) && LINUX_VERSION_CODE < _KVC(2,4,0)
-#error	"Linux 2.3 kernels are not supported"
-#endif
 
-#ifndef	UNUSED_PARAMETER
-#define	UNUSED_PARAMETER(x)	(void) x
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
+#error	"Linux 2.3 and earlier kernels are not supported"
 #endif
 
 #include <linux/autoconf.h>
-#ifdef	CONFIG_SMP
-#define	__SMP__	1
-#endif
-
 
 /*
  * Be nice and get ourselves out of the way of other drivers.
@@ -103,12 +81,8 @@
 #include <asm/dma.h>
 #include <asm/io.h>
 #include <asm/irq.h>
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 #include <linux/smp.h>
 #include <linux/spinlock.h>
-#else
-#include <asm/spinlock.h>
-#endif
 #include <asm/system.h>
 #include <asm/byteorder.h>
 #include <linux/interrupt.h>
@@ -141,15 +115,15 @@
 #undef	__pa
 #define	__pa(x)	x
 #endif
-#if defined (__i386__) && LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
+#if defined (__i386__)
 #undef	__pa
 #define	__pa(x)	x
 #endif
-#if defined (__sparc__) && LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
+#if defined (__sparc__)
 #undef	__pa
 #define	__pa(x)	x
 #endif
-#if defined (__alpha__) && LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
+#if defined (__alpha__)
 #undef	__pa
 #define	__pa(x)	x
 #endif
@@ -180,11 +154,6 @@
 #define	BYTE_ORDER	LITTLE_ENDIAN
 #endif
 
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-#define	DMA_ADDR_T		unsigned long
-#define	QLA_SG_C(sg)		sg->length
-#define	QLA_SG_A(sg)		virt_to_bus(sg->address)
-#else
 #define	DMA_ADDR_T		dma_addr_t
 #define	QLA_SG_C(sg)		sg_dma_len(sg)
 #define	QLA_SG_A(sg)		sg_dma_address(sg)
@@ -195,7 +164,6 @@
 #define	DMA_HTYPE_T		dma_addr_t
 #define	QLA_HANDLE(cmd)		(cmd)->SCp.dma_handle
 #endif
-#endif
 
 #define	HANDLE_LOOPSTATE_IN_OUTER_LAYERS	1
 #ifdef	min
@@ -398,20 +366,6 @@
 #define	ISP_IUNLK_SOFTC			ISP_UNLK_SOFTC
 #define	ISP_IGET_LK_SOFTC		ISP_LOCK_SOFTC
 #define	ISP_DROP_LK_SOFTC		ISP_UNLK_SOFTC
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-#define	ISP_LOCK_SCSI_DONE(isp)		{				\
-		unsigned long _flags;					\
-		spin_lock_irqsave(&io_request_lock, _flags);		\
-		isp->iflags = _flags;					\
-	}
-#define	ISP_UNLK_SCSI_DONE(isp)		{				\
-		unsigned long _flags = isp->iflags;			\
-		spin_unlock_irqrestore(&io_request_lock, _flags);	\
-	}
-#else
-#define	ISP_LOCK_SCSI_DONE(isp)		do { } while(0)
-#define	ISP_UNLK_SCSI_DONE(isp)		do { } while(0)
-#endif
 #define	ISP_LOCKU_SOFTC			ISP_ILOCK_SOFTC
 #define	ISP_UNLKU_SOFTC			ISP_IUNLK_SOFTC
 #define	ISP_TLOCK_INIT(isp)		spin_lock_init(&isp->isp_osinfo.tlock)
@@ -448,13 +402,8 @@
 
 #define	ISP2100_SCRLEN		0x800
 
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-#define	MEMZERO			_isp_memzero
-#define	MEMCPY			_isp_memcpy
-#else
 #define	MEMZERO(b, a)		memset(b, 0, a)
 #define	MEMCPY			memcpy
-#endif
 #define	SNPRINTF		isp_snprintf
 #define	USEC_DELAY		_isp_usec_delay
 #define	USEC_SLEEP(isp, x)						\
@@ -476,11 +425,7 @@
   ((u_int64_t) ((((u_int64_t)(x)->tv_sec) * 1000000 + (x)->tv_usec)))
 #define	NANOTIME_SUB		_isp_microtime_sub
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-#define	MAXISPREQUEST(isp)	256
-#else
 #define	MAXISPREQUEST(isp)	((IS_FC(isp) || IS_ULTRA2(isp))? 1024 : 256)
-#endif
 
 #if	defined(__i386__)
 #define	MEMORYBARRIER(isp, type, offset, size)	barrier()
@@ -739,10 +684,6 @@
 int isp_drain_reset(struct ispsoftc *, char *);
 int isp_drain(struct ispsoftc *, char *);
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-static INLINE void _isp_memcpy(void *, void *, size_t);
-static INLINE void _isp_memzero(void *,  size_t);
-#endif
 static INLINE u_int64_t _isp_microtime_sub(struct timeval *, struct timeval *);
 static INLINE void _isp_usec_delay(unsigned int);
 static INLINE unsigned long _usec_to_jiffies(unsigned int);
@@ -816,37 +757,6 @@
 #define	_SBSWAP(a, b, c)
 #endif
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-static INLINE void
-_isp_memcpy(void *to, void *from, size_t amt)
-{
-	unsigned char *x = to; unsigned char *y = from;
-	while (amt-- != 0) *x++ = *y++;
-}
-
-static INLINE void
-_isp_memzero(void *to, size_t amt)
-{
-	unsigned char *x = to;
-	while (amt-- != 0) *x++ = 0;
-}
-
-static INLINE unsigned long IspOrder(int);
-static INLINE unsigned long
-IspOrder(int nelem)
-{
-    unsigned long order, rsize;
-
-    order = 0;
-    rsize = PAGE_SIZE;
-    while (rsize < (unsigned long) ISP_QUEUE_SIZE(nelem)) {
-	order++;
-	rsize <<= 1;
-    }
-    return (order);
-}
-#endif
-
 static INLINE u_int64_t
 _isp_microtime_sub(struct timeval *b, struct timeval *a)
 {
--- linux-2.5.70-mm6/drivers/scsi/isp/isp_linux.c.old	2003-06-09 00:41:58.000000000 +0200
+++ linux-2.5.70-mm6/drivers/scsi/isp/isp_linux.c	2003-06-09 01:05:32.000000000 +0200
@@ -91,11 +91,6 @@
 static int isp_en_dis_lun(struct ispsoftc *, int, int, int, int);
 #endif
 
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,3,27)
-struct proc_dir_entry proc_scsi_qlc = {
-    PROC_SCSI_QLOGICISP, 3, "isp", S_IFDIR | S_IRUGO | S_IXUGO, 2
-};
-#endif
 static const char *class3_roles[4] = {
     "None", "Target", "Initiator", "Target/Initiator"
 };
@@ -404,11 +399,7 @@
 isplinux_detect(Scsi_Host_Template *tmpt)
 {
     int rval;
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-    tmpt->proc_dir = &proc_scsi_qlc;
-#else
     tmpt->proc_name = "isp";
-#endif
 #if	LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
     spin_unlock_irq(&io_request_lock);
     rval = isplinux_pci_detect(tmpt);
@@ -511,9 +502,7 @@
 	scsi_add_timer(Cmnd, Cmnd->timeout_per_command, Cmnd->done);
 	isp_prt(isp, ISP_LOGDEBUG0, "giving up on target %d", XS_TGT(Cmnd));
 	ISP_UNLK_SOFTC(isp);
-	ISP_LOCK_SCSI_DONE(isp);
 	(*Cmnd->scsi_done)(Cmnd);
-	ISP_UNLK_SCSI_DONE(isp);
 	ISP_LOCK_SOFTC(isp);
 	return;
     }
@@ -650,9 +639,7 @@
 	 * Add back a timer else scsi_done drops this on the floor.
 	 */
 	scsi_add_timer(Cmnd, Cmnd->timeout_per_command, Cmnd->done);
-	ISP_LOCK_SCSI_DONE(isp);
 	(*Cmnd->scsi_done)(Cmnd);
-	ISP_UNLK_SCSI_DONE(isp);
     } while ((Cmnd = Ncmnd) != NULL);
     ISP_LOCK_SOFTC(isp);
 }
@@ -2466,7 +2453,6 @@
     }
     ISP_IUNLK_SOFTC(isp);
     if (Cmnd) {
-	ISP_LOCK_SCSI_DONE(isp);
         while (Cmnd) {
 	    Scsi_Cmnd *f = (Scsi_Cmnd *) Cmnd->host_scribble;
 	    Cmnd->host_scribble = NULL;
@@ -2479,7 +2465,6 @@
 	    (*Cmnd->scsi_done)(Cmnd);
 	    Cmnd = f;
 	}
-	ISP_UNLK_SCSI_DONE(isp);
     }
 }
 
@@ -2532,7 +2517,6 @@
     ISP_IUNLK_SOFTC(isp);
 #endif
     if (Cmnd) {
-	ISP_LOCK_SCSI_DONE(isp);
         while (Cmnd) {
 	    Scsi_Cmnd *f = (Scsi_Cmnd *) Cmnd->host_scribble;
 	    Cmnd->host_scribble = NULL;
@@ -2545,7 +2529,6 @@
 	    (*Cmnd->scsi_done)(Cmnd);
 	    Cmnd = f;
 	}
-	ISP_UNLK_SCSI_DONE(isp);
     }
     return IRQ_HANDLED;
 }
@@ -2839,11 +2822,7 @@
     }
     isp->isp_state = ISP_RUNSTATE;
 
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     isp->isp_osinfo.host->can_queue = isp->isp_maxcmds;
-#else
-    isp->isp_osinfo.host->can_queue = min(255, isp->isp_maxcmds);
-#endif
     if (isp->isp_osinfo.host->can_queue == 0)
 	isp->isp_osinfo.host->can_queue = 1;
 
@@ -2947,12 +2926,6 @@
     return (0);
 }
 
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-#define	ISP_THREAD_CAN_EXIT	isp->isp_host->loaded_as_module
-#else
-#define	ISP_THREAD_CAN_EXIT	0
-#endif
-
 static int
 isp_task_thread(void *arg)
 {
@@ -2964,11 +2937,7 @@
 
