]> err.no Git - linux-2.6/log
linux-2.6
16 years agoGRU: fix preprocessor symbol for sparse
Randy Dunlap [Tue, 12 Aug 2008 22:08:51 +0000 (15:08 -0700)]
GRU: fix preprocessor symbol for sparse

Fix preprocessor symbol so that sparse sees it and does not generate
errors:

  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"
  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"
  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"
  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"
  drivers/misc/sgi-gru/grutlbpurge.c:185:11: error: undefined identifier 'GRUREGION'
  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"
  drivers/misc/sgi-gru/grutables.h:286:2: error: "Unsupported architecture"

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Jack Steiner <steiner@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoradeonfb: fix accel engine hangs
David Miller [Tue, 12 Aug 2008 22:08:51 +0000 (15:08 -0700)]
radeonfb: fix accel engine hangs

Some chips appear to have the 2D engine hang during screen redraw,
typically in a sequence of copyarea operations. This appear to be
solved by adding a flush of the engine destination pixel cache
and waiting for the engine to be idle before issuing the accel
operation. The performance impact seems to be fairly small.

Here is a trace on an RV370 (PCI device ID 0x5b64), it records the
RBBM_STATUS register, then the source x/y, destination x/y, and
width/height used for the copy:

----------------------------------------
radeonfb_prim_copyarea: STATUS[00000140] src[210:70] dst[210:60] wh[a0:10]
radeonfb_prim_copyarea: STATUS[00000140] src[2b8:70] dst[2b8:60] wh[88:10]
radeonfb_prim_copyarea: STATUS[00000140] src[348:70] dst[348:60] wh[40:10]
radeonfb_prim_copyarea: STATUS[80020140] src[390:70] dst[390:60] wh[88:10]
radeonfb_prim_copyarea: STATUS[8002613f] src[40:80] dst[40:70] wh[28:10]
radeonfb_prim_copyarea: STATUS[80026139] src[a8:80] dst[a8:70] wh[38:10]
radeonfb_prim_copyarea: STATUS[80026133] src[e8:80] dst[e8:70] wh[80:10]
radeonfb_prim_copyarea: STATUS[8002612d] src[170:80] dst[170:70] wh[30:10]
radeonfb_prim_copyarea: STATUS[80026127] src[1a8:80] dst[1a8:70] wh[8:10]
radeonfb_prim_copyarea: STATUS[80026121] src[1b8:80] dst[1b8:70] wh[88:10]
radeonfb_prim_copyarea: STATUS[8002611b] src[248:80] dst[248:70] wh[68:10]
----------------------------------------

When things are going fine the copies complete before the next ROP is
even issued, but all of a sudden the 2D unit becomes active (bit 17 in
RBBM_STATUS) and the FIFO retry (bit 13) and FIFO pipeline busy (bit
14) are set as well.  The FIFO begins to backup until it becomes full.

What happens next is the radeon_fifo_wait() times out, and we access
the chip illegally leading to a bus error which usually wedges the
box.  None of this makes it to the console screen, of course :-)
radeon_fifo_wait() should be modified to reset the accelerator when
this timeout happens instead of programming the chip anyways.

----------------------------------------
radeonfb: FIFO Timeout !
ERROR(0): Cheetah error trap taken afsr[0010080005000000] afar[000007f900800e40] TL1(0)
ERROR(0): TPC[595114] TNPC[595118] O7[459788] TSTATE[11009601]
ERROR(0): TPC<radeonfb_copyarea+0xfc/0x248>
ERROR(0): M_SYND(0),  E_SYND(0), Privileged
ERROR(0): Highest priority error (0000080000000000) "Bus error response from system bus"
ERROR(0): D-cache idx[0] tag[0000000000000000] utag[0000000000000000] stag[0000000000000000]
ERROR(0): D-cache data0[0000000000000000] data1[0000000000000000] data2[0000000000000000] data3[0000000000000000]
ERROR(0): I-cache idx[0] tag[0000000000000000] utag[0000000000000000] stag[0000000000000000] u[0000000000000000] l[00\

ERROR(0): I-cache INSN0[0000000000000000] INSN1[0000000000000000] INSN2[0000000000000000] INSN3[0000000000000000]
ERROR(0): I-cache INSN4[0000000000000000] INSN5[0000000000000000] INSN6[0000000000000000] INSN7[0000000000000000]
ERROR(0): E-cache idx[800e40] tag[000000000e049f4c]
ERROR(0): E-cache data0[fffff8127d300180] data1[00000000004b5384] data2[0000000000000000] data3[0000000000000000]
Ker:xnel panic - not syncing: Irrecoverable deferred error trap.
----------------------------------------

Another quirk is that these copyarea calls will not happen until the
first drivers/char/vt.c:redraw_screen() occurs.  This will only happen
if you 1) VC switch or 2) run "consolechars" or 3) unblank the screen.

This seems to happen because until a redraw_screen() the screen scrolling
method used by fbcon is not finalized yet.  I've seen this with other fb
drivers too.

So if all you do is boot straight into X you will never see this bug on
the relevant chips.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoallocate structures for reservation tracking in hugetlbfs outside of spinlocks v2
Andy Whitcroft [Tue, 12 Aug 2008 22:08:49 +0000 (15:08 -0700)]
allocate structures for reservation tracking in hugetlbfs outside of spinlocks v2

[Andrew this should replace the previous version which did not check
the returns from the region prepare for errors.  This has been tested by
us and Gerald and it looks good.

Bah, while reviewing the locking based on your previous email I spotted
that we need to check the return from the vma_needs_reservation call for
allocation errors.  Here is an updated patch to correct this.  This passes
testing here.]

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Tested-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agohugetlbfs: allocate structures for reservation tracking outside of spinlocks
Andy Whitcroft [Tue, 12 Aug 2008 22:08:47 +0000 (15:08 -0700)]
hugetlbfs: allocate structures for reservation tracking outside of spinlocks

In the normal case, hugetlbfs reserves hugepages at map time so that the
pages exist for future faults.  A struct file_region is used to track when
reservations have been consumed and where.  These file_regions are
allocated as necessary with kmalloc() which can sleep with the
mm->page_table_lock held.  This is wrong and triggers may-sleep warning
when PREEMPT is enabled.

Updates to the underlying file_region are done in two phases.  The first
phase prepares the region for the change, allocating any necessary memory,
without actually making the change.  The second phase actually commits the
change.  This patch makes use of this by checking the reservations before
the page_table_lock is taken; triggering any necessary allocations.  This
may then be safely repeated within the locks without any allocations being
required.

Credit to Mel Gorman for diagnosing this failure and initial versions of
the patch.

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Tested-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoisight_firmware: fix a leak and double kfree()
Parag Warudkar [Tue, 12 Aug 2008 22:08:46 +0000 (15:08 -0700)]
isight_firmware: fix a leak and double kfree()

Signed-off-by: Parag Warudkar <parag.warudkar@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocpuidle: make sysfs attributes sysdev class attributes
Rabin Vincent [Tue, 12 Aug 2008 22:08:45 +0000 (15:08 -0700)]
cpuidle: make sysfs attributes sysdev class attributes

These attributes are really sysdev class attributes.  The incorrect
definition leads to an oops because of recent changes which make sysdev
attributes use a different prototype.

Based on Andi's f718cd4add5aea9d379faff92f162571e356cc5f ("sched: make
scheduler sysfs attributes sysdev class devices")

Reported-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Acked-by: Andi Kleen <ak@linux.intel.com>
Cc: "Li, Shaohua" <shaohua.li@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc-isl1208: fix double removal of a sysfs entry
Alessandro Zummo [Tue, 12 Aug 2008 22:08:44 +0000 (15:08 -0700)]
rtc-isl1208: fix double removal of a sysfs entry

Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: Herbert Valerio Riedel <hvr@gnu.org>
Cc: Hartley Sweeten <hartleys@visionengravers.com>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoh8300: fix section mismatches
Yoshinori Sato [Tue, 12 Aug 2008 22:08:43 +0000 (15:08 -0700)]
h8300: fix section mismatches

WARNING: vmlinux.o(.text+0x2fdf): Section mismatch in reference from the variable .LM3 to the variable .init.text:___alloc_bootmem
The function .LM3() references
the variable __init ___alloc_bootmem.
This is often because .LM3 lacks a __init
annotation or the annotation of ___alloc_bootmem is wrong.

WARNING: vmlinux.o(.text+0x2ff5): Section mismatch in reference from the variable .LM4 to the variable .init.text:___alloc_bootmem
The function .LM4() references
the variable __init ___alloc_bootmem.
This is often because .LM4 lacks a __init
annotation or the annotation of ___alloc_bootmem is wrong.

WARNING: vmlinux.o(.text+0x300b): Section mismatch in reference from the variable .LM5 to the variable .init.text:___alloc_bootmem
The function .LM5() references
the variable __init ___alloc_bootmem.
This is often because .LM5 lacks a __init
annotation or the annotation of ___alloc_bootmem is wrong.

WARNING: vmlinux.o(.text+0x304b): Section mismatch in reference from the variable .LM10 to the variable .init.text:_free_area_init
The function .LM10() references
the variable __init _free_area_init.
This is often because .LM10 lacks a __init
annotation or the annotation of _free_area_init is wrong.

