14 years ago8250: fix boot hang with serial console when using with Serial Over Lan port
Mauro Carvalho Chehab [Fri, 20 Feb 2009 23:38:52 +0000 (15:38 -0800)]
8250: fix boot hang with serial console when using with Serial Over Lan port

Intel 8257x Ethernet boards have a feature called Serial Over Lan.

This feature works by emulating a serial port, and it is detected by
kernel as a normal 8250 port.  However, this emulation is not perfect, as
also noticed on changeset 7500b1f602aad75901774a67a687ee985d85893f.

Before this patch, the kernel were trying to check if the serial TX is
capable of work using IRQ's.

This were done with a code similar this:

        serial_outp(up, UART_IER, UART_IER_THRI);
        lsr = serial_in(up, UART_LSR);
        iir = serial_in(up, UART_IIR);
        serial_outp(up, UART_IER, 0);

        if (lsr & UART_LSR_TEMT && iir & UART_IIR_NO_INT)
up->bugs |= UART_BUG_TXEN;

This works fine for other 8250 ports, but, on 8250-emulated SoL port, the
chip is a little lazy to down UART_IIR_NO_INT at UART_IIR register.

Due to that, UART_BUG_TXEN is sometimes enabled.  However, as TX IRQ keeps
working, and the TX polling is now enabled, the driver miss-interprets the
IRQ received later, hanging up the machine until a key is pressed at the
serial console.

This is the 6 version of this patch.  Previous versions were trying to
introduce a large enough delay between serial_outp and serial_in(up,
UART_IIR), but not taking forever.  However, the needed delay couldn't be
safely determined.

At the experimental tests, a delay of 1us solves most of the cases, but
still hangs sometimes.  Increasing the delay to 5us was better, but still
doesn't solve.  A very high delay of 50 ms seemed to work every time.

However, poking around with delays and pray for it to be enough doesn't
seem to be a good approach, even for a quirk.

So, instead of playing with random large arbitrary delays, let's just
disable UART_BUG_TXEN for all SoL ports.

[ fix warnings]
Signed-off-by: Mauro Carvalho Chehab <>
Cc: Alan Cox <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoatyfb: remove unused local variable `pwr_command'
Yang Hongyang [Fri, 20 Feb 2009 23:38:51 +0000 (15:38 -0800)]
atyfb: remove unused local variable `pwr_command'

Signed-off-by: Yang Hongyang<>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agospi_bitbang: add more lowlevel function documentation
Michael Buesch [Fri, 20 Feb 2009 23:38:49 +0000 (15:38 -0800)]
spi_bitbang: add more lowlevel function documentation

This adds more documentation of the lowlevel API to avoid future bugs.

Signed-off-by: Michael Buesch <>
Acked-by: David Brownell <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agovmalloc: call flush_cache_vunmap() from unmap_kernel_range()
Tejun Heo [Fri, 20 Feb 2009 23:38:48 +0000 (15:38 -0800)]
vmalloc: call flush_cache_vunmap() from unmap_kernel_range()

Impact: proper vcache flush on unmap_kernel_range()

flush_cache_vunmap() should be called before pages are unmapped.  Add
a call to it in unmap_kernel_range().

Signed-off-by: Tejun Heo <>
Acked-by: Nick Piggin <>
Acked-by: David S. Miller <>
Cc: <> [2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agocpuset: various documentation fixes and updates
Li Zefan [Fri, 20 Feb 2009 23:38:48 +0000 (15:38 -0800)]
cpuset: various documentation fixes and updates

I noticed the old commit 8f5aa26c75b7722e80c0c5c5bb833d41865d7019
("cpusets: update_cpumask documentation fix") is not a complete fix,
resulting in inconsistent paragraphs.  This patch fixes it and does other
fixes and updates:

- s/migrate_all_tasks()/migrate_live_tasks()/
- describe more cpuset control files
- s/cpumask_t/struct cpumask/
- document cpu hotplug and change of 'sched_relax_domain_level' may cause
  domain rebuild
- document various ways to query and modify cpusets
- the equivalent of "mount -t cpuset" is "mount -t cgroup -o cpuset,noprefix"

Signed-off-by: Li Zefan <>
Acked-by: Randy Dunlap <>
Cc: Paul Menage <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agodocsrc: use config instead of menuconfig
Randy Dunlap [Fri, 20 Feb 2009 23:38:47 +0000 (15:38 -0800)]
docsrc: use config instead of menuconfig

BUILD_DOCSRC should be controlled by "config" instead of "menuconfig".
I have no idea how I managed to use "menuconfig" here.

Signed-off-by: Randy Dunlap <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agosx.c: avoid referencing freed memory if copy_from_user() fails
Dan Carpenter [Fri, 20 Feb 2009 23:38:46 +0000 (15:38 -0800)]
sx.c: avoid referencing freed memory if copy_from_user() fails

The "break" would just result in reusing a free'd pointer.  I don't have
the cards myself to test it though.  :/

Signed-off-by: Dan Carpenter <>
Cc: Ilpo Järvinen <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agosx.c: fix dbl statement if - add missing braces
Ilpo Järvinen [Fri, 20 Feb 2009 23:38:45 +0000 (15:38 -0800)]
sx.c: fix dbl statement if - add missing braces

Caused by 736d54533aed (sx.c: fix missed unlock_kernel() on error path in
sx_fw_ioctl()).  You guys keep breaking things this way in every single
kernel release in at least couple of places...  :-(

Signed-off-by: Ilpo Järvinen <>
Acked-by: Dan Carpenter <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoREADME: fix a wrong filename
Li Zefan [Fri, 20 Feb 2009 23:38:43 +0000 (15:38 -0800)]
README: fix a wrong filename

It should be Documentation/build/kconfig.txt.

Introduced by commit 2af238e455ef5fd31c2f7a06c2db3f13d843b9bf
("kbuild: make *config usage docs").

Signed-off-by: Li Zefan <>
Cc: Sam Ravnborg <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoslab: introduce kzfree()
Johannes Weiner [Fri, 20 Feb 2009 23:38:41 +0000 (15:38 -0800)]
slab: introduce kzfree()

kzfree() is a wrapper for kfree() that additionally zeroes the underlying
memory before releasing it to the slab allocator.

Currently there is code which memset()s the memory region of an object
before releasing it back to the slab allocator to make sure
security-sensitive data are really zeroed out after use.

These callsites can then just use kzfree() which saves some code, makes
users greppable and allows for a stupid destructor that isn't necessarily
aware of the actual object size.

Signed-off-by: Johannes Weiner <>
Reviewed-by: Pekka Enberg <>
Cc: Matt Mackall <>
Acked-by: Christoph Lameter <>
Cc: Nick Piggin <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agomn10300: fix typo && -> || in arch/mn10300/unit-asb2305/pci.c
Wei Yongjun [Fri, 20 Feb 2009 23:38:40 +0000 (15:38 -0800)]
mn10300: fix typo && -> || in arch/mn10300/unit-asb2305/pci.c

Fix the typo && -> ||.

Signed-off-by: Wei Yongjun <>
Signed-off-by: David Howells <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agomn10300: fix oprofile
David Howells [Fri, 20 Feb 2009 23:38:38 +0000 (15:38 -0800)]
mn10300: fix oprofile

oprofile for MN10300 seems to have been broken by the advent of the new
tracing framework.

Signed-off-by: David Howells <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agouml: fix vde network backend in user mode linux
Luca Bigliardi [Fri, 20 Feb 2009 23:38:36 +0000 (15:38 -0800)]
uml: fix vde network backend in user mode linux

* Replace kmalloc() with uml_kmalloc() (fix build failure)

* Remove unnecessary UM_KERN_INFO in printk() (don't display '<6>' while
  printing info)

Signed-off-by: Luca Bigliardi <>
Cc: Jiri Kosina <>
Reviewed-by: WANG Cong <>
Cc: Jeff Dike <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 21:09:32 +0000 (13:09 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  Revert "[XFS] remove old vmap cache"
  Revert "[XFS] use scalable vmap API"

14 years agoMerge branch 'release' of git://
Linus Torvalds [Thu, 19 Feb 2009 21:09:20 +0000 (13:09 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://
  [IA64] xen_domu build fix
  [IA64] fixes configs and add default config for ia64 xen domU
  [IA64] Remove redundant cpu_clear() in __cpu_disable path
  [IA64] Revert "prevent ia64 from invoking irq handlers on offline CPUs"
  [IA64] bte_copy of BTE_MAX_XFER trips BUG_ON.
  [IA64] Build fix for __early_pfn_to_nid() undefined link error

14 years ago[IA64] xen_domu build fix
Tony Luck [Thu, 19 Feb 2009 20:05:00 +0000 (12:05 -0800)]
[IA64] xen_domu build fix

arch/ia64/xen/xen_pv_ops.c:156: error: xen_init_ops causes a section type conflict
arch/ia64/xen/xen_pv_ops.c:340: error: xen_iosapic_ops causes a section type conflict

Signed-off-by: Tony Luck <>
14 years ago[IA64] fixes configs and add default config for ia64 xen domU
Isaku Yamahata [Fri, 16 Jan 2009 03:17:30 +0000 (12:17 +0900)]
[IA64] fixes configs and add default config for ia64 xen domU

This patch fixes xen related Kconfigs and add default config
file for ia64 xen domU.

Signed-off-by: Isaku Yamahata <>
Signed-off-by: Tony Luck <aegl@agluck-desktop.(none)>
14 years ago[IA64] Remove redundant cpu_clear() in __cpu_disable path
Alex Chiang [Mon, 9 Feb 2009 18:16:57 +0000 (11:16 -0700)]
[IA64] Remove redundant cpu_clear() in __cpu_disable path

The second call to cpu_clear() is redundant, as we've already removed
the CPU from cpu_online_map before calling migrate_platform_irqs().

