/home/lenb/src/to-linus branch 'acpi-2.6.12'

[ACPI] suspend/resume ACPI PCI Interrupt Links

Add reference count and disable ACPI PCI Interrupt Link
when no device still uses it.

Warn when drivers have not released Link at suspend time.

http://bugzilla.kernel.org/show_bug.cgi?id=3469

Signed-off-by: David Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] delete boot-time printk()s from processor_idle.c

http://bugzilla.kernel.org/show_bug.cgi?id=4401

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] fix resume issues on Asus L5D

http://bugzilla.kernel.org/show_bug.cgi?id=4416

Signed-off-by: Rafael J. Wysocki <rjwysocki@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] Always set P-state on initialization

Otherwise a platform that supports ACPI based cpufreq
and boots up at lowest possible speed could stay there
forever.  This because the governor may request max speed,
but the code doesn't update if there is no change in
speed, and it assumed the initial state of max speed.

http://bugzilla.kernel.org/show_bug.cgi?id=4634

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] Add "ec_polling" boot option

EC burst mode benefits many machines, some of
them significantly.  However, our current
implementation fails on some machines such
as Rafael's Asus L5D.

This patch restores the alternative EC polling code,
which can be enabled at boot time via "ec_polling"

http://bugzilla.kernel.org/show_bug.cgi?id=4665

Signed-off-by: Luming Yu <luming.yu@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] address boot-freeze with updated DMI blacklist for c-states

http://bugzilla.kernel.org/show_bug.cgi?id=4763

Signed-off-by: David Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] Fix memset arguments in acpi processor_idle.c

http://bugzilla.kernel.org/show_bug.cgi?id=4954

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[ACPI] Fix the regression with c1_default_handler on some systems
where C-states come from FADT.

Thanks to Kevin Radloff for identifying the issue and
isolating it to exact line of code that is causing the issue.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

[PATCH] More qla2xxx configuration fixes

This adds the appropriate FW_LOADER pre-requisite and a separate entry
for ISP24xx support.

Thanks to Adrian Bunk and Jesper Juhl for their efforts in fixing this
quirk.

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

[IB/ucm]: Clean up userspace CM

Only print debug messages when debug_level is set.
Eliminate NULL checks prior to calling kfree.

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Signed-off-by: Libor Michalek <libor@topspin.com>

Merge /scratch/Ksrc/linux-git/

[PATCH] new alpha syscalls

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

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

Fix up qla2xxx configuration bogosity

If we haven't configured the qla24xx driver, then the Makefile shouldn't
do it for us.

This also means that we can avoid the unnecessary selection of FC_ATTRS.

Debugged by James Bottomley

[PATCH] Avoid device suspend on reboot

My fairly ordinary x86 test box gets stuck during reboot on the
wait_for_completion() in ide_do_drive_cmd():

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

Merge head 'for-linus' of /linux/kernel/git/shaggy/jfs-2.6

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge /pub/scm/linux/kernel/git/aegl/linux-2.6

Merge kernel.org:/home/rmk/linux-2.6-arm-smp

[PATCH] ppc32: 8xx remove BROKEN Kconfig entry

8xx is not BROKEN anymore.

Signed-off-by: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc32: 8xx avoid icbi misbehaviour in __flush_dcache_icache_phys

On 8xx, in the case where a pagefault happens for a process who's not
the owner of the vma in question (ptrace for instance), the flush
operation is performed via the physical address.

Unfortunately, that results in a strange, unexplainable "icbi"
instruction fault, most likely due to a CPU bug (see oops below).

Avoid that by flushing the page via its kernel virtual address.

