into home.transmeta.com:/home/torvalds/v2.5/linux

This patch provides the ability for a block driver to signal it's too
busy to receive more work and temporarily halt the request queue. In
concept it's similar to the networking netif_{start,stop}_queue helpers.

To do this cleanly, I've ripped out the old tq_disk task queue. Instead
an internal list of plugged queues is maintained which will honor the
current queue state (see QUEUE_FLAG_STOPPED bit). Execution of
request_fn has been moved to tasklet context. blk_run_queues() provides
similar functionality to the old run_task_queue(&tq_disk).

Now, this only works at the request_fn level and not at the
make_request_fn level. This is on purpose: drivers working at the
make_request_fn level are essentially providing a piece of the block
level infrastructure themselves. There are basically two reasons for
doing make_request_fn style setups:

o block remappers. start/stop functionality will be done at the target
  device in this case, which is the level that will signal hardware full
  (or continue) anyways.

o drivers who wish to receive single entities of "buffers" and not
  merged requests etc. This could use the start/stop functionality. I'd
  suggest _still_ using a request_fn for these, but set the queue
  options so that no merging etc ever takes place. This has the added
  bonus of providing the usual request depletion throttling at the block
  level.

It looks like a race betwen exec_mmap and access_process_vm in
proc_pid_cmdline (or any other procfs functions that uses
access_process_vm).

o blk_get_request() and blk_put_request() needs exporting
o blk_max_pfn is used by BLOCK_BOUNCE_ANY, which modular SCSI needs

> Hmm. That patch does not compile. "p->cpu" does not exist, it's
> "p->thread_info->cpu". Tssk.

Ouch, I am bad.  Sorry.

Make the ChangeLog entry something really defamatory.

	Robert Love

This is William Irwin's algorithmically O(1) version of
count_active_tasks (which is currently O(n) for n total tasks on the
system).

I like it a lot: we become O(1) because now we count uninterruptible
tasks, so we can return (nr_uninterruptible + nr_running).  It does not
introduce any overhead or hurt the case for small n, so I have no
complaints.

This copy has a small optimization over the original posting, but is
otherwise the same thing wli posted earlier.  I have tested to make sure
this returns accurate results and that the kernel profile improves.

Previously assigned resources are perfectly valid - just silently
ignore them.

Attached patch adds output of rt_priority and policy to
/proc/<pid>/stats.

This will not break compatibility with existing applications and will
allow ps(1) and friends to display pertinent scheduling information.

Please apply this patch to let binfmt_em86.c compile again.

As pointed out by Russell King, resource name pointers
of the secondary PCI buses are left uninitialized in the
non-x86 PCI allocation path.

Assigning these pointers in pci_add_new_bus() fixes the problem.

into home.transmeta.com:/home/torvalds/v2.5/linux

 - Eliminate all usages of the obscure QUEUE_EMPTY macro.

 - Eliminate all unneccessary checks for RQ_INACTIVE, this can't happen during
   the time we run the request strategy routine of a single major number block
   device. Perhaps the still remaining usage in scsi and i2o_block.c should be
   killed as well, since the upper ll_rw_blk layer shouldn't pass inactive
   requests down.

Those are all places where we have deeply burried and hidden major number
indexed arrays. Let's deal with them slowly...

Since apparently no body else did care thus far, and since I'm using
this driver, well here it comes:

 - Adjust the airo wireless LAN card driver for the fact that modules
   don't export symbols by default any longer.

 - Make some stuff which obivously should be static there static as well.
   (Plenty of code in Linux actually deserves a review for this
   far too common bug...)

 - Replace ide_delay_50m with mdelay(50). There is absolutely no reason we
   should behave different behaviors whatever IDECS support is enabled or not.

 - Kill last parameter of ide_register_hw(). It should return a pointer to the
   interface registered later.

 - pdc202xx patches by Bartomiej onierkiewicz.

 - ServerWorks chi pset support cleanup by Andrej Panin.

 - Move temporarily ide_setup_ports to main.c unfold it in ide-pnp.c.

This fixes three locations in net/ where per-CPU data could bite us
under preemption.  This is the result of an audit I did and should
constitute all of the unsafe code in net/.

In net/core/skbuff.c I did not have to introduce any code - just
rearrange the grabbing of smp_processor_id() to be in the interrupt off
region.  Pretty clean fixes.

Note in the future we can use put_cpu() and get_cpu() to grab the CPU#
safely.  I will send a patch to Marcelo so we can have a 2.4 version
(which doesn't do the preempt stuff), too...