Signed-off-by: Alex Chiang <>
Signed-off-by: Tony Luck <aegl@agluck-desktop.(none)>
14 years ago[IA64] Revert "prevent ia64 from invoking irq handlers on offline CPUs"
Alex Chiang [Mon, 9 Feb 2009 18:16:16 +0000 (11:16 -0700)]
[IA64] Revert "prevent ia64 from invoking irq handlers on offline CPUs"

This reverts commit e7b140365b86aaf94374214c6f4e6decbee2eb0a.

Commit e7b14036 removes the targetted disabled CPU from the
cpu_online_map after calls to migrate_platform_irqs and fixup_irqs.

Paul McKenney states that the reasoning behind the patch was to
prevent irq handlers from running on CPUs marked offline because:

RCU happily ignores CPUs that don't have their bits set in
cpu_online_map, so if there are RCU read-side critical sections
in the irq handlers being run, RCU will ignore them.  If the
other CPUs were running, they might sequence through the RCU
state machine, which could result in data structures being
yanked out from under those irq handlers, which in turn could
result in oopses or worse.

Unfortunately, both ia64 functions above look at cpu_online_map to find
a new CPU to migrate interrupts onto. This means we can potentially
migrate an interrupt off ourself back to... ourself. Uh oh.

This causes an oops when we finally try to process pending interrupts on
the CPU we want to disable. The oops results from calling __do_IRQ with
a NULL pt_regs:

Unable to handle kernel NULL pointer dereference (address 0000000000000040)
Call Trace:
 [<a000000100016930>] show_stack+0x50/0xa0
                                sp=e0000009c922fa00 bsp=e0000009c92214d0
 [<a0000001000171a0>] show_regs+0x820/0x860
                                sp=e0000009c922fbd0 bsp=e0000009c9221478
 [<a00000010003c700>] die+0x1a0/0x2e0
                                sp=e0000009c922fbd0 bsp=e0000009c9221438
 [<a0000001006e92f0>] ia64_do_page_fault+0x950/0xa80
                                sp=e0000009c922fbd0 bsp=e0000009c92213d8
 [<a00000010000c7a0>] ia64_native_leave_kernel+0x0/0x270
                                sp=e0000009c922fc60 bsp=e0000009c92213d8
 [<a0000001000ecdb0>] profile_tick+0xd0/0x1c0
                                sp=e0000009c922fe30 bsp=e0000009c9221398
 [<a00000010003bb90>] timer_interrupt+0x170/0x3e0
                                sp=e0000009c922fe30 bsp=e0000009c9221330
 [<a00000010013a800>] handle_IRQ_event+0x80/0x120
                                sp=e0000009c922fe30 bsp=e0000009c92212f8
 [<a00000010013aa00>] __do_IRQ+0x160/0x4a0
                                sp=e0000009c922fe30 bsp=e0000009c9221290
 [<a000000100012290>] ia64_process_pending_intr+0x2b0/0x360
                                sp=e0000009c922fe30 bsp=e0000009c9221208
 [<a0000001000112d0>] fixup_irqs+0xf0/0x2a0
                                sp=e0000009c922fe30 bsp=e0000009c92211a8
 [<a00000010005bd80>] __cpu_disable+0x140/0x240
                                sp=e0000009c922fe30 bsp=e0000009c9221168
 [<a0000001006c5870>] take_cpu_down+0x50/0xa0
                                sp=e0000009c922fe30 bsp=e0000009c9221148
 [<a000000100122610>] stop_cpu+0xd0/0x200
                                sp=e0000009c922fe30 bsp=e0000009c92210f0
 [<a0000001000e0440>] kthread+0xc0/0x140
                                sp=e0000009c922fe30 bsp=e0000009c92210c8
 [<a000000100014ab0>] kernel_thread_helper+0xd0/0x100
                                sp=e0000009c922fe30 bsp=e0000009c92210a0
 [<a00000010000a4c0>] start_kernel_thread+0x20/0x40
                                sp=e0000009c922fe30 bsp=e0000009c92210a0

I don't like this revert because it is fragile. ia64 is getting lucky
because we seem to only ever process timer interrupts in this path, but
if we ever race with an IPI here, we definitely use RCU and have the
potential of hitting an oops that Paul describes above.

Patching ia64's timer_interrupt() to check for NULL pt_regs is
insufficient though, as we still hit the above oops.

As a short term solution, I do think that this revert is the right
answer. The revert hold up under repeated testing (24+ hour test runs)
with this setup:

- 8-way rx6600
- randomly toggling CPU online/offline state every 2 seconds
- running CPU exercisers, memory hog, disk exercisers, and
  network stressors
- average system load around ~160

In the long term, we really need to figure out why we set pt_regs = NULL
in ia64_process_pending_intr(). If it turns out that it is unnecessary
to do so, then we could safely re-introduce e7b14036 (along with some
other logic to be smarter about migrating interrupts).

One final note: x86 also removes the disabled CPU from cpu_online_map
and then re-enables interrupts for 1ms, presumably to handle any pending

arch/x86/kernel/irq_32.c (and irq_64.c):
[remove cpu from cpu_online_map]

[break CPU affinities]


So they are doing implicitly what ia64 is doing explicitly.

Signed-off-by: Alex Chiang <>
Signed-off-by: Tony Luck <aegl@agluck-desktop.(none)>
14 years ago[IA64] bte_copy of BTE_MAX_XFER trips BUG_ON.
Robin Holt [Wed, 4 Feb 2009 00:40:59 +0000 (18:40 -0600)]
[IA64] bte_copy of BTE_MAX_XFER trips BUG_ON.

BTE_MAX_XFER is wrong.  It is one greater than the number of cache
lines the BTE is actually able to transfer.  If you request a transfer
of exactly BTE_MAX_XFER size, you trip a very cryptic BUG_ON() which
should certainly be made more clear.

This patch fixes that constant and also cleans up the BUG_ON()s in
arch/ia64/sn/kernel/bte.c to test one condition per line.

Signed-off-by: Robin Holt <>
Signed-off-by: Tony Luck <aegl@agluck-desktop.(none)>
14 years ago[IA64] Build fix for __early_pfn_to_nid() undefined link error
Tony Luck [Thu, 19 Feb 2009 19:22:36 +0000 (11:22 -0800)]
[IA64] Build fix for __early_pfn_to_nid() undefined link error

ia64 only defines __early_pfn_to_nid() for SPARSEMEM && NUMA configurations,
so the recent:

commit: f2dbcfa738368c8a40d4a5f0b65dc9879577cb21
mm: clean up for early_pfn_to_nid()

ends up with some link problems for certain configuration files.

Fix arch/ia64/Kconfig to only define HAVE_ARCH_EARLY_PFN_TO_NID in the
cases where we do provide this function.

Signed-off-by: Tony Luck <>
14 years agoRevert "[XFS] remove old vmap cache"
Felix Blyakher [Wed, 18 Feb 2009 21:56:51 +0000 (15:56 -0600)]
Revert "[XFS] remove old vmap cache"

This reverts commit d2859751cd0bf586941ffa7308635a293f943c17.

This commit caused regression. We'll try to fix use of new
vmap API for next release.

Signed-off-by: Christoph Hellwig <>
Signed-off-by: Felix Blyakher <>
14 years agoRevert "[XFS] use scalable vmap API"
Felix Blyakher [Wed, 18 Feb 2009 21:41:28 +0000 (15:41 -0600)]
Revert "[XFS] use scalable vmap API"

This reverts commit 95f8e302c04c0b0c6de35ab399a5551605eeb006.

This commit caused regression. We'll try to fix use of new
vmap API for next release.

Signed-off-by: Christoph Hellwig <>
Signed-off-by: Felix Blyakher <>
14 years agoMerge
Linus Torvalds [Thu, 19 Feb 2009 17:52:12 +0000 (09:52 -0800)]

  [ARM] 5405/1: ep93xx: remove unused gesbc9312.h header
  [ARM] 5404/1: Fix condition in arm_elf_read_implies_exec() to set READ_IMPLIES_EXEC
  [ARM] omap: fix clock reparenting in omap2_clk_set_parent()
  [ARM] 5403/1: pxa25x_ep_fifo_flush() *ep->reg_udccs always set to 0
  [ARM] 5402/1: fix a case of wrap-around in sanity_check_meminfo()
  [ARM] 5401/1: Orion: fix edge triggered GPIO interrupt support
  [ARM] 5400/1: Add support for inverted rdy_busy pin for Atmel nand device controller
  [ARM] 5391/1: AT91: Enable GPIO clocks earlier
  [ARM] 5390/1: AT91: Watchdog fixes
  [ARM] 5398/1: Add Wan ZongShun to MAINTAINERS for W90P910
  [ARM] omap: fix _omap2_clksel_get_src_field()
  [ARM] omap: fix omap2_divisor_to_clksel() error return value

14 years agoMerge branch 'x86-fixes-for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 17:14:35 +0000 (09:14 -0800)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://
  x86, mce: fix ifdef for 64bit thermal apic vector clear on shutdown
  x86, mce: use force_sig_info to kill process in machine check
  x86, mce: reinitialize per cpu features on resume
  x86, rcu: fix strange load average and ksoftirqd behavior

14 years agoMerge branch 'tracing-fixes-for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 17:14:22 +0000 (09:14 -0800)]
Merge branch 'tracing-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://
  tracing: limit the number of loops the ring buffer self test can make
  tracing: have function trace select kallsyms
  tracing: disable tracing while testing ring buffer
  tracing/function-graph-tracer: trace the idle tasks

14 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 16:35:52 +0000 (08:35 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  [S390] fix "mem=" handling in case of standby memory
  [S390] Fix timeval regression on s390
  [S390] sclp: handle empty event buffers

14 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 16:35:29 +0000 (08:35 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://
  sound: virtuoso: revert "do not overwrite EEPROM on Xonar D2/D2X"
  ALSA: jack - Use card->shortname for input name
  ALSA: usb-audio - Workaround for misdetected sample rate with CM6207
  ALSA: usb-audio - Fix non-continuous rate detection
  sound: usb-audio: fix uninitialized variable with M-Audio MIDI interfaces
  Revert "Sound: hda - Restore PCI configuration space with interrupts off"

14 years ago[ARM] 5405/1: ep93xx: remove unused gesbc9312.h header
Hartley Sweeten [Thu, 19 Feb 2009 16:09:04 +0000 (17:09 +0100)]
[ARM] 5405/1: ep93xx: remove unused gesbc9312.h header

Remove the gesbc9312.h header since it is unused.

