This function was recently converted to use rwsem locking.  But it is called
from interrupts in (at least) buffer_io_error().

And we do want a function like this to be robust and atomic.  So convert it
to use spinlocking.

Filesystems which are using generic_file_llseek() do not need lock_kernel()
in their readir implementations.  All operations (including llseek) are
serialised by the directory's i_sem.

Just fix ext2 and ext3 for now.  Others may need locking between readdir and
who-knows-what.

This function is called a lot.  Every brk().  The atomic_add() against a
global counter hurts on large SMP machines.

The patch simply reduces the rate at which that atomic operation is
performed, by accumulating a per-cpu count which is spilled into the global
counter when the local counter overflows.

It trades off efficiency for a little inaccuracy.

I tried various implementations involving kmalloc_percpu() and open-coded
per-cpu arrays in a generic "per-cpu counter" thing.  They all were
surprisingly sucky - the additional cache misses involved in walking the more
complex data structures really showed up.

vmtruncate() does not need lock_kernel().  And lock_kernel() is not taken by
other vmtruncate() callers.

Turn on MS_ONE_SECOND in ext2 and ext3.

For some filesystems (ext3, reiserfs at least), ->dirty_inode() is very
expensive.  The kernel is currently calling mark_inode_dirty() at up to 1000
times/sec/inode.  But there is no need to do this if the filesystem cannot
store high-resolution times on-disk.

This patch restores the optimisation of only dirtying the filesystem inode
when its on-disk representation has actually changed.

The filesystem will set the MS_ONE_SECOND flag in sb->s_flags to indicate
that it wishes to receive this treatment.

The patch does reduce the call rate to ext3_mark_inode_dirty() from 1000/sec
to 1/sec, but it doesn't make much difference at all to performance because
we're calling ext3_mark_inode_dirty() from other callsites as well.  Those
can be optimised too.

ppc64 support for file file-offset-in-pte

Path from Andi Kleen <ak@muc.de>

Add x86_64 support for file offsets in pte's.

filemap_populate() is currently doing page-at-a-time synchronous I/O.  Add a
call to do_page_cache_readahead() in there so we do a big slurp of IO first.

This is minimal - a lot of the filemap_populate() code can be
rationalised yet.

This patch requires arch support.  I have patches for ia32, ppc64 and x86_64.
Other architectures will break.  It is a five-minute fix.  See

	http://mail.nl.linux.org/linux-mm/2003-03/msg00174.html

for implementation details.


Patch from: Ingo Molnar <mingo@elte.hu>

the attached patch, against BK-curr, is a preparation to make
remap_file_pages() usable on swappable vmas as well.  When 'swapping out'
shared-named mappings the page offset is written into the pte.

it takes one bit from the swap-type bits, otherwise it does not change the
pte layout - so it should be easy to adapt any other architecture to this
change as well.  (this patch does not introduce the protection-bits-in-pte
approach used in my previous patch.)

On 32-bit pte sizes with an effective usable pte range of 29 bits, this
limits mmap()-able file size to 4096 * 2^29 == 2 TBs.  If the usable range is
smaller, then the maximum mmap() size is reduced as well.  The worst-case i
found (PPC) was 2 hw-reserved bits in the swap-case, which limits us to 1 TB
filesize.  Is there any other hw that has an even worse ratio of sw-usable
pte bits?

this mmap() limit can be eliminated by simply not converting the swapped out
pte to a file-pte, but clearning it and falling back to the linear mapping
upon swapin.  This puts the limit into remap_file_pages() alone, but i really
hope no-one wants to use remap_file_pages() on a 32-bit platform, on a larger
than 1-2 TB file.

sys_remap_file_pages() is now enforcing the 'prot' parameter to be zero.
This restriction might be lifted in the future - i really hope we can have
more flexible remapping once 64-bit platforms are commonplace - eg.  things
like memory debuggers could just use the permission bits directly, instead of
creating many small vmas.

i've tested swappable nonlinear ptes and they are swapped out/in
correctly.

some other changes in -A0 relative to 2.5.63-BK:

 - slightly smarter TLB flushing in install_page(). This is still only a
   stupid helper functions - a more efficient 'walk the pagecache tree
   and pagetable at once and use TLB-gather' implementation is preferred.

 - cleanup: pass on pgprot_t instead of unsigned long prot.

 - some sanity checks to make sure file_pte() rules are followed.

 - do not reduce the vma's default protection to PROT_NONE when using
   remap_file_pages() on it. With swappable ptes this is now safe.

This patch fixes a couple of bugs and compile errors in the powerbook
media bay driver.  It was getting initialized after the IDE subsystem,
whereas it needs to be initialized before so that the IDE subsystem
can see the CD-ROM drive in the bay.

This patch updates the mac53c94 scsi HBA driver, used on older
powermacs, to correspond with the recent scsi subsystem changes, to
use the PCI DMA API, to not panic, and to use a spinlock instead of
save_flags/restore_flags/cli/sti.

This patch updates the `mesh' scsi driver used on older powermacs to
correspond with recent changes in the scsi subsystem (things like
using cmd->device->id instead of cmd->target).

This patch forward-ports various fixes to the driver for the PMU
(power manager unit) on powermacs and powerbooks from 2.4, and in
particular, some improvements to the battery charge calculations.

From Ben Herrenschmidt.

The patch below removes the uses of save_flags/restore_flags/cli
etc. from the macserial driver and replaces them with a spinlock.

This patch updates the CUDA driver (the power/reset/ADB controller on
older powermacs) to fix some SMP issues and to match the 2.4 version
of the driver.

From Ben Herrenschmidt.

This patch updates include/linux/adb.h and include/linux/pmu.h with
some additional definitions that we need on powermacs and powerbooks.

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

into samba.org:/home/paulus/kernel/for-linus-ppc

into samba.org:/home/paulus/kernel/for-linus-ppc

Convert the pcnet_cs driver to use the new registration call.

Remove the linked list of pcmcia_drivers. It didn't even handle removal of a
driver properly, so it won't be missed all that much.

Register all pcmcia drivers with the pcmcia bus within the old
register_pccard_driver() function. Alternatively, a new
registration function "pcmcia_register_driver()" (and its
counterpart,  "pcmcia_unregister_driver()") can be used.

Register a bus_type pcmcia_bus_type. This means the initialization of
the ds module needs to be done in two levels: one quite early
(subsys_initcall) so that drivers may use the bus_type; the other one
must stay that late (late_initcall). As only one initcall can be
specified within one module, some tweaking is needed.

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

Since it gets 1 device right it wasnt hard to fix 8)