This adds an optimization to set_cpus_allowed: if the task is not
running, there is no sense in kicking the migration_threads into action,
we just need to update task->cpu.  This was suggested by Mike Kravetz.

Besides being an optimization, this would prevent any future race
between set_cpus_allowed and the migration_threads.

This adds documentation about the O(1) scheduler to Documentation/.  The
new scheduler is complicated and providing future scheduler hackers some
background seems a Good Thing to me.

Specifically:

- add Documentation/sched-coding.txt: an overview of the functions,
  magic numbers, and variables in the scheduler as well as (most
  importantly) a review of the locking semantics.

- add Documentation/sched-design.txt: an edited version of Ingo's
  initial email to lkml about his scheduler.  Goes over the design,
  implementation, and goals of the scheduler.  I tried to edit it where
  needed to bring it in line with the scheduler as it is today.

- modify kernel/sched.c: update your copyright and add a change entry
  for the new scheduler.

The attached trivial patch simply changes the printk debug statement in
do_exit when preempt_count!=0 to say "note" instead of "error" and log
at KERN_INFO in lieu of KERN_ERR.

I want to keep the message around a bit, but people get too paranoid
when things like nfsd legitimately exit with a preempt_count=1.

This makes sure that sys_sync() will terminate.  It counts up the
number of dirty pages in the machine and will refuse to write out more
than 1.25 times this number of pages.  This function is called twice
on the sys_sync() path, so the kernel will actually write 2.5x the number
of initially-dirty pages before giving up.

It might reduce pagemap_lru_lock hold times a little, and is more
consistent.  I think all global page accounting is now inside
page_states[].

Factor out some similar code in page_alloc.c

For historical reasons, ext3 has a private BH state bit which has
global scope.  This patch moves it inside ext3.

Patch from Anton Blanchard which replaces

	printk(KERN_FOO __FUNCTION__ ": msg");

with
	printk(KERN_FOO "%s: msg", __FUNCTION__);

in ext3.

Fixes all the goto spaghetti in generic_file_write() and turns it into
something which humans can understand.

Andi tells me that gcc3 does a decent job of relocating blocks out of
line anyway.  This patch gives the compiler a helping hand with
appropriate use of likely() and unlikely().

Random cleanup: remove the mem_map_t typedef.  Just use 'struct page'
everywhere.

An implementation of directory-synchronous mounts.

I sent this out some months ago and it didn't generate a lot of
interest.  Later we had one of the usual cheery exchanges with Wietse
Venema (postfix development) and he agreed that directory synchronous
mounts were something that he could use, and that there was benefit in
implementing them in Linux.  If you choose to apply this I'll push the
2.4 patch.



Patch against e2fsprogs-1.26:
        http://www.zip.com.au/~akpm/linux/dirsync/e2fsprogs-1.26.patch

Patch against util-linux-2.11n:
        http://www.zip.com.au/~akpm/linux/dirsync/util-linux-2.11n.patch


The kernel patch includes implementations for ext2 and ext3. It's
pretty simple.

- When dirsync is in operation against a directory, the following operations
  are synchronous within that directory:  create, link, unlink, symlink,
  mkdir, rmdir, mknod, rename (synchronous if either the source or dest
  directory is dirsync).

- dirsync is a subset of sync.  So `mount -o sync' or `chattr +S'
  give you everything which `mount -o dirsync' or `chattr +D' gives,
  plus synchronous file writes.

- ext2's inode.i_attr_flags is unused, and is removed.

- mount /dev/foo /mnt/bar -o dirsync  works as expected.

- An ext2 or ext3 directory tree can be set dirsync with `chattr +D -R'.

- dirsync is maintained as new directories are created under
  a `chattr +D' directory.  Like `chattr +S'.

- Other filesystems can trivially be taught about dirsync.  It's just
  a matter of replacing `IS_SYNC(inode)' with `IS_DIRSYNC(inode)' in
  the directory update functions.  IS_SYNC will still be honoured when
  IS_DIRSYNC is used.

- Non-directory files do not have their dirsync flag propagated.  So
  an S_ISREG file which is created inside a dirsync directory will not
  have its dirsync bit set.  chattr needs to do this as well.

- There was a bit of version skew between e2fsprogs' idea of the
  inode flags and the kernel's.  That is sorted out here.

- `lsattr' shows the dirsync flag as "D".  The letter "D" was
  previously being used for Compressed_Dirty_File.  I changed
  Compressed_Dirty_File to use "Z".  Is that OK?

The mount(2) manpage needs to be taught about MS_DIRSYNC.