Signed-off-by: H Hartley Sweeten <>
Signed-off-by: Russell King <>
14 years ago[ARM] 5404/1: Fix condition in arm_elf_read_implies_exec() to set READ_IMPLIES_EXEC
Makito SHIOKAWA [Thu, 19 Feb 2009 14:34:59 +0000 (15:34 +0100)]
[ARM] 5404/1: Fix condition in arm_elf_read_implies_exec() to set READ_IMPLIES_EXEC

READ_IMPLIES_EXEC must be set when:
o binary _is_ an executable stack (i.e. not EXSTACK_DISABLE_X)
o processor architecture is _under_ ARMv6 (XN bit is supported from ARMv6)

Signed-off-by: Makito SHIOKAWA <>
Signed-off-by: Russell King <>
14 years ago[S390] fix "mem=" handling in case of standby memory
Heiko Carstens [Thu, 19 Feb 2009 14:19:01 +0000 (15:19 +0100)]
[S390] fix "mem=" handling in case of standby memory

Standby memory detected with the sclp interface gets always registered
with add_memory calls without considering the limitationt that the
"mem=" kernel paramater implies.
So fix this and only register standby memory that is below the specified
This fixes zfcpdump since it uses "mem=32M". In case there is appr.
2GB standby memory present all of usable memory would be used for the
struct pages needed for standby memory.

Signed-off-by: Heiko Carstens <>
Signed-off-by: Martin Schwidefsky <>
14 years ago[S390] Fix timeval regression on s390
Christian Borntraeger [Thu, 19 Feb 2009 14:19:00 +0000 (15:19 +0100)]
[S390] Fix timeval regression on s390

commit aa5e97ce4bbc9d5daeec16b1d15bb3f6b7b4f4d4
[PATCH] improve precision of process accounting.

Introduced a timing regression:
-bash-3.2# time ls
real    0m0.006s
user    0m1.754s
sys     0m1.094s

The problem was introduced by an error in cputime_to_timeval.
Cputime is now 1/4096 microsecond, therefore, we have to divide
the remainder with 4096 to get the microseconds.

Signed-off-by: Christian Borntraeger <>
Signed-off-by: Martin Schwidefsky <>
14 years ago[S390] sclp: handle empty event buffers
Peter Oberparleiter [Thu, 19 Feb 2009 14:18:59 +0000 (15:18 +0100)]
[S390] sclp: handle empty event buffers

Handle a malformed hardware response which some versions of the
Support Element (SE) may present during SE restart and which otherwise
would result in an endless loop in function sclp_dispatch_evbufs.

Signed-off-by: Peter Oberparleiter <>
Signed-off-by: Martin Schwidefsky <>
14 years ago[ARM] omap: fix clock reparenting in omap2_clk_set_parent()
Russell King [Thu, 19 Feb 2009 13:25:16 +0000 (13:25 +0000)]
[ARM] omap: fix clock reparenting in omap2_clk_set_parent()

When changing the parent of a clock, it is necessary to keep the
clock use counts balanced otherwise things the parent state will
get corrupted.  Since we already disable and re-enable the clock,
we might as well use the recursive versions instead.

Signed-off-by: Russell King <>
14 years agoMerge branch 'fix/usb-audio' into for-linus
Takashi Iwai [Thu, 19 Feb 2009 12:58:05 +0000 (13:58 +0100)]
Merge branch 'fix/usb-audio' into for-linus

14 years agoMerge branch 'fix/misc' into for-linus
Takashi Iwai [Thu, 19 Feb 2009 12:58:01 +0000 (13:58 +0100)]
Merge branch 'fix/misc' into for-linus

14 years agoMerge branch 'fix/hda' into for-linus
Takashi Iwai [Thu, 19 Feb 2009 12:57:52 +0000 (13:57 +0100)]
Merge branch 'fix/hda' into for-linus

14 years ago[ARM] 5403/1: pxa25x_ep_fifo_flush() *ep->reg_udccs always set to 0
Roel Kluin [Thu, 19 Feb 2009 10:57:46 +0000 (11:57 +0100)]
[ARM] 5403/1: pxa25x_ep_fifo_flush() *ep->reg_udccs always set to 0

*ep->reg_udccs is always set to 0.

Signed-off-by: Roel Kluin <>
Acked-by: Eric Miao <>
Signed-off-by: Russell King <>
14 years ago[ARM] 5402/1: fix a case of wrap-around in sanity_check_meminfo()
Nicolas Pitre [Wed, 18 Feb 2009 21:29:22 +0000 (22:29 +0100)]
[ARM] 5402/1: fix a case of wrap-around in sanity_check_meminfo()

In the non highmem case, if two memory banks of 1GB each are provided,
the second bank would evade suppression since its virtual base would
be 0.  Fix this by disallowing any memory bank which virtual base
address is found to be lower than PAGE_OFFSET.

Reported-by: Lennert Buytenhek <>
Signed-off-by: Nicolas Pitre <>
Signed-off-by: Russell King <>
14 years agoMerge branch 'tip/tracing/urgent' of git://
Ingo Molnar [Thu, 19 Feb 2009 09:16:45 +0000 (10:16 +0100)]
Merge branch 'tip/tracing/urgent' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into tracing/urgent

14 years agosound: virtuoso: revert "do not overwrite EEPROM on Xonar D2/D2X"
Clemens Ladisch [Tue, 17 Feb 2009 08:50:30 +0000 (09:50 +0100)]
sound: virtuoso: revert "do not overwrite EEPROM on Xonar D2/D2X"

This reverts commit 7e86c0e6850504ec9516b953f316a47277825e33 ("do not
overwrite EEPROM on Xonar D2/D2X") because it did not actually help with
the problem.

More user reports show that the overwriting of the EEPROM is not
triggered by using this driver but by installing Linux, and that the
installation of any other operating system (even one without any CMI8788
driver) has the same effect.  In other words, the presence of this
driver does not have any effect on the occurrence of the error.  (So
far, the available evidence seems to point to a BIOS bug.)

Furthermore, it turns out that the EEPROM chip is protected against
stray write commands by the command format and by requiring a separate
write-enable command, so the error scenario in the previous commit (that
SPI writes can be misinterpreted as an EEPROM write command) is not even
theoretically possible.

The mixer control that was removed as a consequence of the previous
commit can only be partially emulated in userspace, which also means it
cannot be seen be the in-kernel OSS API emulation, so it is better to
revert that change.

Signed-off-by: Clemens Ladisch <>
Cc: <>
Signed-off-by: Takashi Iwai <>
14 years agotracing: limit the number of loops the ring buffer self test can make
Steven Rostedt [Thu, 19 Feb 2009 03:50:01 +0000 (22:50 -0500)]
tracing: limit the number of loops the ring buffer self test can make

Impact: prevent deadlock if ring buffer gets corrupted

This patch adds a paranoid check to make sure the ring buffer consumer
does not go into an infinite loop. Since the ring buffer has been set
to read only, the consumer should not loop for more than the ring buffer
size. A check is added to make sure the consumer does not loop more than
the ring buffer size.

Signed-off-by: Steven Rostedt <>
14 years agotracing: have function trace select kallsyms
Steven Rostedt [Thu, 19 Feb 2009 03:06:18 +0000 (22:06 -0500)]
tracing: have function trace select kallsyms

Impact: fix output of function tracer to be useful

The function tracer is pretty useless if KALLSYMS is not configured.
Unless you are good at reading hex values, the function tracer should
select the KALLSYMS configuration.

Also, the dynamic function tracer will fail its self test if KALLSYMS
is not selected.

Signed-off-by: Steven Rostedt <>
14 years agotracing: disable tracing while testing ring buffer
Steven Rostedt [Wed, 18 Feb 2009 23:33:57 +0000 (18:33 -0500)]
tracing: disable tracing while testing ring buffer

Impact: fix to prevent hard lockup on self tests

If one of the tracers are broken and is constantly filling the ring
buffer while the test of the ring buffer is running, it will hang
the box. The reason is that the test is a consumer that will not
stop till the ring buffer is empty. But if the tracer is broken and
is constantly producing input to the buffer, this test will never
end. The result is a lockup of the box.

This happened when KALLSYMS was not defined and the dynamic ftrace
test constantly filled the ring buffer, because the filter failed
and all functions were being traced. Something was being called
that constantly filled the buffer.