 #if	LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
     /* XXX: Not really sure why the 2.5.X changes do this */
-    if (ISP_THREAD_CAN_EXIT) {
-	siginitsetinv(&current->blocked, sigmask(SIGHUP));
-    } else {
-	siginitsetinv(&current->blocked, 0);
-    }
+    siginitsetinv(&current->blocked, 0);
     lock_kernel();
     daemonize();
     sprintf(current->comm, "isp_thrd%d", isp->isp_unit);
@@ -2988,10 +2957,6 @@
     while (exit_thread == 0) {
 	isp_prt(isp, ISP_LOGDEBUG1, "isp_task_thread sleeping");
 	down_interruptible(&thread_sleep_semaphore);
-	if (ISP_THREAD_CAN_EXIT) {
-	    if (signal_pending(current))
-		break;
-	}
 	isp_prt(isp, ISP_LOGDEBUG1, "isp_task_thread running");
 
 	spin_lock_irqsave(&isp->isp_osinfo.tlock, flags);
@@ -3063,9 +3028,6 @@
 		ISP_UNLKU_SOFTC(isp);
 		break;
 	    case ISP_THREAD_EXIT:
-		if (ISP_THREAD_CAN_EXIT) {
-		    exit_thread = 1;
-		}
 		break;
 	   default:
 		break;
@@ -3146,10 +3108,8 @@
 MODULE_PARM(isp_default_frame_size, "i");
 MODULE_PARM(isp_default_exec_throttle, "i");
 #endif
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) || defined(MODULE)
 Scsi_Host_Template driver_template = QLOGICISP;
 #include "scsi_module.c"
-#endif
 /*
  * mode: c
  * Local variables:
--- linux-2.5.70-mm6/drivers/scsi/isp/isp_pci.c.old	2003-06-09 00:59:34.000000000 +0200
+++ linux-2.5.70-mm6/drivers/scsi/isp/isp_pci.c	2003-06-09 01:02:38.000000000 +0200
@@ -63,13 +63,8 @@
 static int
 isp_pci_dmasetup(struct ispsoftc *, XS_T *, ispreq_t *, u_int16_t *, u_int16_t);
 
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 static void isp_pci_dmateardown(struct ispsoftc *, XS_T *, u_int16_t);
 #define	IS_HIGH_ISP_ADDR(addr)	(addr >= (u_int64_t) 0x100000000)
-#else
-#define	isp_pci_dmateardown	NULL
-#define	IS_HIGH_ISP_ADDR(addr)	0
-#endif
 
 #if	ISP_DAC_SUPPORTED == 1
 #define	ISP_A64			1
@@ -544,7 +539,6 @@
 	pcs->port = 0;
     }
     kfree(isp->isp_xflist);
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     pci_free_consistent(pcs->pci_dev, RQUEST_QUEUE_LEN(isp) * QENTRY_LEN,
 	isp->isp_rquest, isp->isp_rquest_dma);
     pci_free_consistent(pcs->pci_dev, RESULT_QUEUE_LEN(isp) * QENTRY_LEN,
@@ -553,13 +547,6 @@
 	pci_free_consistent(pcs->pci_dev, ISP2100_SCRLEN,
 	    FCPARAM(isp)->isp_scratch, FCPARAM(isp)->isp_scdma);
     }
-#else
-    RlsPages(isp->isp_rquest, IspOrder(RQUEST_QUEUE_LEN(isp)));
-    RlsPages(isp->isp_result, IspOrder(RESULT_QUEUE_LEN(isp)));
-    if (IS_FC(isp) && FCPARAM(isp)->isp_scratch) {
-	RlsPages(FCPARAM(isp)->isp_scratch, 1);
-    }
-#endif
 #if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,18) && \
     	LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
     if (--isp_nfound <= 0) {
@@ -594,7 +581,6 @@
     vid = isp_pci->pci_dev->vendor;
     did = isp_pci->pci_dev->device;
 
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     io_base = pci_resource_start(isp_pci->pci_dev, 0);
     if (pci_resource_flags(isp_pci->pci_dev, 0) & PCI_BASE_ADDRESS_MEM_TYPE_64)
 	irq = 2;
@@ -609,17 +595,6 @@
 	isp_pci_mapmem &= ~(1 << isp->isp_unit);
 #endif
     }
-#else	/* LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) */
-    io_base = isp_pci->pci_dev->base_address[0];
-    mem_base = isp_pci->pci_dev->base_address[1];
-    if (mem_base & PCI_BASE_ADDRESS_MEM_TYPE_64) {
-#if	BITS_PER_LONG == 64
-	mem_base |= isp_pci->pci_dev->base_address[2] << 32;
-#else
-	isp_pci_mapmem &= ~(1 << isp->isp_unit);
-#endif
-    }
-#endif	/* LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) */
     irq = isp_pci->pci_dev->irq;
 
     if (vid != PCI_VENDOR_ID_QLOGIC) {
@@ -675,13 +650,11 @@
 	return (1);
     }
 
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     if (pci_enable_device(isp_pci->pci_dev)) {
 	printk("%s: fails to be PCI_ENABLEd\n", loc);
 	return (1);
     }
     (void) PRDW(isp_pci, PCI_COMMAND, &cmd);
-#endif
 
     if ((cmd & PCI_CMD_ISP) != pci_cmd_isp) {
 	if (isp_debug & ISP_LOGINFO)
@@ -1165,7 +1138,6 @@
 	MEMZERO(isp->isp_xflist, amt);
     }
     if (isp->isp_rquest == NULL) {
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 	struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
 	dma_addr_t busaddr;
 	isp->isp_rquest =
@@ -1176,17 +1148,6 @@
 	    return (1);
 	}
 	isp->isp_rquest_dma = busaddr;
-#else
-	isp->isp_rquest = (caddr_t) GetPages(IspOrder(RQUEST_QUEUE_LEN(isp)));
-	if (isp->isp_rquest == NULL) {
-	    isp_prt(isp, ISP_LOGERR, "unable to allocate request queue");
-	    return (1);
-	}
-	/*
-	 * Map the Request queue.
-	 */
-	isp->isp_rquest_dma = virt_to_bus(isp->isp_rquest);
-#endif
 	if (isp->isp_rquest_dma & 0x3f) {
 	    isp_prt(isp, ISP_LOGERR, "Request Queue not on 64 byte boundary");
 	    return (1);
@@ -1195,7 +1156,6 @@
     }
 
     if (isp->isp_result == NULL) {
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 	struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
 	dma_addr_t busaddr;
 	isp->isp_result =
@@ -1206,19 +1166,6 @@
 	    return (1);
 	}
 	isp->isp_result_dma = busaddr;
-#else
-	isp->isp_result = (caddr_t) GetPages(IspOrder(RESULT_QUEUE_LEN(isp)));
-	if (isp->isp_result == NULL) {
-	    isp_prt(isp, ISP_LOGERR, "unable to allocate result queue");
-	    free_pages((unsigned long)isp->isp_rquest,
-		IspOrder(RQUEST_QUEUE_LEN(isp)));
-	    return (1);
-	}
-	/*
-	 * Map the result queue.
-	 */
-	isp->isp_result_dma = virt_to_bus(isp->isp_result);
-#endif
 	if (isp->isp_rquest_dma & 0x3f) {
 	    isp_prt(isp, ISP_LOGERR, "Result Queue not on 64 byte boundary");
 	    return (1);
@@ -1229,7 +1176,6 @@
     if (IS_FC(isp)) {
 	fcparam *fcp = isp->isp_param;
 	if (fcp->isp_scratch == NULL) {
-#if	LINUX_VERSION_CODE > KERNEL_VERSION(2,3,92)
 	    struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
 	    dma_addr_t busaddr;
 	    fcp->isp_scratch =
@@ -1239,17 +1185,6 @@
 		return (1);
 	    }
 	    fcp->isp_scdma = busaddr;
-#else
-	   /*
-	    * Just get a page....
-	    */
-	    fcp->isp_scratch = (void *) GetPages(1);
-	    if (fcp->isp_scratch == NULL) {
-		isp_prt(isp, ISP_LOGERR, "unable to allocate scratch space");
-		return (1);
-	    }
-	    fcp->isp_scdma = virt_to_bus((void *)fcp->isp_scratch);
-#endif
 	    MEMZERO(fcp->isp_scratch, ISP2100_SCRLEN);
 	    if (fcp->isp_scdma & 0x7) {
 		isp_prt(isp, ISP_LOGERR, "scratch space not 8 byte aligned");
@@ -1325,7 +1260,6 @@
 	sg++;
     }
     sg = tcmd->cd_data;
-#if	 LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     {
 	struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
 	int new_seg_cnt;
@@ -1342,7 +1276,6 @@
 	    return (CMD_COMPLETE);
 	}
     }
-#endif
     nctios = nseg / ISP_RQDSEG;
     if (nseg % ISP_RQDSEG) {
 	nctios++;
@@ -1402,11 +1335,7 @@
 		 * Unlike normal initiator commands, we don't do
 		 * any swizzling here.
 		 */
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-		cto->ct_dataseg[seg].ds_base = virt_to_bus(sg->address);
-#else
 		cto->ct_dataseg[seg].ds_base = (u_int32_t) sg_dma_address(sg);
-#endif
 		cto->ct_dataseg[seg].ds_count = (u_int32_t) sg->length;
 		cto->ct_xfrlen += sg->length;
 		sg++;
@@ -1586,7 +1515,6 @@
 	sg++;
     }
     sg = tcmd->cd_data;
-#if	 LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
     {
 	struct isp_pcisoftc *pcs = (struct isp_pcisoftc *) isp;
 	int new_seg_cnt;
@@ -1603,7 +1531,6 @@
 	    return (CMD_COMPLETE);
 	}
     }
-#endif
     nctios = nseg / ISP_RQDSEG_T2;
     if (nseg % ISP_RQDSEG_T2) {
 	nctios++;
@@ -1675,12 +1602,8 @@
 		 * Unlike normal initiator commands, we don't do
 		 * any swizzling here.
 		 */
-#if	LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
-		cto->rsp.m0.ct_dataseg[seg].ds_base = virt_to_bus(sg->address);
-#else
 		cto->rsp.m0.ct_dataseg[seg].ds_base =
 		    (u_int32_t) sg_dma_address(sg);
-#endif
 		cto->rsp.m0.ct_dataseg[seg].ds_count = (u_int32_t) sg->length;
 		cto->rsp.m0.ct_xfrlen += sg->length;
 		sg++;
@@ -1818,7 +1741,6 @@
 }
 #endif
 
