I noticed that our PPPoE/PPtP access concentrator leaks pty devices.
When all 4096 indices are leaked, there was need to reboot it.

The following patch fixes this problem.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

rcu_offline_cpu and rcu_move_batch have been broken since the list_head's
in struct rcu_head and struct rcu_data were replaced with singly-linked
lists:

  CC      kernel/rcupdate.o
kernel/rcupdate.c: In function `rcu_move_batch':
kernel/rcupdate.c:222: warning: passing arg 2 of `list_add_tail' from
incompatible pointer type
kernel/rcupdate.c: In function `rcu_offline_cpu':
kernel/rcupdate.c:239: warning: passing arg 1 of `rcu_move_batch' from
incompatible pointer type
kernel/rcupdate.c:240: warning: passing arg 1 of `rcu_move_batch' from
incompatible pointer type
kernel/rcupdate.c:236: warning: label `unlock' defined but not used

Kernel crashes when you try to offline a cpu, not surprisingly.

It also looks like rcu_move_batch isn't preempt-safe so I touched that up,
and got rid of an unused label in rcu_offline_cpu.
Signed-off-by: Nathan Lynch <nathanl@austin.ibm.com>
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

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

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

This patch fixes the error in the subject line.

Attached is a second patch to account for (most of) Herbert Xu's
comments.

I have left-out the part about changing state->card to a
local variable where it is used a lot.  Unfortunately, that usage is
somewhat pervasive and I would prefer to make those changes in a separate
patch -- after I have had a chance to do some testing.

If you'd prefer one patch to account for the original plus these
changes, let me know and I'll be happy to provide it.

Enclosed is a patch for the i810_audio OSS driver to support using
memory-mapped I/O for those chipsets that support it.

 o Added a family of macros -- I810_IOREADx() and I810_IOWRITEx() -- that
key off the existing card->use_mmio flag to select between using readx/writex
or inx/outx for I/O operations.

 o Converted existing inx/outx invocations to use
I810_IOREADx/I810_IOWRITEx instead.

 o Changed GET_CIV(), GET_LVI, and CIV_TO_LVI() not only to use
I810_IOREADx/I810_IOWRITEx but also to take "card" (i.e. struct i810_card)
paramter.

 o Removed check for "Pure MMIO interfaces" in i810_probe() -- replaced w/
(relocated) check for no I/O resources available.

The macros ata_id_has_lba() and ata_id_has_dma() seem to have their bits
reversed.  LBA support is bit 9 of word 49 in the identify page, whereas
DMA support is bit 8 of word 49 in the identify page.

I see we lost this source line in the changes since 2.6.7.  But if I try
SATAPI expecting no data without this source line, then my ata_piix.ko
promptly dies with such complaints as:

kernel: irq 18: nobody cared!

With this source line, I find I can expect no data, though I still have
to add auto sense to overcome:

kernel: ata2: BUG: timeout without command

This fixes several small but serious x86-64 bugs in 2.6.8rc2:

 - Fix array overflow in PCI bus checking (Travis Betak) 
 - Fix broken pci_map_sg in swiotlb (Suresh B. Siddha)
 - Remove bogus bus check in IOMMU code
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

* prism54_get/set_debug_oid are missing checks for a null pointer.
  Reported in Bugzilla number 100.

* In the card initialization routine, we try to set the
  output power. For firmware < 1.0.4.3, this leads to a
  worrying "mgt_commit has failed .." in the log although
  the device continues to react normally.
  Fix is simple, do not try to configure output power.
  (which I believe we should not be doing anyway as it is
   probably against local country regulations)

* Reintroduce pushing 0 into the TRDY_TIMEOUT and RETRY_TIMEOUT
  registers. Make this configurable with module parameter
  init_pcitm.

* We now have the ludicrous situation that some hardware setups
  require this (not even pushing 0xFF helps), whilst others
  don't care either way (the majority), and yet others bork
  if anything is pushed into these regs.
  If anybody can explain this (including Conexant :-) ), my ears
  are open.

* oid_mgt.c is calling islpci_mgt_transaction passing the address
  of a pointer to the management frame. This is not being
  initialized by the caller. The callee only updates this
  pointer when successful. When not, boom.

* Being ultracautious again, not only initialize in the caller,
  also null out the pointer unconditionally in the callee.

into tiger.hpl.hp.com:/data1/bk/lia64/to-linus-2.5

	Wrong ifdef around ppc nvram_read_byte() - it should be
CONFIG_GENERIC_NVRAM.

[that's it for today; there's more (fb stuff), but I'd rather make that
a separate series]
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

	compile fixes in open_pic.c - it missed cpumask conversion in several
places.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

	usual "shift definition of inlines before their uses" in a couple of
places in drivers/*
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

	a couple of "replace char with explicit s8 to make sure it's really
signed" fixes.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

	a) ppc KGDB builds only on two subarch; say so in Kconfig
	b) ide-pmac.c can't be a module; make it dependent on IDE=y.
	c) a bunch of isdn drivers is broken on ppc; marked as such
	d) oaknet is broken; marked as such (includes on files that do not
exist, etc.)
	e) buslogic and cyber2000 are broken on sparc64
	f) s3trio framebuffer is broken
	g) usual 8250-based serial is broken on sparc64; in principle it might
be worth supporting (there are PCI cards of that sort), for now marked as
broken (no asm/serial.h, to start with).
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Usual foo ? void : non-void  in reiserfs; gets triggered on big-endian
platforms.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

	asm-ppc/reg.h defines pollutes namespace like there's no tomorrow;
among other things there are
	LR
	SR0...SR15
	DEC
etc.

That breaks quite a few drivers that should otherwise work on ppc.  SR...
and LR were never used by ppc code; commented out in reg.h (note that
they are just aliases for SPRN_SR... and SPRN_LR).  RPA commented out and
its instances in ppc code (all four of them) replaced with SPRN_RPA.  Ditto
for DEC.  The rest of the offenders (DAR and PVR) are left alone for now and
#undef'ed in drivers in question.

We probably want to rip all these guys out of reg.h and convert ppc users
to SPRN_... forms - file ends up included in alot of places and namespace
pollution like that is a Bad Thing(tm).
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Assorted pointer-to-int fixes:
	a) some places want to take pointer modulo alignment or extract
integer that was cast to pointer (which is legitimate), but do that via
wrong cast, triggering sparse warnings.
	b) usual %x (int)ptr -> %p ptr fixes
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Either it got missed back when if_mii() was introduced, or it got merged
that way later.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

zoran procfs code switched to seq_file and cleaned up; a bunch of homegrown
procfs glue eliminated.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

cciss compat ioctls were broken (mixing kernel and userland pointers);
fixed (and cleaned up) by switch to compat_alloc_user_space() and elimination
of set_fs().
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Usual set - explicit -D<arch>=1 for sparse, __sighandler_t / __sigrestore_t
annotated, annotations in arch/ppc/kernel/*, etc. - same as for other platforms
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

CONFIG_MCOUNT on sparc64 is currently broken; we have both mcount and _mcount
declared (and aliased to each other), but end up exporting the wrong one.  As
the result, with that option all modules get unresolved symbol.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

it could be split (all chunks are obviously independent), but I don't think
that it's worth doing.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

blind dereferencing of userland pointer
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

0->NULL in arch/ppc and include/asm-ppc.

[That's it for this series - there's still ~200Kb to merge, but I'd rather
take a break for a couple of hours before dealing with the rest]
Signed-off-by: Linus Torvalds <torvalds@osdl.org>