Signed-off-by: Steven Rostedt <>
14 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 02:33:04 +0000 (18:33 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  block: fix deadlock in blk_abort_queue() for drivers that readd to timeout list
  block: fix booting from partitioned md array
  block: revert part of 18ce3751ccd488c78d3827e9f6bf54e6322676fb
  cciss: PCI power management reset for kexec
  paride/pg.c: xs(): &&/|| confusion
  fs/bio: bio_alloc_bioset: pass right object ptr to mempool_free
  block: fix bad definition of BIO_RW_SYNC
  bsg: Fix sense buffer bug in SG_IO

14 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 19 Feb 2009 01:55:15 +0000 (17:55 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://
  omap_hsmmc: Change while(); loops with finite version
  omap_hsmmc: recover from transfer failures
  omap_hsmmc: only MMC1 allows HCTL.SDVS != 1.8V
  omap_hsmmc: card detect irq bugfix
  sdhci: fix led naming
  mmc_test: fix basic read test
  s3cmci: Fix hangup in do_pio_write()
  Revert "sdhci: force high speed capability on some controllers"
  MMC: fix bug - SDHC card capacity not correct

14 years agoinotify: fix GFP_KERNEL related deadlock
Ingo Molnar [Wed, 18 Feb 2009 22:48:43 +0000 (14:48 -0800)]
inotify: fix GFP_KERNEL related deadlock

Enhanced lockdep coverage of __GFP_NOFS turned up this new lockdep

[ 1093.677775]
[ 1093.677781] =================================
[ 1093.680031] [ INFO: inconsistent lock state ]
[ 1093.680031] 2.6.29-rc5-tip-01504-gb49eca1-dirty #1
[ 1093.680031] ---------------------------------
[ 1093.680031] inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage.
[ 1093.680031] kswapd0/308 [HC0[0]:SC0[0]:HE1:SE1] takes:
[ 1093.680031]  (&inode->inotify_mutex){+.+.?.}, at: [<c0205942>] inotify_inode_is_dead+0x20/0x80
[ 1093.680031] {RECLAIM_FS-ON-W} state was registered at:
[ 1093.680031]   [<c01696b9>] mark_held_locks+0x43/0x5b
[ 1093.680031]   [<c016baa4>] lockdep_trace_alloc+0x6c/0x6e
[ 1093.680031]   [<c01cf8b0>] kmem_cache_alloc+0x20/0x150
[ 1093.680031]   [<c040d0ec>] idr_pre_get+0x27/0x6c
[ 1093.680031]   [<c02056e3>] inotify_handle_get_wd+0x25/0xad
[ 1093.680031]   [<c0205f43>] inotify_add_watch+0x7a/0x129
[ 1093.680031]   [<c020679e>] sys_inotify_add_watch+0x20f/0x250
[ 1093.680031]   [<c010389e>] sysenter_do_call+0x12/0x35
[ 1093.680031]   [<ffffffff>] 0xffffffff
[ 1093.680031] irq event stamp: 60417
[ 1093.680031] hardirqs last  enabled at (60417): [<c018d5f5>] call_rcu+0x53/0x59
[ 1093.680031] hardirqs last disabled at (60416): [<c018d5b9>] call_rcu+0x17/0x59
[ 1093.680031] softirqs last  enabled at (59656): [<c0146229>] __do_softirq+0x157/0x16b
[ 1093.680031] softirqs last disabled at (59651): [<c0106293>] do_softirq+0x74/0x15d
[ 1093.680031]
[ 1093.680031] other info that might help us debug this:
[ 1093.680031] 2 locks held by kswapd0/308:
[ 1093.680031]  #0:  (shrinker_rwsem){++++..}, at: [<c01b0502>] shrink_slab+0x36/0x189
[ 1093.680031]  #1:  (&type->s_umount_key#4){+++++.}, at: [<c01e6d77>] shrink_dcache_memory+0x110/0x1fb
[ 1093.680031]
[ 1093.680031] stack backtrace:
[ 1093.680031] Pid: 308, comm: kswapd0 Not tainted 2.6.29-rc5-tip-01504-gb49eca1-dirty #1
[ 1093.680031] Call Trace:
[ 1093.680031]  [<c016947a>] valid_state+0x12a/0x13d
[ 1093.680031]  [<c016954e>] mark_lock+0xc1/0x1e9
[ 1093.680031]  [<c016a5b4>] ? check_usage_forwards+0x0/0x3f
[ 1093.680031]  [<c016ab74>] __lock_acquire+0x2c6/0xac8
[ 1093.680031]  [<c01688d9>] ? register_lock_class+0x17/0x228
[ 1093.680031]  [<c016b3d3>] lock_acquire+0x5d/0x7a
[ 1093.680031]  [<c0205942>] ? inotify_inode_is_dead+0x20/0x80
[ 1093.680031]  [<c08824c4>] __mutex_lock_common+0x3a/0x4cb
[ 1093.680031]  [<c0205942>] ? inotify_inode_is_dead+0x20/0x80
[ 1093.680031]  [<c08829ed>] mutex_lock_nested+0x2e/0x36
[ 1093.680031]  [<c0205942>] ? inotify_inode_is_dead+0x20/0x80
[ 1093.680031]  [<c0205942>] inotify_inode_is_dead+0x20/0x80
[ 1093.680031]  [<c01e6672>] dentry_iput+0x90/0xc2
[ 1093.680031]  [<c01e67a3>] d_kill+0x21/0x45
[ 1093.680031]  [<c01e6a46>] __shrink_dcache_sb+0x27f/0x355
[ 1093.680031]  [<c01e6dc5>] shrink_dcache_memory+0x15e/0x1fb
[ 1093.680031]  [<c01b05ed>] shrink_slab+0x121/0x189
[ 1093.680031]  [<c01b0d12>] kswapd+0x39f/0x561
[ 1093.680031]  [<c01ae499>] ? isolate_pages_global+0x0/0x233
[ 1093.680031]  [<c0157eae>] ? autoremove_wake_function+0x0/0x43
[ 1093.680031]  [<c01b0973>] ? kswapd+0x0/0x561
[ 1093.680031]  [<c0157daf>] kthread+0x41/0x82
[ 1093.680031]  [<c0157d6e>] ? kthread+0x0/0x82
[ 1093.680031]  [<c01043ab>] kernel_thread_helper+0x7/0x10

inotify_handle_get_wd() does idr_pre_get() which does a
kmem_cache_alloc() without __GFP_FS - and is hence deadlockable under
extreme MM pressure.

Signed-off-by: Ingo Molnar <>
Acked-by: Peter Zijlstra <>
Cc: MinChan Kim <>
Cc: Nick Piggin <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agospi-gpio: sanitize MISO bitvalue
Michael Buesch [Wed, 18 Feb 2009 22:48:41 +0000 (14:48 -0800)]
spi-gpio: sanitize MISO bitvalue

gpio_get_value() returns 0 or nonzero, but getmiso() expects 0 or 1.
Sanitize the value to a 0/1 boolean.

Signed-off-by: Michael Buesch <>
Acked-by: David Brownell <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoBernhard has moved
Bernhard Walle [Wed, 18 Feb 2009 22:48:40 +0000 (14:48 -0800)]
Bernhard has moved

Since I don't work for SUSE any more and the address is
invalid, correct it in the copyright headers and documentation.

Signed-off-by: Bernhard Walle <>
Cc: Greg KH <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agox86: dell-laptop: depends on POWER_SUPPLY
Randy Dunlap [Wed, 18 Feb 2009 22:48:39 +0000 (14:48 -0800)]
x86: dell-laptop: depends on POWER_SUPPLY

Build breaks when DELL_LAPTOP=y and POWER_SUPPLY=m.  DELL_LAPTOP needs to
depend on POWER_SUPPLY.

dell-laptop.c:(.text+0x1ef3c4): undefined reference to `power_supply_is_system_supplied'
dell-laptop.c:(.text+0x1ef45e): undefined reference to `power_supply_is_system_supplied'

Signed-off-by: Randy Dunlap <>
Cc: Matthew Garrett <>
Cc: Ingo Molnar <>
Cc: Thomas Gleixner <>
Cc: "H. Peter Anvin" <>
Cc: Len Brown <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agovt: Declare PIO_CMAP/GIO_CMAP as compatbile ioctls.
Bill Nottingham [Wed, 18 Feb 2009 22:48:39 +0000 (14:48 -0800)]
vt: Declare PIO_CMAP/GIO_CMAP as compatbile ioctls.

Otherwise, these don't work when called from 32-bit userspace on 64-bit

Cc: Jiri Kosina <>
Cc: Alan Cox <>
Cc: <> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agofbdev/drm: fix Kconfig submenu mess in "Graphics support"
Krzysztof Helt [Wed, 18 Feb 2009 22:48:38 +0000 (14:48 -0800)]
fbdev/drm: fix Kconfig submenu mess in "Graphics support"

Submenus of the graphics support "Support for frame buffer devices" and
"Direct Rendering Manager (XFree86 4.1.0 and higher DRI support)" are
broken in half after latest changes for Intel 915 mode setting support.

The DRM subsection is broken because one option is put outside the choice
section it depends on.

The frame buffers part is broken then due to circular dependency.  Fix
this by make Intel frame buffers depend on CONFIG_INTEL_AGP.

Kconfigs are broken by d2f59357700487a8b944f4f7777d1e97cf5ea2ed
("drm/i915: select framebuffer support automatically").

This is probably not only way to fix this.

Signed-off-by: Krzysztof Helt <>
Cc: Ingo Molnar <>
Cc: Dave Airlie <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agofloppy: request and release only the ports we actually use
Philippe De Muyter [Wed, 18 Feb 2009 22:48:36 +0000 (14:48 -0800)]
floppy: request and release only the ports we actually use

The floppy driver requests an I/O port it doesn't need, and sometimes this
causes a conflict with a motherboard device reported by PNPBIOS.