Oops: kernel access of bad area, sig: 11 [#2]
NIP: C000543C LR: C000B060 SP: C0F35DF0 REGS: c0f35d40 TRAP: 0300 Not tainted
MSR: 00009022 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 10
DAR: 00000010, DSISR: C2000000
TASK = c0ea8430[761] 'gdbserver' THREAD: c0f34000
Last syscall: 26
GPR00: 00009022 C0F35DF0 C0EA8430 00F59000 00000100 FFFFFFFF 00F58000
00000001
GPR08: C021DAEF C0270000 00009032 C0270000 22044024 10025428 01000800
00000001
GPR16: 007FFF3F 00000001 00000000 7FBC6AC0 00F61022 00000001 C0839300
C01E0000
GPR24: 00CD0889 C082F568 3000AC18 C02A7A00 C0EA15C8 00F588A9 C02ACB00
C02ACB00
NIP [c000543c] __flush_dcache_icache_phys+0x38/0x54
LR [c000b060] flush_dcache_icache_page+0x20/0x30
Call trace:
[c000b154] update_mmu_cache+0x7c/0xa4
[c005ae98] do_wp_page+0x460/0x5ec
[c005c8a0] handle_mm_fault+0x7cc/0x91c
[c005ccec] get_user_pages+0x2fc/0x65c
[c0027104] access_process_vm+0x9c/0x1d4
[c00076e0] sys_ptrace+0x240/0x4a4
[c0002bd0] ret_from_syscall+0x0/0x44

Signed-off-by: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Update CREDITS entry and listings in source files for Jesper Juhl

a) update entry in CREDITS for Jesper Juhl
b) remove email address from source files so it's only listed in credits.

Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] clean up inline static vs static inline

`gcc -W' likes to complain if the static keyword is not at the beginning of
the declaration.  This patch fixes all remaining occurrences of "inline
static" up with "static inline" in the entire kernel tree (140 occurrences in
47 files).

While making this change I came across a few lines with trailing whitespace
that I also fixed up, I have also added or removed a blank line or two here
and there, but there are no functional changes in the patch.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbdev: update info->cmap when setting cmap from user-/kernelspace.

The fb_info struct, as defined in include/linux/fb.h, contains an element
that is supposed to hold the current color map:
  struct fb_cmap cmap;            /* Current cmap */

This cmap is currently never updated when either fb_set_cmap() or
fb_set_user_cmap() are called.  As a result, info->cmap contains the
default cmap that was set by a device driver/fbcon and a userspace
application using the FBIOGETCMAP ioctl will not always get the *currently*
used color map.

The patch fixes this by making sure the cmap is copied to info->cmap after
it is set correctly.  It moves most of the code that is responsible for
setting the cmap to fb_set_cmap() and out of fb_set_user_cmap() to avoid
code-duplication.

Signed-off-by: Michal Januszewski <spock@gentoo.org>
Cc: <linux-fbdev-devel@lists.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbcon: don't repaint the cursor when it is disabled.

Currently even when the cursor is disabled (`setterm -cursor off`), it is
still repainted as a black rectangle the size of a single char.  This can
be seen, for example, by chvt'ing to a free tty, disabling the cursor and
doing `dd if=3D/dev/urandom of=3D/dev/fb0`.

The patch changes this behaviour by avoiding painting anything when the
cursor is disabled.

Signed-off-by: Michal Januszewski <spock@gentoo.org>
Cc: <linux-fbdev-devel@lists.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbdev: colormap fixes

Color maps have up to 256 entries.  4096/256 allows for 16 characters per
line.  The format for a cmap entry is "%02x%c%4x%4x%4x\n" %02x entry %c
transp %4x red %4x blue %4x green

You can read the color_map with cat fb0/color_map.

Signed-off-by: Jon Smirl <jonsmirl@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] radeonfb: clean up EDID sysfs attribute

radeonfb does not clean up EDID sysfs attribute

Signed-off-by: Jon Smirl <jonsmirl@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbmem: use unregister_chrdev() on unload

fbdev is missing unregister_chrdev() on unload.

Signed-off-by: Jon Smirl <jonsmirl@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbmon: horizontal frequency rounding fix

Fix rounding error when mode frequency is very close to monitor limit

Signed-off-by: Jon Smirl <jonsmirl@gmail.com>
Acked-by: James Simmons <jsimmons@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] cpm_uart: use DPRAM for early console

m8xx_cpm_hostalloc() can't rely on using the coherent DMA allocator early
on boot because the VM is not fully up yet.  Change it to use the on-board
DPRAM instead.

The current code relies on the "bootmem_page" allocated by
m8xx_cpm_reset(), which must be killed.

This is done in v2.4 but has never been forward ported to v2.6.