WARNING: vmlinux.o(.text+0x30a3): Section mismatch in reference from the variable .LM17 to the variable .init.text:_free_all_bootmem
The function .LM17() references
the variable __init _free_all_bootmem.
This is often because .LM17 lacks a __init
annotation or the annotation of _free_all_bootmem is wrong.

Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agorevert "rtc: cdev lock_kernel() pushdown"
David Brownell [Tue, 12 Aug 2008 22:08:41 +0000 (15:08 -0700)]
revert "rtc: cdev lock_kernel() pushdown"

Revert commit 51a776fa7a7997e726d4a478eda0854c6f9143bd ("rtc: cdev
lock_kernel() pushdown").  The RTC framework does not need BKL
protection.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomemcg: fix oops in mem_cgroup_shrink_usage
Hugh Dickins [Tue, 12 Aug 2008 22:08:41 +0000 (15:08 -0700)]
memcg: fix oops in mem_cgroup_shrink_usage

Got an oops in mem_cgroup_shrink_usage() when testing loop over tmpfs:
yes, of course, loop0 has no mm: other entry points check but this didn't.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocpu hotplug: s390 doesn't support additional_cpus anymore.
Heiko Carstens [Tue, 12 Aug 2008 22:08:40 +0000 (15:08 -0700)]
cpu hotplug: s390 doesn't support additional_cpus anymore.

s390 doesn't support the additional_cpus kernel parameter anymore since a
long time.  So we better update the code and documentation to reflect
that.

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopage allocator: use no-panic variant of alloc_bootmem() in alloc_large_system_hash()
Jan Beulich [Tue, 12 Aug 2008 22:08:39 +0000 (15:08 -0700)]
page allocator: use no-panic variant of alloc_bootmem() in alloc_large_system_hash()

..  since a failed allocation is being (initially) handled gracefully, and
panic()-ed upon failure explicitly in the function if retries with smaller
sizes failed.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoquota: documentation for sending "below quota" messages via netlink and tiny doc...
Jan Kara [Tue, 12 Aug 2008 22:08:39 +0000 (15:08 -0700)]
quota: documentation for sending "below quota" messages via netlink and tiny doc update

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agohugetlb: call arch_prepare_hugepage() for surplus pages
Gerald Schaefer [Tue, 12 Aug 2008 22:08:38 +0000 (15:08 -0700)]
hugetlb: call arch_prepare_hugepage() for surplus pages

The s390 software large page emulation implements shared page tables by
using page->index of the first tail page from a compound large page to
store page table information.  This is set up in arch_prepare_hugepage(),
which is called from alloc_fresh_huge_page_node().

A similar call to arch_prepare_hugepage() is missing for surplus large
pages that are allocated in alloc_buddy_huge_page(), which breaks the
software emulation mode for (surplus) large pages on s390.  This patch
adds the missing call to arch_prepare_hugepage().  It will have no effect
on other architectures where arch_prepare_hugepage() is a nop.

Also, use the correct order in the error path in alloc_fresh_huge_page_node().

Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Acked-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agofeature-removal-schedule.txt: remove the NCR53C9x entry
Adrian Bunk [Tue, 12 Aug 2008 22:08:37 +0000 (15:08 -0700)]
feature-removal-schedule.txt: remove the NCR53C9x entry

Now that the driver is removed we should also remove the entry in
Documentation/feature-removal-schedule.txt

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Christoph Hellwig <hch@lst.de>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Tue, 12 Aug 2008 16:14:24 +0000 (09:14 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: hda - support new AMD HDMI Audio (1002:970f)
  ALSA: hda_intel: ALSA HD Audio patch for Intel Ibex Peak DeviceIDs
  ALSA: wm8750: add missing VREF output
  ALSA: spitz: MONO -> MONO1
  ALSA: wm8750: it's MONO1, not MONO

16 years agoMerge branch 'core-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 12 Aug 2008 15:49:53 +0000 (08:49 -0700)]
Merge branch 'core-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  generic-ipi: fix stack and rcu interaction bug in smp_call_function_mask(), fix

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
Linus Torvalds [Tue, 12 Aug 2008 15:40:19 +0000 (08:40 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  fix spinlock recursion in hvc_console
  stop_machine: remove unused variable
  modules: extend initcall_debug functionality to the module loader
  export virtio_rng.h
  lguest: use get_user_pages_fast() instead of get_user_pages()
  mm: Make generic weak get_user_pages_fast and EXPORT_GPL it
  lguest: don't set MAC address for guest unless specified

16 years agoMerge branch 'agp-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Tue, 12 Aug 2008 15:28:32 +0000 (08:28 -0700)]
Merge branch 'agp-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/agp-2.6

* 'agp-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/agp-2.6:
  agp: fix SIS 5591/5592 wrong PCI id
  intel/agp: rewrite GTT on resume
  agp: use dev_printk when possible
  amd64-agp: run fallback when no bridges found, not when driver registration fails
  intel_agp: official name for GM45 chipset

16 years agoALSA: hda - support new AMD HDMI Audio (1002:970f)
Libin Yang [Tue, 12 Aug 2008 10:25:46 +0000 (12:25 +0200)]
ALSA: hda - support new AMD HDMI Audio (1002:970f)

Signed-off-by: Libin Yang <libin.yang@amd.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
16 years agoALSA: hda_intel: ALSA HD Audio patch for Intel Ibex Peak DeviceIDs
Seth Heasley [Fri, 8 Aug 2008 22:56:39 +0000 (15:56 -0700)]
ALSA: hda_intel: ALSA HD Audio patch for Intel Ibex Peak DeviceIDs

This patch adds the Intel Ibex Peak (PCH) HD Audio Controller DeviceIDs.

Signed-off by: Seth Heasley <seth.heasley@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
16 years agoALSA: wm8750: add missing VREF output
Dmitry Baryshkov [Mon, 11 Aug 2008 22:45:31 +0000 (02:45 +0400)]
ALSA: wm8750: add missing VREF output

Add missing output VREF. After a65f0568f6cc8433877fb71dd7d36b551854b0bc
it's critical, since it makes chip routing initialisation to fail.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
16 years agoALSA: spitz: MONO -> MONO1
Dmitry Baryshkov [Mon, 11 Aug 2008 22:45:30 +0000 (02:45 +0400)]
ALSA: spitz: MONO -> MONO1

Correct route name to be MONO1 instead of MONO to follow
recent fix in wm8750.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
16 years agogeneric-ipi: fix stack and rcu interaction bug in smp_call_function_mask(), fix
Nick Piggin [Tue, 12 Aug 2008 08:05:13 +0000 (18:05 +1000)]
generic-ipi: fix stack and rcu interaction bug in smp_call_function_mask(), fix

> > Nick Piggin (1):
> >       generic-ipi: fix stack and rcu interaction bug in
> > smp_call_function_mask()
>
> I'm still not 100% sure that I have this patch right... I might have seen
> a lockup trace implicating the smp call function path... which may have
> been due to some other problem or a different bug in the new call function
> code, but if some more people can take a look at it before merging?

OK indeed it did have a couple of bugs. Firstly, I wasn't freeing the
data properly in the alloc && wait case. Secondly, I wasn't resetting
CSD_FLAG_WAIT in the for each cpu loop (so only the first CPU would
wait).

After those fixes, the patch boots and runs with the kmalloc commented
out (so it always executes the slowpath).

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agofix spinlock recursion in hvc_console
Christian Borntraeger [Thu, 7 Aug 2008 07:18:34 +0000 (09:18 +0200)]
fix spinlock recursion in hvc_console

commit 611e097d7707741a336a0677d9d69bec40f29f3d
Author: Christian Borntraeger <borntraeger@de.ibm.com>
hvc_console: rework setup to replace irq functions with callbacks
introduced a spinlock recursion problem.

request_irq tries to call the handler if the IRQ is shared.
The irq handler of hvc_console calls hvc_poll and hvc_kill
which might take the hvc_struct spinlock. Therefore, we have
to call request_irq outside the spinlock.

We can move the notifier_add safely outside the spinlock as ->data must
not be changed by the backend. Otherwise, tty_hangup would fail anyway.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agostop_machine: remove unused variable
Li Zefan [Thu, 31 Jul 2008 02:31:02 +0000 (10:31 +0800)]
stop_machine: remove unused variable

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agomodules: extend initcall_debug functionality to the module loader
Arjan van de Ven [Wed, 30 Jul 2008 19:49:02 +0000 (12:49 -0700)]
modules: extend initcall_debug functionality to the module loader

The kernel has this really nice facility where if you put "initcall_debug"
on the kernel commandline, it'll print which function it's going to
execute just before calling an initcall, and then after the call completes
it will

1) print if it had an error code

2) checks for a few simple bugs (like leaving irqs off)
and

3) print how long the init call took in milliseconds.

While trying to optimize the boot speed of my laptop, I have been loving
number 3 to figure out what to optimize...  ...  and then I wished that
the same thing was done for module loading.

