into ppc970.osdl.org:/home/torvalds/v2.6/linux

The NFS readlink() methods all take a buffer length argument.  Use that
instead of assuming PAGE_SIZE... 

We need to return ENAMETOOLONG rather than EIO.

Make sure NFS client doesn't see errors from mode setting on new symlinks.

When nfsd creates a symlink, it tries to set the mode as the mode is 
carried in the NFS request and some filesystems store a mode.

If the filesystem refuses to set the mode (e.g. -EOPNOTSUPP), this
error should not be returned to the client.

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into nuts.davemloft.net:/disk1/BK/sparc-2.6

This adds support for the CP-104 Moxa Smartio serial cards.  Just add
the PCI ID information.

When the kmem_bufctl_t typedef got added to include/asm-ppc64/types.h,
it got added outside the #ifndef __ASSEMBLY__ section, causing
assembler errors.  This patch, from David Gibson, moves it inside the
#ifndef __ASSEMBLY__ region.

Add newlines to some printk debug trace of comempci.c driver.

Remove a lot of un-used and un-needed libgcc funstions from export list
for m68knommu syms.

Correct build lines for Motorola Dragonball 68x328 frame buffer driver. 

Patch from Georges Menie <georges@menie.org>

Remove ColdFire specific code sections for atomic_add and atomc_sub.

These are not needed, the m68k asm code for these functions is ColdFire
clean.

We should un-define all the x86 style IO routines when redefining local
versions.

Create init points for the Motorola Dragonball 68x328 frame
buffer driver.

Patch from Georges Menie <georges@menie.org>

A couple of fixups for asm-m68knommu/bitops.h:

. re-order definition of fls(), to be outside __KERNEL__
. add code for find_next_bit()

Big cleanup of the Motorola DragonBall 68x328 frame buffer.  It was
quite broken before. 

Patch from Georges Menie <georges@menie.org>.

Fix the cache flushing code for the ColdFire 5407 CPU.

The cpushl instruction arguments are wrong, causing it to miss some
cache lines.

Added a nommu version of sysctl_max_map_count.

Fix return value from do_munmap(), it should return 0 on success not
EINVAL.

From: Neil Brown <neilb@cse.unsw.edu.au>

Fix bug #2661

Raid currently calls ->unplug_fn under spin_lock_irqsave(), but unplug_fns
can sleep.


After a morning of scratching my head and trying to come up with some that
does less locking, the following is the best I can come up with.  I'm not
proud of it but it should work.

If I move "nr_pending" out or rdev into the per-personality structures
(e.g.  mirror_info), and if I had "atomic_inc_if_nonzero" I could do with
without locking so much, but random atomic* functions don't seem trivial

into ppc970.osdl.org:/home/torvalds/v2.6/linux

        spinlock_t *lp = &((struct netdev_private *)dev->priv)->lock;
doesn't mix with
	spin_unlock_irqrestore(&lp, flags);

From: David Mosberger <davidm@napali.hpl.hp.com>

From: Martin Diehl <lists@mdiehl.de>

There was a spin_unlock missing in the raw mode tx-completion path.  Probably
it slipped through because the raw mode stuff is never reached with my Actisys
hardware.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Network driver changes:
 - iucv: Make grab_param function SMP safe.
 - lcs: Fix null-pointer dereference after unsuccessful set_online.
 - qeth: Fix kmalloc flags in qeth_alloc_reply.
 - qeth: Show broadcase capability also in route4/6 sysfs attributes.
 - qeth: Remove debug code.
 - qeth: Add option to qetharp user space interface to strip unused
         fields from query arp records.
 - qeth: Add shortcut in outbound path for HiperSockets.
 - qeth: Add more info to qeth_perf_stats.
 - qeth: Add support for direct SNMP interface to OSA express cards.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

zfcp host adapter change:
 - Remove misplaced dot in error message.
 - Remove unused performance statistics code.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

dasd device driver changes:
 - Reset pointer from ccw device to dasd_devmap on device removal.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

s390 core changes:
 - Fix system call trace / audit interface.
 - Fix find_first_bit / find_next_bit inlines assembly constraints.