-#if	LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 #define	FOURG_SEG(x)	(((u64) (x)) & 0xffffffff00000000ULL)
 
 static int
@@ -2171,176 +2093,6 @@
     }
 }
 
-#else
-
-static int
-isp_pci_dmasetup(struct ispsoftc *isp, Scsi_Cmnd *Cmnd, ispreq_t *rq,
-    u_int16_t *nxi, u_int16_t optr)
-{
-    struct scatterlist *sg;
-    DMA_ADDR_T one_shot_addr;
-    unsigned int one_shot_length;
-    int segcnt, seg, ovseg, seglim;
-    void *h;
-    u_int16_t nxti;
-
-#ifdef	LINUX_ISP_TARGET_MODE
-    if (rq->req_header.rqs_entry_type == RQSTYPE_CTIO ||
-        rq->req_header.rqs_entry_type == RQSTYPE_CTIO2) {
-	int s;
-	if (IS_SCSI(isp))
-	    s = tdma_mk(isp, (tmd_cmd_t *)Cmnd, (ct_entry_t *)rq, nxi, optr);
-	else
-	    s = tdma_mkfc(isp, (tmd_cmd_t *)Cmnd, (ct2_entry_t *)rq, nxi, optr);
-	return (s);
-   }
-#endif
-
-    nxti = *nxi;
-    h = (void *) ISP_QUEUE_ENTRY(isp->isp_rquest, isp->isp_reqidx);
-
-    if (Cmnd->request_bufflen == 0) {
-	rq->req_seg_count = 1;
-	goto mbxsync;
-    }
-
-    if (IS_FC(isp)) {
-	if (rq->req_header.rqs_entry_type == RQSTYPE_T3RQS)
-	    seglim = ISP_RQDSEG_T3;
-	else
-	    seglim = ISP_RQDSEG_T2;
-	((ispreqt2_t *)rq)->req_totalcnt = Cmnd->request_bufflen;
-	/*
-	 * Linux doesn't make it easy to tell which direction
-	 * the data is expected to go, and you really need to
-	 * know this for FC. We'll have to assume that some
-	 * of these commands that might be used for writes
-	 * our outbounds and all else are inbound.
-	 */
-	switch (Cmnd->cmnd[0]) {
-	case FORMAT_UNIT:
-	case WRITE_6:
-	case MODE_SELECT:
-	case SEND_DIAGNOSTIC:
-	case WRITE_10:
-	case WRITE_BUFFER:
-	case WRITE_LONG:
-	case WRITE_SAME:
-	case MODE_SELECT_10:
-	case WRITE_12:
-	case WRITE_VERIFY_12:
-	case SEND_VOLUME_TAG:
-	    ((ispreqt2_t *)rq)->req_flags |= REQFLAG_DATA_OUT;
-	    break;
-	default:
-	    ((ispreqt2_t *)rq)->req_flags |= REQFLAG_DATA_IN;
-	}
-    } else {
-	if (Cmnd->cmd_len > 12)
-	    seglim = 0;
-	else
-	    seglim = ISP_RQDSEG;
-	rq->req_flags |=  REQFLAG_DATA_OUT | REQFLAG_DATA_IN;
-    }
-
-    one_shot_addr = (DMA_ADDR_T) 0;
-    one_shot_length = 0;
-    if ((segcnt = Cmnd->use_sg) == 0) {
-	segcnt = 1;
-	sg = NULL;
-	one_shot_length = Cmnd->request_bufflen;
-	one_shot_addr = virt_to_bus(Cmnd->request_buffer);
-    } else {
-	sg = (struct scatterlist *) Cmnd->request_buffer;
-    }
-    if (segcnt == 0) {
-	isp_prt(isp, ISP_LOGWARN, "unable to dma map request");
-	XS_SETERR(Cmnd, HBA_BOTCH);
-	return (CMD_EAGAIN);
-    }
-
-    for (seg = 0, rq->req_seg_count = 0;
-	 seg < segcnt && rq->req_seg_count < seglim;
-	 seg++, rq->req_seg_count++) {
-	DMA_ADDR_T addr;
-	unsigned int length;
-
-	if (sg) {
-		length = QLA_SG_C(sg);
-		addr = QLA_SG_A(sg);
-		sg++;
-	} else {
-		length = one_shot_length;
-		addr = one_shot_addr;
-	}
-
-	if (rq->req_header.rqs_entry_type == RQSTYPE_T2RQS) {
-	    ispreqt2_t *rq2 = (ispreqt2_t *)rq;
-	    rq2->req_dataseg[rq2->req_seg_count].ds_count = length;
-	    rq2->req_dataseg[rq2->req_seg_count].ds_base = addr;
-	} else {
-	    rq->req_dataseg[rq->req_seg_count].ds_count = length;
-	    rq->req_dataseg[rq->req_seg_count].ds_base = addr;
-	}
-	isp_prt(isp, ISP_LOGDEBUG1, "seg0[%d]%llx:%u from %p", seg,
-	    (long long)addr, length, sg? sg->address : Cmnd->request_buffer);
-    }
-
-    if (seg == segcnt) {
-	goto mbxsync;
-    }
-
-    do {
-	int lim;
-	u_int16_t curip;
-	ispcontreq_t local, *crq = &local, *qep;
-
-	curip = nxti;
-	qep = (ispcontreq_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, curip);
-	nxti = ISP_NXT_QENTRY((curip), RQUEST_QUEUE_LEN(isp));
-	if (nxti == optr) {
-	    isp_prt(isp, ISP_LOGDEBUG0, "out of space for continuations");
-	    XS_SETERR(Cmnd, HBA_BOTCH);
-	    return (CMD_EAGAIN);
-	}
-	rq->req_header.rqs_entry_count++;
-	MEMZERO((void *)crq, sizeof (*crq));
-	crq->req_header.rqs_entry_count = 1;
-	if (rq->req_header.rqs_entry_type == RQSTYPE_T3RQS) {
-	    lim = ISP_CDSEG64;
-	    crq->req_header.rqs_entry_type = RQSTYPE_A64_CONT;
-	} else {
-	    lim = ISP_CDSEG;
-	    crq->req_header.rqs_entry_type = RQSTYPE_DATASEG;
-	}
-
-	for (ovseg = 0; seg < segcnt && ovseg < lim;
-	     rq->req_seg_count++, seg++, ovseg++, sg++) {
-	    if (sg->length == 0) {
-		panic("zero length s-g element at line %d", __LINE__);
-	    }
-	    crq->req_dataseg[ovseg].ds_count = QLA_SG_C(sg);
-	    crq->req_dataseg[ovseg].ds_base = QLA_SG_A(sg);
-	    isp_prt(isp, ISP_LOGDEBUG1, "seg%d[%d]%llx:%u from %p",
-		rq->req_header.rqs_entry_count-1, ovseg,
-	    	(unsigned long long) QLA_SG_A(sg), QLA_SG_C(sg), sg->address);
-	}
-	MEMORYBARRIER(isp, SYNC_REQUEST, curip, QENTRY_LEN);
-	isp_put_cont_req(isp, crq, qep);
-    } while (seg < segcnt);
-mbxsync:
-    if (rq->req_header.rqs_entry_type == RQSTYPE_T3RQS) {
-	isp_put_request_t3(isp, (ispreqt3_t *) rq, (ispreqt3_t *) h);
-    } else if (rq->req_header.rqs_entry_type == RQSTYPE_T2RQS) {
-	isp_put_request_t2(isp, (ispreqt2_t *) rq, (ispreqt2_t *) h);
-    } else {
-	isp_put_request(isp, (ispreq_t *) rq, (ispreq_t *) h);
-    }
-    *nxi = nxti;
-    return (CMD_QUEUED);
-}
-#endif
-
 static void
 isp_pci_reset1(struct ispsoftc *isp)
 {

Re: 2.5.70-mm6

[Joe]

All in all -mm6 seems fine here but for two small
problems, one of which is the continuing issue with
gdm which first surfaced in -mm5 IIRC -

Jun  7 18:32:01 jyro kernel: ip_tables: (C) 2000-2002 Netfilter core team
Jun  7 18:32:01 jyro kernel: ip_conntrack version 2.1 (4086 buckets, 
32688 max)
- 324 bytes per conntrack
Jun  7 18:32:06 jyro gdm[1282]: gdm_slave_xioerror_handler: Fatal X 
error - Restarting :0
Jun  7 18:32:07 jyro gdm[1293]: gdm_slave_xioerror_handler: Fatal X 
error - Restarting :0
Jun  7 18:32:12 jyro gdm[1305]: gdm_slave_xioerror_handler: Fatal X 
error - Restarting :0
Jun  7 18:32:16 jyro gdm[1311]: gdm_slave_xioerror_handler: Fatal X 
error - Restarting :0
Jun  7 18:32:16 jyro gdm[1237]: deal_with_x_crashes: Running the 
XKeepsCrashing script
Jun  7 18:32:54 jyro gdm[1237]: Failed to start X server several times 
in a short time period; disabling display :0

Re: [patch] 2.5.70-mm6: isp driver cleanups

[Andrew Morton]

Adrian Bunk <bunk@fs.tum.de> wrote:
>
> isp_linux.h already states kernels < 2.4 are not supported. The patch 
>  below removes #ifdef'd code for kernels < 2.4.

I shall forward this along to Matthew, but there is not a lot of point in
working against the -mm version of this driver: it is just there for people
to test.

Thanks.

Re: 2.5.70-mm6

[Martin J. Bligh]

--On Monday, June 09, 2003 19:45:58 +0200 Maciej Soltysiak <solt@dns.toxicfilms.tv> wrote:

>> . -mm kernels will be running at HZ=100 for a while.  This is because
>>   the anticipatory scheduler's behaviour may be altered by the lower
>>   resolution.  Some architectures continue to use 100Hz and we need the
>>   testing coverage which x86 provides.
>
> The interactivity seems to have dropped. Again, with common desktop
> applications: xmms playing with ALSA, when choosing navigating through
> evolution options or browsing with opera, music skipps.
> X is running with nice -10, but with mm5 it ran smoothly.

If you don't nice the hell out of X, does it work OK?

M.

Re: 2.5.70-mm6

[Maciej Soltysiak]

> . -mm kernels will be running at HZ=100 for a while.  This is because
>   the anticipatory scheduler's behaviour may be altered by the lower
>   resolution.  Some architectures continue to use 100Hz and we need the
>   testing coverage which x86 provides.
The interactivity seems to have dropped. Again, with common desktop
applications: xmms playing with ALSA, when choosing navigating through
evolution options or browsing with opera, music skipps.
X is running with nice -10, but with mm5 it ran smoothly.

Regards,
Maciej

Re: 2.5.70-mm6

[Alistair J Strachan]

On Monday 09 June 2003 18:45, Maciej Soltysiak wrote:
> > . -mm kernels will be running at HZ=100 for a while.  This is because
> >   the anticipatory scheduler's behaviour may be altered by the lower
> >   resolution.  Some architectures continue to use 100Hz and we need the
> >   testing coverage which x86 provides.
>
> The interactivity seems to have dropped. Again, with common desktop
> applications: xmms playing with ALSA, when choosing navigating through
> evolution options or browsing with opera, music skipps.
> X is running with nice -10, but with mm5 it ran smoothly.

[alistair] 07:02 PM [~] uname -r
2.5.70-mm6

For what it's worth, I'm running an LFS base system with very few packages 
installed over the top. X is as packaged, it is not reniced. I am, however, 
running setiathome constantly in the background, which seems to pound the 
scheduler.

As Maciej reported, this seems to be significantly better with -mm5 (HZ = 
1000?). Amusingly, doing a renice -20 `pidof xmms` seems to make absolutely 
no difference to the scheduler in 2.5-mm.

This kernel does not have preempt enabled.

Cheers,
Alistair.

Re: 2.5.70-mm6

[Maciej Soltysiak]

> If you don't nice the hell out of X, does it work OK?
No.

The way I reproduce the sound skips:
run xmms, run evolution, compose a mail with gpg.
on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
on mm5 xmms plays without stops. (with X -10)

Regards,
Maciej

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Mon, 2003-06-09 at 19:45, Maciej Soltysiak wrote:
> > . -mm kernels will be running at HZ=100 for a while.  This is because
> >   the anticipatory scheduler's behaviour may be altered by the lower
> >   resolution.  Some architectures continue to use 100Hz and we need the
> >   testing coverage which x86 provides.
> The interactivity seems to have dropped. Again, with common desktop
> applications: xmms playing with ALSA, when choosing navigating through
> evolution options or browsing with opera, music skipps.
> X is running with nice -10, but with mm5 it ran smoothly.

Sadly, I must agree with you... Sound with XMMS and Mplayer is chunky
when switching between virtual desktops, or even dragging windows. Is
this caused by latest scheduler patches, or has something to with
HZ=100?

