GitHub/exynos8895/android_kernel_samsung_universal8895.git
17 years agoIB/ipath: Remove unsafe fastrcvint code from interrupt handler
Dave Olson [Fri, 20 Jul 2007 19:50:55 +0000 (12:50 -0700)]
IB/ipath: Remove unsafe fastrcvint code from interrupt handler

The fastrcvint code's purpose was to avoid reading the interrupt
status if kernel packets were in the receive queue (to reduce
overhead).  Because intstatus was not read, we could miss the error
interrupt bit indicating freeze mode, since it only delivers a single
interrupt, even if still pending after intclear is written.

This patch removes that unsafe optimization.

Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/ehca: Move extern declarations from .c files to .h files
Hoang-Nam Nguyen [Sun, 29 Jul 2007 04:47:53 +0000 (21:47 -0700)]
IB/ehca: Move extern declarations from .c files to .h files

Make sure declarations stay in sync with definitions by keeping all
extern declarations in common .h files.

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/mlx4: Whitespace fix
Roland Dreier [Sun, 29 Jul 2007 03:52:44 +0000 (20:52 -0700)]
IB/mlx4: Whitespace fix

Remove extra dumb-looking blank line that snuck in somehow.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoIB/ehca: Fix include order to better match kernel style
Hoang-Nam Nguyen [Sat, 28 Jul 2007 15:36:32 +0000 (08:36 -0700)]
IB/ehca: Fix include order to better match kernel style

Include <rdma/...> headers after <asm/...> headers.

Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agomlx4_core: Remove kfree() in mlx4_mr_alloc() error flow
Jack Morgenstein [Thu, 26 Jul 2007 08:16:58 +0000 (11:16 +0300)]
mlx4_core: Remove kfree() in mlx4_mr_alloc() error flow

mlx4_mr_alloc() doesn't actually allocate mr (it just initializes the
pointer that the caller passes in), so it shouldn't free it if an
error occurs.

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years agoRDMA/amso1100: Initialize the wait_queue_head_t in the c2_qp structure
Tom Tucker [Tue, 24 Jul 2007 19:31:52 +0000 (14:31 -0500)]
RDMA/amso1100: Initialize the wait_queue_head_t in the c2_qp structure

Fix a crash if the driver has to wait for a QP reference to be dropped
when destroying the QP.

Signed-off-by: Ethan Burns <eaburns@iol.unh.edu>
Acked-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
17 years ago[IA64] Compare pointer against NULL, not '0'
Yoann Padioleau [Fri, 27 Jul 2007 09:44:42 +0000 (11:44 +0200)]
[IA64] Compare pointer against NULL, not '0'

When comparing a pointer, it's clearer to compare it to NULL than to 0.

Signed-off-by: Yoann Padioleau <padator@wanadoo.fr>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Fix build failure in fs/quota.c
Tony Luck [Fri, 27 Jul 2007 22:35:43 +0000 (15:35 -0700)]
[IA64] Fix build failure in fs/quota.c

b716395e2b8e450e294537de0c91476ded2f0395 added code to handle
a compatability issue with 32bit quota tools, but the new compat
routines are only needed when CONFIG_COMPAT=y (and with this set
to 'n' there are compilation problems since some new typedefs are
not visible).

Reported by Doug Chapman.  Fix tuned by a cast of thousands (Andi,
Andreas, Arthur, HPA, Willy)

Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Use new acpi_device_id struct for HID init in ia64/hp/common/sba_iommu.c
Thomas Renninger [Fri, 27 Jul 2007 22:38:31 +0000 (15:38 -0700)]
[IA64] Use new acpi_device_id struct for HID init in ia64/hp/common/sba_iommu.c

Forgot to adjust this one with the acpi autoloading patches
in commit 8c8eb78f673c07b60f31751e1e47ac367c60c6b7

Acked-by: Myron Stowe <myron.stowe@hp.com>
Acked-by: Len Brown <len.brown@intel.com>
Signed-off-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Fix wrong return value in parse_vector_domain
Kenji Kaneshige [Thu, 26 Jul 2007 06:32:38 +0000 (15:32 +0900)]
[IA64] Fix wrong return value in parse_vector_domain

Fix wrong return value in parse_vector_domain().

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Fix wrong assumption in acpi_gsi_to_irq
Kenji Kaneshige [Thu, 26 Jul 2007 06:31:34 +0000 (15:31 +0900)]
[IA64] Fix wrong assumption in acpi_gsi_to_irq

The ia64's acpi_gsi_to_irq() function assumes irq == vector. But in
fact irq can be different from vector. This patch fix this wrong
assumption.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years ago[IA64] Add sanity check into __bind_irq_vector
Kenji Kaneshige [Thu, 26 Jul 2007 06:30:45 +0000 (15:30 +0900)]
[IA64] Add sanity check into __bind_irq_vector

Add some sanity checks into __bind_irq_vector().