Signed-off-by: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Cc: Pantelis Antoniou <panto@intracom.gr>
Cc: Kumar Gala <kumar.gala@freescale.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] dvb: rename lgdt3302 frontend module to lgdt330x

Rename lgdt3302 to lgdt330x, to make way for the addition of lgdt3303
support in future revisions.

I am changing the name of this module now so that hopefully the name will
be changed before the release of 2.6.13 ...  It wouldn't make sense to
release 2.6.13 with the name lgdt3302 in it, which will only be renamed to
lgdt330x in later versions.

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v4l: fix tuning with MXB driver

I noticed that some past changes to the gerneric Video4Linux tuner module
for analog tuners broke my "Multimedia eXtension Board" driver.

The tuner driver was made aware of Video4Linux2 tuning ioctls, but my
driver was not ported and still uses the Video4Linux1 ioctls.  This does
not work anymore as intendend, the tuning is currently broken.

The attached patch fixes non-working tuning in MXB driver introduced by
some recent generic tuner changes by replacing Video4Linux1 tuner ioctls
with proper Video4Linux2 tuner ioctls.

- fix non-working tuning in MXB driver introduced by some recent generic
  tuner changes by replacing Video4Linux1 tuner ioctls with proper
  Video4Linux2 tuner ioctls

Signed-off-by: Michael Hunold <hunold@linuxtv.org>
Cc: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v4l: hybrid dvb: rename CFLAGS from CONFIG_DVB_xxxx back to original HAVE_xxxx

The #define CONFIG_DVB_* are actually CFLAGS set by Makefile.  CONFIG_*
namespace is reserved for Kconfig.  This renames them back to HAVE_*

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v4l: hybrid dvb: move #defines to Makefile

This patch moves #define from cx88-dvb.c and saa7134-dvb.c into Makefile as
CFLAGS, allowing code compatability with video4linux cvs.

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v4l: hybrid dvb: fix warnings with -Wundef

This patch adds a missing #ifdef to saa7134-dvb.c (thanks to Mauro Carvalho
Chehab) and changes #if to #ifdef in both files.

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] dvb/v4l: cx88 cleanup

Remove unneeded comment.

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] lgdt3302: warning fix

warning: `i2c_readbytes' defined but not used

This code will either be re-enabled or deleted in a future patch.

Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] dvb/4vl: RF input selection fir

Select the RF input connector based upon the type of demodulation selected.
 ANT RF connector is selected for 8-VSB and CABLE RF connector is selected
for QAM64/QAM256.  This only affects the cards that use the Microtune 4042
tuner.

Signed-off-by: Mac Michaels <wmichaels1@earthlink.net>
Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] dvb/v4l: lgdt3302: isolate tuner

Remove the dvb_pll_desc from the frontend and replace with a
pll_set-callback to isolate the tuner programming from the frontend.

Signed-off-by: Mac Michaels <wmichaels1@earthlink.net>
Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v4l: fix regression modprobe bttv freezes the computer

Remove redundant bttv_reset_audio() which caused the computer to freeze
with some bt8xx based DVB cards when loading the bttv driver.

Signed-off-by: Johannes Stezenbach <js@linuxtv.org>
Signed-off-by: Michael Krufky <mkrufky@m1k.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] VIDEO_SAA7134 must depend on SOUND

VIDEO_SAA7134=y and SOUND=n results in the following compile error:

  LD      .tmp_vmlinux1
drivers/built-in.o(.text+0x4fafcb): In function `saa7134_initdev':
: undefined reference to `unregister_sound_dsp'
drivers/built-in.o(.text+0x4fb141): In function `saa7134_initdev':
: undefined reference to `register_sound_dsp'
drivers/built-in.o(.text+0x4fb17c): In function `saa7134_initdev':
: undefined reference to `register_sound_mixer'
drivers/built-in.o(.text+0x4fb339): In function `saa7134_finidev':
: undefined reference to `unregister_sound_mixer'
drivers/built-in.o(.text+0x4fb341): In function `saa7134_finidev':
: undefined reference to `unregister_sound_dsp'
make: *** [.tmp_vmlinux1] Error 1

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Mauro Carvalho Chehab <mchehab@brturbo.com.br>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] drivers/media/video/tveeprom.c: possible cleanups

This patch contains the following possible cleanups:
- make two needlessly global structs static
- #if 0 the EXPORT_SYMBOL'ed but unused function tveeprom_dump

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] DVICO Fusion DVB-T1 Tuner (LG-Z201) fix

It is a small modification to the table that defines the way that the
LG-Z201 tuner is controlled for the DVICO Fusion DVB-T1 tuner card.