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Mon, 2003-06-09 at 19:39, Martin J. Bligh wrote:
> --On Monday, June 09, 2003 19:45:58 +0200 Maciej Soltysiak <solt@dns.toxicfilms.tv> wrote:
> 
> >> . -mm kernels will be running at HZ=100 for a while.  This is because
> >>   the anticipatory scheduler's behaviour may be altered by the lower
> >>   resolution.  Some architectures continue to use 100Hz and we need the
> >>   testing coverage which x86 provides.
> >
> > The interactivity seems to have dropped. Again, with common desktop
> > applications: xmms playing with ALSA, when choosing navigating through
> > evolution options or browsing with opera, music skipps.
> > X is running with nice -10, but with mm5 it ran smoothly.
> 
> If you don't nice the hell out of X, does it work OK?

I can't appreciate much different. I've assigned shortcuts to switch
between desktops easily. Switching between desktops very fast causes
XMMS to skip sound. This also happens when dragging windows.

Re: 2.5.70-mm6

[Martin J. Bligh]

>> If you don't nice the hell out of X, does it work OK?
> No.
> 
> The way I reproduce the sound skips:
> run xmms, run evolution, compose a mail with gpg.
> on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
> on mm5 xmms plays without stops. (with X -10)

Does this (from Ingo?) do anything useful to it?

diff -urpN -X /home/fletch/.diff.exclude 400-reiserfs_dio/kernel/sched.c 420-sched_interactive/kernel/sched.c
--- 400-reiserfs_dio/kernel/sched.c	Fri May 30 19:26:34 2003
+++ 420-sched_interactive/kernel/sched.c	Fri May 30 19:28:06 2003
@@ -89,6 +89,8 @@ int node_threshold = 125;
 #define STARVATION_LIMIT	(starvation_limit)
 #define NODE_THRESHOLD		(node_threshold)
 
+#define TIMESLICE_GRANULARITY (HZ/20 ?: 1)
+
 /*
  * If a task is 'interactive' then we reinsert it in the active
  * array after it has expired its current timeslice. (it will not
@@ -1365,6 +1367,27 @@ void scheduler_tick(int user_ticks, int 
 			enqueue_task(p, rq->expired);
 		} else
 			enqueue_task(p, rq->active);
+	} else {
+		/*
+		 * Prevent a too long timeslice allowing a task to monopolize
+		 * the CPU. We do this by splitting up the timeslice into
+		 * smaller pieces.
+		 *
+		 * Note: this does not mean the task's timeslices expire or
+		 * get lost in any way, they just might be preempted by
+		 * another task of equal priority. (one with higher
+		 * priority would have preempted this task already.) We
+		 * requeue this task to the end of the list on this priority
+		 * level, which is in essence a round-robin of tasks with
+		 * equal priority.
+		 */
+		if (!(p->time_slice % TIMESLICE_GRANULARITY) &&
+			       		(p->array == rq->active)) {
+			dequeue_task(p, rq->active);
+			set_tsk_need_resched(p);
+			p->prio = effective_prio(p);
+			enqueue_task(p, rq->active);
+		}
 	}
 out_unlock:
 	spin_unlock(&rq->lock);

Re: 2.5.70-mm6

[Pasi Savolainen]

* Maciej Soltysiak <solt@dns.toxicfilms.tv>:
>> . -mm kernels will be running at HZ=100 for a while.  This is because
>>   the anticipatory scheduler's behaviour may be altered by the lower
>>   resolution.  Some architectures continue to use 100Hz and we need the
>>   testing coverage which x86 provides.
> The interactivity seems to have dropped. Again, with common desktop
> applications: xmms playing with ALSA, when choosing navigating through
> evolution options or browsing with opera, music skipps.
> X is running with nice -10, but with mm5 it ran smoothly.

I see that idle() is called much less often than before (1000
calls/second down to 150 calls/second, estimated and non-scientifical).

non-linear scale down is most probably because processes get more done
and don't wait so much.

idle() is also get called more when there is some load.

There is something weird though, I have this constant 0.8 load which I
can't pinpoint, in -mm4 fully idle machine was at about 0.1 load.

Regarding my stupidly reported Xfree86 -problem, it was PEBKAC, though I
can't tell what exactly that was. Only one module changed way to iterate
pci_find_device between boots.


-- 
   Psi -- <http://www.iki.fi/pasi.savolainen>

Re: 2.5.70-mm6

[Maciej Soltysiak]

> > on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
> > on mm5 xmms plays without stops. (with X -10)
>
> Does this (from Ingo?) do anything useful to it?
No, maybe a little bit, the skips are still there.

Regards,
Maciej

Re: 2.5.70-mm6

[William Lee Irwin III]

[-- Attachment #1: Type: text/plain, Size: 375 bytes --]

At some point in the past, mbligh wrote:
>> Does this (from Ingo?) do anything useful to it?

On Mon, Jun 09, 2003 at 09:42:36PM +0200, Maciej Soltysiak wrote:
> No, maybe a little bit, the skips are still there.

How about one or the other of these two? (not both at once, though,
they appear to clash).

I apologize in advance if MIME attachments are bad for you?


-- wli

[-- Attachment #2: galbraith.patch --]
[-- Type: text/plain, Size: 7376 bytes --]

diff -prauN linux-2.5.70-bk8/include/linux/sched.h galbraith-2.5.70-bk8-1/include/linux/sched.h
--- linux-2.5.70-bk8/include/linux/sched.h	Mon Jun  2 02:15:11 2003
+++ galbraith-2.5.70-bk8-1/include/linux/sched.h	Tue Jun  3 07:10:17 2003
@@ -336,7 +336,9 @@
 	prio_array_t *array;
 
 	unsigned long sleep_avg;
-	unsigned long last_run;
+	unsigned long long last_run;
+	unsigned int run_nsecs;
+	unsigned int sleep_nsecs;
 
 	unsigned long policy;
 	unsigned long cpus_allowed;
diff -prauN linux-2.5.70-bk8/kernel/sched.c galbraith-2.5.70-bk8-1/kernel/sched.c
--- linux-2.5.70-bk8/kernel/sched.c	Tue May 27 04:05:34 2003
+++ galbraith-2.5.70-bk8-1/kernel/sched.c	Tue Jun  3 07:10:18 2003
@@ -74,6 +74,12 @@
 #define MAX_SLEEP_AVG		(10*HZ)
 #define STARVATION_LIMIT	(10*HZ)
 #define NODE_THRESHOLD		125
+#define SCHED_NANOSECOND	1
+#define SCHED_SECOND		(1000000000 * SCHED_NANOSECOND)
+#define SCHED_TICK		(SCHED_SECOND / HZ)
+#define TICKS_PER_SECOND	(SCHED_SECOND / SCHED_TICK)
+
+extern unsigned long long monotonic_clock(void);
 
 /*
  * If a task is 'interactive' then we reinsert it in the active
@@ -342,9 +348,23 @@
  */
 static inline void activate_task(task_t *p, runqueue_t *rq)
 {
-	long sleep_time = jiffies - p->last_run - 1;
+	unsigned long long now = monotonic_clock();
+	long long sleep =  now - p->last_run + p->sleep_nsecs;
+	int ticks = 0, requeue_waker = 0;
+
+	if (sleep >= SCHED_TICK) {
+		while (sleep >= SCHED_SECOND) {
+			sleep -= SCHED_SECOND;
+			ticks += TICKS_PER_SECOND;
+		}
+		while (sleep >= SCHED_TICK) {
+			sleep -= SCHED_TICK;
+			ticks++;
+		}
+		p->sleep_nsecs = sleep;
+	} else p->sleep_nsecs += sleep;
 
-	if (sleep_time > 0) {
+	if (ticks > 0) {
 		int sleep_avg;
 
 		/*
@@ -355,7 +375,7 @@
 		 * spends sleeping, the higher the average gets - and the
 		 * higher the priority boost gets as well.
 		 */
-		sleep_avg = p->sleep_avg + sleep_time;
+		sleep_avg = p->sleep_avg + ticks;
 
 		/*
 		 * 'Overflow' bonus ticks go to the waker as well, so the
@@ -363,8 +383,10 @@
 		 * boosting tasks that are related to maximum-interactive
 		 * tasks.
 		 */
-		if (sleep_avg > MAX_SLEEP_AVG)
+		if (sleep_avg > MAX_SLEEP_AVG) {
 			sleep_avg = MAX_SLEEP_AVG;
+			p->sleep_nsecs = 0;
+		}
 		if (p->sleep_avg != sleep_avg) {
 			p->sleep_avg = sleep_avg;
 			p->prio = effective_prio(p);
@@ -548,6 +570,8 @@
 	current->sleep_avg = current->sleep_avg * PARENT_PENALTY / 100;
 	p->sleep_avg = p->sleep_avg * CHILD_PENALTY / 100;
 	p->prio = effective_prio(p);
+	p->run_nsecs = 0;
+	p->sleep_nsecs = 0;
 	set_task_cpu(p, smp_processor_id());
 
 	if (unlikely(!current->array))
@@ -1147,6 +1171,49 @@
 		(jiffies - (rq)->expired_timestamp >= \
 			STARVATION_LIMIT * ((rq)->nr_running) + 1)))
 
+inline void __scheduler_tick(runqueue_t *rq, task_t *p)
+{
+	unsigned long long now = monotonic_clock();
+	prio_array_t *array = rq->active;
+	int ticks;
+
+	p->run_nsecs += now - p->last_run;
+	/* Task might have expired already, but not scheduled off yet */
+	if (p->array != array) {
+		set_tsk_need_resched(p);
+		goto abort;
+	}
+	if (p->run_nsecs < SCHED_TICK || p->policy == SCHED_FIFO )
+		goto abort;
+
+	for (ticks = 0; p->run_nsecs >= SCHED_TICK; ticks++)
+		p->run_nsecs -= SCHED_TICK;
+	if (ticks > p->time_slice)
+		show_task(p);
+	if (p->sleep_avg > ticks)
+		p->sleep_avg -= ticks;
+	else
+		p->sleep_avg = 0;
+	p->time_slice -= ticks;
+
+	if (p->time_slice <= 0) {
+		dequeue_task(p, p->array);
+		p->prio = effective_prio(p);
+		p->time_slice = task_timeslice(p);
+		p->first_time_slice = 0;
+		set_tsk_need_resched(p);
+		if ((EXPIRED_STARVING(rq) && !rt_task(p)) ||
+				!TASK_INTERACTIVE(p)) {
+			array = rq->expired;
+			if (!rq->expired_timestamp)
+				rq->expired_timestamp = jiffies;
+		}
+		enqueue_task(p, array);
+	}
+abort:
+	p->last_run = monotonic_clock();
+}
+
 /*
  * This function gets called by the timer code, with HZ frequency.
  * We call it with interrupts disabled.
@@ -1159,11 +1226,12 @@
 	int cpu = smp_processor_id();
 	runqueue_t *rq = this_rq();
 	task_t *p = current;
+	int idle = p == rq->idle;
 
 	if (rcu_pending(cpu))
 		rcu_check_callbacks(cpu, user_ticks);
 