Spot the difference:

aops.readpage
aops.readpages
aops.writepage
aops.writeback_mapping

The patch renames `writeback_mapping' to `writepages'

Turn on multipage no-buffers reads for ext3.

Multipage BIO writeout from the pagecache.

It's pretty much the same as multipage reads.  It falls back to buffers
if things got complex.

The write case is a little more complex because it handles pages which
have buffers and pages which do not.  If the page didn't have buffers
this code does not add them.

Implements BIO-based multipage reads into the pagecache, and turns this
on for ext2.

CPU load for `cat large_file > /dev/null' is reduced by approximately
15%.  Similar reductions for tiobench with a single thread.  (Earlier
claims of 25% were exaggerated - they were measured with slab debug
enabled.  But 15% isn't bad for a load which is dominated by copy_*_user
costs).

With 2, 4 and 8 tiobench threads, throughput is increased as well, which was
unexpected.  It's due to request queue weirdness.  (Generally the
request queueing is doing bad things under certain workloads - that's a
separate issue.)

BIOs of up to 64 kbytes are assembled and submitted for readahead and
for single-page reads.  So the work involved in reading 32 pages has gone
from:

	- allocate and attach 32 buffer_heads
	- submit 32 buffer_heads
	- allocate 32 bios
	- submit 32 bios

to:

	- allocate 2 bios
	- submit 2 bios

These pages never have buffers attached.  Buffers will be attached
later if the application writes to these pages (file overwrite).

The first version of this code (in the "delayed allocation" patches)
tries to handle everything - bios which start mid-page, bios which end
mid-page and pages which are covered by multiple bios.  It is very
complex code and in fact appears to be incorrect: out-of-order BIO
completion could cause a page to come unlocked at the wrong time.

This implementation is much simpler: if things get complex, it just
falls back to the buffer-based block_read_full_page(), which isn't
going away, and which understands all that complexity.  There's no
point in doing this in two places.

This code will bypass the buffer layer for

 - fully-mapped pages which are on-disk contiguous.

 - fully unmapoped pages (holes)

 - partially unmapped pages, where the unmappedness is at the end of
   the page (end-of-file).

and everything else falls back to buffers.

This means that with blocksize == PAGE_CACHE_SIZE, 100% of pages are
handed direct to BIO.  With a heavy 10-minute dbench run on 4k
PAGE_CACHE_SIZE and 1k blocks, 95% of pages were handed direct to BIO.
Almost all of the other 5% were passed to block_read_full_page()
because they were already partially uptodate from an earlier sub-page
write().  This ratio will fall if PAGE_CACHE_SIZE/blocksize is greater
than four.  But if that's the case, CPU efficiency is far from the main
concern - there are significant seek and bandwidth problems just at 4
blocks per page.

This code will stress out the block layer somewhat - RAID0 doesn't like
multipage BIOs, and there are probably others.  RAID0 seems to struggle
along - readahead fails but read falls back to single-page reads, which
succeed.  Such problems may be worked around by setting MPAGE_BIO_MAX_SIZE
to PAGE_CACHE_SIZE in fs/mpage.c.

It is trivial to enable multipage reads for many other filesystems.  We
can do that after completion of external testing of ext2.

Relax the requirements on the writeback_mapping a_op.

This function is passed the number of pages which it should write.  The
current fs-writeback.c code will get confused if the address_space
writes back more pages than it was asked to.

With this change the address_space may write more pages than required
if that is convenient.  Extent-based fileystems may wish to do this.

Pages which are under writeout to swap are locked, and not
PageWriteback().  So page allocators do not throttle against them in
shrink_caches().

This causes enormous list scans and general coma under really heavy
swapout loads.

One fix would be to teach shrink_cache() to wait on PG_locked for swap
pages.  The other approach is to set both PG_locked and PG_writeback
for swap pages so they can be handled in the same manner as file-backed
pages in shrink_cache().

This patch takes the latter approach.

Fix bug in the loop driver.

When presented with a multipage BIO, loop is overindexing the first
page in the BIO rather than advancing to the second page.  It scribbles
on the backing file and/or on kernel memory.

This happens with multipage BIO-based pagecache I/O and presumably with
O_DIRECT also.

The fix is much-needed with the multipage-BIO patches - using that code
on loop-backed filesystems has rather messy results.

The set_page_dirty() in the ext3_writepage() failure path isn't right.
set_page_dirty() will alter buffer states - it's a "whole page"
dirtying.

__set_page_dirty_buffers() is emitting warnings when it refuses to set
dirty a non-uptodate buffer against a partially-mapped page.

All we want to do in there is to move the page back onto
mapping->dirty_pages, without altering the state of its buffers.