This patch makes the module loader use this exact same functionality; it's
a logical extension in my view (since modules are just sort of late
binding initcalls anyway) and so far I've found it quite useful in finding
where things are too slow in my boot.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agoexport virtio_rng.h
Christian Borntraeger [Fri, 8 Aug 2008 09:15:07 +0000 (11:15 +0200)]
export virtio_rng.h

Hello Rusty,

The entropy device was added after we exported all virtio headers. This
patch adds virtio_rng.h to the exportable userspace headers.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agolguest: use get_user_pages_fast() instead of get_user_pages()
Rusty Russell [Tue, 12 Aug 2008 22:52:53 +0000 (17:52 -0500)]
lguest: use get_user_pages_fast() instead of get_user_pages()

Using a simple page table thrashing program I measure a slight
improvement.  The program creates five processes.  Each touches 1000
pages then schedules the next process.  We repeat this 1000 times.  As
lguest only caches 4 cr3 values, this rebuilds a lot of shadow page
tables requiring virt->phys mappings.

Before: 5.93 seconds
After: 5.40 seconds

(Counts of slow vs fastpath in this usage are 6092 and 2852462 respectively.)

And more importantly for lguest, the code is simpler.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agomm: Make generic weak get_user_pages_fast and EXPORT_GPL it
Rusty Russell [Tue, 12 Aug 2008 22:52:52 +0000 (17:52 -0500)]
mm: Make generic weak get_user_pages_fast and EXPORT_GPL it

Out of line get_user_pages_fast fallback implementation, make it a weak
symbol, get rid of CONFIG_HAVE_GET_USER_PAGES_FAST.

Export the symbol to modules so lguest can use it.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agolguest: don't set MAC address for guest unless specified
Rusty Russell [Tue, 12 Aug 2008 22:52:51 +0000 (17:52 -0500)]
lguest: don't set MAC address for guest unless specified

This shows up when trying to bridge:
tap0: received packet with  own address as source address

As Max Krasnyansky points out, there's no reason to give the guest the
same mac address as the TUN device.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Max Krasnyansky <maxk@qualcomm.com>
16 years agoagp: fix SIS 5591/5592 wrong PCI id
Krzysztof Helt [Wed, 6 Aug 2008 16:48:45 +0000 (18:48 +0200)]
agp: fix SIS 5591/5592 wrong PCI id

The correct id is the id of the main host (5591) not
the id of the PCI-to-PCI bridge AGP (0001).
Output from "lspci -nv" shows that only the former
has AGP capabilities flag set:

00:00.0 0600: 1039:5591 (rev 02)
        Flags: bus master, medium devsel, latency 64
        Memory at ec000000 (32-bit, non-prefetchable) [size=32M]
        Capabilities: [c0] AGP version 1.0

00:02.0 0604: 1039:0001 (prog-if 00 [Normal decode])
        Flags: bus master, fast devsel, latency 0
        Bus: primary=00, secondary=01, subordinate=01, sec-latency=0
        I/O behind bridge: 0000c000-0000cfff
        Memory behind bridge: eb500000-eb5fffff
        Prefetchable memory behind bridge: eb300000-eb3fffff

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agointel/agp: rewrite GTT on resume
Keith Packard [Thu, 31 Jul 2008 05:48:07 +0000 (15:48 +1000)]
intel/agp: rewrite GTT on resume

On my Intel chipset (965GM), the GTT is entirely erased across
suspend/resume.  This patch simply re-plays the current mapping at resume
time to restore the table.=20

I noticed this once I started relying on persistent GTT mappings across VT
switch in our GEM work -- the old X server and DRM code carefully unbind
all memory from the GTT on VT switch, but GEM does not bother.

I placed the list management and rewrite code in the generic layer on the
assumption that it will be needed on other hardware, but I did not add the
rewrite call to anything other than the Intel resume function.

Keep a list of current GATT mappings.  At resume time, rewrite them into
the GATT.  This is needed on Intel (at least) as the entire GATT is
cleared across suspend/resume.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Keith Packard <keithp@keithp.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agoagp: use dev_printk when possible
Bjorn Helgaas [Wed, 30 Jul 2008 19:26:51 +0000 (12:26 -0700)]
agp: use dev_printk when possible

Convert printks to use dev_printk().

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agoamd64-agp: run fallback when no bridges found, not when driver registration fails
Bjorn Helgaas [Wed, 30 Jul 2008 19:26:51 +0000 (12:26 -0700)]
amd64-agp: run fallback when no bridges found, not when driver registration fails

I think the intent was that if no bridges matched agp_amd64_pci_table[],
we would fall back to checking for any bridge with the AGP capability.
But in the current code, we execute the fallback path only when
pci_register_driver() itself fails, which is unrelated to whether any
matching devices were found.

This patch counts the AGP bridges found in the probe() method and executes
the fallback path when none is found.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agointel_agp: official name for GM45 chipset
Zhenyu Wang [Wed, 30 Jul 2008 19:26:50 +0000 (12:26 -0700)]
intel_agp: official name for GM45 chipset

Signed-off-by: Zhenyu Wang <zhenyu.z.wang@intel.com>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 11 Aug 2008 23:46:31 +0000 (16:46 -0700)]
Merge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched, cpu hotplug: fix set_cpus_allowed() use in hotplug callbacks
  sched: fix mysql+oltp regression
  sched_clock: delay using sched_clock()
  sched clock: couple local and remote clocks
  sched clock: simplify __update_sched_clock()
  sched: eliminate scd->prev_raw
  sched clock: clean up sched_clock_cpu()
  sched clock: revert various sched_clock() changes
  sched: move sched_clock before first use
  sched: test runtime rather than period in global_rt_runtime()
  sched: fix SCHED_HRTICK dependency
  sched: fix warning in hrtick_start_fair()

16 years agoMerge branch 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 11 Aug 2008 23:46:11 +0000 (16:46 -0700)]
Merge branch 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  posix-timers: fix posix_timer_event() vs dequeue_signal() race
  posix-timers: do_schedule_next_timer: fix the setting of ->si_overrun

16 years agoMerge branch 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 11 Aug 2008 23:45:46 +0000 (16:45 -0700)]
Merge branch 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  lockdep: fix debug_lock_alloc
  lockdep: increase MAX_LOCKDEP_KEYS
  generic-ipi: fix stack and rcu interaction bug in smp_call_function_mask()
  lockdep: fix overflow in the hlock shrinkage code
  lockdep: rename map_[acquire|release]() => lock_map_[acquire|release]()
  lockdep: handle chains involving classes defined in modules
  mm: fix mm_take_all_locks() locking order
  lockdep: annotate mm_take_all_locks()
  lockdep: spin_lock_nest_lock()
  lockdep: lock protection locks
  lockdep: map_acquire
  lockdep: shrink held_lock structure
  lockdep: re-annotate scheduler runqueues
  lockdep: lock_set_subclass - reset a held lock's subclass
  lockdep: change scheduler annotation
  debug_locks: set oops_in_progress if we will log messages.
  lockdep: fix combinatorial explosion in lock subgraph traversal

16 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 11 Aug 2008 23:44:35 +0000 (16:44 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: fix 2.6.27rc1 cannot boot more than 8CPUs
  x86: make "apic" an early_param() on 32-bit, NULL check
  EFI, x86: fix function prototype
  x86, pci-calgary: fix function declaration
  x86: work around gcc 3.4.x bug
  x86: make "apic" an early_param() on 32-bit
  x86, debug: tone down arch/x86/kernel/mpparse.c debugging printk
  x86_64: restore the proper NR_IRQS define so larger systems work.
  x86: Restore proper vector locking during cpu hotplug
  x86: Fix broken VMI in 2.6.27-rc..
  x86: fdiv bug detection fix

16 years agoMerge branch 'core/locking' into core/urgent
Ingo Molnar [Mon, 11 Aug 2008 22:11:49 +0000 (00:11 +0200)]
Merge branch 'core/locking' into core/urgent

16 years agoMerge branch 'sched/clock' into sched/urgent
Ingo Molnar [Mon, 11 Aug 2008 22:07:02 +0000 (00:07 +0200)]
Merge branch 'sched/clock' into sched/urgent

16 years agolockdep: fix debug_lock_alloc
Peter Zijlstra [Mon, 11 Aug 2008 20:45:51 +0000 (22:45 +0200)]
lockdep: fix debug_lock_alloc

When we enable DEBUG_LOCK_ALLOC but do not enable PROVE_LOCKING and or
LOCK_STAT, lock_alloc() and lock_release() turn into nops, even though
we should be doing hlock checking (check=1).

This causes a false warning and a lockdep self-disable.

