This function tries to allocate increasingly large buffers, but it gets the
bounds wrong by a factor of PAGE_SIZE.  It causes boot-time devfs mounting
to fail.

We're getting asserion failures in commit in data=journal mode.

journal_unmap_buffer() has unexpectedly donated this buffer to the committing
transaction, and the commit-time assertion doesn't expect that to happen.  It
doesn't happen in 2.4 because both paths are under lock_journal().

Simply remove the assertion: the commit code will uncheckpoint the buffer and
then recheckpoint it if needed.

From: Felipe Alfaro Solana <felipe_alfaro@linuxmail.org>

Fix compile error in 2.6.0-test2 when Netfilter IP connection tracking
is enabled.

From: Manfred Spraul <manfred@colorfullife.com>

Eli Barzilay noticed that select() for tty devices is broken: For
stopped tty devices, select says POLLOUT and write fails with -EAGAIN.

    http://marc.theaimsgroup.com/?l=linux-kernel&m=105902461110282&w=2

I've tracked this back to normal_poll in drivers/char/n_tty.c:

 > if (tty->driver->chars_in_buffer(tty) < WAKEUP_CHARS)
 >                mask |= POLLOUT | POLLWRNORM;

It assumes that a following write will succeed if less than 256 bytes
are in the write buffer right now. This assumption is wrong for
con_write_room: if the console is stopped, it returns 0 bytes buffer
size (con_write_room()). Ditto for pty_write_room.

From: florin@iucha.net (Florin Iucha)

Fix compilation of net/ipv4/netfilter/ipt_helper.c by including the
proper header files.

From: Nathan Scott <nathans@sgi.com>

This patch adds a mechanism by which a filesystem can register an interest in
the completion of direct I/O.  The completion routine will be given the
inode, an offset and a length, and an optional filesystem-private field.

We have extended the use of the buffer_head-based interface (i.e.
get_block_t) for direct I/O such that the b_private field is now utilised.
It is defined to be initially zero at the start of I/O, and will be passed
into the filesystem unmodified by the VFS with each map request, while
setting up the direct I/O.  Once I/O has completed the final value of this
pointer will be passed into a filesystems I/O completion handler.  This
mechanism can be used to keep track of all of the mapping requests which
encompass an individual direct I/O request.

This has been implemented specifically for XFS, but is done so as to be as
generic as possible.  XFS uses this mechanism to provide support for
unwritten extents - these are file extents which have been pre-allocated
on-disk, but not yet written to (once written, these become regular file
extents, but only once I/O is complete).

From: Nikita Danilov <Nikita@Namesys.COM>

Use zone->pressure (rathar than scanning priority) to determine when to
start reclaiming mapped pages in refill_inactive_zone().  When using
priority every call to try_to_free_pages() starts with scanning parts of
active list and skipping mapped pages (because reclaim_mapped evaluates to
0 on low priorities) no matter how high memory pressure is.

From: Nikita Danilov <Nikita@Namesys.COM>

The vmscan logic at present will scan the inactive list with increasing
priority until a threshold is triggered.  At that threshold we start
unmapping pages from pagetables.

The problem is that each time someone calls into this code, the priority is
initially low, so some mapped pages will be refiled event hough we really
should be unmapping them now.

Nikita's patch adds the `pressure' field to struct zone.  it is a decaying
average of the zone's memory pressure and allows us to start unmapping pages
immediately on entry to page reclaim, based on measurements which were made
in earlier reclaim attempts.

kswapd currently takes a throttling nap even if it freed all the pages it
was asked to free.

Change it so we only throttle if reclaim is not being sufficiently
successful.

We're currently just setting the referenced bit when modifying pagecache in
write().

Consequently overwritten (and redirtied) pages are remaining on the inactive
list.  The net result is that a lot of dirty pages are reaching the tail of
the LRU in page reclaim and are getting written via the writepage() in there.

But a core design objective is to minimise the amount of IO via that path,
and to maximise the amount of IO via balance_dirty_pages().  Because the
latter has better IO patterns.

This may explain the bad IO patterns which Gerrit talked about at KS.

From: Andrey Borzenkov <arvidjaar@mail.ru>

A while back Andrey fixed a devfs bug in which we were running
remove_wait_queue() against a wait_queue_head which was on another process's
stack, and which had gone out of scope.

The patch reverts that fix and does it the same way as 2.4: just leave the
waitqueue struct dangling on the waitqueue_head: there is no need to touch it
at all.

It adds a big comment explaining why we are doing this nasty thing.

From: Hans-Joachim Hetscher <me@privacy.net>

the Hamradio 6pack driver wasn't modified to work with the 1000 HZ
internal kernel timebase.

It is using "snd".  It should be using "sound".

From: jbarnes@sgi.com (Jesse Barnes)

This patch is needed for some discontig boxes since the memory maps may
be built out-of-order.

From: bzzz@tmi.comex.ru

Now we have sync_fs(), the kludge of using write_super() to detect when the
VFS is trying to sync the fs is unneeded.