This patch makes the floppy driver request and release only the ports it
actually uses.  It also factors out the request/release stuff and the
io-ports list so they're all in one place now.

The current floppy driver uses only these ports:

    0x3f2 (FD_DOR)
    0x3f4 (FD_STATUS)
    0x3f5 (FD_DATA)
    0x3f7 (FD_DCR/FD_DIR)

but it requests 0x3f2-0x3f5 and 0x3f7, which includes the unused port

Some BIOSes report 0x3f3 as a motherboard resource.  The PNP system driver
reserves that, which causes a conflict when the floppy driver requests
0x3f2-0x3f5 later.

Philippe reported that this conflict broke the floppy driver between
2.6.11 and 2.6.22.  His PNPBIOS reports these devices:

    $ cat 00:07/id 00:07/resources # motherboard device
    state = active
    io 0x80-0x80
    io 0x10-0x1f
    io 0x22-0x3f
    io 0x44-0x5f
    io 0x90-0x9f
    io 0xa2-0xbf
    io 0x3f0-0x3f1
    io 0x3f3-0x3f3

    $ cat 00:03/id 00:03/resources # floppy device
    state = active
    io 0x3f4-0x3f5
    io 0x3f2-0x3f2


Signed-off-by: Bjorn Helgaas <>
Signed-off-by: Philippe De Muyter <>
Reported-by: Philippe De Muyter <>
Tested-by: Philippe De Muyter <>
Cc: Adam M Belay <>
Cc: Robert Hancock <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agojsm: additional device support
Adam Lackorzynski [Wed, 18 Feb 2009 22:48:34 +0000 (14:48 -0800)]
jsm: additional device support

I have a Digi Neo 8 PCI card (114f:00b1) Serial controller: Digi
International Digi Neo 8 (rev 05)

that works with the jsm driver after using the following patch.

Signed-off-by: Adam Lackorzynski <>
Cc: Scott H Kilau <>
Cc: Wendy Xiong <>
Acked-by: Alan Cox <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agomm: fix memmap init for handling memory hole
KAMEZAWA Hiroyuki [Wed, 18 Feb 2009 22:48:33 +0000 (14:48 -0800)]
mm: fix memmap init for handling memory hole

Now, early_pfn_in_nid(PFN, NID) may returns false if PFN is a hole.
and memmap initialization was not done. This was a trouble for
sparc boot.

To fix this, the PFN should be initialized and marked as PG_reserved.
This patch changes early_pfn_in_nid() return true if PFN is a hole.

Signed-off-by: KAMEZAWA Hiroyuki <>
Reported-by: David Miller <>
Tested-by: KOSAKI Motohiro <>
Cc: Mel Gorman <>
Cc: Heiko Carstens <>
Cc: <> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agomm: clean up for early_pfn_to_nid()
KAMEZAWA Hiroyuki [Wed, 18 Feb 2009 22:48:32 +0000 (14:48 -0800)]
mm: clean up for early_pfn_to_nid()

What's happening is that the assertion in mm/page_alloc.c:move_freepages()
is triggering:

BUG_ON(page_zone(start_page) != page_zone(end_page));

Once I knew this is what was happening, I added some annotations:

if (unlikely(page_zone(start_page) != page_zone(end_page))) {
printk(KERN_ERR "move_freepages: Bogus zones: "
       "start_page[%p] end_page[%p] zone[%p]\n",
       start_page, end_page, zone);
printk(KERN_ERR "move_freepages: "
       "start_zone[%p] end_zone[%p]\n",
       page_zone(start_page), page_zone(end_page));
printk(KERN_ERR "move_freepages: "
       "start_pfn[0x%lx] end_pfn[0x%lx]\n",
       page_to_pfn(start_page), page_to_pfn(end_page));
printk(KERN_ERR "move_freepages: "
       "start_nid[%d] end_nid[%d]\n",
       page_to_nid(start_page), page_to_nid(end_page));

And here's what I got:

move_freepages: Bogus zones: start_page[2207d0000] end_page[2207dffc0] zone[fffff8103effcb00]
move_freepages: start_zone[fffff8103effcb00] end_zone[fffff8003fffeb00]
move_freepages: start_pfn[0x81f600] end_pfn[0x81f7ff]
move_freepages: start_nid[1] end_nid[0]

My memory layout on this box is:

[    0.000000] Zone PFN ranges:
[    0.000000]   Normal   0x00000000 -> 0x0081ff5d
[    0.000000] Movable zone start PFN for each node
[    0.000000] early_node_map[8] active PFN ranges
[    0.000000]     0: 0x00000000 -> 0x00020000
[    0.000000]     1: 0x00800000 -> 0x0081f7ff
[    0.000000]     1: 0x0081f800 -> 0x0081fe50
[    0.000000]     1: 0x0081fed1 -> 0x0081fed8
[    0.000000]     1: 0x0081feda -> 0x0081fedb
[    0.000000]     1: 0x0081fedd -> 0x0081fee5
[    0.000000]     1: 0x0081fee7 -> 0x0081ff51
[    0.000000]     1: 0x0081ff59 -> 0x0081ff5d

So it's a block move in that 0x81f600-->0x81f7ff region which triggers
the problem.

This patch:

Declaration of early_pfn_to_nid() is scattered over per-arch include
files, and it seems it's complicated to know when the declaration is used.
 I think it makes fix-for-memmap-init not easy.

This patch moves all declaration to include/linux/mm.h

After this,
     -> Use static definition in include/linux/mm.h
     -> Use generic definition in mm/page_alloc.c
     -> per-arch back end function will be called.

Signed-off-by: KAMEZAWA Hiroyuki <>
Tested-by: KOSAKI Motohiro <>
Reported-by: David Miller <>
Cc: Mel Gorman <>
Cc: Heiko Carstens <>
Cc: <> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agofs/super.c: add lockdep annotation to s_umount
Peter Zijlstra [Wed, 18 Feb 2009 22:48:30 +0000 (14:48 -0800)]
fs/super.c: add lockdep annotation to s_umount

Li Zefan said:

Thread 1:
  for ((; ;))
      mount -t cpuset xxx /mnt > /dev/null 2>&1
      cat /mnt/cpus > /dev/null 2>&1
      umount /mnt > /dev/null 2>&1

Thread 2:
  for ((; ;))
      mount -t cpuset xxx /mnt > /dev/null 2>&1
      umount /mnt > /dev/null 2>&1

(Note: It is irrelevant which cgroup subsys is used.)

After a while a lockdep warning showed up:

[ INFO: possible recursive locking detected ]
2.6.28 #479
mount/13554 is trying to acquire lock:
 (&type->s_umount_key#19){--..}, at: [<c049d888>] sget+0x5e/0x321

but task is already holding lock:
 (&type->s_umount_key#19){--..}, at: [<c049da0c>] sget+0x1e2/0x321

other info that might help us debug this:
1 lock held by mount/13554:
 #0:  (&type->s_umount_key#19){--..}, at: [<c049da0c>] sget+0x1e2/0x321

stack backtrace:
Pid: 13554, comm: mount Not tainted 2.6.28-mc #479
Call Trace:
 [<c044ad2e>] validate_chain+0x4c6/0xbbd
 [<c044ba9b>] __lock_acquire+0x676/0x700
 [<c044bb82>] lock_acquire+0x5d/0x7a
 [<c049d888>] ? sget+0x5e/0x321
 [<c061b9b8>] down_write+0x34/0x50
 [<c049d888>] ? sget+0x5e/0x321
 [<c049d888>] sget+0x5e/0x321
 [<c045a2e7>] ? cgroup_set_super+0x0/0x3e
 [<c045959f>] ? cgroup_test_super+0x0/0x2f
 [<c045bcea>] cgroup_get_sb+0x98/0x2e7
 [<c045cfb6>] cpuset_get_sb+0x4a/0x5f
 [<c049dfa4>] vfs_kern_mount+0x40/0x7b
 [<c049e02d>] do_kern_mount+0x37/0xbf
 [<c04af4a0>] do_mount+0x5c3/0x61a
 [<c04addd2>] ? copy_mount_options+0x2c/0x111
 [<c04af560>] sys_mount+0x69/0xa0
 [<c0403251>] sysenter_do_call+0x12/0x31

The cause is after alloc_super() and then retry, an old entry in list
fs_supers is found, so grab_super(old) is called, but both functions hold
s_umount lock:

struct super_block *sget(...)
if (test) {
list_for_each_entry(old, &type->fs_supers, s_instances) {
if (!test(old, data))
if (!grab_super(old))  <--- 2nd: down_write(&old->s_umount);
goto retry;
if (s)
return old;
if (!s) {
s = alloc_super(type);   <--- 1th: down_write(&s->s_umount)
if (!s)
return ERR_PTR(-ENOMEM);
goto retry;

It seems like a false positive, and seems like VFS but not cgroup needs to
be fixed.

Peter said:

We can simply put the new s_umount instance in a but lockdep doesn't
particularly cares about subclass order.

If there's any issue with the callers of sget() assuming the s_umount lock
being of sublcass 0, then there is another annotation we can use to fix
that, but lets not bother with that if this is sufficient.


Signed-off-by: Peter Zijlstra <>
Tested-by: Li Zefan <>
Reported-by: Li Zefan <>
Cc: Al Viro <>
Cc: Paul Menage <>
Cc: Arjan van de Ven <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoatmel_serial might lose modem status change
Atsushi Nemoto [Wed, 18 Feb 2009 22:48:28 +0000 (14:48 -0800)]
atmel_serial might lose modem status change

I found a problem of handling of modem status of atmel_serial driver.