Rectify this.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: fix 2.6.27rc1 cannot boot more than 8CPUs
Yinghai Lu [Mon, 11 Aug 2008 20:36:04 +0000 (13:36 -0700)]
x86: fix 2.6.27rc1 cannot boot more than 8CPUs

Jeff Chua reported that booting a !bigsmp kernel on a 16-way box
hangs silently.

this is a long-standing issue, smp start AP cpu could check the
apic id >=8 etc before trying to start it.

achieve this by moving the def_to_bigsmp check later and skip the
apicid id > 8

[ mingo@elte.hu: clean up the message that is printed. ]

Reported-by: "Jeff Chua" <jeff.chua.linux@gmail.com>
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
 arch/x86/kernel/setup.c   |    6 ------
 arch/x86/kernel/smpboot.c |   10 ++++++++++
 2 files changed, 10 insertions(+), 6 deletions(-)

16 years agomake struct scsi_dh_devlist's static
Adrian Bunk [Mon, 11 Aug 2008 18:59:21 +0000 (11:59 -0700)]
make struct scsi_dh_devlist's static

This patch makes several needlessly global struct scsi_dh_devlist's
static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'for-linus' of git://git.o-hand.com/linux-mfd
Linus Torvalds [Mon, 11 Aug 2008 17:44:43 +0000 (10:44 -0700)]
Merge branch 'for-linus' of git://git.o-hand.com/linux-mfd

* 'for-linus' of git://git.o-hand.com/linux-mfd:
  mfd: tc6393 cleanup and update
  mfd: have TMIO drivers and subdevices depend on ARM
  mfd: TMIO MMC driver
  mfd: driver for the TMIO NAND controller
  mfd: t7l66 MMC platform data
  mfd: tc6387 MMC platform data
  mfd: Fix 7l66 and 6387 according to the new mfd-core API
  mfd: Fix tc6393 according to the new tmio.h
  mfd: driver for the TC6387XB TMIO controller.
  mfd: driver for the T7L66XB TMIO SoC
  mfd: TMIO MMC structures and accessors.

16 years agoMerge branch 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Mon, 11 Aug 2008 17:44:13 +0000 (10:44 -0700)]
Merge branch 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'i2c-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  hwmon: (lm75) Drop legacy i2c driver
  i2c: correct some size_t printk formats
  i2c: Check for address business before creating clients
  i2c: Let users select algorithm drivers manually again
  i2c: Fix NULL pointer dereference in i2c_new_probed_device
  i2c: Fix oops on bus multiplexer driver loading

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Mon, 11 Aug 2008 17:42:43 +0000 (10:42 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] pcwd.c - fix open_allowed type.
  [WATCHDOG] fix watchdog/ixp4xx_wdt.c compilation
  [WATCHDOG] fix watchdog/wdt285.c compilation
  [WATCHDOG] fix watchdog/at91rm9200_wdt.c compilation
  [WATCHDOG] fix watchdog/shwdt.c compilation
  [WATCHDOG] fix watchdog/txx9wdt.c compilation
  [WATCHDOG] MAINTAINERS: remove ZF MACHZ WATCHDOG entry
  [WATCHDOG] Fix build with CONFIG_ITCO_VENDOR_SUPPORT=n

16 years agox86: make "apic" an early_param() on 32-bit, NULL check
Rene Herman [Mon, 11 Aug 2008 17:20:17 +0000 (19:20 +0200)]
x86: make "apic" an early_param() on 32-bit, NULL check

Cyrill Gorcunov observed:

> you turned it into early_param so now it's NULL injecting vulnerabled.
> Could you please add checking for NULL str param?

fix that.

Also, change the name of 'str' into 'arg', to make it more apparent
that this is an optional argument that can be NULL, not a string
parameter that is empty when unset.

Reported-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Rene Herman <rene.herman@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Mon, 11 Aug 2008 17:40:28 +0000 (10:40 -0700)]
Merge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  powerpc: Remove include/linux/harrier_defs.h
  powerpc: Do not ignore arch/powerpc/include
  powerpc: Delete completed "ppc removal" task from feature removal file
  powerpc/mm: Fix attribute confusion with htab_bolt_mapping()
  powerpc/pci: Don't keep ISA memory hole resources in the tree
  powerpc: Zero fill the return values of rtas argument buffer
  powerpc/4xx: Update defconfig files for 2.6.27-rc1
  powerpc/44x: Incorrect NOR offset in Warp DTS
  powerpc/44x: Warp DTS changes for board updates
  powerpc/4xx: Cleanup Warp for i2c driver changes.
  powerpc/44x: Adjust warp-nand resource end address

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Mon, 11 Aug 2008 17:38:36 +0000 (10:38 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  PCI: Limit VPD length for Broadcom 5708S
  PCI PM: Export pci_pme_active to drivers
  PCI: remove duplicate symbol from pci_ids.h
  PCI: check the return value of device_create_bin_file() in pci_create_bus()
  PCI: fully restore MSI state at resume time
  DMA: make dma-coherent.c documentation kdoc-friendly
  PCI: make pci_register_driver() a macro
  PCI: add Broadcom 5708S to VPD length quirk

16 years agoFix race/oops in tty layer after BKL pushdown
Christian Borntraeger [Mon, 11 Aug 2008 08:02:49 +0000 (09:02 +0100)]
Fix race/oops in tty layer after BKL pushdown

While testing our KVM code for s390 (starting and killall kvm in a loop)
I can reproduce the following oops:

  Unable to handle kernel pointer dereference at virtual kernel address 6b6b6b6b6b6b6000 Oops: 0038 [#1] SMP
  Modules linked in: dm_multipath sunrpc qeth_l3 qeth_l2 dm_mod qeth
  ccwgroup CPU: 1 Not tainted 2.6.27-rc1 #54
  Process kuli (pid: 4409, task: 00000000b6aa5940, ksp: 00000000b7343e10)
  Krnl PSW : 0704e00180000000 00000000002e0b8c
  (disassociate_ctty+0x1c0/0x288) R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:3
  CC:2 PM:0 EA:3 Krnl GPRS: 0000000000000000 6b6b6b6b6b6b6b6b
  0000000000000001 00000000000003a6 00000000002e0a46 00000000004b4160
  0000000000000001 00000000bbd79758 00000000b7343e58 00000000b8854148
  00000000bd34dea0 00000000b7343c20 0000000000000001 00000000004b6d08
  00000000002e0a46 00000000b7343c20 Krnl Code: 00000000002e0b7e:
  eb9fb0a00004 lmg %r9,%r15,160(%r11) 00000000002e0b84:
  07f4 bcr 15,%r4 00000000002e0b86:
  e31090080004 lg %r1,8(%r9) >00000000002e0b8c:
  d501109cd000 clc 156(2,%r1),0(%r13) 00000000002e0b92:
  a784ff5d brc 8,2e0a4c 00000000002e0b96:
  b9040029 lgr %r2,%r9 00000000002e0b9a:
  c0e5fffff9c3 brasl %r14,2dff20 00000000002e0ba0:
  a7f4ff56 brc 15,2e0a4c Call Trace:
  ([<00000000002e0a46>] disassociate_ctty+0x7a/0x288)
   [<0000000000141fe6>] do_exit+0x212/0x8d4
   [<0000000000142708>] do_group_exit+0x60/0xcc
   [<0000000000150660>] get_signal_to_deliver+0x270/0x3ac
   [<000000000010bfd6>] do_signal+0x8e/0x8dc
   [<0000000000113772>] sysc_sigpending+0xe/0x22
   [<000001ff0000b134>] 0x1ff0000b134
  INFO: lockdep is turned off.
  Last Breaking-Event-Address:
   [<00000000002e0a48>] disassociate_ctty+0x7c/0x288
  Kernel panic - not syncing: Fatal exception: panic_on_oops

It seems that tty was already free in disassocate_ctty when it tries
to dereference tty->driver.

After moving the lock_kernel before the mutex_unlock, I can no longer
reproduce the problem.

[ This is a temporary partial fix for the documented and long standing
  race in disassociate_tty.  This stops most problem cases for now.

  For the next release the -next tree has an initial implementation of
  kref counting for tty structures and this quickfix will be dropped.

                                                              - Alan ]

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by; Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k{,nommu}: Wire up new system calls
Geert Uytterhoeven [Mon, 11 Aug 2008 07:00:30 +0000 (09:00 +0200)]
m68k{,nommu}: Wire up new system calls

Wire up for m68k{,nommu} the system calls that were added in the last merge
window:

 - 4006553b06306b34054529477b06b68a1c66249b ("flag parameters: inotify_init")
 - ed8cae8ba01348bfd83333f4648dd807b04d7f08 ("flag parameters: pipe")
 - 336dd1f70ff62d7dd8655228caed4c5bfc818c56 ("flag parameters: dup2")
 - a0998b50c3f0b8fdd265c63e0032f86ebe377dbf ("flag parameters: epoll_create")
 - 9fe5ad9c8cef9ad5873d8ee55d1cf00d9b607df0 ("flag parameters add-on: remove
 epoll_create size param")
 - b087498eb5605673b0f260a7620d91818cd72304 ("flag parameters: eventfd")
 - 9deb27baedb79759c3ab9435a7d8b841842d56e9 ("flag parameters: signalfd")

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoRevert "fbcon: bgcolor fix"
Linus Torvalds [Mon, 11 Aug 2008 17:29:11 +0000 (10:29 -0700)]
Revert "fbcon: bgcolor fix"

This reverts commit 2d04a4a72d7e1519b4838f24bdd4b5d0f3f426dc, which made
it impossible to make the softcursor use the highlight colors.

Yes, the fourth bit should be "blinking", but since we cannot reasonably
blink in fbcon, highlighting it with a bright background is preferable.

