This patch takes the BKL out of vfs_readdir() and moves it into the
individual filesystems, all 35 of them.  I have the feeling that this
wasn't done before because there are a lot of these to change and it was
a pain to find them all.  I definitely got all of those that were
defined in the in the structure declaration like this "readdir:
fs_readdir;" vxfs_readdir was assigned strangely, but I found it anyway.
I also left devfs out of this one.  Richard seems confident that devfs
has no need for the BKL.

Tue Apr 30 13:23:13 CEST 2002 ide-clean-48

This fixes the "performance" degradation partially, becouse we don't
miss that many jiffies in choose_urgent_device() anymore.  However
choose_urgent_device has to be fixed for the off by one error to don't
loop for a whole 1/100 second before submitting the next request.

- Include small declaration bits for Jens. (WIN_NOP fix in esp.)

- Fix ide-pmac to conform to the recent API changes.

- Prepare and improve the handling of the request queue. It sucks now as many
   request as possible. This is improving the performance.

- Rewrite choose_drive() to iterate explicitely over the channels and devices
    on them. It is not performance critical to iterate over this typically quite
    small array of disks and allows us to let them act on the natural entity,
    namely the channel as well as to remove the drive->next field from struct
    ata_device.  Make the device eviction code in ide_do_request() more
    intelliglible.  Add some comments explaining the reasoning behind the code
    there.

- Now finally since the code for choosing the drive which will be serviced next
    is intelliglibly it became obvious that the attempt to choose the next drive
    based on the duration of the last request was entierly bogous. (Because for
    example wakeups can take a long time, but this doesn't indicate that the
    drive is slow.) Remove this criterium and the corresponding accounting
    therefore. Threat all drives fairly right now.

Surprise surprise the overall system throughput increased :-).

- Remove the specific CONFIG_IDEDMA_PCI_WIP in favor of using the generic
   CONFIG_EXPERIMENTAL tag. (Pointed out by Vojtech Pavlik).

- Change the signature of the IRQ handler to take the request directly as a
   parameter. This doesn't blow the code up but makes it much more obvious and
   finally it's reducing the number of side effects of the hwgroup->rq field.

- A second sharp look after the above change allowed us to remove the wrq field
   from the hwgroup struct. It's just not used at all.

- Change the signature of the end_request member of struct ata_operations to
   take the request as a second argument. Similar for __ide_end_request()
   and ide_end_request().

- Remove BUG_ON() items just before ide_set_handler(). The check in
   ide_set_handler is clever enough now.

- Remove the rq subfield from ide-scsi packet structure. We have now the
   request context always in place. Same for floppy.

- Let the timer expiry function take the request as a direct argument.

Yes I know those changes are extensive. But they are a necessary step
in between for the following purposes:

- Consolidate the whole ATA/ATAPI stuff on passing a single unified request
   handling object. Because after eliminating those side effects it's far easier
   to see what's passed where.

- Minimizing the amount of side effects in the overall code. That's a good
   thing anyway and it *doesn't* cost us neither performance nor space, since
   the stack depths are small anyway here.

- Minimizing the usage of hwgroup - which should go away if possible.

Renames sync_buffers() to sync_blockdev() and removes its (never used)
second argument.

Removes fsync_no_super() in favour of direct calls to sync_blockdev().

- Fixes a performance problem - callers of
  prepare_write/commit_write, etc are locking pages, which synchronises
  them behind writeback, which also locks these pages.  Significant
  slowdowns for some workloads.

- So pages are no longer locked while under writeout.  Introduce a
  new PG_writeback and associated infrastructure to support this design
  change.

- Pages which are under read I/O still use PageLocked.  Pages which
  are under write I/O have PageWriteback() true.

  I considered creating Page_IO instead of PageWriteback, and marking
  both readin and writeout pages as PageIO().  So pages are unlocked
  during both read and write.  There just doesn't seem a need to do
  this - nobody ever needs unblocking access to a page which is under
  read I/O.

- Pages under swapout (brw_page) are PageLocked, not PageWriteback.
  So their treatment is unchangeded.

  It's not obvious that pages which are under swapout actually need
  the more asynchronous behaviour of PageWriteback.

  I was setting the swapout pages PageWriteback and unlocking them
  prior to submitting the buffers in brw_page().  This led to deadlocks
  on the exit_mmap->zap_page_range->free_swap_and_cache path.  These
  functions call block_flushpage under spinlock.  If the page is
  unlocked but has locked buffers, block_flushpage->discard_buffer()
  sleeps.  Under spinlock.  So that will need fixing if for some reason
  we want swapout to use PageWriteback.

  Kernel has called block_flushpage() under spinlock for a long time.
   It is assuming that a locked page will never have locked buffers.
  This appears to be true, but it's ugly.

