The packet scheduler simulates 64 bit jiffies on 32 bit platforms by running
a timer keeping a mark and and offset.  Since there is no locking and this is
racy and doesn't handle jiffie wrap real well.

We can use get_jiffies_64 on 2.6 do get what is needed.
The downside is the overhead of a function call, and a cache miss in
get_jiffies_64.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

The last argument in the PSCHED_TDIFF_SAFE is no longer used;
only usage eliminated by previous patch. It gets rid of a bad macro
usage.

Also, can use the standard min_t macro which also eliminates the
macro problem of double evaluation of bound.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

This code in the csz scheduler, is just plain broken.  The TDIFF_SAFE 
effectively expands to:
	unsigned long delay = now - q->t_c;
	if (delay > 0) {
		delay = 0;
		goto do_reset;
	}
	if (delay >> q->delta_log)

So delay is always 0!  I assume that what was originally intended
is the to keep delay bounded to 1<<q->delta_log. 
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

The packet scheduling code has some ugly define's which were to deal with
configuration possibilities and the old style module exports.  With the current
2.6 method, this is unnecessary.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

When CONFIG_NET_CLS_ACT is not set 'result' variable in prio_classify is
unused. Also I was looking over the rest of the module and had hard time
understanding the logic in prio_enqueue - I rearranged it a bit for better
readability. Plus there are some formatting changes.
Signed-off-by: Dmitry Torokhov <dtor_core@ameritech.net>
Signed-off-by: David S. Miller <davem@redhat.com>

Another patchlet after turning on preempt and rerunning tests.
- against dev.c is a genuine bug - i wish preempt would have
given a more sane hint. Strange on how its a non issue when preempt is
off. 
- against act_api.c is just an annoyance. I removed the kmod because it
will only make sense to list actions if policies exist for that action.
And if policies exist, the module would be loaded.
Signed-off-by: Jamal Hadi Salim <hadi@zynx.com>
Signed-off-by: David S. Miller <davem@redhat.com>

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

I noticed that fs/buffer.c::drop_buffers() contains some code that
AFAICS doesn't actually do anything other than waste cpu cycles so here
is patch to remove it...  The local variable was_uptodate is being
messed with but it is not being read anywhere so it seems entirely
pointless.

I assume this must be a remainder from old code which mucked around with
the page uptodateness but which has since been (re-)moved.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

efi.h declares a function type, and then uses that as an argument to
another function, and expects the compiler to magically demote the
function to a function pointer.

Even a gcc person (rth) was surprised that this was legal, and it
doesn't match any other use of a function pointer in the kernel, and
sparse doesn't like the implicit type-conversion.

So make the type sane in the first place, instead of depending on
a very weird corner case of the C language.

This fixes a bug in the ppc64 memset where the code that gets the
destination address aligned (or is supposed to) was looking at the
bottom 3 bits of the count rather than the destination address.  The
result of this was that the kernel wouldn't boot on POWER3 machines. 
The patch also removes an unnecessary duplicate instruction.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

Verify at least a full header is there before derefencing
length struct member.

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

There are currently two files besides mm/vmalloc.c that make use of that
function:

- arch/sh/kernel/cpu/sh4/sq.c
  which defined its own prototype locally risking not being in sync with
  the real function, and

- arch/arm/kernel/module.c which has no prototype at all and cause
  build warnings.

This fixes those issues

remove extraneous security_inode_setattr call in hugetlbfs, it's already
done by VFS.
Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Al's current changes to struct nameidata broke nfsroot for my discless
clients (oops in nfs_fill_super).  The patch below fixes this problem
for me.

Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

char can be either signed or unsigned, depending on the target system.
Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: David S. Miller <davem@redhat.com>

sizeof(struct kiocb) is dangerously large for a structure commonly
allocated on-stack. This patch converts the 24*sizeof(long) field,
->private, to a void pointer for use by file_operations entrypoints.
A ->dtor() method is added to the kiocb in order to support the release
of dynamically allocated structures referred to by ->private.