Reported-by: Pavel Machek <pavel@suse.cz>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Antonino A. Daplas <adaplas@pol.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoEFI, x86: fix function prototype
Randy Dunlap [Thu, 7 Aug 2008 22:12:39 +0000 (15:12 -0700)]
EFI, x86: fix function prototype

Fix function prototype in header file to match source code:

linux-next-20080807/arch/x86/kernel/efi_64.c:100:14: error: symbol 'efi_ioremap' redeclared with different type (originally declared at include2/asm/efi.h:89) - different address spaces

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86, pci-calgary: fix function declaration
Randy Dunlap [Thu, 7 Aug 2008 22:14:55 +0000 (15:14 -0700)]
x86, pci-calgary: fix function declaration

Fix function declaration:

 linux-next-20080807/arch/x86/kernel/pci-calgary_64.c:1353:36: warning: non-ANSI function declaration of function 'get_tce_space_from_tar'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: work around gcc 3.4.x bug
Jeremy Fitzhardinge [Fri, 8 Aug 2008 20:46:07 +0000 (13:46 -0700)]
x86: work around gcc 3.4.x bug

Simon Horman reported that gcc-3.4.x crashes when compiling
pgd_prepopulate_pmd() when PREALLOCATED_PMDS == 0 and CONFIG_DEBUG_INFO
is enabled.

Adding an extra check for PREALLOCATED_PMDS == 0 [which is compiled out
by gcc] seems to avoid the problem.

Reported-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: make "apic" an early_param() on 32-bit
Rene Herman [Mon, 11 Aug 2008 15:45:53 +0000 (17:45 +0200)]
x86: make "apic" an early_param() on 32-bit

On 32-bit, "apic" is a __setup() param meaning it is parsed rather
late in the game. Make it an early_param() for apic_printk() use
by arch/x86/kernel/mpparse.c.

On 64-bit, it already is an early_param().

Signed-off-by: Rene Herman <rene.herman@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86, debug: tone down arch/x86/kernel/mpparse.c debugging printk
Rene Herman [Mon, 11 Aug 2008 15:44:57 +0000 (17:44 +0200)]
x86, debug: tone down arch/x86/kernel/mpparse.c debugging printk

commit 11a62a056093a7f25f1595fbd8bd5f93559572b6 turns some formerly
nopped debugging printks in arch/x86/kernel/mppparse.c into regular
ones. The one at the top of smp_scan_config() in particular also
prints on !CONFIG_SMP/CONFIG_X86_LOCAL_APIC kernels and UP machines
without anything resembling MP tables which makes their lowly UP
owners wonder...

Turn the former Dprintk()s into apic_printk()s instead meaning that
their printing is dependent on passing the apic=verbose (or =debug)
command line param.

On 32-bit, "apic" is a __setup() param which isn't early enough
for this code and therefore needs a followup changing it into an
early_param(). On 64-bit, it already is.

Signed-off-by: Rene Herman <rene.herman@gmail.com>
Cc: Andrew Morton <akpm@osdl.org>
Cc: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched, cpu hotplug: fix set_cpus_allowed() use in hotplug callbacks
Dmitry Adamushko [Wed, 30 Jul 2008 10:34:04 +0000 (12:34 +0200)]
sched, cpu hotplug: fix set_cpus_allowed() use in hotplug callbacks

Mark Langsdorf reported:

> One of my co-workers noticed that the powernow-k8
> driver no longer restarts when a CPU core is
> hot-disabled and then hot-enabled on AMD quad-core
> systems.
>
> The following comands work fine on 2.6.26 and fail
> on 2.6.27-rc1:
>
> echo 0 > /sys/devices/system/cpu/cpu3/online
> echo 1 > /sys/devices/system/cpu/cpu3/online
> find /sys -name cpufreq
>
> For 2.6.26, the find will return a cpufreq
> directory for each processor.  In 2.6.27-rc1,
> the cpu3 directory is missing.
>
> After digging through the code, the following
> logic is failing when the core is hot-enabled
> at runtime.  The code works during the boot
> sequence.
>
>       cpumask_t = current->cpus_allowed;
>       set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
>       if (smp_processor_id() != cpu)
>               return -ENODEV;

So set the CPU active before calling the CPU_ONLINE notifier chain,
there are a handful of notifiers that use set_cpus_allowed().

This fix also solves the problem with x86-microcode. I've sent
alternative patches for microcode, but as this "rely on
set_cpus_allowed_ptr() being workable in cpu-hotplug(CPU_ONLINE, ...)"
assumption seems to be more broad than what we thought, perhaps this fix
should be applied.

With this patch we define that by the moment CPU_ONLINE is being sent,
a 'cpu' is online and ready for tasks to be migrated onto it.

Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Reported-by: Mark Langsdorf <mark.langsdorf@amd.com>
Tested-by: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: increase MAX_LOCKDEP_KEYS
Ingo Molnar [Mon, 11 Aug 2008 10:37:27 +0000 (12:37 +0200)]
lockdep: increase MAX_LOCKDEP_KEYS

certain configs produce:

 [   70.076229] BUG: MAX_LOCKDEP_KEYS too low!
 [   70.080230] turning off the locking correctness validator.

tune them up.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agogeneric-ipi: fix stack and rcu interaction bug in smp_call_function_mask()
Nick Piggin [Mon, 11 Aug 2008 03:49:30 +0000 (13:49 +1000)]
generic-ipi: fix stack and rcu interaction bug in smp_call_function_mask()

* Venki Pallipadi <venkatesh.pallipadi@intel.com> wrote:

> Found a OOPS on a big SMP box during an overnight reboot test with
> upstream git.
>
> Suresh and I looked at the oops and looks like the root cause is in
> generic_smp_call_function_interrupt() and smp_call_function_mask() with
> wait parameter.
>
> The actual oops looked like
>
> [   11.277260] BUG: unable to handle kernel paging request at ffff8802ffffffff
> [   11.277815] IP: [<ffff8802ffffffff>] 0xffff8802ffffffff
> [   11.278155] PGD 202063 PUD 0
> [   11.278576] Oops: 0010 [1] SMP
> [   11.279006] CPU 5
> [   11.279336] Modules linked in:
> [   11.279752] Pid: 0, comm: swapper Not tainted 2.6.27-rc2-00020-g685d87f #290
> [   11.280039] RIP: 0010:[<ffff8802ffffffff>]  [<ffff8802ffffffff>] 0xffff8802ffffffff
> [   11.280692] RSP: 0018:ffff88027f1f7f70  EFLAGS: 00010086
> [   11.280976] RAX: 00000000ffffffff RBX: 0000000000000000 RCX: 0000000000000000
> [   11.281264] RDX: 0000000000004f4e RSI: 0000000000000001 RDI: 0000000000000000
> [   11.281624] RBP: ffff88027f1f7f98 R08: 0000000000000001 R09: ffffffff802509af
> [   11.281925] R10: ffff8800280c2780 R11: 0000000000000000 R12: ffff88027f097d48
> [   11.282214] R13: ffff88027f097d70 R14: 0000000000000005 R15: ffff88027e571000
> [   11.282502] FS:  0000000000000000(0000) GS:ffff88027f1c3340(0000) knlGS:0000000000000000
> [   11.283096] CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
> [   11.283382] CR2: ffff8802ffffffff CR3: 0000000000201000 CR4: 00000000000006e0
> [   11.283760] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [   11.284048] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> [   11.284337] Process swapper (pid: 0, threadinfo ffff88027f1f2000, task ffff88027f1f0640)
> [   11.284936] Stack:  ffffffff80250963 0000000000000212 0000000000ee8c78 0000000000ee8a66
> [   11.285802]  ffff88027e571550 ffff88027f1f7fa8 ffffffff8021adb5 ffff88027f1f3e40
> [   11.286599]  ffffffff8020bdd6 ffff88027f1f3e40 <EOI>  ffff88027f1f3ef8 0000000000000000
> [   11.287120] Call Trace:
> [   11.287768]  <IRQ>  [<ffffffff80250963>] ? generic_smp_call_function_interrupt+0x61/0x12c
> [   11.288354]  [<ffffffff8021adb5>] smp_call_function_interrupt+0x17/0x27
> [   11.288744]  [<ffffffff8020bdd6>] call_function_interrupt+0x66/0x70
> [   11.289030]  <EOI>  [<ffffffff8024ab3b>] ? clockevents_notify+0x19/0x73
> [   11.289380]  [<ffffffff803b9b75>] ? acpi_idle_enter_simple+0x18b/0x1fa
> [   11.289760]  [<ffffffff803b9b6b>] ? acpi_idle_enter_simple+0x181/0x1fa
> [   11.290051]  [<ffffffff8053aeca>] ? cpuidle_idle_call+0x70/0xa2
> [   11.290338]  [<ffffffff80209f61>] ? cpu_idle+0x5f/0x7d
> [   11.290723]  [<ffffffff8060224a>] ? start_secondary+0x14d/0x152
> [   11.291010]
> [   11.291287]
> [   11.291654] Code:  Bad RIP value.
> [   11.292041] RIP  [<ffff8802ffffffff>] 0xffff8802ffffffff
> [   11.292380]  RSP <ffff88027f1f7f70>
> [   11.292741] CR2: ffff8802ffffffff
> [   11.310951] ---[ end trace 137c54d525305f1c ]---
>
> The problem is with the following sequence of events:
>
> - CPU A calls smp_call_function_mask() for CPU B with wait parameter
> - CPU A sets up the call_function_data on the stack and does an rcu add to
>   call_function_queue
> - CPU A waits until the WAIT flag is cleared
> - CPU B gets the call function interrupt and starts going through the
>   call_function_queue
> - CPU C also gets some other call function interrupt and starts going through
>   the call_function_queue
> - CPU C, which is also going through the call_function_queue, starts referencing
>   CPU A's stack, as that element is still in call_function_queue
> - CPU B finishes the function call that CPU A set up and as there are no other
>   references to it, rcu deletes the call_function_data (which was from CPU A
>   stack)
> - CPU B sees the wait flag and just clears the flag (no call_rcu to free)
> - CPU A which was waiting on the flag continues executing and the stack
>   contents change
>
> - CPU C is still in rcu_read section accessing the CPU A's stack sees
>   inconsistent call_funation_data and can try to execute
>   function with some random pointer, causing stack corruption for A
>   (by clearing the bits in mask field) and oops.

