http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.59/2.5.59-mm1/
I've stripped these patches down to just those things which seriously should
go into Linus's tree.
I'm mainly interested in gathering bugfixes across the rest of this month,
but if people have more substantial things which are clearly 2.6 material and
need some testing then OK.
Changes since 2.5.58-mm1:
-deadline-np.patch
-saner-readahead.patch
-ext3-leak-fix.patch
-hugetlbfs-read-write.patch
-ext3-ino_t-cleanup.patch
Merged
-rcf.patch
run-child-first didn't seem to help anything, and an alarming number of
cleanups and fixes were needed to get it working right. Later.
-cputimes_stat.patch
Not to Linus's taste.
-smaller-head-arrays.patch
This was just for trolling Manfred.
+deadline-np-42.patch
+deadline-np-43.patch
I/O scheduler changes from Nick.
-ptrace-flush.patch
Wasn't a serious patch anyway
-rcu-stats.patch
Not sure we need this.
+ext3-scheduling-storm.patch
Fix the bug wherein ext3 sometimes shows blips of 100k context
switches/sec.
-ext3-fsync-speedup.patch
Haven't tested this yet.
-lockless-current_kernel_time.patch
Is ia32-only.
-honour-vm_reserved.patch
Unnecessary
+mixer-bounds-check.patch
sound/oss/sb_mixer.c fixlets
+kirq.patch
Nitin Kamble's rework of the ia32 SMP interrupt distribution logic.
+ext3-truncate-ordered-pages.patch
Free truncated pages earlier from ext3 - should fix the strange "cannot
fork" thing which Con reports.
+read_cache_pages-cleanup.patch
+remove-GFP_HIGHIO.patch
Cleanups.
-wli-02_do_sak.patch
-wli-03_proc_super.patch
-wli-06_uml_get_task.patch
-wli-07_numaq_mem_map.patch
-wli-08_numaq_pgdat.patch
-wli-09_has_stopped_jobs.patch
-wli-10_inode_wait.patch
-wli-13_rmap_nrpte.patch
Not convinced about these.
-dcache_rcu-2.patch
-dcache_rcu-3.patch
Looks like nothing is happening here until Mr Viro returns.
-page-walk-api.patch
-page-walk-api-2.5.53-mm2-update.patch
-page-walk-scsi.patch
-page-walk-scsi-2.5.53-mm2.patch
This is 2.7 material.
+stack-overflow-fix.patch
Fix thinko in the ia32 stack overflow checker.
+Richard_Henderson_for_President.patch
It's all right for him - he lives in Australia.
+parenthesise-pgd_index.patch
+macro-double-eval-fix.patch
+mmzone-parens.patch
+blkdev-fixes.patch
Header file safety.
All 34 patches:
kgdb.patch
devfs-fix.patch
deadline-np-42.patch
(undescribed patch)
deadline-np-43.patch
(undescribed patch)
setuid-exec-no-lock_kernel.patch
remove lock_kernel() from exec of setuid apps
buffer-debug.patch
buffer.c debugging
warn-null-wakeup.patch
reiserfs-readpages.patch
reiserfs v3 readpages support
fadvise.patch
implement posix_fadvise64()
ext3-scheduling-storm.patch
ext3: fix scheduling storm and lockups
auto-unplug.patch
self-unplugging request queues
less-unplugging.patch
Remove most of the blk_run_queues() calls
scheduler-tunables.patch
scheduler tunables
htlb-2.patch
hugetlb: fix MAP_FIXED handling
mixer-bounds-check.patch
OSS ioctl bounds checking fix
kirq.patch
ext3-truncate-ordered-pages.patch
ext3: explicitly free truncated pages
prune-icache-stats.patch
add stats for page reclaim via inode freeing
vma-file-merge.patch
mmap-whitespace.patch
read_cache_pages-cleanup.patch
cleanup in read_cache_pages()
remove-GFP_HIGHIO.patch
remove __GFP_HIGHIO
quota-lockfix.patch
quota locking fix
quota-offsem.patch
quota semaphore fix
oprofile-p4.patch
op4-fix.patch
wli-11_pgd_ctor.patch
(undescribed patch)
wli-11_pgd_ctor-update.patch
pgd_ctor update
stack-overflow-fix.patch
stack overflow checking fix
Richard_Henderson_for_President.patch
Subject: [PATCH] Richard Henderson for President!
parenthesise-pgd_index.patch
Subject: i386 pgd_index() doesn't parenthesize its arg
macro-double-eval-fix.patch
Subject: Re: i386 pgd_index() doesn't parenthesize its arg
mmzone-parens.patch
asm-i386/mmzone.h macro paren/eval fixes
blkdev-fixes.patch
blkdev.h fixes
Andrew Morton <[email protected]> wrote:
>
> -wli-02_do_sak.patch
> -wli-03_proc_super.patch
> -wli-09_has_stopped_jobs.patch
Actually, these were merged into 2.5.59.