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Thu, 26 Jul 2007 21:00:56 +0000 (14:00 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  drivers/mmc/core/: make 3 functions static
  mmc: add missing printk levels
  mmc: remove redundant debug information from sdhci and wbsd
  mmc: proper debugging output in core
  mmc: be more verbose about card insertions/removal
  mmc: Don't hold lock when releasing an added card
  mmc: add a might_sleep() to mmc_claim_host()
  mmc: update kerneldoc
  mmc: update header file paths
  sdhci: add support to ENE-CB714
  mmc: check error bits before command completion

17 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Thu, 26 Jul 2007 21:00:09 +0000 (14:00 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc: (21 commits)
  [POWERPC] spusched: Fix initial timeslice calculation
  [POWERPC] spufs: Fix incorrect initialization of cbe_spu_info.spus
  [POWERPC] Fix Maple platform ISA bus
  [POWERPC] Make pci_iounmap actually unmap things
  [POWERPC] Add function to check if address is an IO port
  [POWERPC] Fix Pegasos keyboard detection
  [POWERPC] iSeries: Fix section mismatch warning in lpevents
  [POWERPC] iSeries: Fix section mismatch warnings
  [POWERPC] iSeries: We need vio_enable_interrupts
  [POWERPC] Fix RTC and device tree on linkstation machines
  [POWERPC] Add of_register_i2c_devices()
  [POWERPC] Fix loop with unsigned long counter variable
  [POWERPC] Fix register labels on show_regs() message for 4xx/Book-E
  [POWERPC] Only allow building of BootX text support on PPC_MULTIPLATFORM
  [POWERPC] Fix the ability to reset on MPC8544 DS and MPC8568 MDS boards
  [POWERPC] Fix mpc7448hpc2 tsi108 device_type bug
  [POWREPC] Fixup a number of modpost warnings on ppc32
  [POWERPC] Fix ethernet PHY support on MPC8544 DS
  [POWERPC] Don't try to allocate resources for a Freescale POWERPC PHB
  Revert "[POWERPC] Don't complain if size-cells == 0 in prom_parse()"
  ...

17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched
Linus Torvalds [Thu, 26 Jul 2007 20:59:59 +0000 (13:59 -0700)]
Merge git://git./linux/kernel/git/mingo/linux-2.6-sched

* git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched:
  [PATCH] sched: debug feature - make the sched-domains tree runtime-tweakable
  [PATCH] sched: add above_background_load() function
  [PATCH] sched: update Documentation/sched-stats.txt
  [PATCH] sched: mark sysrq_sched_debug_show() static
  [PATCH] sched: make cpu_clock() not use the rq clock
  [PATCH] sched: remove unused rq->load_balance_class
  [PATCH] sched: arch preempt notifier mechanism
  [PATCH] sched: increase SCHED_LOAD_SCALE_FUZZ

17 years agoDon't force-enable suspend/hibernate support just for ACPI
Linus Torvalds [Thu, 26 Jul 2007 20:44:58 +0000 (13:44 -0700)]
Don't force-enable suspend/hibernate support just for ACPI

It's a totally independent decision for the user whether he wants
suspend and/or hibernation support, and ACPI shouldn't care.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoFix ThinkPad T42 poweroff failure introduced by by "PM: Introduce pm_power_off_prepare"
Rafael J. Wysocki [Thu, 26 Jul 2007 14:29:55 +0000 (16:29 +0200)]
Fix ThinkPad T42 poweroff failure introduced by by "PM: Introduce pm_power_off_prepare"

Commit bd804eba1c8597cbb7cd5a5f9fe886aae16a079a ("PM: Introduce
pm_power_off_prepare") caused problems in the poweroff path, as reported by
YOSHIFUJI Hideaki / 吉藤英明.

Generally, sysdev_shutdown() should be called after the ACPI preparation for
powering the system off.  To make it happen, we can separate sysdev_shutdown()
from device_shutdown() and call it directly wherever necessary.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Tested-by: YOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@linux-ipv6.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoRevert most of "x86: Fix alternatives and kprobes to remap write-protected kernel...
Linus Torvalds [Thu, 26 Jul 2007 19:07:21 +0000 (12:07 -0700)]
Revert most of "x86: Fix alternatives and kprobes to remap write-protected kernel text"

This reverts most of commit 19d36ccdc34f5ed444f8a6af0cbfdb6790eb1177.

The way to DEBUG_RODATA interactions with KPROBES and CPU hotplug is to
just not mark the text as being write-protected in the first place.
Both of those facilities depend on rewriting instructions.

Having "helpful" debug facilities that just cause more problem is not
being helpful.  It just adds complexity and bugs. Not worth it.

Reported-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Andi Kleen <ak@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoPNP: fix up after Lindent
Bjorn Helgaas [Thu, 26 Jul 2007 17:41:21 +0000 (10:41 -0700)]
PNP: fix up after Lindent

These are manual fixups after running Lindent.  No functional change.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Adam Belay <ambx1@neo.rr.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoPNP: Lindent all source files
Bjorn Helgaas [Thu, 26 Jul 2007 17:41:20 +0000 (10:41 -0700)]
PNP: Lindent all source files

Run Lindent on all PNP source files.

Produced by:

    $ quilt new pnp-lindent
    $ find drivers/pnp -name \*.[ch] | xargs quilt add
    $ quilt add include/linux/{pnp.h,pnpbios.h}
    $ scripts/Lindent drivers/pnp/*.c drivers/pnp/*/*.c include/linux/pnp*.h
    $ quilt refresh --sort

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Adam Belay <ambx1@neo.rr.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: cleanup tsc.c merge artifact
Thomas Gleixner [Thu, 26 Jul 2007 17:41:19 +0000 (10:41 -0700)]
x86_64: cleanup tsc.c merge artifact

tsc_unstable is declared twice.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom32r: Fix ei_tx_timeout() in drivers/net/lib8390.c
Hirokazu Takata [Thu, 26 Jul 2007 17:41:19 +0000 (10:41 -0700)]
m32r: Fix ei_tx_timeout() in drivers/net/lib8390.c

Change INT0 trigger mode from edge-sense mode to level-sense mode,
in order to fix the following timeout error:
  'NETDEV WATCHDOG: eth0: transmit timed out'.