- Adds new function wait_on_page_writeback().  Renames wait_on_page()
  to wait_on_page_locked() to remind people that they need to call the
  appropriate one.

- Renames filemap_fdatasync() to filemap_fdatawrite().  It's more
  accurate - "sync" implies, if anything, writeout and wait.  (fsync,
  msync) Or writeout.  it's not clear.

- Subtly changes the filemap_fdatawrite() internals - this function
  used to do a lock_page() - it waited for any other user of the page
  to let go before submitting new I/O against a page.  It has been
  changed to simply skip over any pages which are currently under
  writeback.

  This is the right thing to do for memory-cleansing reasons.

  But it's the wrong thing to do for data consistency operations (eg,
  fsync()).  For those operations we must ensure that all data which
  was dirty *at the time of the system call* are tight on disk before
  the call returns.

  So all places which care about this have been converted to do:

	filemap_fdatawait(mapping);	/* Wait for current writeback */
	filemap_fdatawrite(mapping);	/* Write all dirty pages */
	filemap_fdatawait(mapping);	/* Wait for I/O to complete */

- Fixes a truncate_inode_pages problem - truncate currently will
  block when it hits a locked page, so it ends up getting into lockstep
  behind writeback and all of the file is pointlessly written back.

  One fix for this is for truncate to simply walk the page list in the
  opposite direction from writeback.

  I chose to use a separate cleansing pass.  It is more
  CPU-intensive, but it is surer and clearer.  This is because there is
  no reason why the per-address_space ->vm_writeback and
  ->writeback_mapping functions *have* to perform writeout in
  ->dirty_pages order.  They may choose to do something totally
  different.

  (set_page_dirty() is an a_op now, so address_spaces could almost
  privatise the whole dirty-page handling thing.  Except
  truncate_inode_pages and invalidate_inode_pages assume that the pages
  are on the address_space lists.  hmm.  So making truncate_inode_pages
  and invalidate_inode_pages a_ops would make some sense).

Implements hashed waitqueues for buffer_heads.  Drops twelve bytes from
struct buffer_head.

Moves all buffer_head-related stuff out of linux/fs.h and into
linux/buffer_head.h.  buffer_head.h is currently included at the very
end of fs.h.  So it is possible to include buffer_head directly from
all .c files and remove this nested include.

Also rationalises all the set_buffer_foo() and mark_buffer_bar()
functions.  We have:

	set_buffer_foo(bh)
	clear_buffer_foo(bh)
	buffer_foo(bh)

and, in some cases, where needed:

	test_set_buffer_foo(bh)
	test_clear_buffer_foo(bh)

And that's it.

BUFFER_FNS() and TAS_BUFFER_FNS() macros generate all the above real
inline functions.  Normally not a big fan of cpp abuse, but in this
case it fits.  These function-generating macros are available to
filesystems to expand their own b_state functions.  JBD uses this in
one case.

Remove show_buffers().  It really has nothing to show any more.  just
buffermem_pages() - move that out into the callers.

There's a lot of duplication in this code.  better approach would be to
remove all the duplicated code out in the architectures and implement
generic show_memory_state().  Later.

Remove PG_skip.  Nothing is using it (the change was acked by rmk a
while back)

Removes inode.i_dirty_data_buffers.  It's no longer used - all dirty
buffers have their pages marked dirty and filemap_fdatasync() /
filemap_fdatawait() catches it all.

Updates all callers.

This required a change in JFS - it has "metapages" which
are a container around a page which holds metadata.  They
were holding these pages locked and were relying on fsync_inode_data_buffers
for writing them out.  So fdatasync() deadlocked.

I've changed JFS to not lock those pages.  Change was acked
by Dave Kleikamp <shaggy@austin.ibm.com> as the right
thing to do, but may not be complete.  Probably igrab()
against ->host is needed to pin the address_space down.

Removal of buffer_head.b_inode.  The list_emptiness of b_inode_buffers
is used to indicate whether the buffer is on an inode's
i_dirty_buffers.

Remove writeout_one_page(), waitfor_one_page() and the now-unused
generic_buffer_fdatasync().

Add new

	write_one_page(struct page *page, int wait)

which is exported to modules.  Update callers to use that.  It's only
used for IS_SYNC operations.

Removes some redundant BUG checks - trueness of PageSwapCache() implies
that page->mapping is non-NULL, and we've already checked that.

page->flags cleanup.

Moves the definitions of the page->flags bits and all the PageFoo
macros into linux/page-flags.h.  That file is currently included from
mm.h, but the stage is set to remove that and include page-flags.h
direct in all .c files which require that.  (120 of them).