> -cputimes_stat.patch
>
> Not to Linus's taste.
Pity, I just had another reason to use this today (Checking if a network
app was locking itself down to a cpu)
> -lockless-current_kernel_time.patch
>
> Is ia32-only.
We can fix that. Ive been avoiding it because it will take some non
trivial cleanup of our ppc64 time.c. Based on how often get_current_time
is appearing in profiles and also how gettimeofday has been known to
cause problems on large SMP (due to read_lock on xtime starving
write_lock in the timer irq) I think this should get merged.
Anton
Anton Blanchard <[email protected]> wrote:
>
>
> > -cputimes_stat.patch
> >
> > Not to Linus's taste.
>
> Pity, I just had another reason to use this today (Checking if a network
> app was locking itself down to a cpu)
You can query that with sched_getaffinity()
> > -lockless-current_kernel_time.patch
> >
> > Is ia32-only.
>
> We can fix that. Ive been avoiding it because it will take some non
> trivial cleanup of our ppc64 time.c. Based on how often get_current_time
> is appearing in profiles
Have you some numbers handy?
> and also how gettimeofday has been known to
> cause problems on large SMP (due to read_lock on xtime starving
> write_lock in the timer irq) I think this should get merged.
>
OK. I've had basically zero success getting non-ia32 people to test these
patches, and breaking their build won't help that.
But yes, this code needs to go ahead.
--- a/arch/i386/kernel/io_apic.c 2003-01-17 13:11:25.000000000 +0200
+++ b/arch/i386/kernel/io_apic.c 2003-01-17 13:11:04.000000000 +0200
@@ -589,6 +589,8 @@
__initcall(balanced_irq_init);
+#else /* !SMP */
+static inline void move_irq(int irq) { }
#endif /* defined(CONFIG_SMP) */
On Fri, 17 Jan 2003, Andrew Morton wrote:
> http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.59/2.5.59-mm1/
> -rcf.patch
>
> run-child-first didn't seem to help anything, and an alarming number of
> cleanups and fixes were needed to get it working right. Later.
I don't know about right, it seems to make threaded applications
originally developed on BSD work better (much lower context switching).
Anyone know if BSD does rcf? This may be an artifact of...
> +ext3-scheduling-storm.patch
>
> Fix the bug wherein ext3 sometimes shows blips of 100k context
> switches/sec.
Is this a 2.5 bug only? Does this need to be back ported to 2.4? Perhaps
this is why I have ctx rate problems and some other sites don't with a
certain application. Very commercial, unfortunately.
--
bill davidsen <[email protected]>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
Bill Davidsen <[email protected]> wrote:
>
> On Fri, 17 Jan 2003, Andrew Morton wrote:
>
> > http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.59/2.5.59-mm1/
>
> > -rcf.patch
> >
> > run-child-first didn't seem to help anything, and an alarming number of
> > cleanups and fixes were needed to get it working right. Later.
>
> I don't know about right, it seems to make threaded applications
> originally developed on BSD work better (much lower context switching).
> Anyone know if BSD does rcf? This may be an artifact of...
"seems to make"? This is too vague for me to comment on, unfortunately.
What applications? What measurements have been made?
It can only affect creation of new threads, not the switching between extant
ones.
> > +ext3-scheduling-storm.patch
> >
> > Fix the bug wherein ext3 sometimes shows blips of 100k context
> > switches/sec.
>
> Is this a 2.5 bug only? Does this need to be back ported to 2.4? Perhaps
> this is why I have ctx rate problems and some other sites don't with a
> certain application. Very commercial, unfortunately.
>
The problem has existed in 2.4 since 2.4.20-pre5. The context switch
problem will only exhibit for small periods of time (say, 10's to 100's of
milliseconds) when the filesystem is under heavy write load.
A patch for 2.4 is at
http://www.zip.com.au/~akpm/linux/patches/2.4/2.4.20/ext3-scheduling-storm.patch