The sole in-tree users of ->private are async network read/write,
which are not, in fact, async, and so need not handle preallocated
->private as they would need to if ->ki_retry were ever used. The sole
truly async operations are direct IO pread()/pwrite() which do not
now use ->ki_retry(). All they would need to do in that case is to
check for ->private already being allocated for async kiocbs.

This rips 88B off the stack on 32-bit in the common case.
Signed-off-by: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: David S. Miller <davem@redhat.com>

CAP_DAC_OVERRIDE fails to give search permission in a directory with
mode 0.

Patch acked by Chris Wright <chrisw@osdl.org>
Signed-off-by: Andries Brouwer <aeb@cwi.nl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

There's no reason to keep files that
a) nobody #include's
b) produce #error 
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

For each node there are a defined list of MAX_NR_ZONES zones.  These are
selected as a result of the __GFP_DMA and __GFP_HIGHMEM zone modifier flags
being passed to the memory allocator as part of the GFP mask.  Each node
has a set of zone lists, node_zonelists, which defines the list and order
of zones to scan for each flag combination.  When initialising these lists
we iterate over modifier combinations 0 ..  MAX_NR_ZONES.  However, this is
only correct when there are at most ZONES_SHIFT flags.  If another flag is
introduced zonelists for it would not be initialised.

This patch introduces GFP_ZONETYPES (based on GFP_ZONEMASK) as a bound for
the number of modifier combinations.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Like a few other drivers the vc driver is doing unlocked careless
references to file offsets.  Its got kind of 2.2 locking that hasnt been
updated in the lseek function so someone at least tried. 

Fortunately we have a real lock for this and can just make lseek use that
too.

The 'assume 64bit load is atomic' bug seems to be very widely spread

akpm@osdl.org:

   Move declarations to vt_kern.h
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Here's the accel capabilities patch for radeonfb.  It updates radeonfb to
advertise its acceleration capabilities via fbinfo.flags.  I've tested this
on my box, and it gives me a nice fast console.

defect: "$ fbset -accel 0" doesn't work for radeonfb -- disabling accel
will only work from the kernel command line :-/

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: David Eger <eger@havoc.gtf.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This has been reported a couple of times and is consistently causing some
folks grief, so Urban, would you mind terribly if i send this patch to at
least clear current bug reports.  If there is additional stuff you want
ontop of this let me know and i can send a follow up patch.

The bug is that at times we haven't completed setting up the smb_ops so we
have a temporary 'null' ops in place until the connection is completely up.
 With this setup it's possible to hit ->readdir() whilst the null ops are
still in place, so we put the process to sleep until the connection setup
is complete and then call the real ->readdir().

This patch addresses the bugzilla report at
http://bugzilla.kernel.org/show_bug.cgi?id=1671Signed-off-by: Zwane Mwaikambo <zwane@linuxpower.ca>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

- Fix the !CONFIG_FRAME_POINTER && !CONFIG_KALLSYMS output formatting.

- Make print_context_stack() static
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

If a fault in the kernel leads to an unexpected protection fault whilst in
a code path which holds mmap_sem we will deadlock in do_page_fault() while
trying to classify the fault.  By carefully testing the source of the fault
we can detect and OOPS on the vast majority of these, greatly enhancing
diagnosis of such bugs.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

We need setup.h for COMMAND_LINE_SIZE.
Signed-off-by: Joseph Fannin <jhf@rivenstone.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

A microsecond is 1 millionth of a second not
1 thousandth of a second.

Noticed by Markus Mottl <markus@oefai.at>
Signed-off-by: David S. Miller <davem@redhat.com>

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: David S. Miller <davem@redhat.com>

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

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

Also uninline pfn_to_page and page_to_pfn.
Struct page is now 8 bytes smaller.

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

Jeremy Katz of Red Hat reported that his iSeries machine would not boot
with 2.6.7 based kernels.

It appears that with the inclusion of Paul Mackerras' patch "Optimize
exception/syscall entry/exit" a small previous patch got reverted.  Here
is that patch again.  The lack of this patch does not stop all iSeries
machines booting, but it does stop some.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

affs is failing to propagate the inode_setattr() return value.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>