Nice debugging work.

I'd suggest something like the attached (boot tested) patch as the simple
fix for now.

I expect the benefits from the less synchronized, multiple-in-flight-data
global queue will still outweigh the costs of dynamic allocations. But
if worst comes to worst then we just go back to a globally synchronous
one-at-a-time implementation, but that would be pretty sad!

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: fix mysql+oltp regression
Mike Galbraith [Mon, 11 Aug 2008 11:32:02 +0000 (13:32 +0200)]
sched: fix mysql+oltp regression

Defer commit 6d299f1b53b84e2665f402d9bcc494800aba6386 to the next release.

Testing of the tip/sched/clock tree revealed a mysql+oltp regression
which bisection eventually traced back to this commit in mainline.

Pertinent test results:  Three run sysbench averages, throughput units
in read/write requests/sec.

clients         1     2     4     8    16    32    64
6e0534f      9646 17876 34774 33868 32230 30767 29441
2.6.26.1     9112 17936 34652 33383 31929 30665 29232
6d299f1      9112 14637 28370 33339 32038 30762 29204

Note: subsequent commits hide the majority of this regression until you
apply the clock fixes, at which time it reemerges at full magnitude.

We cannot see anything bad about the change itself so we defer it to the
next release until this problem is fully analysed.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'linus' into sched/urgent
Ingo Molnar [Mon, 11 Aug 2008 11:40:56 +0000 (13:40 +0200)]
Merge branch 'linus' into sched/urgent

16 years agopowerpc: Remove include/linux/harrier_defs.h
Paul Mackerras [Mon, 11 Aug 2008 10:59:59 +0000 (20:59 +1000)]
powerpc: Remove include/linux/harrier_defs.h

It was only used by code in arch/ppc, and arch/ppc is gone, so remove
the unused harrier_defs.h as well.

Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years agolockdep: fix overflow in the hlock shrinkage code
Peter Zijlstra [Mon, 11 Aug 2008 10:34:42 +0000 (12:34 +0200)]
lockdep: fix overflow in the hlock shrinkage code

There is a overflow by 1 case in the new shrunken hlock code.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86_64: restore the proper NR_IRQS define so larger systems work.
Eric W. Biederman [Sun, 10 Aug 2008 07:35:50 +0000 (00:35 -0700)]
x86_64: restore the proper NR_IRQS define so larger systems work.

As pointed out and tracked by Yinghai Lu <yhlu.kernel@gmail.com>:

 Dhaval Giani got:
 kernel BUG at arch/x86/kernel/io_apic_64.c:357!
 invalid opcode: 0000 [1] SMP
 CPU 24
 ...

his system (x3950) has 8 ioapic, irq > 256

This was caused by:

       commit 9b7dc567d03d74a1fbae84e88949b6a60d922d82
       Author: Thomas Gleixner <tglx@linutronix.de>
       Date:   Fri May 2 20:10:09 2008 +0200

          x86: unify interrupt vector defines

          The interrupt vector defines are copied 4 times around with minimal
          differences. Move them all into asm-x86/irq_vectors.h

It appears that Thomas did not notice that x86_64 does something
completely different when he merge irq_vectors.h

We can solve this for 2.6.27 by simply reintroducing the old heuristic
for setting NR_IRQS on x86_64 to a usable value, which trivially removes
the regression.

Long term it would be nice to harmonize the handling of ioapic interrupts
of x86_32 and x86_64 so we don't have this kind of confusion.

Dhaval Giani <dhaval@linux.vnet.ibm.com> tested an earlier version of
this patch by YH which confirms simply increasing NR_IRQS fixes the
problem.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Cc: Mike Travis <travis@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: Restore proper vector locking during cpu hotplug
Eric W. Biederman [Sat, 9 Aug 2008 22:09:02 +0000 (15:09 -0700)]
x86: Restore proper vector locking during cpu hotplug

Having cpu_online_map change during assign_irq_vector can result
in some really nasty and weird things happening.  The one that
bit me last time was accessing non existent per cpu memory for non
existent cpus.

This locking was removed in a sloppy x86_64 and x86_32 merge patch.

Guys can we please try and avoid subtly breaking x86 when we are
merging files together?

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
16 years agolockdep: rename map_[acquire|release]() => lock_map_[acquire|release]()
Ingo Molnar [Mon, 11 Aug 2008 08:30:30 +0000 (10:30 +0200)]
lockdep: rename map_[acquire|release]() => lock_map_[acquire|release]()

the names were too generic:

 drivers/uio/uio.c:87: error: expected identifier or '(' before 'do'
 drivers/uio/uio.c:87: error: expected identifier or '(' before 'while'
 drivers/uio/uio.c:113: error: 'map_release' undeclared here (not in a function)

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoALSA: wm8750: it's MONO1, not MONO
Dmitry Baryshkov [Sat, 9 Aug 2008 11:05:28 +0000 (15:05 +0400)]
ALSA: wm8750: it's MONO1, not MONO

Since first commit wm8750 contained output named MONO, but
all routes mentioned MONO1. Correct MONO to be MONO1.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
16 years agolockdep: handle chains involving classes defined in modules
Rabin Vincent [Mon, 11 Aug 2008 07:30:26 +0000 (09:30 +0200)]
lockdep: handle chains involving classes defined in modules

Solve this by marking the classes as unused and not printing information
about the unused classes.

Reported-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Acked-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agomm: fix mm_take_all_locks() locking order
Peter Zijlstra [Mon, 11 Aug 2008 07:30:25 +0000 (09:30 +0200)]
mm: fix mm_take_all_locks() locking order

Lockdep spotted:

=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.27-rc1 #270
-------------------------------------------------------
qemu-kvm/2033 is trying to acquire lock:
 (&inode->i_data.i_mmap_lock){----}, at: [<ffffffff802996cc>] mm_take_all_locks+0xc2/0xea

but task is already holding lock:
 (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&anon_vma->lock){----}:
       [<ffffffff8025cd37>] __lock_acquire+0x11be/0x14d2
       [<ffffffff8025d0a9>] lock_acquire+0x5e/0x7a
       [<ffffffff804c655b>] _spin_lock+0x3b/0x47
       [<ffffffff8029a2ef>] vma_adjust+0x200/0x444
       [<ffffffff8029a662>] split_vma+0x12f/0x146
       [<ffffffff8029bc60>] mprotect_fixup+0x13c/0x536
       [<ffffffff8029c203>] sys_mprotect+0x1a9/0x21e
       [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b
       [<ffffffffffffffff>] 0xffffffffffffffff

-> #0 (&inode->i_data.i_mmap_lock){----}:
       [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2
       [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219
       [<ffffffff8025d515>] lock_release+0x127/0x14a
       [<ffffffff804c6403>] _spin_unlock+0x1e/0x50
       [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0
       [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112
       [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10
       [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm]
       [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78
       [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274
       [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78
       [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b
       [<ffffffffffffffff>] 0xffffffffffffffff

other info that might help us debug this:

5 locks held by qemu-kvm/2033:
 #0:  (&mm->mmap_sem){----}, at: [<ffffffff802a95d0>] do_mmu_notifier_register+0x55/0x112
 #1:  (mm_all_locks_mutex){--..}, at: [<ffffffff8029963e>] mm_take_all_locks+0x34/0xea
 #2:  (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea
 #3:  (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea
 #4:  (&anon_vma->lock){----}, at: [<ffffffff8029967a>] mm_take_all_locks+0x70/0xea

stack backtrace:
Pid: 2033, comm: qemu-kvm Not tainted 2.6.27-rc1 #270

Call Trace:
 [<ffffffff8025b7c7>] print_circular_bug_tail+0xb8/0xc3
 [<ffffffff8025ca54>] __lock_acquire+0xedb/0x14d2
 [<ffffffff80259bb1>] ? add_lock_to_list+0x7e/0xad
 [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea
 [<ffffffff8029967a>] ? mm_take_all_locks+0x70/0xea
 [<ffffffff8025d397>] lock_release_non_nested+0x1c2/0x219
 [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea
 [<ffffffff802996cc>] ? mm_take_all_locks+0xc2/0xea
 [<ffffffff8025b202>] ? trace_hardirqs_on_caller+0x4d/0x115
 [<ffffffff802995d9>] ? mm_drop_all_locks+0x7f/0xb0
 [<ffffffff8025d515>] lock_release+0x127/0x14a
 [<ffffffff804c6403>] _spin_unlock+0x1e/0x50
 [<ffffffff802995d9>] mm_drop_all_locks+0x7f/0xb0
 [<ffffffff802a965d>] do_mmu_notifier_register+0xe2/0x112
 [<ffffffff802a96a8>] mmu_notifier_register+0xe/0x10
 [<ffffffffa0043b6b>] kvm_dev_ioctl+0x11e/0x287 [kvm]
 [<ffffffff8033f9f2>] ? file_has_perm+0x83/0x8e
 [<ffffffff802bd0ca>] vfs_ioctl+0x2a/0x78
 [<ffffffff802bd36f>] do_vfs_ioctl+0x257/0x274
 [<ffffffff802bd3e1>] sys_ioctl+0x55/0x78
 [<ffffffff8020c0db>] system_call_fastpath+0x16/0x1b

Which the locking hierarchy in mm/rmap.c confirms as valid.