I believe that a mistake was made when the dvb tuner code was reorganised
(to use a generic table for the tuner information instead of inline code)
and as a result, the DVICO card doesn't tune properly.

The modification I have made to the table makes it behave like it did with
the old inline tuner code that worked.  The patch is on top of the 2.6.12
kernel.

Signed-off-by: Gregory B Frost <frosts1@hotkey.net.au>
Signed-off-by: Johannes Stezenbach <js@linuxtv.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add core locking documentation to Infiniband

Add core locking documentation to Infiniband

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Eliminate sparse warnings in SA client

Eliminate sparse warnings in SA client

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Hook up userspace CM to the make system

Hook up userspace CM to the make system

Signed-off-by: Libor Michalek <libor@topspin.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add kernel portion of user CM implementation (fix)

Include the patch openib-general changing class_simple to class.

Signed-off-by: Tom Duffy <tduffy@sun.com>
Cc: Hal Rosenstock <halr@voltaire.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add kernel portion of user CM implementation

Add kernel portion of user CM implementation

Signed-off-by: Libor Michalek <libor@topspin.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add the header file for user space CM

Add the header file for user space CM.  This file defines the ABI used by the
CM for kernel/user communication.

Signed-off-by: Libor Michalek <libor@topspin.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Implementation for RMPP support in user MAD

Implementation for RMPP support in user MAD

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: User MAD ABI changes to support RMPP

User MAD ABI changes to support RMPP

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add the kernel CM implementation

Add the kernel CM implementation

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add the header file for kernel CM (Communications Manager)

Add the header file for kernel CM (Communications Manager)

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add Service Record support to SA client

Add Service Record support to SA client

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add RMPP implementation

Add RMPP implementation.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Introduce RMPP APIs

Introduce RMPP APIs

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: A couple of IB core bug fixes

Replace be32_to_cpup with be32_to_cpu and fix bug referencing pointer rather
than value in ib_create_ah_from_wc().

Signed-off-by: Tom Duffy <tduffy@sun.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add ib_create_ah_from_wc to IB verbs

Added new call: ib_create_ah_from_wc.  Call will allocate an address handle
given work completion information, including any received GRH.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Fix a couple of MAD code paths

Fixed locking to handle error posting MAD send work requests.  Fixed handling
canceling a MAD with an active work request.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Optimize canceling a MAD

Optimize canceling a MAD.
- Eliminate searching timeout list in cancel case.
- Remove duplicate calls to queue work item.
- Eliminate resending a MAD before MAD is completed.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add ib_modify_mad API to MAD

Add new MAD layer call to modify (ib_modify_mad) the timeout of a sent MAD,
and simplify cancel code.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Eliminate MAD cache leak associated with local completions

Eliminate MAD cache leak associated with local completions.  Also, when
canceling MAD, empty local completion list as well.

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Simplify calling of list_del in MAD

Simplify calling of list_del.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add automatic retries to MAD layer

Add automatic retries to MAD layer.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add ib_coalesce_recv_mad to MAD

Add implementation for ib_coalesce_recv_mad.  Also, clear allocated MAD data
buffer in ib_create_send_mad.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Minor cleanup during MAD startup and shutdown

Minor cleanup during startup and shutdown

Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Fix timeout/cancelled MAD handling

Fixes an issue processing a sent MAD after it has timed out or been canceled.
The race occurs when a response MAD matches with the send request.  The
request could time out or be canceled after the response MAD matches with the
request, but before the request completion can be processed.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Change ib_mad_send_wr_private struct

Have ib_mad_send_wr_private reference the private agent structure directly,
rather than the exposed agent definition.  Remove unneeded parameters to
functions and simplify code were possible from this change.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Change saving of user's send wr_id in MAD

Move saving of user's send wr_id to better match layering of received response
handling.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Combine some MAD routines

Combine response_mad() and solicited_mad() routines into a single function and
simplify/encapsulate its usage.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Add MAD helper functions

Add new helper routines for allocating MADs for sending and formatting a send
WR.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Update MAD client API

Automatically allocate a MR when registering a MAD agent.
MAD clients are modified to use this updated API.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] IB: Update FMR functions

Change some functions to return void rather than an int since they are always
returning 0, thus making checking return values rather pointless.