With the commit 1ecc26 ("atmel_serial: split the interrupt handler"),
handling of modem status signal was splitted into two parts.  The
atmel_tasklet_func() compares new status with irq_status_prev, but
irq_status_prev is not correct if signal status was changed while the port
is closed.

Here is a sequence to cause problem:

1. Remote side sets CTS (and DSR).
2. Local side close the port.
3. Local side clears RTS and DTR.
4. Remote side clears CTS and DSR.
5. Local side reopen the port.  hw_stopped becomes 1.
6. Local side sets RTS and DTR.
7. Remote side sets CTS and DSR.

Then CTS change interrupt can be received, but since CTS bit in
irq_status_prev and new status is same, uart_handle_cts_change() will not
be called (so hw_stopped will not be cleared, i.e.  cannot send any data).

I suppose irq_status_prev should be initialized at somewhere in open

Itai Levi pointed out that we need to initialize atmel_port->irq_status
as well here. His analysis is as follows:

> Regarding the second part of the patch (which resets irq_status_prev),
> it turns out that both versions of the patch (mine and Atsushi's)
> still leave enough room for faulty behavior when opening the port.
> This is because we are not resetting both irq_status_prev and
> irq_status in atmel_startup() to CSR, which leads faulty behavior in
> the following sequences:
> First case:
> 1. closing the port while CTS line = 1 (TX not allowed)
> 2. setting CTS line = 0 (TX allowed)
> 3. opening the port
> 4. transmitting one char
> 5. Cannot transmit more chars, although CTS line is 0
> Second case:
> 1. closing the port while CTS line = 0 (TX allowed)
> 2. setting CTS line = 1 (TX not allowed)
> 3. opening the port
> 4. receiving some chars
> 5. Now we can transmit, although CTS line is 1
> This reason for this is that the tasklet is scheduled as a result of
> TX or RX interrupts (not a status change!), in steps 4 above. Inside
> the tasklet, the atmel_port->irq_status (which holds the value from
> the previous session) is compared to atmel_port->irq_status_prev.
> Hence, a status-change of the CTS line is faultily detected.
> Both cases were verified on 9260 hardware.

[ folded with patch from Itai Levi]
Signed-off-by: Atsushi Nemoto <>
Signed-off-by: Haavard Skinnemoen <>
Cc: Remy Bohmer <>
Cc: Marc Pignat <>
Cc: Itai Levi <>
Cc: Alan Cox <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoatmel-mci: fix initialization of dma slave data
Dan Williams [Wed, 18 Feb 2009 22:48:26 +0000 (14:48 -0800)]
atmel-mci: fix initialization of dma slave data

The conversion of atmel-mci to dma_request_channel missed the
initialization of the channel dma_slave information.  The filter_fn passed
to dma_request_channel is responsible for initializing the channel's
private data.  This implementation has the additional benefit of enabling
a generic client-channel data passing mechanism.

Reviewed-by: Atsushi Nemoto <>
Signed-off-by: Dan Williams <>
Acked-by: Haavard Skinnemoen <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agolis3lv02d: add axes knowledge of HP Pavilion dv5 models
Giuseppe Bilotta [Wed, 18 Feb 2009 22:48:25 +0000 (14:48 -0800)]
lis3lv02d: add axes knowledge of HP Pavilion dv5 models

Add support for HP Pavilion dv5.

Since Intel-based models have an inverted x axis, while AMD-based models
have an inverted y axis, we introduce a new macro that special-cases axis
orientation based on two DMI entries: HP dv5 axis configuration is then
based on both the PRODUCT and BOARD name.

Signed-off-by: Giuseppe Bilotta <>
Cc: Eric Piel <>
Cc: Pavel Machek <>
Tested-by: Palatis Tseng <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agolis3lv02d: support both one- and two-byte sensors
Giuseppe Bilotta [Wed, 18 Feb 2009 22:48:24 +0000 (14:48 -0800)]
lis3lv02d: support both one- and two-byte sensors

Sensors responding with 0x3B to WHO_AM_I only have one data register per
direction, thus returning a signed byte from the position which is
occupied by the MSB in sensors responding with 0x3A.

Since multiple sensors share the reply to WHO_AM_I, we rename the defines
to better indicate what they identify (family of single and double
precision sensors).

We support both kind of sensors by checking for the sensor type on init
and defining appropriate data-access routines and sensor limits (for the
joystick) depending on what we find.

[ coding-style fixes]
Signed-off-by: Giuseppe Bilotta <>
Acked-by: Eric Piel <>
Cc: Pavel Machek <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agohp accelerometer: add freefall detection
Pavel Machek [Wed, 18 Feb 2009 22:48:23 +0000 (14:48 -0800)]
hp accelerometer: add freefall detection

This adds freefall handling to hp_accel driver.  According to HP, it
should just work, without us having to set the chip up by hand.

hpfall.c is example .c program that parks the disk when accelerometer
detects free fall.  It should work; for now, it uses fixed 20seconds
protection period.

Signed-off-by: Pavel Machek <>
Cc: Thomas Renninger <>
Cc: Éric Piel <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoeeepc: should depend on INPUT
Alexey Dobriyan [Wed, 18 Feb 2009 22:48:22 +0000 (14:48 -0800)]
eeepc: should depend on INPUT

Otherwise with INPUT=m, EEEPC_LAPTOP=y one gets

drivers/built-in.o: In function `input_sync':
eeepc-laptop.c:(.text+0x18ce51): undefined reference to `input_event'
drivers/built-in.o: In function `input_report_key':
eeepc-laptop.c:(.text+0x18ce73): undefined reference to `input_event'
drivers/built-in.o: In function `eeepc_hotk_check':
eeepc-laptop.c:(.text+0x18d05f): undefined reference to `input_allocate_device'
eeepc-laptop.c:(.text+0x18d10f): undefined reference to `input_register_device'
eeepc-laptop.c:(.text+0x18d131): undefined reference to `input_free_device'
drivers/built-in.o: In function `eeepc_backlight_exit':
eeepc-laptop.c:(.text+0x18d546): undefined reference to `input_unregister_device'

Signed-off-by: Alexey Dobriyan <>
Cc: Len Brown <>
Cc: Ingo Molnar <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agopm: fix build for CONFIG_PM unset
Rafael J. Wysocki [Wed, 18 Feb 2009 22:48:21 +0000 (14:48 -0800)]
pm: fix build for CONFIG_PM unset

Compilation of kprobes.c with CONFIG_PM unset is broken due to some broken
config dependncies.  Fix that.

Signed-off-by: Rafael J. Wysocki <>
Reported-by: Ingo Molnar <>
Tested-by: Masami Hiramatsu <>
Cc: Len Brown <>
Acked-by: Pavel Machek <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agocgroups: fix possible use after free
Li Zefan [Wed, 18 Feb 2009 22:48:20 +0000 (14:48 -0800)]
cgroups: fix possible use after free

In cgroup_kill_sb(), root is freed before sb is detached from the list, so
another sget() may find this sb and call cgroup_test_super(), which will
access the root that has been freed.

Reported-by: Al Viro <>
Signed-off-by: Li Zefan <>
Acked-by: Paul Menage <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agomm: task dirty accounting fix
Nick Piggin [Wed, 18 Feb 2009 22:48:18 +0000 (14:48 -0800)]
mm: task dirty accounting fix

YAMAMOTO-san noticed that task_dirty_inc doesn't seem to be called properly for
cases where set_page_dirty is not used to dirty a page (eg. mark_buffer_dirty).

Additionally, there is some inconsistency about when task_dirty_inc is
called.  It is used for dirty balancing, however it even gets called for

So rather than increment it in a set_page_dirty wrapper, move it down to
exactly where the dirty page accounting stats are incremented.

Cc: YAMAMOTO Takashi <>
Signed-off-by: Nick Piggin <>
Acked-by: Peter Zijlstra <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agotimerfd: add flags check
Davide Libenzi [Wed, 18 Feb 2009 22:48:18 +0000 (14:48 -0800)]
timerfd: add flags check

As requested by Michael, add a missing check for valid flags in
timerfd_settime(), and make it return EINVAL in case some extra bits are

Michael said:
If this is to be any use to userland apps that want to check flag
support (perhaps it is too late already), then the sooner we get it
into the kernel the better: 2.6.29 would be good; earlier stables as
well would be even better.

[ remove unused TFD_FLAGS_SET]
Acked-by: Michael Kerrisk <>
Signed-off-by: Davide Libenzi <>
Cc: <> [2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoPavel has moved
Pavel Machek [Wed, 18 Feb 2009 22:48:16 +0000 (14:48 -0800)]
Pavel has moved

My address will stop working some day, so put working one into
MAINTAINERS/CREDITS.  It would be cool to get this to 2.6.29...  it should
not really break anything.

Signed-off-by: Pavel Machek <>
Cc: Jiri Kosina <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoseq_file: properly cope with pread
Eric Biederman [Wed, 18 Feb 2009 22:48:16 +0000 (14:48 -0800)]
seq_file: properly cope with pread

Currently seq_read assumes that the offset passed to it is always the
offset it passed to user space.  In the case pread this assumption is
broken and we do the wrong thing when presented with pread.

To solve this I introduce an offset cache inside of struct seq_file so we
know where our logical file position is.  Then in seq_read if we try to
read from another offset we reset our data structures and attempt to go to
the offset user space wanted.