The patch also makes all the page flag macros and functions consistent:

For PG_foo, the following functions are defined:

	SetPageFoo
	ClearPageFoo
	TestSetPageFoo
	TestClearPageFoo
	PageFoo

and that's it.

- Page_Uptodate is renamed to PageUptodate

- LockPage is removed.  All users updated to use SetPageLocked

- UnlockPage is removed.  All callers updated to use unlock_page().
  it's a real function - there's no need to hide that fact.

- PageTestandClearReferenced renamed to TestClearPageReferenced

- PageSetSlab renamed to SetPageSlab

- __SetPageReserved is removed.  It's an infinitesimally small
   microoptimisation, and is inconsistent.

- TryLockPage is renamed to TestSetPageLocked

- PageSwapCache() is renamed to page_swap_cache(), so it doesn't
  pretend to be a page->flags bit test.

Convert minixfs directory code to not rely on the state of data outside
i_size.

Removes the buffer_head unused list.  Use a mempool instead.

The reduced lock contention provided about a 10% boost on ANton's
12-way.

[ I reversed the order in which writeback walks the superblock's
  dirty inodes.  It sped up dbench's unlink phase greatly.  I'm
  such a sleaze ]

The core writeback patch.  Switches file writeback from the dirty
buffer LRU over to address_space.dirty_pages.

- The buffer LRU is removed

- The buffer hash is removed (uses blockdev pagecache lookups)

- The bdflush and kupdate functions are implemented against
  address_spaces, via pdflush.

- The relationship between pages and buffers is changed.

  - If a page has dirty buffers, it is marked dirty
  - If a page is marked dirty, it *may* have dirty buffers.
  - A dirty page may be "partially dirty".  block_write_full_page
    discovers this.

- A bunch of consistency checks of the form

	if (!something_which_should_be_true())
		buffer_error();

  have been introduced.  These fog the code up but are important for
  ensuring that the new buffer/page code is working correctly.

- New locking (inode.i_bufferlist_lock) is introduced for exclusion
  from try_to_free_buffers().  This is needed because set_page_dirty
  is called under spinlock, so it cannot lock the page.  But it
  needs access to page->buffers to set them all dirty.

  i_bufferlist_lock is also used to protect inode.i_dirty_buffers.

- fs/inode.c has been split: all the code related to file data writeback
  has been moved into fs/fs-writeback.c

- Code related to file data writeback at the address_space level is in
  the new mm/page-writeback.c

- try_to_free_buffers() is now non-blocking

- Switches vmscan.c over to understand that all pages with dirty data
  are now marked dirty.

- Introduces a new a_op for VM writeback:

	->vm_writeback(struct page *page, int *nr_to_write)

  this is a bit half-baked at present.  The intent is that the address_space
  is given the opportunity to perform clustered writeback.  To allow it to
  opportunistically write out disk-contiguous dirty data which may be in other zones.
  To allow delayed-allocate filesystems to get good disk layout.

- Added address_space.io_pages.  Pages which are being prepared for
  writeback.  This is here for two reasons:

  1: It will be needed later, when BIOs are assembled direct
     against pagecache, bypassing the buffer layer.  It avoids a
     deadlock which would occur if someone moved the page back onto the
     dirty_pages list after it was added to the BIO, but before it was
     submitted.  (hmm.  This may not be a problem with PG_writeback logic).

  2: Avoids a livelock which would occur if some other thread is continually
     redirtying pages.

- There are two known performance problems in this code:

  1: Pages which are locked for writeback cause undesirable
     blocking when they are being overwritten.  A patch which leaves
     pages unlocked during writeback comes later in the series.

  2: While inodes are under writeback, they are locked.  This
     causes namespace lookups against the file to get unnecessarily
     blocked in wait_on_inode().  This is a fairly minor problem.

     I don't have a fix for this at present - I'll fix this when I
     attach dirty address_spaces direct to super_blocks.

- The patch vastly increases the amount of dirty data which the
  kernel permits highmem machines to maintain.  This is because the
  balancing decisions are made against the amount of memory in the
  machine, not against the amount of buffercache-allocatable memory.

  This may be very wrong, although it works fine for me (2.5 gigs).

  We can trivially go back to the old-style throttling with
  s/nr_free_pagecache_pages/nr_free_buffer_pages/ in
  balance_dirty_pages().  But better would be to allow blockdev
  mappings to use highmem (I'm thinking about this one, slowly).  And
  to move writer-throttling and writeback decisions into the VM (modulo
  the file-overwriting problem).

- Drops 24 bytes from struct buffer_head.  More to come.

- There's some gunk like super_block.flags:MS_FLUSHING which needs to
  be killed.  Need a better way of providing collision avoidance
  between pdflush threads, to prevent more than one pdflush thread
  working a disk at the same time.