Fix this by first taking all the mapping->i_mmap_lock instances and then
take all anon_vma->lock instances.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: annotate mm_take_all_locks()
Peter Zijlstra [Mon, 11 Aug 2008 07:30:25 +0000 (09:30 +0200)]
lockdep: annotate mm_take_all_locks()

The nesting is correct due to holding mmap_sem, use the new annotation
to annotate this.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: spin_lock_nest_lock()
Peter Zijlstra [Mon, 11 Aug 2008 07:30:24 +0000 (09:30 +0200)]
lockdep: spin_lock_nest_lock()

Expose the new lock protection lock.

This can be used to annotate places where we take multiple locks of the
same class and avoid deadlocks by always taking another (top-level) lock
first.

NOTE: we're still bound to the MAX_LOCK_DEPTH (48) limit.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: lock protection locks
Peter Zijlstra [Mon, 11 Aug 2008 07:30:24 +0000 (09:30 +0200)]
lockdep: lock protection locks

On Fri, 2008-08-01 at 16:26 -0700, Linus Torvalds wrote:

> On Fri, 1 Aug 2008, David Miller wrote:
> >
> > Taking more than a few locks of the same class at once is bad
> > news and it's better to find an alternative method.
>
> It's not always wrong.
>
> If you can guarantee that anybody that takes more than one lock of a
> particular class will always take a single top-level lock _first_, then
> that's all good. You can obviously screw up and take the same lock _twice_
> (which will deadlock), but at least you cannot get into ABBA situations.
>
> So maybe the right thing to do is to just teach lockdep about "lock
> protection locks". That would have solved the multi-queue issues for
> networking too - all the actual network drivers would still have taken
> just their single queue lock, but the one case that needs to take all of
> them would have taken a separate top-level lock first.
>
> Never mind that the multi-queue locks were always taken in the same order:
> it's never wrong to just have some top-level serialization, and anybody
> who needs to take <n> locks might as well do <n+1>, because they sure as
> hell aren't going to be on _any_ fastpaths.
>
> So the simplest solution really sounds like just teaching lockdep about
> that one special case. It's not "nesting" exactly, although it's obviously
> related to it.

Do as Linus suggested. The lock protection lock is called nest_lock.

Note that we still have the MAX_LOCK_DEPTH (48) limit to consider, so anything
that spills that it still up shit creek.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: map_acquire
Peter Zijlstra [Mon, 11 Aug 2008 07:30:23 +0000 (09:30 +0200)]
lockdep: map_acquire

Most the free-standing lock_acquire() usages look remarkably similar, sweep
them into a new helper.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: shrink held_lock structure
Dave Jones [Mon, 11 Aug 2008 07:30:23 +0000 (09:30 +0200)]
lockdep: shrink held_lock structure

struct held_lock {
        u64                        prev_chain_key;       /*     0     8 */
        struct lock_class *        class;                /*     8     8 */
        long unsigned int          acquire_ip;           /*    16     8 */
        struct lockdep_map *       instance;             /*    24     8 */
        int                        irq_context;          /*    32     4 */
        int                        trylock;              /*    36     4 */
        int                        read;                 /*    40     4 */
        int                        check;                /*    44     4 */
        int                        hardirqs_off;         /*    48     4 */

        /* size: 56, cachelines: 1 */
        /* padding: 4 */
        /* last cacheline: 56 bytes */
};

struct held_lock {
        u64                        prev_chain_key;       /*     0     8 */
        long unsigned int          acquire_ip;           /*     8     8 */
        struct lockdep_map *       instance;             /*    16     8 */
        unsigned int               class_idx:11;         /*    24:21  4 */
        unsigned int               irq_context:2;        /*    24:19  4 */
        unsigned int               trylock:1;            /*    24:18  4 */
        unsigned int               read:2;               /*    24:16  4 */
        unsigned int               check:2;              /*    24:14  4 */
        unsigned int               hardirqs_off:1;       /*    24:13  4 */

        /* size: 32, cachelines: 1 */
        /* padding: 4 */
        /* bit_padding: 13 bits */
        /* last cacheline: 32 bytes */
};

[mingo@elte.hu: shrunk hlock->class too]
[peterz@infradead.org: fixup bit sizes]
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agolockdep: re-annotate scheduler runqueues
Peter Zijlstra [Mon, 11 Aug 2008 07:30:22 +0000 (09:30 +0200)]
lockdep: re-annotate scheduler runqueues

Instead of using a per-rq lock class, use the regular nesting operations.

However, take extra care with double_lock_balance() as it can release the
already held rq->lock (and therefore change its nesting class).

So what can happen is:

 spin_lock(rq->lock); // this rq subclass 0

 double_lock_balance(rq, other_rq);
   // release rq
   // acquire other_rq->lock subclass 0
   // acquire rq->lock subclass 1

 spin_unlock(other_rq->lock);

leaving you with rq->lock in subclass 1

So a subsequent double_lock_balance() call can try to nest a subclass 1
lock while already holding a subclass 1 lock.

Fix this by introducing double_unlock_balance() which releases the other
rq's lock, but also re-sets the subclass for this rq's lock to 0.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agolockdep: lock_set_subclass - reset a held lock's subclass
Peter Zijlstra [Mon, 11 Aug 2008 07:30:21 +0000 (09:30 +0200)]
lockdep: lock_set_subclass - reset a held lock's subclass

this can be used to reset a held lock's subclass, for arbitrary-depth
iterated data structures such as trees or lists which have per-node
locks.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'linus' into sched/clock
Ingo Molnar [Mon, 11 Aug 2008 06:59:21 +0000 (08:59 +0200)]
Merge branch 'linus' into sched/clock

16 years agosched_clock: delay using sched_clock()
Peter Zijlstra [Mon, 11 Aug 2008 06:59:03 +0000 (08:59 +0200)]
sched_clock: delay using sched_clock()

Some arch's can't handle sched_clock() being called too early - delay
this until sched_clock_init() has been called.

Reported-by: Bill Gatliff <bgat@billgatliff.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Nishanth Aravamudan <nacc@us.ibm.com>
CC: Russell King - ARM Linux <linux@arm.linux.org.uk>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agopowerpc: Do not ignore arch/powerpc/include
Junio C Hamano [Fri, 8 Aug 2008 01:45:08 +0000 (18:45 -0700)]
powerpc: Do not ignore arch/powerpc/include

Back when .gitignore file was added to arch/powerpc/ in 06f2138 ([POWERPC]
Add files build to .gitignore, 2006-11-26), there indeed was nothing
tracked in the ignored hierarchy and ignoring everything made sense.  But
we have very many tracked files there these days, and having a higher
level .gitignore that ignores everything is asking for future troubles..

This should have been part of b8b572e (powerpc: Move include files to
arch/powerpc/include/asm, 2008-08-01).

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years agopowerpc: Delete completed "ppc removal" task from feature removal file
Robert P. J. Day [Wed, 6 Aug 2008 17:58:44 +0000 (03:58 +1000)]
powerpc: Delete completed "ppc removal" task from feature removal file

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years agopowerpc/mm: Fix attribute confusion with htab_bolt_mapping()
Benjamin Herrenschmidt [Tue, 5 Aug 2008 06:19:56 +0000 (16:19 +1000)]
powerpc/mm: Fix attribute confusion with htab_bolt_mapping()

The function htab_bolt_mapping() is used to create permanent
mappings in the MMU hash table, for example, in order to create
the linear mapping of vmemmap.  It's also used by early boot
ioremap (before mem_init_done).

However, the way ioremap uses it is incorrect as it passes it the
protection flags in the "linux PTE" form while htab_bolt_mapping()
expects them in the hash table format.  This is made more confusing by
the fact that some of those flags are actually in the same position in
both cases.

This fixes it all by making htab_bolt_mapping() take normal linux
protection flags instead, and use a little helper to convert them to
htab flags. Callers can now use the usual PAGE_* definitions safely.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
 arch/powerpc/include/asm/mmu-hash64.h |    2 -
 arch/powerpc/mm/hash_utils_64.c       |   65 ++++++++++++++++++++--------------
 arch/powerpc/mm/init_64.c             |    9 +---
 3 files changed, 44 insertions(+), 32 deletions(-)
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years agopowerpc/pci: Don't keep ISA memory hole resources in the tree
Benjamin Herrenschmidt [Thu, 31 Jul 2008 05:24:13 +0000 (15:24 +1000)]
powerpc/pci: Don't keep ISA memory hole resources in the tree

When we have an ISA memory hole (ie, a PCI window that allows us to
generate PCI memory cycles at low PCI address) mixed with other
resources using a different CPU <=> PCI mapping, we must not keep
the ISA hole in the bridge resource list.

