stat64 has been changed to return jiffies granuality as nsec in previously
unused fields. This allows make to make better decisions on when
to recompile a file. Follows losely the Solaris API.

CURRENT_TIME has been redefined to return struct timespec.  The users
who don't use it in a inode/attr context have been changed to use a new
get_seconds() function.  CURRENT_TIME is implemented by an out-of-line
function.

There is a small performance penalty in this patch.  The previous
filemap code had an optimization to flush atime only once a second.
This is currently gone, which will increase flushes a bit.  I believe
the correct solution if it should be a problem is to have per super
block fields that give an arbitary atime flush granuality - so that you
can set it to be only flushed once a hour if you prefer that.  I will
work on that later in separate patches if the need should arise.

struct inode and the attr struct has been changed to store struct
timespec instead of time_t for [cma]time.  Not all file systems support
this granuality, but some like XFS,NFSv3,CIFS,JFS do.  The others will
currently truncate the nsec part on flushing to disk.  There was some
discussion on this rounding on l-k previously.  I went for simple
truncation because there is not much evidence IMHO that the more
complicated roundings have any advantages.  In practice application will
be rather unlikely to notice the rounding anyways - they can only see a
difference when an inode is flush from memory and reloaded in less than
a second, which is rather unlikely.

This patch will break some userspace monitoring apps in the name of
having sane disk statistics in 2.6.x.

Patch from Rick Lindsley <ricklind@us.ibm.com>

In 2.5.46, there are now disk statistics being collected twice: once
for gendisk/hd_struct, and once for dkstat.  They are collecting the
same thing.  This patch removes dkstat, which also had the disadvantage
of being limited by DK_MAX_MAJOR and DK_MAX_DISK. (Those #defines are
removed too.)

In addition, this patch removes disk statistics from /proc/stat since
they are now available via sysfs and there seems to have been a general
preference in previous discussions to "clean up" /proc/stat.  Too many
disks being reported in /proc/stat also caused buffer overflows when
trying to print out the data.

The code in led.c from the parisc architecture has not apparently been
recompiled under recent versions of 2.5, since it references
kstat.dk_drive which doesn't exist in later versions.  Accordingly,
I've added an #ifdef 0 and a comment to that code so that it may at
least compile, albeit without one feature -- a step up from its state
now.  If it is preferable to keep the broken code in, that patch may
easily be excised from below.

We're currently incrementing /proc/vmstat:pgalloc in front of the
per-cpu page queues, and incrementing /proc/vmstat:pgfree behind the
per-cpu queues.  So they get out of whack.

Change it so that we increment the counters each time someone requests
a page.  ie: they're both in front of the queues.

Also, remove a duplicated prep_new_page() call and as a consequence,
drop the whole additional list walk in rmqueue_bulk().

There was some strange code in the __getblk()/__find_get_block()/
__bread() area which was performimg multiple bh_lru_install() calls as
well as multiple touch_buffer() calls.

Fix all that up.  We only need to run bh_lru_install() and
touch_buffer() in __find_get_block().  Because if the block wasn't
found, __getblk() will create it and will re-run __find_get_block().

Also document a few things and make a couple of internal symbols static
to buffer.c

Also, don't run __find_get_block() from within
unmap_underlying_metadata().  We hardly expect to find that block
inside the LRU.  And we hardly expect to use it as metadata in the near
future so there's no point in letting it evict another buffer if we
found it.  So just go straight into the pagecache lookup for
unmap_underlying_metadata().

Patch from Lev Makhlis <mlev@despammed.com>

The disk accounting will overflow after 4,000,000 seconds.  Extend that
by a factor of 1000.

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch makes various private structures and procedures static.

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch removes hugetlb's intrusion into /proc/

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch removes hugetlb's intrusion into kernel/sysctl.c

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch internalizes hugetlb initialization, implementing a command-line
option in the process.

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch removes the unused function unlink_vma().

Patch from William Lee Irwin III <wli@holomorphy.com>

This patch eliminates zap_hugetlb_resources, along with its usages.
This actually fixes bugs, as zap_hugetlb_resources was itself buggy.

Patch from William Lee Irwin III <wli@holomorphy.com>

Idle time accounting is disturbed by the iowait statistics, for several
reasons:

(1) iowait time is not subdivided among cpus.
        The only way the distinction between idle time subtracted from
        cpus (in order to be accounted as iowait) can be made is by
        summing counters for a total and dividing the individual tick
        counters by the proportions. Any tick type resolution which is
        not properly per-cpu breaks this, meaning that cpus which are
        entirely idle, when any iowait is present on the system, will
        have all idle ticks accounted to iowait instead of true idle time.

(2) kstat_read_proc() misreports iowait time
        The idle tick counter is passed twice to the sprintf(), once
        in the idle tick position, and once in the iowait tick position.