This patch is required only for the Mappi platform.

Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Cc: Hitoshi Yamamoto <hitoshiy@linux-m32r.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoAllow nodes to exist that only contain ZONE_MOVABLE
Mel Gorman [Thu, 26 Jul 2007 17:41:18 +0000 (10:41 -0700)]
Allow nodes to exist that only contain ZONE_MOVABLE

With the introduction of kernelcore=, a configurable zone is created on
request.  In some cases, this value will be small enough that some nodes
contain only ZONE_MOVABLE.  On some NUMA configurations when this occurs,
arch-independent zone-sizing will get the size of the memory holes within
the node incorrect.  The value of present_pages goes negative and the boot
fails.

This patch fixes the bug in the calculation of the size of the hole.  The
test case is to boot test a NUMA machine with a low value of kernelcore=
before and after the patch is applied.  While this bug exists in early
kernel it cannot be triggered in practice.

This patch has been boot-tested on a variety machines with and without
kernelcore= set.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoChar: cyclades, select FW_LOADER
Jiri Slaby [Thu, 26 Jul 2007 17:41:17 +0000 (10:41 -0700)]
Char: cyclades, select FW_LOADER

Fix link errors below by selecting FW_LOADER
  LD      .tmp_vmlinux1
drivers/built-in.o: In function `cyz_load_fw':
drivers/char/cyclades.c:4908: undefined reference to `request_firmware'
drivers/char/cyclades.c:4979: undefined reference to `release_firmware'

Cc: <toralf.foerster@gmx.de>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: fix section mismatch warnings in tce
Sam Ravnborg [Thu, 26 Jul 2007 17:41:16 +0000 (10:41 -0700)]
x86_64: fix section mismatch warnings in tce

Fix the following two section mismatch warnings:

WARNING: vmlinux.o(.text+0x1ce84): Section mismatch: reference to .init.text:free_bootmem (between 'free_tce_table' and 'build_tce_table')
WARNING: vmlinux.o(.text+0x1d04d): Section mismatch: reference to .init.text:__alloc_bootmem_low (between 'alloc_tce_table' and 'kretprobe_trampoline_holder')

In both cases the functions was used only from __init
context so mark them __init.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodrivers/edac: fix pasemi kconfig depends
Doug Thompson [Thu, 26 Jul 2007 17:41:16 +0000 (10:41 -0700)]
drivers/edac: fix pasemi kconfig depends

Fixed 'depends on PPC_PASEMI' in EDAC Kconfig.  Module PASEMI depends ONLY on
the PASEMI on PPC.

Was previously enabled for ALL PPC

Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Egor N. Martovetsky <egor@pasemi.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoinclude/asm-:mips add missing edac h file
Doug Thompson [Thu, 26 Jul 2007 17:41:15 +0000 (10:41 -0700)]
include/asm-:mips add missing edac h file

EDAC has a foundation to perform software memory scrubbing, but it requires a
per architecture (atomic_scrub) function for performing an atomic update
operation.  Under X86, this is done with a

lock:  add  [addr],0

in the file asm-x86/edac.h

This patch provides the MIPS arch with that atomic function, atomic_scrub() in

asm-mips/edac.h

Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodrivers/edac: fix edac_pci sysfs
Doug Thompson [Thu, 26 Jul 2007 17:41:15 +0000 (10:41 -0700)]
drivers/edac: fix edac_pci sysfs

This patch fixes sysfs exit code for the EDAC PCI device in a similiar manner
and the previous fixes for EDAC_MC and EDAC_DEVICE.

It removes the old (and incorrect) completion model and uses reference counts
on per instance kobjects and on the edac core module.

This pattern was applied to the edac_mc and edac_device code, but the EDAC PCI
code was missed.  In addition, this fixes a system hang after a low level
driver was unloaded.  (A cleanup function was called twice, which really
screwed things up)

Cc: Greg KH <greg@kroah.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodrivers/edac: fix reset edac_mc pollmsec
Doug Thompson [Thu, 26 Jul 2007 17:41:14 +0000 (10:41 -0700)]
drivers/edac: fix reset edac_mc pollmsec

This fixes a deadlock that could occur on a 'setup' and 'teardown' sequence of
the workq for a edac_mc control structure instance.  A similiar fix was
previously implemented for the edac_device code.

In addition, the edac_mc device code there was missing code to allow the workq
period valu to be altered via sysfs control.

This patch adds that fix on the code, and allows for the changing of the
period value as well.

Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofix 'dynreloc miscount' link error on Powerpc
Sam Ravnborg [Thu, 26 Jul 2007 17:41:13 +0000 (10:41 -0700)]
fix 'dynreloc miscount' link error on Powerpc

Nathan Lynch <ntl@pobox.com> reported:
2.6.23-rc1 breaks the build for 64-bit powerpc for me (using
maple_defconfig):

  LD      vmlinux.o
powerpc64-unknown-linux-gnu-ld: dynreloc miscount for
kernel/built-in.o, section .opd
powerpc64-unknown-linux-gnu-ld: can not edit opd Bad value
make: *** [vmlinux.o] Error 1

However, I see a possibly related binutils patch:
http://article.gmane.org/gmane.comp.gnu.binutils/33650

It was tracked down to be caused by the weak prototype
declaration in mm.h:
__attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma);

But there is no need to make the declaration weak - only the definition
needs to be marked weak.  So drop the weak declaration.  And in the process
drop the duplicate definition in page.h for powerpc.