Signed-off-by: Tom Duffy <tduffy@sun.com>
Signed-off-by: Libor Michalek <libor@topspin.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add text for dealing with "dot releases" to README

The emergence of so-called "dot releases" that are non-incremental patches
against a base kernel requires different handling of patches (revert
previous patches before applying the newest one).  This patch adds a
paragrach to $TOPDIR/README explaining how to do deal with dot release
patches.

Signed-off-by: Kurt Wall <kwall@kurtwerks.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] serial_core whitespace fix

Use tabs for formatting like anywhere else in this file.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] sys_get_thread_area does not clear the returned argument

sys_get_thread_area does not memset to 0 its struct user_desc info before
copying it to user space...  since sizeof(struct user_desc) is 16 while the
actual datas which are filled are only 12 bytes + 9 bits (across the
bitfields), there is a (small) information leak.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] turn many #if $undefined_string into #ifdef $undefined_string

turn many #if $undefined_string into #ifdef $undefined_string to fix some
warnings after -Wno-def was added to global CFLAGS

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] reiserfs doesn't use mbcache

reiserfs doesn't use the mbcache, so this can go.

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

[PATCH] Documentation/Changes: document the required udev version

Document that udev 058 is required.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mbcache: Remove unused mb_cache_shrink parameter

The cache parameter to mb_cache_shrink isn't used.  We may as well remove
it.

Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc64: tpm_infineon build fix

ppc64 uses symbol `DAR', as does the TPM driver, causing a build failure.
Change the TPM name.

Cc: Marcel Selhorst <selhorst@crypto.rub.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] tpm: Support for Infineon TPM

This patch provides a new device driver for the Infineon SLD 9630 TT Trusted
Platform Module (TPM 1.1b) [1] which is embedded on Intel- mainboards or in
HP/ Fujitsu-Siemens / Toshiba-Notebooks.  A nearly complete list where this
module is integrated in can be found in [2].

This kernel module acts as a communication gateway between the linux kernel
and the hardware chip and fits the TPM-specific interfaces created by IBM in
drivers/char/tpm/tpm.h

Further information about this module and a list of succesfully tested and
therefore supported hardware can be found at our project page [3].

[1]
http://www.infineon.com/cgi/ecrm.dll/ecrm/scripts/public_download.jsp?oid=114135&parent_oid=29049

[2]
http://www.tonymcfadden.net/tpmvendors.htm

[3]
http://www.prosec.rub.de/tpm

Signed-off-by: Marcel Selhorst <selhorst@crypto.rub.de>
Acked-by: Kylene Jo Hall <kjhall@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] kernel/crash_dump.c: add kerneldoc

Add kerneldoc to kernel/crash_dump.c

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] kernel/cpuset.c: add kerneldoc, fix typos

Add kerneldoc to kernel/cpuset.c

Fix cpuset typos in init/Kconfig

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Acked-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] kernel/capability.c: add kerneldoc

Add kerneldoc to kernel/capability.c

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] stale POSIX lock handling

I believe that there is a problem with the handling of POSIX locks, which
the attached patch should address.

The problem appears to be a race between fcntl(2) and close(2).  A
multithreaded application could close a file descriptor at the same time as
it is trying to acquire a lock using the same file descriptor.  I would
suggest that that multithreaded application is not providing the proper
synchronization for itself, but the OS should still behave correctly.

SUS3 (Single UNIX Specification Version 3, read: POSIX) indicates that when
a file descriptor is closed, that all POSIX locks on the file, owned by the
process which closed the file descriptor, should be released.

The trick here is when those locks are released.  The current code releases
all locks which exist when close is processing, but any locks in progress
are handled when the last reference to the open file is released.

There are three cases to consider.

One is the simple case, a multithreaded (mt) process has a file open and
races to close it and acquire a lock on it.  In this case, the close will
release one reference to the open file and when the fcntl is done, it will
release the other reference.  For this situation, no locks should exist on
the file when both the close and fcntl operations are done.  The current
system will handle this case because the last reference to the open file is
being released.