  The correct way to do that is to put a flag in the request queue to
  say "there's a pdlfush thread working this disk".  This is easy to
  do: just generalise the "ra_pages" pointer to point at a struct which
  includes ra_pages and the new collision-avoidance flag.

Changes the way in which the readahead code locates the readahead
setting for the underlying device.

- struct block_device and struct address_space gain a *pointer* to the
  current readahead tunable.

- The tunable lives in the request queue and is altered with the
  traditional ioctl.

- The value gets *copied* into the struct file at open() time.  So a
  fcntl() mode to modify it per-fd is simple.

- Filesystems which are not request_queue-backed get the address of the
  global `default_ra_pages'.  If we want, this can become a tunable.

- Filesystems are at liberty to alter address_space.ra_pages to point
  at some other fs-private default at new_inode/read_inode/alloc_inode
  time.

- The ra_pages pointer can become a structure pointer if, at some time
  in the future, high-level code needs more detailed information about
  device characteristics.

  In fact, it'll need to become a struct pointer for use by
  writeback: my current writeback code has the problem that multiple
  pdflush threads can get stuck on the same request queue.  That's a
  waste of resources.  I currently have a silly flag in the superblock
  to try to avoid this.

  The proper way to get this exclusion is for the high-level
  writeback code to be able to do a test-and-set against a
  per-request_queue flag.  That flag can live in a structure alongside
  ra_pages, conveniently accessible at the pagemap level.

One thing still to-be-done is going into all callers of blk_init_queue
and blk_queue_make_request and making sure that they're setting up a
sensible default.  ATA wants 248 sectors, and floppy drives don't want
128kbytes, I suspect.  Later.

This patch provides global accounting of locked and dirty pages.  It
does this via lightweight per-CPU data structures.  The page_cache_size
accounting has been changed to use this facility as well.

Locked and dirty page accounting is needed for making writeback and
throttling decisions.

The patch also starts to move code which is related to page->flags
out of linux/mm.h and into linux/page-flags.h

Convert ext2 directory handling to not rely on the contents of pages
outside i_size.

This is because block_write_full_page (which is used for all writeback)
zaps the page outside i_size.

Emit a printk when a page allocation fails.  Considered useful for
diagnosing crashes.

	OK, here comes.  Patch below is an attempt to do the fastwalk
stuff in right way and so far it seems to be working.

 - dentry leak is plugged
 - locked/unlocked state of nameidata doesn't depend on history - it
   depends only on point in code.
 - LOOKUP_LOCKED is gone.
 - following mounts and .. doesn't drop dcache_lock
 - light-weight permission check distinguishes between "don't know" and
   "permission denied", so we don't call full-blown permission() unless
   we have to.
 - code that changes root/pwd holds dcache_lock _and_ write lock on
   current->fs->lock.  I.e. if we hold dcache_lock we can safely
   access our ->fs->{root,pwd}{,mnt}
 - __d_lookup() does not increment refcount; callers do dget_locked()
   if they need it (behaviour of d_lookup() didn't change, obviously).
 - link_path_walk() logics had been (somewhat) cleaned up.

- Fix bogus set_multimode() change. I tough I had reverted it before diff-ing.
   This was causing hangs of /dev/hdparm -m8 /dev/hda and similar commands.

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

The following patch is a _very minor_ cleanup in the ext3 code for
ext3_prepare_write().  It simply removes the setting of "handle" from
the current transaction handle, because "handle" is actually set again
a couple of lines later, where it properly allocates a new transaction
handle for this write.  The code removed in this patch is probably left
over from some development version of ext3 where the transaction handle
was started before ext3_prepare_write was called.

The only reason I saw this was because I was trying to find where the
handle was allocated for an ext3 file write, and at first glance it
didn't appear to be allocated anywhere...

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

Yes, gdb and strace are broken on alpha since 2.5.6, IIRC.
Some necessary 'thread_info' changes still are missing in ptrace.c.

Fixed thus.

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

The patch below fixes sem_exit() so that the BKL is always released.

Here's a simple patch to allow the bttv driver to be build with newer
binutils...

In conjunction with some of the earlier problems found in /proc code,
now it turns out that snprintf doesn't work correctly in the kernel...
Without the following patch, snprintf can return lengths greater than
the size argument passed.  The net effect is that code using the return
value from snprintf can still buffer overrun.  This is fixed by not
updating the pointer in the buffer unless there is actually space.

The following patch drops the above functions from the ARM port; we've
already removed them from the syscall table on ARM, so we can safely
remove these from the ARM build.

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

I added ->readdir(), and a new section referring to the sb_op BKL shift
as well.