-	if (p == rq->idle) {
+	if (idle) {
 		/* note: this timer irq context must be accounted for as well */
 		if (irq_count() - HARDIRQ_OFFSET >= SOFTIRQ_OFFSET)
 			kstat_cpu(cpu).cpustat.system += sys_ticks;
@@ -1171,8 +1239,7 @@
 			kstat_cpu(cpu).cpustat.iowait += sys_ticks;
 		else
 			kstat_cpu(cpu).cpustat.idle += sys_ticks;
-		rebalance_tick(rq, 1);
-		return;
+		goto out;
 	}
 	if (TASK_NICE(p) > 0)
 		kstat_cpu(cpu).cpustat.nice += user_ticks;
@@ -1180,11 +1247,6 @@
 		kstat_cpu(cpu).cpustat.user += user_ticks;
 	kstat_cpu(cpu).cpustat.system += sys_ticks;
 
-	/* Task might have expired already, but not scheduled off yet */
-	if (p->array != rq->active) {
-		set_tsk_need_resched(p);
-		goto out;
-	}
 	spin_lock(&rq->lock);
 	/*
 	 * The task was running during this tick - update the
@@ -1194,42 +1256,10 @@
 	 * it possible for interactive tasks to use up their
 	 * timeslices at their highest priority levels.
 	 */
-	if (p->sleep_avg)
-		p->sleep_avg--;
-	if (unlikely(rt_task(p))) {
-		/*
-		 * RR tasks need a special form of timeslice management.
-		 * FIFO tasks have no timeslices.
-		 */
-		if ((p->policy == SCHED_RR) && !--p->time_slice) {
-			p->time_slice = task_timeslice(p);
-			p->first_time_slice = 0;
-			set_tsk_need_resched(p);
-
-			/* put it at the end of the queue: */
-			dequeue_task(p, rq->active);
-			enqueue_task(p, rq->active);
-		}
-		goto out_unlock;
-	}
-	if (!--p->time_slice) {
-		dequeue_task(p, rq->active);
-		set_tsk_need_resched(p);
-		p->prio = effective_prio(p);
-		p->time_slice = task_timeslice(p);
-		p->first_time_slice = 0;
-
-		if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) {
-			if (!rq->expired_timestamp)
-				rq->expired_timestamp = jiffies;
-			enqueue_task(p, rq->expired);
-		} else
-			enqueue_task(p, rq->active);
-	}
-out_unlock:
+	 __scheduler_tick(rq, p);
 	spin_unlock(&rq->lock);
 out:
-	rebalance_tick(rq, 0);
+	rebalance_tick(rq, idle);
 }
 
 void scheduling_functions_start_here(void) { }
@@ -1264,8 +1294,8 @@
 	rq = this_rq();
 
 	release_kernel_lock(prev);
-	prev->last_run = jiffies;
 	spin_lock_irq(&rq->lock);
+	__scheduler_tick(rq, prev);
 
 	/*
 	 * if entering off of a kernel preemption go straight
@@ -1320,6 +1350,7 @@
 	if (likely(prev != next)) {
 		rq->nr_switches++;
 		rq->curr = next;
+		next->last_run = prev->last_run;
 
 		prepare_arch_switch(rq, next);
 		prev = context_switch(rq, prev, next);
diff -prauN linux-2.5.70-bk8/arch/i386/kernel/timers/timer_tsc.c galbraith-2.5.70-bk8-1/arch/i386/kernel/timers/timer_tsc.c
--- linux-2.5.70-bk8/arch/i386/kernel/timers/timer_tsc.c	Mon Apr 21 08:11:07 2003
+++ galbraith-2.5.70-bk8-1/arch/i386/kernel/timers/timer_tsc.c	Tue Jun  3 07:10:18 2003
@@ -102,12 +102,13 @@
 static unsigned long long monotonic_clock_tsc(void)
 {
 	unsigned long long last_offset, this_offset, base;
+	unsigned long flags;
 	
 	/* atomically read monotonic base & last_offset */
-	read_lock_irq(&monotonic_lock);
+	read_lock_irqsave(&monotonic_lock, flags);
 	last_offset = ((unsigned long long)last_tsc_high<<32)|last_tsc_low;
 	base = monotonic_base;
-	read_unlock_irq(&monotonic_lock);
+	read_unlock_irqrestore(&monotonic_lock, flags);
 
 	/* Read the Time Stamp Counter */
 	rdtscll(this_offset);

[-- Attachment #3: galbraith_thud.diff --]
[-- Type: text/plain, Size: 1162 bytes --]

--- linux-2.5.70.virgin/kernel/sched.c.org	Tue Jun  3 06:44:48 2003
+++ linux-2.5.70.virgin/kernel/sched.c	Tue Jun  3 17:28:10 2003
@@ -66,7 +66,7 @@
  */
 #define MIN_TIMESLICE		( 10 * HZ / 1000)
 #define MAX_TIMESLICE		(200 * HZ / 1000)
-#define CHILD_PENALTY		50
+#define CHILD_PENALTY		80
 #define PARENT_PENALTY		100
 #define EXIT_WEIGHT		3
 #define PRIO_BONUS_RATIO	25
@@ -355,6 +355,7 @@
 		 * spends sleeping, the higher the average gets - and the
 		 * higher the priority boost gets as well.
 		 */
+		sleep_time = min(sleep_time, (long) p->time_slice);
 		sleep_avg = p->sleep_avg + sleep_time;
 
 		/*
@@ -545,8 +546,10 @@
 	 * and children as well, to keep max-interactive tasks
 	 * from forking tasks that are max-interactive.
 	 */
-	current->sleep_avg = current->sleep_avg * PARENT_PENALTY / 100;
-	p->sleep_avg = p->sleep_avg * CHILD_PENALTY / 100;
+	if (likely(current->parent->pid > 1)) {
+		current->sleep_avg = current->sleep_avg * PARENT_PENALTY / 100;
+		p->sleep_avg = p->sleep_avg * CHILD_PENALTY / 100;
+	} else current->sleep_avg = p->sleep_avg = MAX_SLEEP_AVG;
 	p->prio = effective_prio(p);
 	set_task_cpu(p, smp_processor_id());
 

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Mon, 2003-06-09 at 20:51, Martin J. Bligh wrote:
> >> If you don't nice the hell out of X, does it work OK?
> > No.
> > 
> > The way I reproduce the sound skips:
> > run xmms, run evolution, compose a mail with gpg.
> > on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
> > on mm5 xmms plays without stops. (with X -10)
> 
> Does this (from Ingo?) do anything useful to it?

I can confirm that 2.5.70-mm6 with Ingo's patch and HZ set back to 1000
is nearly perfect (it still takes some seconds for the scheduler to
adjust dynamic priorities).

Re: 2.5.70-mm6

[Martin J. Bligh]

--On Monday, June 09, 2003 22:08:42 +0200 Felipe Alfaro Solana <felipe_alfaro@linuxmail.org> wrote:

> On Mon, 2003-06-09 at 20:51, Martin J. Bligh wrote:
>> >> If you don't nice the hell out of X, does it work OK?
>> > No.
>> > 
>> > The way I reproduce the sound skips:
>> > run xmms, run evolution, compose a mail with gpg.
>> > on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
>> > on mm5 xmms plays without stops. (with X -10)
>> 
>> Does this (from Ingo?) do anything useful to it?
> 
> I can confirm that 2.5.70-mm6 with Ingo's patch and HZ set back to 1000
> is nearly perfect (it still takes some seconds for the scheduler to
> adjust dynamic priorities).

OK ... sorry to be pedantic, but I want to nail this down.
It's still broken with HZ=1000, but without Ingo's patch, right?

M.

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Mon, 2003-06-09 at 22:14, Martin J. Bligh wrote:
> --On Monday, June 09, 2003 22:08:42 +0200 Felipe Alfaro Solana <felipe_alfaro@linuxmail.org> wrote:
> 
> > On Mon, 2003-06-09 at 20:51, Martin J. Bligh wrote:
> >> >> If you don't nice the hell out of X, does it work OK?
> >> > No.
> >> > 
> >> > The way I reproduce the sound skips:
> >> > run xmms, run evolution, compose a mail with gpg.
> >> > on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
> >> > on mm5 xmms plays without stops. (with X -10)
> >> 
> >> Does this (from Ingo?) do anything useful to it?
> > 
> > I can confirm that 2.5.70-mm6 with Ingo's patch and HZ set back to 1000
> > is nearly perfect (it still takes some seconds for the scheduler to
> > adjust dynamic priorities).
> 
> OK ... sorry to be pedantic, but I want to nail this down.
> It's still broken with HZ=1000, but without Ingo's patch, right?

I have to try that combination... Please, allow for a few hours and I'll
post the results.

Re: 2.5.70-mm6

[Diego Calleja García]

On Mon, 9 Jun 2003 19:45:58 +0200 (CEST)
Maciej Soltysiak <solt@dns.toxicfilms.tv> wrote:

> The interactivity seems to have dropped. Again, with common desktop
> applications: xmms playing with ALSA, when choosing navigating through
> evolution options or browsing with opera, music skipps.
> X is running with nice -10, but with mm5 it ran smoothly.

Under "heavy" disk usage (when sylpheed finish merging the lkml
messages in the 92M lkml mail folder) X pointer stops moving 
(say, 1/8 or 1/6 seconds, very noticeable, pointer stops, windows stop
redrawing, etc).

System is a dual p3 800; fs is ext3. This odd behaviour
seems to happen since the 2.5.69-mm9 ext3 locking changes.
(well i started testing 2.5.70-mm3 because i'm timid,
but never happened before in mm or mainline)

Sorry that i can't provide really useful information now ;(

Re: 2.5.70-mm6

[Andrew Morton]

Diego Calleja García <diegocg@teleline.es> wrote:
>
> On Mon, 9 Jun 2003 19:45:58 +0200 (CEST)
> Maciej Soltysiak <solt@dns.toxicfilms.tv> wrote:
> 
> > The interactivity seems to have dropped. Again, with common desktop
> > applications: xmms playing with ALSA, when choosing navigating through
> > evolution options or browsing with opera, music skipps.
> > X is running with nice -10, but with mm5 it ran smoothly.
> 
> Under "heavy" disk usage (when sylpheed finish merging the lkml
> messages in the 92M lkml mail folder) X pointer stops moving 
> (say, 1/8 or 1/6 seconds, very noticeable, pointer stops, windows stop
> redrawing, etc).

I've noticed similar.  Just a new vague jerkiness.

> System is a dual p3 800; fs is ext3. This odd behaviour
> seems to happen since the 2.5.69-mm9 ext3 locking changes.
> (well i started testing 2.5.70-mm3 because i'm timid,
> but never happened before in mm or mainline)

Might be.  There were some CPU scheduler changes a week before 2.5.69-mm9. 
If it's in ext3 then profiling will probably shake it out.  I haven't done
a lot of profiling yet.

Re: 2.5.70-mm6

[Maciej Soltysiak]

> How about one or the other of these two? (not both at once, though,
> they appear to clash).
Success, no audio skipps with galbraith.patch and mm6.

> -- wli
Maciej

Re: 2.5.70-mm6

[William Lee Irwin III]

At some point in the past, I wrote:
>> How about one or the other of these two? (not both at once, though,
>> they appear to clash).