[ restore FMODE_PWRITE]
[ seq_open needs its fmode opened up to take advantage of this]
Signed-off-by: Eric Biederman <>
Cc: Alexey Dobriyan <>
Cc: Al Viro <>
Cc: Paul Turner <>
Cc: <> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agovfs: separate FMODE_PREAD/FMODE_PWRITE into separate flags
Paul Turner [Wed, 18 Feb 2009 22:48:15 +0000 (14:48 -0800)]
vfs: separate FMODE_PREAD/FMODE_PWRITE into separate flags

Separate FMODE_PREAD and FMODE_PWRITE into separate flags to reflect the
reality that the read and write paths may have independent restrictions.

A git grep verifies that these flags are always cleared together so this
new behavior will only apply to interfaces that change to clear flags

This is required for "seq_file: properly cope with pread", a post-2.6.25
regression fix.

[ add comment]
Signed-off-by: Paul Turner <>
Cc: Eric Biederman <>
Cc: Alexey Dobriyan <>
Cc: Al Viro <>
Cc: <> [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agocgroups: update documentation about css_set hash table
Li Zefan [Wed, 18 Feb 2009 22:48:14 +0000 (14:48 -0800)]
cgroups: update documentation about css_set hash table

The css_set hash table was introduced in 2.6.26, so update the
documentation accordingly.

Signed-off-by: Li Zefan <>
Acked-by: Paul Menage <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoaoe: ignore vendor extension AoE responses
Ed Cashin [Wed, 18 Feb 2009 22:48:13 +0000 (14:48 -0800)]
aoe: ignore vendor extension AoE responses

The Welland ME-747K-SI AoE target generates unsolicited AoE responses that
are marked as vendor extensions.  Instead of ignoring these packets, the
aoe driver was generating kernel messages for each unrecognized response
received.  This patch corrects the behavior.

Signed-off-by: Ed Cashin <>
Reported-by: <>
Tested-by: <>
Cc: <>
Cc: Alex Buell <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agovmalloc: add __get_vm_area_caller()
Benjamin Herrenschmidt [Wed, 18 Feb 2009 22:48:12 +0000 (14:48 -0800)]
vmalloc: add __get_vm_area_caller()

We have get_vm_area_caller() and __get_vm_area() but not

On powerpc, I use __get_vm_area() to separate the ranges of addresses
given to vmalloc vs.  ioremap (various good reasons for that) so in order
to be able to implement the new caller tracking in /proc/vmallocinfo, I
need a "_caller" variant of it.

(akpm: needed for ongoing powerpc development, so merge it early)

[ coding-style fixes]
Signed-off-by: Benjamin Herrenschmidt <>
Reviewed-by: KOSAKI Motohiro <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
14 years agoomap_hsmmc: Change while(); loops with finite version
Jean Pihet [Fri, 6 Feb 2009 15:42:51 +0000 (16:42 +0100)]
omap_hsmmc: Change while(); loops with finite version

Replace the infinite 'while() ;' loops
with a finite loop version.

Signed-off-by: Jean Pihet <>
Acked-by: Tony Lindgren <>
Signed-off-by: Pierre Ossman <>
14 years agoomap_hsmmc: recover from transfer failures
Jean Pihet [Wed, 11 Feb 2009 21:11:39 +0000 (13:11 -0800)]
omap_hsmmc: recover from transfer failures

Timeouts during a command that has a data phase can result in the next
command issued after the command that failed not being processed, i.e.  no
interrupt ever occurs to indicate the command has completed.  This failure
can result in a deadlock.

This patch resets the data state machine to clear the error in case of a
command timeout.

Tested on OMAP3430 chip and intensive MMC/SD device removal while
transferring data.

Signed-off-by: Andy Lowe <>
Signed-off-by: Jean Pihet <>
Signed-off-by: Adrian Hunter <>
Signed-off-by: Andrew Morton <>
Acked-by: Tony Lindgren <>
Signed-off-by: Pierre Ossman <>
14 years agoomap_hsmmc: only MMC1 allows HCTL.SDVS != 1.8V
David Brownell [Tue, 17 Feb 2009 22:49:01 +0000 (14:49 -0800)]
omap_hsmmc: only MMC1 allows HCTL.SDVS != 1.8V

Based on a patch from Tony Lindgren ... after initialization,
never change HCTL.SDVS except for MMC1.  The other controller
instances only support 1.8V in that field, although they can
suport other card/SDIO/eMMC/... voltages with level shifting
solutions such as external transceivers.

MMC2 behavior sanity tested on Overo/WLAN, OMAP3430 SDP, and
custom hardware.  MMC1 also sanity tested on those platforms
plus Beagle.  This also fixes a bug preventing MMC2 (and also
presumably MMC3) from powering down when requested.

Signed-off-by: David Brownell <>
Acked-by: Tony Lindgren <>
Signed-off-by: Pierre Ossman <>
14 years agoomap_hsmmc: card detect irq bugfix
David Brownell [Wed, 4 Feb 2009 22:42:03 +0000 (14:42 -0800)]
omap_hsmmc: card detect irq bugfix

Work around lockdep issue when card detect IRQ handlers run in
thread context ... it forces IRQF_DISABLED, which prevents all
access to twl4030 card detect signals.

Signed-off-by: David Brownell <>
Acked-by: Tony Lindgren <>
Signed-off-by: Pierre Ossman <>
14 years agosdhci: fix led naming
Helmut Schaa [Sat, 14 Feb 2009 15:22:39 +0000 (16:22 +0100)]
sdhci: fix led naming

Fix the led device naming for the sdhci driver.

The led class documentation defines the led name to have the
form "devicename:colour:function" while not applicable sections
should be left blank.

To comply with the documentation the led device name is changed
from "mmc*" to "mmc*::".

Signed-off-by: Helmut Schaa <>
Signed-off-by: Pierre Ossman <>
14 years agommc_test: fix basic read test
Rabin Vincent [Fri, 13 Feb 2009 17:25:26 +0000 (22:55 +0530)]
mmc_test: fix basic read test

Due to a typo in the Basic Read test, it's currently identical to the
Basic Write test.  Fix this.

Signed-off-by: Rabin Vincent <>
Signed-off-by: Pierre Ossman <>
14 years agos3cmci: Fix hangup in do_pio_write()
Yauhen Kharuzhy [Wed, 11 Feb 2009 21:25:52 +0000 (13:25 -0800)]
s3cmci: Fix hangup in do_pio_write()

This commit fixes the regression what was added by commit
088a78af978d0c8e339071a9b2bca1f4cb368f30 "s3cmci: Support transfers
which are not multiple of 32 bits."

fifo_free() now returns amount of available space in FIFO buffer in
bytes.  But do_pio_write() writes to FIFO 32-bit words.  Condition for
return from cycle is (fifo_free() == 0), but when fifo has 1..3 bytes
of free space then this condition will never be true and system hangs.

This patch changes condition in the while() to (fifo_free() > 3).

Signed-off-by: Yauhen Kharuzhy <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Pierre Ossman <>
14 years agoALSA: jack - Use card->shortname for input name
Takashi Iwai [Wed, 18 Feb 2009 15:46:27 +0000 (16:46 +0100)]
ALSA: jack - Use card->shortname for input name

Currently the jack layer refers to card->longname as a part of
its input device name string.  However, longname is often really long
and way too ugly as an identifier, such as,
"HDA Intel at 0xf8400000 irq 21".

This patch changes the code to use card->shortname instead.
The shortname string contains usually the h/w vendor and product
names but without messy I/O port or IRQ numbers.

Signed-off-by: Takashi Iwai <>
14 years agoblock: fix deadlock in blk_abort_queue() for drivers that readd to timeout list
Hannes Reinecke [Wed, 18 Feb 2009 09:30:15 +0000 (10:30 +0100)]
block: fix deadlock in blk_abort_queue() for drivers that readd to timeout list

blk_abort_queue() iterates the timeout list and aborts each request on the
list, but if the driver error handling readds a request to the timeout list
during this processing, we could be looping forever. Fix this by splicing
current entries to a local list and run over that list instead.

Signed-off-by: Jens Axboe <>
14 years agoblock: fix booting from partitioned md array
Neil Brown [Wed, 18 Feb 2009 09:33:59 +0000 (10:33 +0100)]
block: fix booting from partitioned md array

Hi Tejun,

 it looks like your commit:

   block: don't depend on consecutive minor space

 broke a particular case for booting from partitioned md/raid devices.
 That is the second time this has been broken recently.  The previous
 time was fixed by

   block: do_mounts - accept root=<non-existant partition>

 Because the data isn't available when an md device is first created
 (we add disks and set it up after creation), the initial partition
 scan finds nothing.  It is not until the device is opened that
 another partition scan happens and finds something.

 So at the point where the kernel parameter "root=/dev/md_d0p1" is
 being parsed, md_d0 exists, but md_d0p1 does not.
 However if we let blk_lookup_devt return the correct device number
 even though the device doesn't exist, then the attempt to mount it
 will successfully find the partition.

 I have tried in the past to find a way to get the partition table to
 be read as soon as the array is assembled but that proved impossible
 (at the time).  I don't remember the details, and could possibly
 revisit it.  However it would be really nice if blk_lookup_devt
 could be adjusted to again accept non existant partitions.

Signed-off-by: Jens Axboe <>
14 years agoblock: revert part of 18ce3751ccd488c78d3827e9f6bf54e6322676fb
Jens Axboe [Tue, 17 Feb 2009 12:59:08 +0000 (13:59 +0100)]
block: revert part of 18ce3751ccd488c78d3827e9f6bf54e6322676fb

The above commit added WRITE_SYNC and switched various places to using
that for committing writes that will be waited upon immediately after
submission. However, this causes a performance regression with AS and CFQ
for ext3 at least, since sync_dirty_buffer() will submit some writes with
WRITE_SYNC while ext3 has sumitted others dependent writes without the sync
flag set. This causes excessive anticipation/idling in the IO scheduler
because sync and async writes get interleaved, causing a big performance
regression for the below test case (which is meant to simulate sqlite
like behaviour).