(3) performance enhancement
        The O(1) scheduler was very carefully constructed to perform
        accesses only to localized cachelines whenever possible. The
        global counter violates one of its core design principles,
        and the localization of "most" accesses is in greater harmony
        with its overall design and provides (at the very least) a
        qualitative performance improvement wrt. cache.

The method of correcting this is simple: embed an atomic iowait counter
in the runqueues, find the runqueue being manipulated in io_schedule(),
increment its atomic counter prior to schedule(), and decrement it
after returning from schedule(), which is guaranteed to be the same one,
as the counter incremented is tracked as a variable local to the procedure.
Then simply sum to obtain a global iowait statistic.

(Atomicity is required as the post-wait decrement may occur on a different
cpu from the one owning the counter.)

io_schedule() and io_schedule_timeout() are moved to sched.c as they must
access the runqueues, which are private to sched.c, and nr_iowait() is
created in order to export the sum of all runqueues' nr_iowait().

A patch from Janet Morgan <janetmor@us.ibm.com>

If you feed an iovec with a bad address not at the zeroeth segment into
readv or writev, it returns the wrong value.

	iovec 1:  base is 8050b20 len is 64
	iovec 2:  base is ffffffff len is 64
	iovec 3:  base is 8050ba0 len is 64

The writev should return 64 bytes but is returning 128

This is because we've added the new segment's length into `count'
before running access_ok().

The patch changes it to fix that up on the slow path, if access_ok() fails.

- Revert the 3c59x.c compile warning fixes.  The return type of inl()
  was reverted back to the correct 32 bits.

- Fix an uninitialised timer in ext3 (JBD debug mode only) - run
  setup_ro_after() during initialisation.

- Fix ifdef/endif imbalance in JFS

   - CS4231 - added sparc support to merge sparc/cs4231.c code
   - ICE1712
     - added support for AK4529
     - added support for Midiman M-Audio Delta410
   - USB driver
     - fixed against newer USB API but allow compilation under 2.4

   - CS46xx driver
     - DSP is started after initializing AC97 codecs
     - rewrite SPDIF output stuff
     - variable period size support on playback and capture
     - DAC volume mechanism  rewrite
     - IEC958 input volume mechanism rewrite
     - added "AC3 Mode Switch" in mixer
     - code cleanups
   - ENS1371 driver
     - added definitions for the ES1373 chip
     - added code to control IEC958 (S/PDIF) channel status register

  - Moved initialization of card->id to card_register() function.
    The new default id is composed from the shortname given by driver.
  - ES18xx - Fixed power management defines
  - VIA82xx - The SG table is build inside hw_params (outside spinlock - memory allocation).

  - added kmalloc_nocheck and vmalloc_nocheck macros
  - PCM
    - the page callback returns 'struct page *'
    - fixed delay function (moved put_user call outside spinlock)
  - OSS PCM emulation
    - fixed read() lock when stream was terminated and no data is available
  - EMU8000
   - added 'can schedule' condition to snd_emu8000_write_wait()
  - AC'97
   - added ALC650 support
  - ALI5451
    - removed double free

DOH!!! Somebody clone me a replacement brain: I must have burnt
another fuse.  It turns out the new readpages was evading our
read/write serialization.

This broke things like 'ld' over NFS, which rewrites chunks of
files it has already written.

last direct call into fs code is gone

Sufficient locking for fifo_open is provided by the inode semaphore.

- fix loading of another configuration
- accept longer strings in configuration
- move conf_filename to mconf.c (it's the only user)
- fix off by one error during string scanning

I enhanced the sendfile method so that we could pass a proper actor to
it (which exposes the full power of the internal implementation).

Now knfsd calls the sendfile vector rather than depending on a
readpage() that hasn't been set up fully.

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

into atnf.csiro.au:/workaholix1/kernel/v2.5/rgooch-2.5

Sorry, one small goof...

Use new "nmi_enter/exit()" which acts the same as the regular irq
entries (increases the preempt count appropriately), but doesn't try to
start processing softirqs on nmi exit (it just decreases the count).

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

into samba.org:/scratch/anton/linux-2.5_ppc64_work

into samba.org:/scratch/anton/linux-2.5_ppc64_work

into samba.org:/scratch/anton/linux-2.5_ppc64_work

into samba.org:/home/anton/ppc64/for-linus-ppc64

into samba.org:/home/anton/ppc64/for-linus-ppc64

Given the previous set of patches that integrate NFS with the VM +
pdflush memory control, and add mechanisms to cope with low memory
conditions, the time is now ripe to rip out the 256 outstanding
request limit, as well as the associated LRU list in the superblock,
and the nfs_flushd daemon.

The following patch offers a 30% speed increase on my test setup with
512MB of core memory (iozone using 4 threads each writing a 512MB file
over 100Mbit to a Solaris server). Setting mem=64m, I still see a 2-3%
speed increase.