Note: the arch_vma_name fix for x86_64 needs to be applied first to avoid
breaking x86_64

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Nathan Lynch <ntl@pobox.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agox86_64: fix arch_vma_name
Roland McGrath [Thu, 26 Jul 2007 17:41:12 +0000 (10:41 -0700)]
x86_64: fix arch_vma_name

The function arch_vma_name() is declared weak and thus it was
not noticed that x86_64 had two almost identical implementations.

It was introduced in syscall32.c by: c633090e3105e779c97d4978e5e3d7d66b291cfb
It was introduced in mm/init.c by: 2aae950b21e4bc789d1fc6668faf67e8748300b7

Signed-off-by: Roland McGrath <roland@redhat.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agochipsfb: use correct pm state
Rafael J. Wysocki [Thu, 26 Jul 2007 17:41:12 +0000 (10:41 -0700)]
chipsfb: use correct pm state

chipsfb.c shouldn't use PM_SUSPEND_MEM in there, but PM_EVENT_SUSPEND.

Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofix inode_table test in ext234_check_descriptors
Eric Sandeen [Thu, 26 Jul 2007 17:41:11 +0000 (10:41 -0700)]
fix inode_table test in ext234_check_descriptors

ext[234]_check_descriptors sanity checks block group descriptor geometry at
mount time, testing whether the block bitmap, inode bitmap, and inode table
reside wholly within the blockgroup.  However, the inode table test is off
by one so that if the last block in the inode table resides on the last
block of the block group, the test incorrectly fails.  This is because it
tests the last block as (start + length) rather than (start + length - 1).

This can be seen by trying to mount a filesystem made such as:

 mkfs.ext2 -F -b 1024 -m 0 -g 256 -N 3744 fsfile 1024

which yields:

 EXT2-fs error (device loop0): ext2_check_descriptors: Inode table for group 0 not in group (block 101)!
 EXT2-fs: group descriptors corrupted!

There is a similar bug in e2fsprogs, patch already sent for that.

(I wonder if inside(), outside(), and/or in_range() should someday be
used in this and other tests throughout the ext filesystems...)

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoxenbus_xs.c: fix a use-after-free
Adrian Bunk [Thu, 26 Jul 2007 17:41:10 +0000 (10:41 -0700)]
xenbus_xs.c: fix a use-after-free

This patch fixes an obvious use-after-free spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoedac is bust on mips
Andrew Morton [Thu, 26 Jul 2007 17:41:10 +0000 (10:41 -0700)]
edac is bust on mips

drivers/edac/edac_stub.c:15:22: asm/edac.h: No such file or directory

was it even supposed to work?

Cc: Douglas Thompson <dougthompson@xmission.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoi2c: ds1682 warning fix
Andrew Morton [Thu, 26 Jul 2007 17:41:09 +0000 (10:41 -0700)]
i2c: ds1682 warning fix

ia64:

drivers/i2c/chips/ds1682.c: In function `ds1682_show':
drivers/i2c/chips/ds1682.c:78: warning: long long unsigned int format, long unsigned int arg (arg 3)
drivers/i2c/chips/ds1682.c:78: warning: long long unsigned int format, long unsigned int arg (arg 3)

Cc: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoReorder RTC Makefile
Alessandro Zummo [Thu, 26 Jul 2007 17:41:08 +0000 (10:41 -0700)]
Reorder RTC Makefile

Alphabetic reordering of the drivers in the rtc subsys makefile.

(akpm: merge this asap!  Makefiles are the source of many patch conflicts..)

Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomemory unplug: isolate_lru_page fix
KAMEZAWA Hiroyuki [Thu, 26 Jul 2007 17:41:08 +0000 (10:41 -0700)]
memory unplug: isolate_lru_page fix

release_pages() in mm/swap.c changes page_count() to be 0 without removing
PageLRU flag...

This means isolate_lru_page() can see a page, PageLRU() &&
page_count(page)==0..  This is BUG.  (get_page() will be called against
count=0 page.)

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomemory unplug: migration by kernel
KAMEZAWA Hiroyuki [Thu, 26 Jul 2007 17:41:07 +0000 (10:41 -0700)]
memory unplug: migration by kernel

In usual, migrate_pages(page,,) is called with holding mm->sem by system call.
(mm here is a mm_struct which maps the migration target page.)
This semaphore helps avoiding some race conditions.

But, if we want to migrate a page by some kernel codes, we have to avoid
some races. This patch adds check code for following race condition.

1. A page which page->mapping==NULL can be target of migration. Then, we have
   to check page->mapping before calling try_to_unmap().

2. anon_vma can be freed while page is unmapped, but page->mapping remains as
   it was. We drop page->mapcount to be 0. Then we cannot trust page->mapping.
   So, use rcu_read_lock() to prevent anon_vma pointed by page->mapping from
   being freed during migration.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomake timerfd return a u64 and fix the __put_user
Davide Libenzi [Thu, 26 Jul 2007 17:41:07 +0000 (10:41 -0700)]
make timerfd return a u64 and fix the __put_user

Davi fixed a missing cast in the __put_user(), that was making timerfd
return a single byte instead of the full value.