On Tue, Jun 10, 2003 at 10:54:55AM +0200, Maciej Soltysiak wrote:
> Success, no audio skipps with galbraith.patch and mm6.

Mike, any chance you can turn your series of patches into one that
applies atop mingo's intra-timeslice priority preemption patch? If
not, I suppose someone else could.

There also appears to be some kind of issue with using monotonic_clock()
with timer_pit as well as some locking overhead concerns. Something
should probably be done about those things before trying to merge the
fine-grained time accounting patch.


-- wli

Re: 2.5.70-mm6

[Felipe Alfaro Solana]

On Mon, 2003-06-09 at 22:14, Martin J. Bligh wrote:
> --On Monday, June 09, 2003 22:08:42 +0200 Felipe Alfaro Solana <felipe_alfaro@linuxmail.org> wrote:
> 
> > On Mon, 2003-06-09 at 20:51, Martin J. Bligh wrote:
> >> >> If you don't nice the hell out of X, does it work OK?
> >> > No.
> >> > 
> >> > The way I reproduce the sound skips:
> >> > run xmms, run evolution, compose a mail with gpg.
> >> > on mm6 the gpg part stops the sound for a few seconds. (with X -10 and 0)
> >> > on mm5 xmms plays without stops. (with X -10)
> >> 
> >> Does this (from Ingo?) do anything useful to it?
> > 
> > I can confirm that 2.5.70-mm6 with Ingo's patch and HZ set back to 1000
> > is nearly perfect (it still takes some seconds for the scheduler to
> > adjust dynamic priorities).
> 
> OK ... sorry to be pedantic, but I want to nail this down.
> It's still broken with HZ=1000, but without Ingo's patch, right?

Well, Ingo's patch makes XMMS more resistant to audio skip when HZ=1000.
Anyways, with HZ=1000 interactivity is much better than with HZ=100
(with or without Ingo's patch).

Re: 2.5.70-mm6

[Mike Galbraith]

At 02:20 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>At some point in the past, I wrote:
> >> How about one or the other of these two? (not both at once, though,
> >> they appear to clash).
>
>On Tue, Jun 10, 2003 at 10:54:55AM +0200, Maciej Soltysiak wrote:
> > Success, no audio skipps with galbraith.patch and mm6.

(victim of fast hw methinks.  dog slow old isa card will probably work fine)

>Mike, any chance you can turn your series of patches into one that
>applies atop mingo's intra-timeslice priority preemption patch? If
>not, I suppose someone else could.

I've never seen it.  Is this the test-starve fix I heard mentioned on lkml 
once?

>There also appears to be some kind of issue with using monotonic_clock()
>with timer_pit as well as some locking overhead concerns. Something
>should probably be done about those things before trying to merge the
>fine-grained time accounting patch.

Ingo had me measure impact with lat_ctx, and it wasn't very encouraging 
(and my box is UP).  I'm not sure that I wasn't seeing some cache effects 
though, because the numbers jumped around quite a bit.  Per Ingo, the 
sequence lock change will greatly improve scalability.  Doing anything 
extra in that path is going to cost some pain though, so I'm trying to 
figure out a way to do something ~similar.  (ala perfect is the enemy of 
good mantra).

wrt pit, yeah, that diff won't work if you don't have a tsc.  If something 
like it were used, it'd have to have ifdefs to continue using 
jiffies.  (the other option being only presentable on April 1:)

         -Mike 

Re: 2.5.70-mm6

[William Lee Irwin III]

At 02:20 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> Mike, any chance you can turn your series of patches into one that
>> applies atop mingo's intra-timeslice priority preemption patch? If
>> not, I suppose someone else could.

On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> I've never seen it.  Is this the test-starve fix I heard mentioned on lkml 
> once?

No idea what the posted name was. What it does is obvious enough. It
was posted earlier in this thread.


At 02:20 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> There also appears to be some kind of issue with using monotonic_clock()
>> with timer_pit as well as some locking overhead concerns. Something
>> should probably be done about those things before trying to merge the
>> fine-grained time accounting patch.

On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> Ingo had me measure impact with lat_ctx, and it wasn't very encouraging 
> (and my box is UP).  I'm not sure that I wasn't seeing some cache effects 
> though, because the numbers jumped around quite a bit.  Per Ingo, the 
> sequence lock change will greatly improve scalability.  Doing anything 
> extra in that path is going to cost some pain though, so I'm trying to 

Okay, so mitigating the hit to context switch is ongoing.


On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> figure out a way to do something ~similar.  (ala perfect is the enemy of 
> good mantra).

\vomit{Next you'll be telling me worse is better.}


On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> wrt pit, yeah, that diff won't work if you don't have a tsc.  If something 
> like it were used, it'd have to have ifdefs to continue using 
> jiffies.  (the other option being only presentable on April 1:)

The issue is the driver returning garbage; not having as good of
precision from hardware is no fault of the method. I'd say timer_pit
should just return jiffies converted to nanoseconds.

Also, I posted the "thud" fix earlier in this thread in addition to the
monotonic_clock() bits. AFAICT it mitigates (or perhaps even fixes) an
infinite priority escalation scenario.


-- wli

Re: 2.5.70-mm6

[Maciej Soltysiak]

> How about one or the other of these two? (not both at once, though,
> they appear to clash).
2.5.70-mm7 seems not to cause mentioned problems.

Regards,
Maciej

Re: 2.5.70-mm6

[Grzegorz Jaskiewicz]

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

It is not only ext3 problem, I've got this issue on reiserfs as well. It is 
just coused by schleduer.
Anyway, 2.5.x is now slower than 2.4.21-rc7 about 25% on disc access as well !


- --
Grzegorz Jaskiewicz
K4 Labs
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE+5cq9qu082fCQYIgRAj8FAJ9Rmo8q4VYLjrFt0zvtklw+MUFnPQCaA7VA
SzoDJgprNamOpz0qO+HXXKM=
=8FbC
-----END PGP SIGNATURE-----

Re: 2.5.70-mm6

[Mike Galbraith]

At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>At 02:20 AM 6/10/2003 -0700, William Lee Irwin III wrote:
> >> Mike, any chance you can turn your series of patches into one that
> >> applies atop mingo's intra-timeslice priority preemption patch? If
> >> not, I suppose someone else could.
>
>On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> > I've never seen it.  Is this the test-starve fix I heard mentioned on lkml
> > once?
>
>No idea what the posted name was. What it does is obvious enough. It
>was posted earlier in this thread.
>
>
>At 02:20 AM 6/10/2003 -0700, William Lee Irwin III wrote:
> >> There also appears to be some kind of issue with using monotonic_clock()
> >> with timer_pit as well as some locking overhead concerns. Something
> >> should probably be done about those things before trying to merge the
> >> fine-grained time accounting patch.
>
>On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> > Ingo had me measure impact with lat_ctx, and it wasn't very encouraging
> > (and my box is UP).  I'm not sure that I wasn't seeing some cache effects
> > though, because the numbers jumped around quite a bit.  Per Ingo, the
> > sequence lock change will greatly improve scalability.  Doing anything
> > extra in that path is going to cost some pain though, so I'm trying to
>
>Okay, so mitigating the hit to context switch is ongoing.

If it's used, seems some work will be required to measure the true (and 
practical) impact.

>On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> > figure out a way to do something ~similar.  (ala perfect is the enemy of
> > good mantra).
>
>\vomit{Next you'll be telling me worse is better.}

That's an unearned criticism.

Timeslice management is currently an approximation.  IFF the approximation 
is good enough, it will/must out perform pedantic bean-counting.  Current 
timeslice management apparently isn't quite good enough, so I'm trying to 
find a way to increase it's informational content without the (in general 
case useless) overhead of attempted perfection.  I'm failing miserably btw ;-)

>On Tue, Jun 10, 2003 at 01:31:32PM +0200, Mike Galbraith wrote:
> > wrt pit, yeah, that diff won't work if you don't have a tsc.  If something
> > like it were used, it'd have to have ifdefs to continue using
> > jiffies.  (the other option being only presentable on April 1:)
>
>The issue is the driver returning garbage; not having as good of
>precision from hardware is no fault of the method. I'd say timer_pit
>should just return jiffies converted to nanoseconds.

That's the option that I figured was April 1 material, because of the 
missing precision.  If it could be made (somehow that I don't understand) 
to produce a reasonable approximation of a high resolution clock, sure.

>Also, I posted the "thud" fix earlier in this thread in addition to the
>monotonic_clock() bits. AFAICT it mitigates (or perhaps even fixes) an
>infinite priority escalation scenario.

(yes, mitigates... maybe cures with round down, not really sure)

Couple that change with reintroduction of backboost to offset some of it's 
other effects and a serious reduction of CHILD_PENALTY and you'll have a 
very snappy desktop.  However, there is another side of the 
equation.  Large instantaneous variance in sleep_avg/prio also enables the 
scheduler to react quickly such that tasks which were delayed for whatever 
reason rapidly get a chance collect their ticks and catch up.  It can and 
does cause perceived unfairness... which is in reality quite fair.  It's 
horrible for short period concurrency, but great for long period 
throughput.  AFAIKT, it's a hard problem.

         -Mike 

Re: 2.5.70-mm6

[William Lee Irwin III]

At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> \vomit{Next you'll be telling me worse is better.}

On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> That's an unearned criticism.
> Timeslice management is currently an approximation.  IFF the approximation 
> is good enough, it will/must out perform pedantic bean-counting.  Current 
> timeslice management apparently isn't quite good enough, so I'm trying to 
> find a way to increase it's informational content without the (in general 
> case useless) overhead of attempted perfection.  I'm failing miserably btw 
> ;-)

The criticism was of the maxim invoked, not the specific direction
you're hacking in.


At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> The issue is the driver returning garbage; not having as good of
>> precision from hardware is no fault of the method. I'd say timer_pit
>> should just return jiffies converted to nanoseconds.

On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> That's the option that I figured was April 1 material, because of the 
> missing precision.  If it could be made (somehow that I don't understand) 
> to produce a reasonable approximation of a high resolution clock, sure.

If there's a TSC, it can be used for extra interpolation. But I think
timer_tsc already does that. I don't think it's quite so laughable; the
machines missing reliable time sources are truly crippled in some way,
by obsolescence or misdesign. I wouldn't call this a deficit. The
major platforms will do fine, and the rest will do no worse than now
apart from perhaps some function call and arithmetic overhead.


At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> Also, I posted the "thud" fix earlier in this thread in addition to the
>> monotonic_clock() bits. AFAICT it mitigates (or perhaps even fixes) an
>> infinite priority escalation scenario.

On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> (yes, mitigates... maybe cures with round down, not really sure)
> Couple that change with reintroduction of backboost to offset some of it's 
> other effects and a serious reduction of CHILD_PENALTY and you'll have a 
> very snappy desktop.  However, there is another side of the 
> equation.  Large instantaneous variance in sleep_avg/prio also enables the 
> scheduler to react quickly such that tasks which were delayed for whatever 
> reason rapidly get a chance collect their ticks and catch up.  It can and 
> does cause perceived unfairness... which is in reality quite fair.  It's 
> horrible for short period concurrency, but great for long period 
> throughput.  AFAIKT, it's a hard problem.