From: Mingming Cao <cmm@us.ibm.com>

There is a uninitialized goal value being referenced in both ext3 and ext2
find goal block functions (ext3_find_goal() and ext2_find_goal()).

In the non-sequential write case, these functions check the goal value(non
zero) before calling ext3(2)_find_near() to find the goal block to
allocate.

Since the goal value is uninitialized(non zero), the ext3(2)_find_near() is
never being called in the non-sequential write, thus ext3(2)_find_goal()
failed to guide a goal block in the random write case.

ext3(2)_new_block() takes the junk goal value and will turn it to goal 0
since it's normally beyond the filesystem block number limit.  The fix is
trivial.

From: Rusty Russell <rusty@rustcorp.com.au>

The IA64 hotplug CPU merge seems to have included some core changes: in
particular the recalc_bh_state() needs to sum for all (including offline)
cpus, since we don't empty the counters on CPU down.  The totals printed by
/proc/stat (the first loop) should include offline cpus, too (apparently
printing out the per-cpu lines for offline cpus confuses top).

The recent ARM-specific fix broke ia32.  Hopefully the ARM team can find an
arch-specific implementation of VGA_MAP_MEM() which makes it work.

From: Matt Mackall <mpm@selenic.com>

Doing the algebra:

c = (a - b) * 3/2
a' = a - c = a - 3/2(a - b) = (2a - 3a + 3b)/2 = (3b - a)/2
a' >= 0
3b - a >= 0
3b >= a
b >= a/3
nr_free_pages() >= mempages/3

We can indeed get into trouble if we try to load a large kernel on a very
small box (ie kernel reserves more than 2/3 of usable memory).  Surprisingly I
haven't hit this, but here's a fix.

From: "J. Bruce Fields" <bfields@fieldses.org>

svc_recv may call svc_sock_release before rqstp->rq_res is initialized.

From: NeilBrown <neilb@cse.unsw.edu.au>

From: "J. Bruce Fields" <bfields@fieldses.org>

To help the user diagnose problems caused by user-level daemons not running.

From: NeilBrown <neilb@cse.unsw.edu.au>

From: "J. Bruce Fields" <bfields@fieldses.org>

Currently we are counting the number of threads already asleep and returning
an immediate NFS4ERR_DELAY (==JUKEBOX) error if more than half are already
asleep.

This patch removes that logic, so instead we only return NFS4ERR_DELAY if an
upcall times out (if it takes more than a second to return).

With the thread counting there is the risk that even when all the relevant
subsystems are responsive, the client may still see occasional NFS4ERR_DELAY
returns just because, by coincidence, several upcalls were initiated at the
same time.  I expect clients will delay several seconds before retrying after
NFS4ERR_DELAY, so this will be quite noticeable to users.  Sporadic long
delays like this are likely to lead users to suspect a problem somewhere, when
in fact there is none.

The current scheme ensures that we can still process requests not depending on
upcalls, even when all threads would otherwise be tied up waiting on upcalls. 
However, this is not something that should happen under normal circumstances;
if a server spends a significant portion of its time with all threads waiting
for upcalls, this a sign that something is seriously wrong.

In such a circumstance (e.g., an ldap server dies), we can, at least, bound
the waiting time to a second without the need for counting threads.

In short, removing the thread-counting will allow us to behave predictably
when things are working, while still allowing some progress when they don't.

It would be a worthwhile project to measure the amount of time threads spend
waiting for upcalls (or for reads, for that matter); if a significant portion
of the time they spend handling requests is spent sleeping, then there's an
opportunity to improve nfsd performance: if we can break the one-to-one
mapping between requests and threads, then we can lower the number of threads
required to keep the nfs server busy.

However, both the currently available options for doing this are problematic:
returning JUKEBOX/DELAY errors at random times will lead to unpredictable
performance, and saving a copy of the request to be processed from scratch
again later is wasteful and makes it difficult to provide correct semantics,
especially in the NFSv4 case.

So for now I believe waits with short timeouts are the best option.

From: NeilBrown <neilb@cse.unsw.edu.au>

From: "J. Bruce Fields" <bfields@fieldses.org>

1 second should be plenty of time; if we're going to take longer than that
it's probably better just to return NFS4ERR_DELAY and let the client retry
anyway.