With this change we don't accidentally run commits in response to kupdate
and bdflush activity and it speedup up some heavy workloads significantly.

From: Alex Tomas <bzzz@tmi.comex.ru>

ext3_getblk() memsets a newly allocated buffer, but forgets to check
whether a different thread brought it uptodate while we waited for the
buffer lock.

It's OK normally because we're serialised by the page lock.  But lustre
apparently is doing something different with getblk and hits this race.

Plus I suspect it's racy with competing O_DIRECT writes.

From: Alex Tomas <bzzz@tmi.comex.ru>

ext3_get_inode_loc() read inode's block only if:

  1) this inode has no copy in memory
  2) inode's block has another valid inode(s)

this optimization allows to avoid needless I/O in two cases:

1) just allocated inode is first valid in the inode's block

2) kernel wants to write inode, but buffer in which inode
   belongs to gets freed by VM

Since Jens changed the block layer to fail readahead if the queue has no
requests free, a few changes suggest themselves.

- It's a bit silly to go and alocate a bunch of pages, build BIOs for them,
  submit the IO only to have it fail, forcing us to free the pages again.

  So the patch changes do_page_cache_readahead() to peek at the queue's
  read_congested state.  If the queue is read-congested we abandon the entire
  readahead up-front without doing all that work.

- If the queue is not read-congested, we go ahead and do the readahead,
  after having set PF_READAHEAD.

  The backing_dev_info's read-congested threshold cuts in when 7/8ths of
  the queue's requests are in flight, so it is probable that the readahead
  abandonment code in __make_request will now almost never trigger.

- The above changes make do_page_cache_readahead() "unreliable", in that it
  may do nothing at all.

  However there are some system calls:

	- fadvise(POSIX_FADV_WILLNEED)
	- madvise(MADV_WILLNEED)
	- sys_readahead()

  In which the user has an expectation that the kernel will actually
  perform the IO.

  So the patch creates a new "force_page_cache_readahead()" which will
  perform the IO regardless of the queue's congestion state.

  Arguably, this is the wrong thing to do: even though the application
  requested readahead it could be that the kernel _should_ abandon the user's
  request because the disk is so busy.

  I don't know.  But for now, let's keep the above syscalls behaviour
  unchanged.  It is trivial to switch back to do_page_cache_readahead()
  later.

Since Jens added the pagecache readahead support in the block layer we've
been getting bogus IO error messages from buffer.c due to __make_request
calling end_io against a non-uptodate buffer.

We can just use PF_READAHEAD to shut that up.  But really, we shouldn't even
have allocated all those pages and submittted the readahead IO if the queue
was congested.   We have the infrastructure to do that now.

From: Dave McCracken <dmccr@us.ibm.com>

This patch solves the race between truncate and page in which can cause stray
anon pages to appear in the truncated region.

The race occurs when a process is sleeping in pagein IO during the truncate:
there's a window after checking i_size in which the paging-in process decides
that the page was an OK one.

This leaves an anon page in the pagetables, and if the file is subsequently
extended we have an anon page floating about inside a file-backed mmap - user
modifications will not be written out.

Apparently this is also needed for the implementation of POSIX semantics for
distributed filesystems.

We use a generation counter in the address_space so the paging-in process can
determine whether there was a truncate which might have shot the new page
down.

It's a bit grubby to be playing with files and inodes in do_no_page(), but we
do need the page_table_lock coverage for this, and rearranging thngs to
provide that coverage to filemap_nopage wasn't very nice either.

From: "Paul E. McKenney" <paulmck@us.ibm.com>

The patch reworks and generalises vmtruncate_list() a bit to create an API
which invalidates a specified portion of an address_space, permitting
distributed filesystems to maintain POSIX semantics when a file mmap()ed on
one client is modified on another client.

From: Chris Mason <mason@suse.com>

unlock_buffer() needs a barrier before the waitqueue_active() optimisation.

We need to subtract the number of freed slab pages from the number of pages
to free, not add it.

From: Marc Zyngier <mzyngier@freesurf.fr>

The following patch tries to fix a small bug that crept in at some
point during 2.5.

None of my 3c592 or 3c597 would work if I didn't force media
type. Instead, it would try to probe MII, looking for a suitable
transceiver, and finaly give up, because these cards really do not
have any sort of MII... :

  EISA: Probing bus 0 at Intel Corp. 82375EB
  EISA: Mainboard DEC5000 detected.
  EISA: slot 2 : ADP0001 detected.
  EISA: slot 3 : ADP7771 detected.
  EISA: slot 4 : DPTA401 detected.
  EISA: slot 5 : TCM5920 detected.
  3c59x: Donald Becker and others. www.scyld.com/network/vortex.html
  00:05: 3Com EISA 3c592 EISA 10Mbps Demon/Vortex at 0x5000. Vers LK1.1.19
    ***WARNING*** No MII transceivers found!
  EISA: Detected 4 cards.