Talking with Michael about the timerfd man page, we think it'd be better to
use a u64 for the returned value, to align it with the eventfd
implementation.

This is an ABI change.  The timerfd code is new in 2.6.22 and if we merge this
into 2.6.23 then we should also merge it into 2.6.22.x.  That will leave a few
early 2.6.22 kernels out in the wild which might misbehave when a future
timerfd-enabled glibc is run on them.

mtk says: The difference would be that read() will only return 4 bytes, while
the application will expect 8.  If the application is checking the size of
returned value, as it should, then it will be able to detect the problem (it
could even be sophisticated enough to know that if this is a 4-byte return,
then it is running on an old 2.6.22 kernel).  If the application is not
checking the return from read(), then its 8-byte buffer will not be filled --
the contents of the last 4 bytes will be undefined, so the u64 value as a
whole will be junk.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Cc: Davi Arnaut <davi@haxent.com.br>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoX86_POWERNOW_K8_ACPI must depend on ACPI
Adrian Bunk [Thu, 26 Jul 2007 17:41:05 +0000 (10:41 -0700)]
X86_POWERNOW_K8_ACPI must depend on ACPI

This patch fixes the following compile error introduced by
commit e8666b2718fdb5bf0ea7c3126f7e292bbbf2946b and reported
by Alexey Dobriyan:

<--  snip  -->

   CC      arch/i386/kernel/acpi/cstate.o
In file included from arch/i386/kernel/acpi/cstate.c:17:
include/acpi/processor.h:88: error: expected specifier-qualifier-list before 'acpi_integer'

<--  snip  -->

If you select something you must ensure that the dependencies of what
you are selecting are fulfilled.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Alexey Dobriyan <adobriyan@sw.ru>
Cc: Joshua Hoblitt <jhoblitt@ifa.hawaii.edu>
Cc: Dave Jones <davej@redhat.com>
Cc: Michal Piotrowski <michal.k.k.piotrowski@gmail.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation VII: FIXMEs
Rusty Russell [Thu, 26 Jul 2007 17:41:05 +0000 (10:41 -0700)]
lguest: documentation VII: FIXMEs

Documentation: The FIXMEs

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation VI: Switcher
Rusty Russell [Thu, 26 Jul 2007 17:41:04 +0000 (10:41 -0700)]
lguest: documentation VI: Switcher

Documentation: The Switcher

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation V: Host
Rusty Russell [Thu, 26 Jul 2007 17:41:04 +0000 (10:41 -0700)]
lguest: documentation V: Host

Documentation: The Host

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation IV: Launcher
Rusty Russell [Thu, 26 Jul 2007 17:41:03 +0000 (10:41 -0700)]
lguest: documentation IV: Launcher

Documentation: The Launcher

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation III: Drivers
Rusty Russell [Thu, 26 Jul 2007 17:41:03 +0000 (10:41 -0700)]
lguest: documentation III: Drivers

Documentation: The Drivers

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation II: Guest
Rusty Russell [Thu, 26 Jul 2007 17:41:02 +0000 (10:41 -0700)]
lguest: documentation II: Guest

Documentation: The Guest

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolguest: documentation I: Preparation
Rusty Russell [Thu, 26 Jul 2007 17:41:02 +0000 (10:41 -0700)]
lguest: documentation I: Preparation

The netfilter code had very good documentation: the Netfilter Hacking HOWTO.
Noone ever read it.

So this time I'm trying something different, using a bit of Knuthiness.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoxen: xen/page.h compile fix
Jeremy Fitzhardinge [Thu, 26 Jul 2007 17:41:01 +0000 (10:41 -0700)]
xen: xen/page.h compile fix

Fix:
linux/include/xen/page.h: In function mfn_pte:
linux/include/xen/page.h:149: error: __supported_pte_mask undeclared (first use in this function)
linux/include/xen/page.h:149: error: (Each undeclared identifier is reported only once
linux/include/xen/page.h:149: error: for each function it appears in.)

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agortc-ds1307: typo fix found by coverity
David Brownell [Thu, 26 Jul 2007 17:41:00 +0000 (10:41 -0700)]
rtc-ds1307: typo fix found by coverity

Fix a typo turned up by a Coverity check:  referring to the wrong register,
which could cause problems with DS1338 RTCs whose oscillators halted.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofixup s3c24xx build after arch moves
Ben Dooks [Thu, 26 Jul 2007 17:40:59 +0000 (10:40 -0700)]
fixup s3c24xx build after arch moves

Fix the include files moved around during the s3c24xx arch moves.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoPM: fix compiler error of PPC dart_iommu
Ryusuke Konishi [Thu, 26 Jul 2007 17:40:59 +0000 (10:40 -0700)]
PM: fix compiler error of PPC dart_iommu

A dummy inline function of register_nosave_region_late was accidentally
removed by the recent PM patch that introduced suspend notifiers.
This elimination causes the following compiler error on PPC machines.

  CC      arch/powerpc/sysdev/dart_iommu.o
arch/powerpc/sysdev/dart_iommu.c: In function 'iommu_init_late_dart':
arch/powerpc/sysdev/dart_iommu.c:376: error: implicit declaration of function
'register_nosave_region_late'
make[1]: *** [arch/powerpc/sysdev/dart_iommu.o] Error 1
make: *** [arch/powerpc/sysdev] Error 2

This patch fixes the problem.

Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoSlab maintainer & Credits update
Christoph Lameter [Thu, 26 Jul 2007 17:40:56 +0000 (10:40 -0700)]
Slab maintainer & Credits update

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agokernel-doc fix for kmod.c
Randy Dunlap [Thu, 26 Jul 2007 17:40:56 +0000 (10:40 -0700)]
kernel-doc fix for kmod.c

Fix kmod.c:
Warning(linux-2.6.23-rc1//kernel/kmod.c:364): No description found for parameter 'envp'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agotiny signalfd cleanup
Ulrich Drepper [Thu, 26 Jul 2007 17:40:55 +0000 (10:40 -0700)]
tiny signalfd cleanup

This is probably a leftover from a time when the return wasn't there yet.
Now the extra assignment is just irritating.

Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Cc: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomore reiserfs endianness annotations
Al Viro [Thu, 26 Jul 2007 16:47:03 +0000 (17:47 +0100)]
more reiserfs endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoarm unaligned.h annotations
Al Viro [Thu, 26 Jul 2007 16:46:19 +0000 (17:46 +0100)]
arm unaligned.h annotations

Have put_unaligned() warn if types would be wrong
for assignment, slap force-casts where needed.  Cast the
result of get_unaligned to typeof(*ptr).  With that in
place we get proper typechecking, both from gcc and from sparse,
including that for bitwise types.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k {in,out}_le{16,32} endianness misannotation
Al Viro [Thu, 26 Jul 2007 16:42:59 +0000 (17:42 +0100)]
m68k {in,out}_le{16,32} endianness misannotation

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoax88796 (address space): cast to unsigned long, not long
Al Viro [Thu, 26 Jul 2007 16:36:29 +0000 (17:36 +0100)]
ax88796 (address space): cast to unsigned long, not long

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoraw1394 __user annotation
Al Viro [Thu, 26 Jul 2007 16:36:19 +0000 (17:36 +0100)]
raw1394 __user annotation

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodrivers/ misc __iomem annotations
Al Viro [Thu, 26 Jul 2007 16:36:09 +0000 (17:36 +0100)]
drivers/ misc __iomem annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoxfs ioctl __user annotations
Al Viro [Thu, 26 Jul 2007 16:35:59 +0000 (17:35 +0100)]
xfs ioctl __user annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomake powerpc BUG_ON() OK with pointers and bitwise
Al Viro [Thu, 26 Jul 2007 16:35:49 +0000 (17:35 +0100)]
make powerpc BUG_ON() OK with pointers and bitwise

Since powerpc insists on printing the _value_ of condition
and on casting it to long...  At least let's make it a force-cast.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoamd64: fix get_user() on bitwise
Al Viro [Thu, 26 Jul 2007 16:35:39 +0000 (17:35 +0100)]
amd64: fix get_user() on bitwise

We really need force-cast when converting to final result type;
unsigned long can be silently converted to integer types and
to pointers, but not to bitwise.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomake __chk_{user,io}_ptr() accept pointers to volatile
Al Viro [Thu, 26 Jul 2007 16:35:29 +0000 (17:35 +0100)]
make __chk_{user,io}_ptr() accept pointers to volatile

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agocxgb3 gfp_t annotations
Al Viro [Thu, 26 Jul 2007 16:35:19 +0000 (17:35 +0100)]
cxgb3 gfp_t annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomisannotation in pppol2tp
Al Viro [Thu, 26 Jul 2007 16:35:09 +0000 (17:35 +0100)]
misannotation in pppol2tp

Address of auto variable is not a userland pointer.  A good thing, too,
since if pppol2tp_tunnel_getsockopt() would _really_ get a userland pointer
as argument, it would be an instant roothole...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agocyclone.c: silly use of volatile, __iomem fixes
Al Viro [Thu, 26 Jul 2007 16:34:59 +0000 (17:34 +0100)]
cyclone.c: silly use of volatile, __iomem fixes

u32* volatile cyclone_timer means volatile auto pointer to u32,
which is clearly not what had been intended (we never even take
the address of that variable, let alone pass it to something that
could change it behind our back).  u32 volatile * is what the
authors apparently wanted to say, but in reality we don't need that
qualifier there at all - it's (properly) only passed to iomem helpers
which takes care of that stuff just fine.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agouse CLOCKSOURCE_MASK() instead of too large constant
Al Viro [Thu, 26 Jul 2007 16:34:49 +0000 (17:34 +0100)]
use CLOCKSOURCE_MASK() instead of too large constant

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoalpha: long constant
Al Viro [Thu, 26 Jul 2007 16:34:39 +0000 (17:34 +0100)]
alpha: long constant

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoia64 time.c: ANSIfy
Al Viro [Thu, 26 Jul 2007 16:34:29 +0000 (17:34 +0100)]
ia64 time.c: ANSIfy

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodeal with alpha section warnings
Al Viro [Thu, 26 Jul 2007 16:34:19 +0000 (17:34 +0100)]
deal with alpha section warnings

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoviohs: extern on function definition
Al Viro [Thu, 26 Jul 2007 16:34:09 +0000 (17:34 +0100)]
viohs: extern on function definition

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoinitramfs: missing __init
Al Viro [Thu, 26 Jul 2007 16:33:59 +0000 (17:33 +0100)]
initramfs: missing __init

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agolockd and nfsd endianness annotation fixes
Al Viro [Thu, 26 Jul 2007 16:33:49 +0000 (17:33 +0100)]
lockd and nfsd endianness annotation fixes