The second case is when the mt process has dup(2)'d the file descriptor.
The close will release one reference to the file and the fcntl, when done,
will release another, but there will still be at least one more reference
to the open file.  One could argue that the existence of a lock on the file
after the close has completed is okay, because it was acquired after the
close operation and there is still a way for the application to release the
lock on the file, using an existing file descriptor.

The third case is when the mt process has forked, after opening the file
and either before or after becoming an mt process.  In this case, each
process would hold a reference to the open file.  For each process, this
degenerates to first case above.  However, the lock continues to exist
until both processes have released their references to the open file.  This
lock could block other lock requests.

The changes to release the lock when the last reference to the open file
aren't quite right because they would allow the lock to exist as long as
there was a reference to the open file.  This is too long.

The new proposed solution is to add support in the fcntl code path to
detect a race with close and then to release the lock which was just
acquired when such as race is detected.  This causes locks to be released
in a timely fashion and for the system to conform to the POSIX semantic
specification.

This was tested by instrumenting a kernel to detect the handling locks and
then running a program which generates case #3 above.  A dangling lock
could be reliably generated.  When the changes to detect the close/fcntl
race were added, a dangling lock could no longer be generated.

Cc: Matthew Wilcox <willy@debian.org>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: use __cpcmd in vmcp_write

vmcp_write uses GPF_DMA for the memory allocation of the response buffer, so
it can use the low level function __cpcmd directly, no need to call the
wrapper.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: cpu timer reset in machine check handler

Fix wrong move direction of timer values for cpu accounting in case of a
machine check that indicates a broken cpu timer.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: 31 bit memory size limit

Limit reported memory size to 2GB if running in 31 bit mode.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: channel tape fixes

Tape driver fixes:
 - Added deferred condition handling to tape driver core.
 - Added ability to handle busy conditions.
 - Code cleanup.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: free dasd slab cache

Free dasd slab cache on module unload.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: fba dasd i/o errors

The FBA discipline does not use retries for failed requests.  A request fails
after the first unsuccessful start attempt.  There are some rare conditions
(e.g.  CIO path recovery) in which the start of an i/o on a fba device can
fail.  A tiny amount of retries is therefore reasonable.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: resource accessibility event handling

When processing resource accessibility events, continue searching for further
affected subchannels if a link address is provided in the event information.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: debug data for ifcc/ccc

Fix debug data in case of an interface-control or channel-control check: don't
log the not yet accumulated interrupt-response-block, but the one we just
received.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: external call performance

The kernel uses the SIGP external call order code to signal other CPUs.  When
running with dedicated CPUs external calls don't get delivered immediately but
within a fixed polling invervall.  This can lead to delays where the system
appears to do nothing.  Replace the SIGP external call order with the SIGP
emergency call order since this one gets delivered immediately.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: atomic64 inline functions

The atomic64 primitives are supposed to have 64-bit parameters instead of int.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: find_next_{zero}_bit fixes

The find_next_{zero}_bit primitives on s390* should never return a bit number
bigger then the bit field size.  In the case of a bitfield that doesn't end on
a word boundary, an offset that makes the search start at the last word of the
bit field and the last word doesn't contain any zero/one bits the search is
continued with a call to find_first_bit with a negative size.  The search
normally ends pretty quickly because the words following the bit field contain
a mix of zeros and ones.  But the bit number that is returned in this case is
too big.

To fix this and additional if to check for this case is needed.  To make the
code easier to read I removed the assembler parts from the
find_next_{zero}_bit functions, the C-ified code is as good.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] s390: spin lock retry

Split spin lock and r/w lock implementation into a single try which is done
inline and an out of line function that repeatedly tries to get the lock
before doing the cpu_relax().  Add a system control to set the number of
retries before a cpu is yielded.

The reason for the spin lock retry is that the diagnose 0x44 that is used to
give up the virtual cpu is quite expensive.  For spin locks that are held only
for a short period of time the costs of the diagnoses outweights the savings
for spin locks that are held for a longer timer.  The default retry count is
1000.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] v850: Define L1_CACHE_SHIFT and L1_CACHE_SHIFT_MAX

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>