With the enclosed patch, it just works, at least on my setup (3c592 on
Alpha, and 3c597 on x86). I haven't been able to test it didn't break
cards with MII, because I do not have such cards in my test boxes...

The patch also removes two useless EISA-only #define I introduced some
time ago.

From: Greg KH <greg@kroah.com>

Different architectures use different types for dev_t, so it is hard to
print dev_t variables out correctly.  Quite a lot of code is wrong now, and
will continue to be wrong when 64-bit dev_t is merged.

Greg's patch introduces a little wrapper function which can be used to
safely form a dev_t for printing.  I added the format_dev_t function as
well, which is needed for direct insertion in a printk statement.

Currently, all of the 3c59x power management code is disabled unless the
`enable_wol' module parameter is provided.  This was done because the PM
support was added quite late in the 2.4 cycle.

It was always intended that this conditionality be removed in 2.5.

From: Jeremy Fitzhardinge <jeremy@goop.org>

The 900MHz Pentium M has two spaces before the frequency:
"Intel(R) Pentium(R) M processor  900MHz"

This patch adds a 2nd CPU macro (_CPU) which also takes the
stringified speed so that extra spacing can be added.

We get a bug report about once per month wherein find_get_block_slow() spits
an error message.  For some reason we have buffers against a blockdev page
which have the incorrect b_size.

Probably, an earlier set_blcoksize() failed to invalidate all the apges for
some reason.  I just don't know.

The patch adds a bit of extra debug info to aid in diagnosing this.

From: William Lee Irwin III <wli@holomorphy.com>

The original pgd/pmd slabification patches had a critical bug on
non-PAE where both modifications of pgd entries to remove pagetables
attached for non-PSE mappings back to a PSE state and modifications of
pgd entries to attach pagetables to bring PSE mappings into a non-PSE
state were not propagated to cached pgd's. PAE was immune to it owing
to the shared kernel pmd.

The following patch vs. 2.5.69 restores the slabification done to cache
preconstructed pagetables with the proper propagation of conversions
to and from PSE mappings to cached pgd's for the non-PAE case.

This is an optimization to reduce the bitblitting overhead for spawning
small tasks (for larger ones, bottom-level pagetable copies dominate)
primarily on non-PAE; the PAE code change is largely to remove #ifdefs
and to treat the two cases uniformly, though some positive but small
performance improvement has been observed for PAE in one of mbligh's
posts. The non-PAE performance improvement has been observed on a box
running a script-heavy end-user workload as a large long-term profile
hit count reduction for pgd_alloc() and relatives thereof.

I would very much appreciate outside testers. Even though I've been
able to verify this boots and runs properly and survives several cycles
of restarting X on my non-PAE Thinkpad T21, that environment has never
been able to reproduce the bug. Those with the proper graphics hardware
to prod the affected codepaths into action are the ones best suited to
verify proper functionality. There is also some locking introduced; if
some performance verification on non-PAE SMP i386 targets (my SMP
targets unfortunately all require PAE due to arch code dependencies)
that also have the proper hardware could be done, that would help
determine whether alternative locking schemes that competed against
the one shown here are preferable (in particular, the ticket-based
scheme mentioned in the comments).

From Stephen Smalley <sds@epoch.ncsc.mil>

This has been in -mm for a few weeks and James Morris has been
regression testing each release.

- remove unneeded loglevel manipulation in journal_dirty_metadata()

- remove crud which was acidentally added to blkmtd.c

- remove unused vars in mxser.c (Vinay K Nallamothu <vinay-rc@naturesoft.net>)

- PF_LESS_THROTTLE was using the wrong bit (Joe Korty <joe.korty@ccur.com>)

- unused var in cyclades.c ("Krishnakumar. R" <krishnakumar@naturesoft.net>)

This patch notes that users should get nfs-utils 1.0.5 (1.0.4 had a
memory usage bug), and tells them where to get it.

Here's another MIPS update.  The patch is huge because it completly
folds mips64 into mips, thereby eleminating 41010 lines of code.

Only bug fixes and making it compile again and a few minor features.
Also one security fix that got lost earlier

- Document boot options
- Better cpu local data
- Emulate FIOQSIZE
- Fix return value of 32bit ipccall
- Various minor style fixes
- Save some memory in apic tables
- Merge with 2.6.0test2/i386
- Readd ioport fix
- Sort exception tables at boot time
- Add local.h
- Fix for_each_cpu on UP
- Add utimes and tgkill system calls for 64bit
- Update defconfig

devfs_mk_cdev now only takes 3 parameters (dev_t, umode_t, fmt..), so
update this driver to the new API.

Time to update my CREDITS entry...

This patch updates the PowerMac cpufreq driver so that
it builds & works in current 2.6

This adds proper registration of CPUs on ppc32, without this, accesses
to cpufreq will oops.

[V4L] - set debug verbosity to 0 for both Hexium drivers
[V4L] - declare all local functions and variables static

[DVB] - Hand off all processing of urb data to a tasklet