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agonet/* misc endianness annotations
Al Viro [Thu, 26 Jul 2007 16:33:39 +0000 (17:33 +0100)]
net/* misc endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoip6_tunnel - endianness annotations
Al Viro [Thu, 26 Jul 2007 16:33:29 +0000 (17:33 +0100)]
ip6_tunnel - endianness annotations

Convert rel_info to host-endian before calling ip6_tnl_err().
The things become much more straightforward that way.
The key observation (and the reason why that code actually
worked) is that after ip6_tnl_err() we either immediately
bailed out or had rel_info set to 0 or had it set to host-endian
and guaranteed to hit
(rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED)
case.  So inconsistent endianness didn't really lead to bugs,
but it had been subtle and prone to breakage.  New variant is
saner and obviously safe.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agonetfilter endian regressions
Al Viro [Thu, 26 Jul 2007 16:33:19 +0000 (17:33 +0100)]
netfilter endian regressions

no real bugs, just misannotations cropping up

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agosun userflash is PCI-dependent
Al Viro [Thu, 26 Jul 2007 16:33:09 +0000 (17:33 +0100)]
sun userflash is PCI-dependent

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agomore VIRT_TO_BUS dependencies
Al Viro [Thu, 26 Jul 2007 16:32:59 +0000 (17:32 +0100)]
more VIRT_TO_BUS dependencies

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofix missing arguments in drivers/rtc/rtc-stk17ta8.c
Al Viro [Thu, 26 Jul 2007 16:32:49 +0000 (17:32 +0100)]
fix missing arguments in drivers/rtc/rtc-stk17ta8.c

struct bin_attribute * is needed in bin_attribute ->read()/->write()
now.  Incidentally, could people please run the fscking compiler
before and after applying their patch and compare the build logs?
That (and many, many other) would be caught immediately.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: use setup_irq() in 68328 timer code
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: use setup_irq() in 68328 timer code

Use setup_irq() instead of request_irq() to set up system timer
in 68328 timer code. With the old m68knommu irq code this
was safe, but it is not now within the generic irq framework.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: use setup_irq() in ColdFire PIT timer
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: use setup_irq() in ColdFire PIT timer

Use setup_irq() instead of request_irq() to set up system timer
in ColdFire PIT timer code. With the old m68knommu irq code this
was safe, but it is not now within the generic irq framework.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: remove unused mach_trap_init
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: remove unused mach_trap_init

Remove the unused mach_trap_init function pointer. All use of it
removed with change to using generic irq framework.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: use setup_irq() in ColdFire simple timer
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: use setup_irq() in ColdFire simple timer

Use setup_irq() instead of request_irq() to set up system timer
in ColdFire simple timer code. With the old m68knommu irq code this
was safe, but it is not now within the generic irq framework.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: remove use of colfire_trap_init
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: remove use of colfire_trap_init

The switch to using the generic irq framework removed the
coldfire_trap_init() code, so remove all references to it.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: use setup_irq() in 68360 timer code
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: use setup_irq() in 68360 timer code

Use setup_irq() instead of request_irq() to set up system timer
in 68360 timer code. With the old m68knommu irq code this
was safe, but it is not now within the generic irq framework.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68knommu: add prototype for ack_bad_irq
Greg Ungerer [Thu, 26 Jul 2007 15:09:00 +0000 (01:09 +1000)]
m68knommu: add prototype for ack_bad_irq

Create prototype for ack_bad_irq() for m68knommu.
Compilation of kernel/irq/handle.c fails without it.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years ago[PATCH] sched: debug feature - make the sched-domains tree runtime-tweakable
Nick Piggin [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: debug feature - make the sched-domains tree runtime-tweakable

debugging feature: make the sched-domains tree runtime-tweakable.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ mingo@elte.hu: made it depend on CONFIG_SCHED_DEBUG & small updates ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: add above_background_load() function
Con Kolivas [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: add above_background_load() function

Add an above_background_load() function which can be used by other
subsystems to detect if there is anything besides niced tasks running.

Place it in sched.h to allow it to be compiled out if not used.

Unused for now, but it is a useful hint to the IO scheduler and to
swap-prefetch.

Signed-off-by: Con Kolivas <kernel@kolivas.org>
Cc: Peter Williams <pwil3058@bigpond.net.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: update Documentation/sched-stats.txt
Joachim Deguara [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: update Documentation/sched-stats.txt

While learning about schedstats I found that the documentation in the tree
is old.  I updated it and found some interesting stuff like schedstats
version 14 is the same as version and version 13 never saw a kernel
release!  Also there are 6 fields in the current schedstats that are not
used anymore.  Nick had made them irrelevant in commit
476d139c218e44e045e4bc6d4cc02b010b343939 but never removed them.

Thanks to Rick's perl script who I borrowed some of the updated descriptions
from.

Signed-off-by: Joachim Deguara <joachim.deguara@amd.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: Rick Lindsley <ricklind@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: mark sysrq_sched_debug_show() static
Josh Triplett [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: mark sysrq_sched_debug_show() static

Only sched.c uses sysrq_sched_debug_show, and sched.c includes sched_debug.c,
so all uses of sysrq_sched_debug_show occur in the same source file.

Eliminates a sparse warning:
warning: symbol 'sysrq_sched_debug_show' was not declared. Should it be static?