I don't know that there are answers better than mitigation.

I propose we err on the other side of the fence and back off as cases
where the larger instantaneous variances are more clearly needed arise.


-- wli

Re: 2.5.70-mm6

[Mingming Cao]

Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.70/2.5.70-mm6/
> 

I run 50 fsx tests on ext3 filesystem on 2.5.70-mm6 kernel. Serveral fsx 
tests hang with the status D, after the tests run for a while.  No oops, 
no error messages.  I found same problem on mm5, but 2.5.70 is fine.

Here is the stack info:

fsx           D C3496298   932      1           933   852 (NOTLB)
Call Trace:
  [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D C0109E2E   933      1           934   932 (NOTLB)
Call Trace:
  [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
  [<c02b04f3>] generic_unplug_device+0x23/0x70
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D F6A43DA8   934      1           935   933 (NOTLB)
Call Trace:
  [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D F7B55A00   935      1           936   934 (NOTLB)
Call Trace:
  [<c02aea96>] elv_next_request+0x16/0x100
  [<c02b052a>] generic_unplug_device+0x5a/0x70
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D F6A1E000   936      1           937   935 (NOTLB)
Call Trace:
  [<c011bcc0>] schedule+0x280/0x4f0
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D F6A1DDA8   937      1           938   936 (NOTLB)
Call Trace:
  [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   938      1           939   937 (NOTLB)
Call Trace:
  [<c011c546>] io_schedule+0x26/0x30
  [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0157587>] __wait_on_buffer+0x17/0x20
  [<c01a3018>] journal_invalidatepage+0x88/0x130
  [<c0193e93>] ext3_invalidatepage+0x43/0x50
  [<c01423e7>] do_invalidatepage+0x27/0x30
  [<c014247e>] truncate_complete_page+0x8e/0x90
  [<c0142604>] truncate_inode_pages+0xd4/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D C0109E2E   939      1           940   938 (NOTLB)
Call Trace:
  [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
  [<c011c546>] io_schedule+0x26/0x30
  [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   940      1           941   939 (NOTLB)
Call Trace:
  [<c011c546>] io_schedule+0x26/0x30
  [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0157587>] __wait_on_buffer+0x17/0x20
  [<c01a3018>] journal_invalidatepage+0x88/0x130
  [<c0193e93>] ext3_invalidatepage+0x43/0x50
  [<c01423e7>] do_invalidatepage+0x27/0x30
  [<c014247e>] truncate_complete_page+0x8e/0x90
  [<c0142604>] truncate_inode_pages+0xd4/0x2f0
  [<c014712b>] vmtruncate+0x6b/0x100
  [<c0171e94>] inode_setattr+0x134/0x150
  [<c0195aca>] ext3_setattr+0x7a/0x1a0
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c0172080>] notify_change+0x160/0x19d
  [<c0153f3a>] do_truncate+0x6a/0x90
  [<c0161037>] sys_fstat64+0x37/0x40
  [<c0154228>] sys_ftruncate+0x118/0x1b0
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0155c29>] sys_lseek+0x69/0xb0
  [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   941      1                 940 (NOTLB)
Call Trace:
  [<c011c546>] io_schedule+0x26/0x30
  [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c015af00>] submit_bh+0x90/0x1e0
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c02b1cc4>] submit_bio+0x54/0xa0
  [<c0158acd>] __bread_slow_wq+0xad/0xf0
  [<c0158db6>] __bread+0x46/0x50
  [<c018eb98>] read_block_bitmap+0x58/0xa0
  [<c018f902>] ext3_new_block+0x1a2/0x590
  [<c0158c73>] __find_get_block+0x73/0x100
  [<c01922f7>] ext3_alloc_block+0x37/0x40
  [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
  [<c01924c2>] ext3_get_branch+0x72/0x100
  [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
  [<c015b501>] alloc_buffer_head+0x41/0x50
  [<c0192ef4>] ext3_get_block+0x64/0xb0
  [<c0159827>] __block_prepare_write+0x227/0x4b0
  [<c015a364>] block_prepare_write+0x34/0x50
  [<c0192e90>] ext3_get_block+0x0/0xb0
  [<c01934d1>] ext3_prepare_write+0x71/0x130
  [<c0192e90>] ext3_get_block+0x0/0xb0
  [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
  [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
  [<c0158d97>] __bread+0x27/0x50
  [<c01951f3>] ext3_get_inode_loc+0x103/0x1a0
  [<c013adee>] generic_file_aio_write+0x9e/0x120
  [<c0190bf4>] ext3_file_write+0x44/0xe0
  [<c0156046>] do_sync_write+0xb6/0xf0
  [<c0177a56>] __mark_inode_dirty+0xf6/0x100
  [<c0160f6e>] cp_new_stat64+0xfe/0x110
  [<c011ddf0>] autoremove_wake_function+0x0/0x50
  [<c0127cb6>] update_wall_time+0x16/0x40
  [<c01280f0>] do_timer+0xc0/0xd0
  [<c015613e>] vfs_write+0xbe/0x130
  [<c01558f0>] generic_file_llseek+0x0/0xf0
  [<c0156262>] sys_write+0x42/0x70
  [<c010943f>] syscall_call+0x7/0xb

Re: 2.5.70-mm6

[Andrew Morton]

Mingming Cao <cmm@us.ibm.com> wrote:
>
> I run 50 fsx tests on ext3 filesystem on 2.5.70-mm6 kernel. Serveral fsx 
>  tests hang with the status D, after the tests run for a while.  No oops, 
>  no error messages.  I found same problem on mm5, but 2.5.70 is fine.
> 
>  Here is the stack info:

Thanks for this.

Everything is waiting on I/O.  It looks like either the device driver
failed or the IO scheduler got its state all screwed up.

Which device driver are you using there?

If you could, please retest with "elevator=deadline"?

Re: 2.5.70-mm6

[Mingming Cao]

On Tue, 2003-06-10 at 20:12, Andrew Morton wrote:
> Mingming Cao <cmm@us.ibm.com> wrote:
> >
> > I run 50 fsx tests on ext3 filesystem on 2.5.70-mm6 kernel. Serveral fsx 
> >  tests hang with the status D, after the tests run for a while.  No oops, 
> >  no error messages.  I found same problem on mm5, but 2.5.70 is fine.
> > 
> >  Here is the stack info:
> 
> Thanks for this.
> 
> Everything is waiting on I/O.  It looks like either the device driver
> failed or the IO scheduler got its state all screwed up.
> 
> Which device driver are you using there?

I am usering Qlogic qla2xxx V8 driver
> 
> If you could, please retest with "elevator=deadline"?
> 
Sure.

And I saw sysrq on 2.5.70-mm6 failed on my test machine, 2.5.70-mm5
works.  I could trigger sysrq by send"t", but the stack info for all
threads are the same ---the stack info of the sysrq itself.

Re: 2.5.70-mm6

[Mingming Cao]

> Mingming Cao <cmm@us.ibm.com> wrote:
> >
> > I run 50 fsx tests on ext3 filesystem on 2.5.70-mm6 kernel. Serveral fsx 
> >  tests hang with the status D, after the tests run for a while.  No oops, 
> >  no error messages.  I found same problem on mm5, but 2.5.70 is fine.
Sorry, the tests in 2.5.70 also failed, same problem.

On Tue, 2003-06-10 at 20:12, Andrew Morton wrote
> If you could, please retest with "elevator=deadline"?
> 
Thanks for your feedback.

This time I got more fsx tests hang(about 25).  Before normally I saw 5
or 10 tests fail. Here is the stack info.

kjournald     D 00000001   849      1           900   847 (L-TLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015b14c>] sync_dirty_buffer+0x6c/0xd0
 [<c0157587>] __wait_on_buffer+0x17/0x20
 [<c01a43b7>] journal_commit_transaction+0xbf7/0x122b
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011bc47>] schedule+0x207/0x4f0
 [<c01a7246>] kjournald+0x236/0x260
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0109352>] ret_from_fork+0x6/0x14
 [<c01a6ff0>] commit_timeout+0x0/0x10
 [<c01a7010>] kjournald+0x0/0x260
 [<c010722d>] kernel_thread_helper+0x5/0x18