If we do, things might start trying to allocate device resources
in there and will get the PCI addresses wrong.

This fixes it by arranging to remove the ISA memory hole resource in
this case.  This fixes various cases of PCMCIA breakage on PowerBooks
using the MPC106 "grackle" bridge.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years agopowerpc: Zero fill the return values of rtas argument buffer
Nathan Fontenot [Wed, 30 Jul 2008 16:23:27 +0000 (02:23 +1000)]
powerpc: Zero fill the return values of rtas argument buffer

The kernel copy of the rtas args struct contains the return
value(s) for the specified rtas call.  These are copied back
to user space with the assumption that every value has been
set by the rtas call, which turns out to be not always true.
Thus userspace can see random values and think the call failed
when in fact it succeeded, but for some reason didn't set one
of the return values.

This fixes the problem by zeroing out the return value fields
of the rtas args struct before processing the rtas call.

Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years ago[WATCHDOG] pcwd.c - fix open_allowed type.
Wim Van Sebroeck [Sun, 10 Aug 2008 21:57:03 +0000 (21:57 +0000)]
[WATCHDOG] pcwd.c - fix open_allowed type.

Fix following warnings:
drivers/watchdog/pcwd.c: In function 'pcwd_open':
drivers/watchdog/pcwd.c:703: warning: passing argument 2 of 'test_and_set_bit' from incompatible pointer type
drivers/watchdog/pcwd.c: In function 'pcwd_close':
drivers/watchdog/pcwd.c:723: warning: passing argument 2 of 'clear_bit' from incompatible pointer type

Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
16 years agomfd: tc6393 cleanup and update
Ian Molton [Sun, 10 Aug 2008 21:32:07 +0000 (23:32 +0200)]
mfd: tc6393 cleanup and update

This patchset cleans up the TC6393XB support.

* Add provision for the MMC subdevice
* Disable / enable clocks on suspend / resume
* Remove fragments of badly merged code (eg. linux/fb include etc.)
* Use a device specific clock name to break dependancy on ARM/PXA2XX
* Drop unnecessary resource names
* Switch to tmio_io* accessors

Signed-off-by: Ian Molton <spyro@f2s.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
16 years agomfd: have TMIO drivers and subdevices depend on ARM
Samuel Ortiz [Tue, 5 Aug 2008 17:27:58 +0000 (19:27 +0200)]
mfd: have TMIO drivers and subdevices depend on ARM

The TMIO chips are only found (and thus tested) on ARM machines.
Moreover, we don't want the TMIO cells to be built if one of the TMIO
driver is not selected (which indirectly make the TMIO cells drivers
depend on ARM as well).

Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
16 years agomfd: TMIO MMC driver
Ian Molton [Tue, 15 Jul 2008 15:02:21 +0000 (16:02 +0100)]
mfd: TMIO MMC driver

This patch adds support for the MMC subdevice 'cell' commonly found in
TMIO based MFDs.

Signed-off-by: Ian Molton <spyro@f2s.com>
Acked-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
16 years agomfd: driver for the TMIO NAND controller
Ian Molton [Tue, 15 Jul 2008 15:04:22 +0000 (16:04 +0100)]
mfd: driver for the TMIO NAND controller

This patch adds support for the NAND controller commonly found in
TMIO based MFDs.

Signed-off-by: Ian Molton <spyro@f2s.com>
Acked-By: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Samuel Ortiz <sameo@openedhand.com>
16 years agohwmon: (lm75) Drop legacy i2c driver
Jean Delvare [Sun, 10 Aug 2008 20:56:16 +0000 (22:56 +0200)]
hwmon: (lm75) Drop legacy i2c driver

Drop the legacy lm75 driver, and add a detect callback to the
new-style driver to achieve the same functionality.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: David Brownell <david-b@pacbell.net>
16 years agoi2c: correct some size_t printk formats
David Brownell [Sun, 10 Aug 2008 20:56:16 +0000 (22:56 +0200)]
i2c: correct some size_t printk formats

Fix various printk format strings where %zd was passed a size_t;
those should be %zu instead.  (Courtesy of a version of GCC which
warns when these details are wrong.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Check for address business before creating clients
Jean Delvare [Sun, 10 Aug 2008 20:56:16 +0000 (22:56 +0200)]
i2c: Check for address business before creating clients

We check for address business in i2c_probe_address(),
i2c_detect_address() and i2c_new_probed_device(), but this isn't
sufficient. Drivers can call i2c_attach_client() and
i2c_new_device() on any address, so we must check the address there
as well.

This fixes bug #11239:
http://bugzilla.kernel.org/show_bug.cgi?id=11239

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Let users select algorithm drivers manually again
Jean Delvare [Sun, 10 Aug 2008 20:56:15 +0000 (22:56 +0200)]
i2c: Let users select algorithm drivers manually again

In kernel 2.6.26, the ability to select I2C algorithm drivers manually
was removed, as all in-kernel drivers do that automatically. However
there were some complaints that it was a problem for out-of-tree I2C
bus drivers. In order to address these complaints, let's allow manual
selection of these drivers again, but still hide them by default for
better general user experience.

This closes bug #11140:
http://bugzilla.kernel.org/show_bug.cgi?id=11140

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Fix NULL pointer dereference in i2c_new_probed_device
Hans Verkuil [Sun, 10 Aug 2008 20:56:15 +0000 (22:56 +0200)]
i2c: Fix NULL pointer dereference in i2c_new_probed_device

Fix a NULL pointer dereference that happened when calling
i2c_new_probed_device on one of the addresses for which we use byte
reads instead of quick write for detection purpose (that is: 0x30-0x37
and 0x50-0x5f).

Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Fix oops on bus multiplexer driver loading
Jean Delvare [Sun, 10 Aug 2008 20:56:15 +0000 (22:56 +0200)]
i2c: Fix oops on bus multiplexer driver loading

The two I2C bus multiplexer drivers (i2c-amd756-s4882 and
i2c-nforce2-s4985) make use of the bus they want to multiplex before
checking if it is really present. Swap the instructions to test for
presence first. This fixes a oops reported by Ingo Molnar.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Ingo Molnar <mingo@elte.hu>
16 years ago[WATCHDOG] fix watchdog/ixp4xx_wdt.c compilation
Adrian Bunk [Sun, 10 Aug 2008 11:03:41 +0000 (14:03 +0300)]
[WATCHDOG] fix watchdog/ixp4xx_wdt.c compilation

This patch fixes the following compile error caused by
commit 20d35f3e50ea7e573f9568b9fce4e98523aaee5d
([WATCHDOG 22/57] ixp4xx_wdt: unlocked_ioctl):

<--  snip  -->

...
  CC      drivers/watchdog/ixp4xx_wdt.o
ixp4xx_wdt.c:32: error: expected '=', ',', ';', 'asm' or '__attribute__'
ixp4xx_wdt.c: In function 'wdt_enable':
ixp4xx_wdt.c:41: error: 'wdt_lock' undeclared (first use in this
ixp4xx_wdt.c:41: error: (Each undeclared identifier is reported only
ixp4xx_wdt.c:41: error: for each function it appears in.)
ixp4xx_wdt.c: In function 'wdt_disable':
ixp4xx_wdt.c:52: error: 'wdt_lock' undeclared (first use in this
ixp4xx_wdt.c: In function 'ixp4xx_wdt_init':
ixp4xx_wdt.c:186: error: 'wdt_lock' undeclared (first use in this
make[3]: *** [drivers/watchdog/ixp4xx_wdt.o] Error 1

<--  snip  -->

Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
16 years ago[WATCHDOG] fix watchdog/wdt285.c compilation
Adrian Bunk [Fri, 8 Aug 2008 16:03:46 +0000 (19:03 +0300)]
[WATCHDOG] fix watchdog/wdt285.c compilation

This patch fixes the following compile error caused by
commit d0e58eed05f9baf77c4f75e794ae245f6dae240a
([WATCHDOG 55/57] wdt285: switch to unlocked_ioctl and tidy up ...):

<--  snip  -->

...
  CC [M]  drivers/watchdog/wdt285.o
wdt285.c: In function 'footbridge_watchdog_init':
wdt285.c:211: error: 'KERN_WARN' undeclared (first use in this function)
wdt285.c:211: error: (Each undeclared identifier is reported only once
wdt285.c:211: error: for each function it appears in.)
wdt285.c:212: error: expected ')' before string constant
make[3]: *** [drivers/watchdog/wdt285.o] Error 1

<--  snip  -->

Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
16 years ago[WATCHDOG] fix watchdog/at91rm9200_wdt.c compilation
Adrian Bunk [Fri, 8 Aug 2008 15:57:45 +0000 (18:57 +0300)]
[WATCHDOG] fix watchdog/at91rm9200_wdt.c compilation

This patch fixes the following compile error:

<--  snip  -->

...
  CC      drivers/watchdog/at91rm9200_wdt.o
at91rm9200_wdt.c:188: error: 'at91_wdt_ioctl' undeclared here (not in a
make[3]: *** [drivers/watchdog/at91rm9200_wdt.o] Error 1

<--  snip  -->

Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>