Signed-off-by: Josh Triplett <josh@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: make cpu_clock() not use the rq clock
Ingo Molnar [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: make cpu_clock() not use the rq clock

it is enough to disable interrupts to get the precise rq-clock
of the local CPU.

this also solves an NMI watchdog regression: the NMI watchdog
calls touch_softlockup_watchdog(), which might deadlock on
rq->lock if the NMI hits an rq-locked critical section.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: remove unused rq->load_balance_class
Satoru Takeuchi [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: remove unused rq->load_balance_class

Remove unused rq->load_balance_class.

Signed-off-by: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: arch preempt notifier mechanism
Avi Kivity [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: arch preempt notifier mechanism

This adds a general mechanism whereby a task can request the scheduler to
notify it whenever it is preempted or scheduled back in.  This allows the
task to swap any special-purpose registers like the fpu or Intel's VT
registers.

Signed-off-by: Avi Kivity <avi@qumranet.com>
[ mingo@elte.hu: fixes, cleanups ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years ago[PATCH] sched: increase SCHED_LOAD_SCALE_FUZZ
Ingo Molnar [Thu, 26 Jul 2007 11:40:43 +0000 (13:40 +0200)]
[PATCH] sched: increase SCHED_LOAD_SCALE_FUZZ

increase SCHED_LOAD_SCALE_FUZZ that adds a small amount of
over-balancing: to help distribute CPU-bound tasks more fairly on SMP
systems.

the problem of unfair balancing was noticed and reported by Tong N Li.

10 CPU-bound tasks running on 8 CPUs, v2.6.23-rc1:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 2572 mingo     20   0  1576  244  196 R  100  0.0   1:03.61 loop
 2578 mingo     20   0  1576  248  196 R  100  0.0   1:03.59 loop
 2576 mingo     20   0  1576  248  196 R  100  0.0   1:03.52 loop
 2571 mingo     20   0  1576  244  196 R  100  0.0   1:03.46 loop
 2569 mingo     20   0  1576  244  196 R   99  0.0   1:03.36 loop
 2570 mingo     20   0  1576  244  196 R   95  0.0   1:00.55 loop
 2577 mingo     20   0  1576  248  196 R   50  0.0   0:31.88 loop
 2574 mingo     20   0  1576  248  196 R   50  0.0   0:31.87 loop
 2573 mingo     20   0  1576  248  196 R   50  0.0   0:31.86 loop
 2575 mingo     20   0  1576  248  196 R   50  0.0   0:31.86 loop

v2.6.23-rc1 + patch:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 2681 mingo     20   0  1576  244  196 R   85  0.0   3:51.68 loop
 2688 mingo     20   0  1576  244  196 R   81  0.0   3:46.35 loop
 2682 mingo     20   0  1576  244  196 R   80  0.0   3:43.68 loop
 2685 mingo     20   0  1576  248  196 R   80  0.0   3:45.97 loop
 2683 mingo     20   0  1576  248  196 R   80  0.0   3:40.25 loop
 2679 mingo     20   0  1576  244  196 R   80  0.0   3:33.53 loop
 2680 mingo     20   0  1576  244  196 R   79  0.0   3:43.53 loop
 2686 mingo     20   0  1576  244  196 R   79  0.0   3:39.31 loop
 2687 mingo     20   0  1576  244  196 R   78  0.0   3:33.31 loop
 2684 mingo     20   0  1576  244  196 R   77  0.0   3:27.52 loop

so they now nicely converge to the expected 80% long-term CPU usage.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
17 years agoMerge branch 'fixes-2.6.23' of master.kernel.org:/pub/scm/linux/kernel/git/galak...
Paul Mackerras [Thu, 26 Jul 2007 06:43:24 +0000 (16:43 +1000)]
Merge branch 'fixes-2.6.23' of /linux/kernel/git/galak/powerpc into merge

17 years ago[POWERPC] spusched: Fix initial timeslice calculation
Christoph Hellwig [Wed, 25 Jul 2007 11:31:09 +0000 (21:31 +1000)]
[POWERPC] spusched: Fix initial timeslice calculation

Currently we calculate the first timeslice for every context
incorrectly - alloc_spu_context calls spu_set_timeslice before we set
ctx->prio so we always calculate the longest possible timeslice for the
lowest possible priority.

This patch makes sure to update the schedule-related fields before
calculating the timeslice and also makes sure we update the timeslice for
a non-running context when entering spu_run so a priority change affects
the context as soon as possible.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] spufs: Fix incorrect initialization of cbe_spu_info.spus
Masato Noguchi [Wed, 25 Jul 2007 03:20:15 +0000 (13:20 +1000)]
[POWERPC] spufs: Fix incorrect initialization of cbe_spu_info.spus

We currently initialize cbe_spu_info[].spus in both init_spu_base and
spu_sched_init. The initialise in spu_sched_init clears the SPU list,
so we end up with no physical SPUs. Because of this, the spu_run
syscall will block forever.

This change removes the unnecessary initialization in spu_sched_init.

Signed-off-by: Masato Noguchi <Masato.Noguchi@jp.sony.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
17 years ago[POWERPC] Fix Maple platform ISA bus
Benjamin Herrenschmidt [Thu, 26 Jul 2007 04:07:14 +0000 (14:07 +1000)]
[POWERPC] Fix Maple platform ISA bus

The Maple platform has ISA IOs but didn't call the new functions to
actually map those, thus crashing when trying to access the nvram.

This fixes Maple and JS2x using SLOF.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>