fsx           D 00000001   900      1           901   849 (NOTLB)
Call Trace:
 [<c01a1862>] do_get_write_access+0x522/0x680
 [<c011bf30>] default_wake_function+0x0/0x30
 [<c01951f3>] ext3_get_inode_loc+0x103/0x1a0
 [<c01a19f9>] journal_get_write_access+0x39/0x60
 [<c0195d46>] ext3_reserve_inode_write+0x86/0xe0
 [<c0195dcb>] ext3_mark_inode_dirty+0x2b/0x60
 [<c0195e8c>] ext3_dirty_inode+0x8c/0x90
 [<c0177a56>] __mark_inode_dirty+0xf6/0x100
 [<c0171680>] inode_update_time+0x80/0xb0
 [<c013a455>] generic_file_aio_write_nolock+0x215/0xa50
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c0158d97>] __bread+0x27/0x50
 [<c01951f3>] ext3_get_inode_loc+0x103/0x1a0
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c0177a56>] __mark_inode_dirty+0xf6/0x100
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011b7ee>] scheduler_tick+0x16e/0x3b0
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C3812E48   901      1           902   900 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c015b390>] block_sync_page+0x0/0x10
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   902      1           903   901 (NOTLB)
Call Trace:
 [<c015b390>] block_sync_page+0x0/0x10
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   903      1           904   902 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015af00>] submit_bh+0x90/0x1e0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c02b1cc4>] submit_bio+0x54/0xa0
 [<c0158acd>] __bread_slow_wq+0xad/0xf0
 [<c0158db6>] __bread+0x46/0x50
 [<c018eb98>] read_block_bitmap+0x58/0xa0
 [<c018f902>] ext3_new_block+0x1a2/0x590
 [<c0158c73>] __find_get_block+0x73/0x100
 [<c01922f7>] ext3_alloc_block+0x37/0x40
 [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
 [<c01924c2>] ext3_get_branch+0x72/0x100
 [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
 [<c015b501>] alloc_buffer_head+0x41/0x50
 [<c0192ef4>] ext3_get_block+0x64/0xb0
 [<c0159827>] __block_prepare_write+0x227/0x4b0
 [<c015a364>] block_prepare_write+0x34/0x50
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c01934d1>] ext3_prepare_write+0x71/0x130
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c01715a2>] update_atime+0x92/0xf0
 [<c01393d4>] __generic_file_aio_read+0x1c4/0x210
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c0177a56>] __mark_inode_dirty+0xf6/0x100
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   904      1           925   903 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D EF963DD4   925      1           926   904 (NOTLB)
Call Trace:
 [<c02b0537>] generic_unplug_device+0x67/0x70
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C3810004   926      1           927   925 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c015b390>] block_sync_page+0x0/0x10
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D EF95FDA8   927      1           928   926 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   928      1           929   927 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015af00>] submit_bh+0x90/0x1e0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c02b1cc4>] submit_bio+0x54/0xa0
 [<c0158acd>] __bread_slow_wq+0xad/0xf0
 [<c0158db6>] __bread+0x46/0x50
 [<c018eb98>] read_block_bitmap+0x58/0xa0
 [<c018f902>] ext3_new_block+0x1a2/0x590
 [<c0158c73>] __find_get_block+0x73/0x100
 [<c01922f7>] ext3_alloc_block+0x37/0x40
 [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
 [<c01924c2>] ext3_get_branch+0x72/0x100
 [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
 [<c015b501>] alloc_buffer_head+0x41/0x50
 [<c0192ef4>] ext3_get_block+0x64/0xb0
 [<c0159827>] __block_prepare_write+0x227/0x4b0
 [<c015a364>] block_prepare_write+0x34/0x50
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c01934d1>] ext3_prepare_write+0x71/0x130
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c0158d97>] __bread+0x27/0x50
 [<c01951f3>] ext3_get_inode_loc+0x103/0x1a0
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c0177a56>] __mark_inode_dirty+0xf6/0x100
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   929      1           935   928 (NOTLB)
Call Trace:
 [<c01a1862>] do_get_write_access+0x522/0x680
 [<c011bf30>] default_wake_function+0x0/0x30
 [<c0158c73>] __find_get_block+0x73/0x100
 [<c01a19f9>] journal_get_write_access+0x39/0x60
 [<c0194879>] ext3_free_data+0x39/0x150
 [<c0194a7d>] ext3_free_branches+0xed/0x280
 [<c019508e>] ext3_truncate+0x47e/0x4e0
 [<c014703c>] invalidate_mmap_range+0x7c/0x100
 [<c0194c10>] ext3_truncate+0x0/0x4e0
 [<c014714c>] vmtruncate+0x8c/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   935      1           936   929 (NOTLB)
Call Trace:
 [<c02b0537>] generic_unplug_device+0x67/0x70
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D EF8FBDD4   936      1           937   935 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c02b007b>] ll_front_merge_fn+0x4b/0x120
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   937      1           938   936 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   938      1           939   937 (NOTLB)
Call Trace:
 [<c01a1862>] do_get_write_access+0x522/0x680
 [<c011bf30>] default_wake_function+0x0/0x30
 [<c01a19f9>] journal_get_write_access+0x39/0x60
 [<c0194879>] ext3_free_data+0x39/0x150
 [<c0194e9b>] ext3_truncate+0x28b/0x4e0
 [<c014703c>] invalidate_mmap_range+0x7c/0x100
 [<c0194c10>] ext3_truncate+0x0/0x4e0
 [<c014714c>] vmtruncate+0x8c/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   939      1           940   938 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015af00>] submit_bh+0x90/0x1e0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c02b1cc4>] submit_bio+0x54/0xa0
 [<c0158acd>] __bread_slow_wq+0xad/0xf0
 [<c0158db6>] __bread+0x46/0x50
 [<c018eb98>] read_block_bitmap+0x58/0xa0
 [<c018f902>] ext3_new_block+0x1a2/0x590
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c01922f7>] ext3_alloc_block+0x37/0x40
 [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
 [<c0158bc9>] bh_lru_install+0xb9/0xf0
 [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
 [<c015b501>] alloc_buffer_head+0x41/0x50
 [<c0192ef4>] ext3_get_block+0x64/0xb0
 [<c0159827>] __block_prepare_write+0x227/0x4b0
 [<c015a364>] block_prepare_write+0x34/0x50
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c01934d1>] ext3_prepare_write+0x71/0x130
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c011bf5a>] default_wake_function+0x2a/0x30
 [<c011de17>] autoremove_wake_function+0x27/0x50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c02f689e>] scsi_put_command+0x6e/0x90
 [<c02fb63e>] scsi_end_request+0xae/0xc0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c03332b3>] qla2x00_intr_handler+0x203/0x220
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   940      1           941   939 (NOTLB)
Call Trace:
 [<c02b0537>] generic_unplug_device+0x67/0x70
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C3815338   941      1           942   940 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c015b390>] block_sync_page+0x0/0x10
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   942      1           943   941 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   943      1           944   942 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014703c>] invalidate_mmap_range+0x7c/0x100
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   944      1           945   943 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015af00>] submit_bh+0x90/0x1e0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c02b1cc4>] submit_bio+0x54/0xa0
 [<c0158acd>] __bread_slow_wq+0xad/0xf0
 [<c0158db6>] __bread+0x46/0x50
 [<c018eb98>] read_block_bitmap+0x58/0xa0
 [<c018f902>] ext3_new_block+0x1a2/0x590
 [<c0158c73>] __find_get_block+0x73/0x100
 [<c01922f7>] ext3_alloc_block+0x37/0x40
 [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
 [<c01924c2>] ext3_get_branch+0x72/0x100
 [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
 [<c015b501>] alloc_buffer_head+0x41/0x50
 [<c0192ef4>] ext3_get_block+0x64/0xb0
 [<c0159827>] __block_prepare_write+0x227/0x4b0
 [<c015a364>] block_prepare_write+0x34/0x50
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c01934d1>] ext3_prepare_write+0x71/0x130
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c011de17>] autoremove_wake_function+0x27/0x50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c02fb63e>] scsi_end_request+0xae/0xc0
 [<c02fb95d>] scsi_io_completion+0x14d/0x470
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0342c31>] sd_rw_intr+0x61/0x270
 [<c02f6fdc>] scsi_finish_command+0x8c/0xc0
 [<c03332b3>] qla2x00_intr_handler+0x203/0x220
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C0109E2E   945      1           946   944 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C380CA70   946      1           947   945 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c015b390>] block_sync_page+0x0/0x10
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0138be6>] find_get_pages+0x36/0x60
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c014230e>] pagevec_lookup+0x2e/0x40
 [<c01426fc>] truncate_inode_pages+0x1cc/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D 00000001   947      1           948   946 (NOTLB)
Call Trace:
 [<c011c546>] io_schedule+0x26/0x30
 [<c0157565>] __wait_on_buffer_wq+0xf5/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c015af00>] submit_bh+0x90/0x1e0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c02b1cc4>] submit_bio+0x54/0xa0
 [<c0158acd>] __bread_slow_wq+0xad/0xf0
 [<c0158db6>] __bread+0x46/0x50
 [<c018eb98>] read_block_bitmap+0x58/0xa0
 [<c018f902>] ext3_new_block+0x1a2/0x590
 [<c0195ca9>] ext3_mark_iloc_dirty+0x29/0x40
 [<c01922f7>] ext3_alloc_block+0x37/0x40
 [<c01926ba>] ext3_alloc_branch+0x4a/0x2c0
 [<c0192cbc>] ext3_get_block_handle+0x18c/0x360
 [<c015b501>] alloc_buffer_head+0x41/0x50
 [<c0192ef4>] ext3_get_block+0x64/0xb0
 [<c0159827>] __block_prepare_write+0x227/0x4b0
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c015a364>] block_prepare_write+0x34/0x50
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c01934d1>] ext3_prepare_write+0x71/0x130
 [<c0192e90>] ext3_get_block+0x0/0xb0
 [<c013a5cd>] generic_file_aio_write_nolock+0x38d/0xa50
 [<c0195808>] ext3_do_update_inode+0x1c8/0x3d0
 [<c011bf5a>] default_wake_function+0x2a/0x30
 [<c011de17>] autoremove_wake_function+0x27/0x50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c013adee>] generic_file_aio_write+0x9e/0x120
 [<c0190bf4>] ext3_file_write+0x44/0xe0
 [<c0156046>] do_sync_write+0xb6/0xf0
 [<c02f689e>] scsi_put_command+0x6e/0x90
 [<c02fb63e>] scsi_end_request+0xae/0xc0
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011bf9a>] __wake_up_common+0x3a/0x60
 [<c03332b3>] qla2x00_intr_handler+0x203/0x220
 [<c015613e>] vfs_write+0xbe/0x130
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0156262>] sys_write+0x42/0x70
 [<c010943f>] syscall_call+0x7/0xb

fsx           D C0109E2E   948      1           949   947 (NOTLB)
Call Trace:
 [<c0109e2e>] apic_timer_interrupt+0x1a/0x20
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

fsx           D EF86BDD4   949      1           988   948 (NOTLB)
Call Trace:
 [<c02b0537>] generic_unplug_device+0x67/0x70
 [<c011c546>] io_schedule+0x26/0x30
 [<c01386ac>] wait_on_page_bit_wq+0xcc/0x100
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c011ddf0>] autoremove_wake_function+0x0/0x50
 [<c0142815>] truncate_inode_pages+0x2e5/0x2f0
 [<c014712b>] vmtruncate+0x6b/0x100
 [<c0171e94>] inode_setattr+0x134/0x150
 [<c0195aca>] ext3_setattr+0x7a/0x1a0
 [<c0160f6e>] cp_new_stat64+0xfe/0x110
 [<c0172080>] notify_change+0x160/0x19d
 [<c0153f3a>] do_truncate+0x6a/0x90
 [<c0161037>] sys_fstat64+0x37/0x40
 [<c0154228>] sys_ftruncate+0x118/0x1b0
 [<c01558f0>] generic_file_llseek+0x0/0xf0
 [<c0155c29>] sys_lseek+0x69/0xb0
 [<c010943f>] syscall_call+0x7/0xb

Re: 2.5.70-mm6

[Andrew Morton]

Mingming Cao <cmm@us.ibm.com> wrote:
>
> 
> > Mingming Cao <cmm@us.ibm.com> wrote:
> > >
> > > I run 50 fsx tests on ext3 filesystem on 2.5.70-mm6 kernel. Serveral fsx 
> > >  tests hang with the status D, after the tests run for a while.  No oops, 
> > >  no error messages.  I found same problem on mm5, but 2.5.70 is fine.
>
> Sorry, the tests in 2.5.70 also failed, same problem.

OK.  It would be useful to test ext2 as well.

> On Tue, 2003-06-10 at 20:12, Andrew Morton wrote
> > If you could, please retest with "elevator=deadline"?
> > 
> Thanks for your feedback.
> 
> This time I got more fsx tests hang(about 25).  Before normally I saw 5
> or 10 tests fail. Here is the stack info.

Everything stuck waiting for IO to complete again.

Are you able to try a different qlogic driver?  Or a different HBA?

I tried to reproduce this but I don't have sufficient info.

How much memory does that machine have, and what fsx-linux command lines
are you using?

Thanks.

Re: 2.5.70-mm6

[Mingming Cao]

On Thu, 2003-06-12 at 10:50, Andrew Morton wrote:
> > > Mingming Cao <cmm@us.ibm.com> wrote:
> > Sorry, the tests in 2.5.70 also failed, same problem.
> 
> OK.  It would be useful to test ext2 as well.
I will try ext2.
 
> > On Tue, 2003-06-10 at 20:12, Andrew Morton wrote
> 
> Everything stuck waiting for IO to complete again.
> 
> Are you able to try a different qlogic driver?  Or a different HBA?
I will try.
>
> I tried to reproduce this but I don't have sufficient info.
>
Sorry about this.
 
> How much memory does that machine have, and what fsx-linux command lines
> are you using?
The tests were run on 8 way PIII 700MHZ, with 4G memory.  The fsx
command line is:

/root/fsx -R -W -r 4096 -w 4096 /mnt/mntxx/foo &


Thanks for looking at this.