---- test case ----

int main(int argc, char **argv)

int fdes, i;
FILE *fp;
struct timeval start;
struct timeval end;
struct timeval res;

gettimeofday(&start, NULL);
for (i=0; i<ROWS; i++) {
fp = fopen("test_file", "a");
fprintf(fp, "Some Text Data\n");
fdes = fileno(fp);
gettimeofday(&end, NULL);

timersub(&end, &start, &res);
fprintf(stdout, "time to write %d lines is %ld(msec)\n", ROWS,
(res.tv_sec*1000000 + res.tv_usec)/1000);

return 0;


Thanks to for tracking down this performance
regression and providing a test case.

Signed-off-by: Jens Axboe <>
14 years agocciss: PCI power management reset for kexec
Chip Coldwell [Mon, 16 Feb 2009 12:11:56 +0000 (13:11 +0100)]
cciss: PCI power management reset for kexec

The kexec kernel resets the CCISS hardware in three steps:

1. Use PCI power management states to reset the controller in the
   kexec kernel.

2. Clear the MSI/MSI-X bits in PCI configuration space so that MSI
   initialization in the kexec kernel doesn't fail.

3. Use the CCISS "No-op" message to determine when the controller
   firmware has recovered from the PCI PM reset.

[ cleanups]
Signed-off-by: Mike Miller <>
Cc: Randy Dunlap <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Jens Axboe <>
14 years agoparide/pg.c: xs(): &&/|| confusion
Roel Kluin [Mon, 16 Feb 2009 12:11:55 +0000 (13:11 +0100)]
paride/pg.c: xs(): &&/|| confusion

&&/|| confusion

Signed-off-by: Roel Kluin <>
Cc: Bartlomiej Zolnierkiewicz <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Jens Axboe <>
14 years agofs/bio: bio_alloc_bioset: pass right object ptr to mempool_free
Subhash Peddamallu [Mon, 16 Feb 2009 09:27:07 +0000 (10:27 +0100)]
fs/bio: bio_alloc_bioset: pass right object ptr to mempool_free

When freeing from bio pool use right ptr to account for bs->front_pad,
instead of bio ptr,

Signed-off-by: Subhash Peddamallu <>
Signed-off-by: Jens Axboe <>
14 years agoblock: fix bad definition of BIO_RW_SYNC
Jens Axboe [Mon, 16 Feb 2009 09:25:40 +0000 (10:25 +0100)]
block: fix bad definition of BIO_RW_SYNC

We can't OR shift values, so get rid of BIO_RW_SYNC and use BIO_RW_SYNCIO
and BIO_RW_UNPLUG explicitly. This brings back the behaviour from before

Signed-off-by: Jens Axboe <>
14 years agobsg: Fix sense buffer bug in SG_IO
Boaz Harrosh [Tue, 3 Feb 2009 06:47:29 +0000 (07:47 +0100)]
bsg: Fix sense buffer bug in SG_IO

When submitting requests via SG_IO, which does a sync io, a
bsg_command is not allocated. So an in-Kernel sense_buffer was not
set. However when calling blk_execute_rq() with no sense buffer
one is provided from the stack. Now bsg at blk_complete_sgv4_hdr_rq()
would check if rq->sense_len and a sense was requested by sg_io_v4
the rq->sense was copy_user() back, but by now it is already mangled
stack memory.

I have fixed that by forcing a sense_buffer when calling bsg_map_hdr().
The bsg_command->sense is provided in the write/read path like before,
and on-the-stack buffer is provided when doing SG_IO.

I have also fixed a dprintk message to print rq->errors in hex because
of the scsi bit-field use of this member. For other block devices it
does not matter anyway.

Signed-off-by: Boaz Harrosh <>
Acked-by: FUJITA Tomonori <>
Signed-off-by: Jens Axboe <>
14 years agoRevert "sdhci: force high speed capability on some controllers"
Pierre Ossman [Mon, 2 Feb 2009 20:13:49 +0000 (21:13 +0100)]
Revert "sdhci: force high speed capability on some controllers"

This reverts commit a4b76193774b463b922cab2f92450efb20d29ef0.

It turned out that the controller had problem running at the
higher speed, so go back to trusting the hardware capability

Signed-off-by: Pierre Ossman <>
14 years agoMMC: fix bug - SDHC card capacity not correct
Yi Li [Thu, 5 Feb 2009 07:31:57 +0000 (15:31 +0800)]
MMC: fix bug - SDHC card capacity not correct

Signed-off-by: Yi Li <>
Signed-off-by: Bryan Wu <>
Signed-off-by: Pierre Ossman <>
14 years agoAdd support for VT6415 PCIE PATA IDE Host Controller
Zlatko Calusic [Wed, 18 Feb 2009 00:33:34 +0000 (01:33 +0100)]
Add support for VT6415 PCIE PATA IDE Host Controller

Signed-off-by: Zlatko Calusic <>
Signed-off-by: Linus Torvalds <>
14 years agoUSB/PCI: Fix resume breakage of controllers behind cardbus bridges
Rafael J. Wysocki [Fri, 13 Feb 2009 22:41:12 +0000 (23:41 +0100)]
USB/PCI: Fix resume breakage of controllers behind cardbus bridges

If a USB PCI controller is behind a cardbus bridge, we are trying to
restore its configuration registers too early, before the cardbus
bridge is operational.  To fix this, call pci_restore_state() from
usb_hcd_pci_resume() and remove usb_hcd_pci_resume_early() which is
no longer necessary (the configuration spaces of USB controllers that
are not behind cardbus bridges will be restored by the PCI PM core
with interrupts disabled anyway).

This patch fixes the regression from 2.6.28 tracked as

[ Side note: the proper long-term fix is probably to just force the
  unplug event at suspend time instead of doing a plug/unplug at resume
  time, but this patch is fine regardless  - Linus ]

Signed-off-by: Rafael J. Wysocki <>
Reported-by: Miles Lane <>
Signed-off-by: Linus Torvalds <>
14 years agox86, mce: fix ifdef for 64bit thermal apic vector clear on shutdown
Andi Kleen [Thu, 12 Feb 2009 12:39:35 +0000 (13:39 +0100)]
x86, mce: fix ifdef for 64bit thermal apic vector clear on shutdown

Impact: Bugfix

The ifdef for the apic clear on shutdown for the 64bit intel thermal
vector was incorrect and never triggered. Fix that.

Signed-off-by: Andi Kleen <>
Acked-by: Thomas Gleixner <>
Signed-off-by: H. Peter Anvin <>
14 years agox86, mce: use force_sig_info to kill process in machine check
Andi Kleen [Thu, 12 Feb 2009 12:39:33 +0000 (13:39 +0100)]
x86, mce: use force_sig_info to kill process in machine check

Impact: bug fix (with tolerant == 3)

do_exit cannot be called directly from the exception handler because
it can sleep and the exception handler runs on the exception stack.
Use force_sig() instead.

Based on a earlier patch by Ying Huang who debugged the problem.

Signed-off-by: Andi Kleen <>
Acked-by: Thomas Gleixner <>
Signed-off-by: H. Peter Anvin <>
14 years agox86, mce: reinitialize per cpu features on resume
Andi Kleen [Thu, 12 Feb 2009 12:39:26 +0000 (13:39 +0100)]
x86, mce: reinitialize per cpu features on resume

Impact: Bug fix

This fixes a long standing bug in the machine check code. On resume the
boot CPU wouldn't get its vendor specific state like thermal handling
reinitialized. This means the boot cpu wouldn't ever get any thermal
events reported again.

Call the respective initialization functions on resume

v2: Remove ancient init because they don't have a resume device anyways.
    Pointed out by Thomas Gleixner.
v3: Now fix the Subject too to reflect v2 change

Signed-off-by: Andi Kleen <>
Acked-by: Thomas Gleixner <>
Signed-off-by: H. Peter Anvin <>
14 years ago[ARM] 5401/1: Orion: fix edge triggered GPIO interrupt support
Nicolas Pitre [Tue, 17 Feb 2009 19:45:50 +0000 (20:45 +0100)]
[ARM] 5401/1: Orion: fix edge triggered GPIO interrupt support

The GPIO interrupts can be configured as either level triggered or edge
triggered, with a default of level triggered.  When an edge triggered
interrupt is requested, the gpio_irq_set_type method is called which
currently switches the given IRQ descriptor between two struct irq_chip
instances: orion_gpio_irq_level_chip and orion_gpio_irq_edge_chip. This
happens via __setup_irq() which also calls irq_chip_set_defaults() to
assign default methods to uninitialized ones.  The problem is that
irq_chip_set_defaults() is called before the irq_chip reference is
switched, leaving the new irq_chip (orion_gpio_irq_edge_chip in this
case) with uninitialized methods such as chip->startup() causing a kernel

Many solutions are possible, such as making irq_chip_set_defaults() global
and calling it from gpio_irq_set_type(), or calling __irq_set_trigger()
before irq_chip_set_defaults() in __setup_irq().  But those require
modifications to the generic IRQ code which might have adverse effect on
other architectures, and that would still be a fragile arrangement.
Manually copying the missing methods from within gpio_irq_set_type()
would be really ugly and it would break again the day new methods with
automatic defaults are added.

A better solution is to have a single irq_chip instance which can deal
with both edge and level triggered interrupts.  It is also a good idea
to switch the IRQ handler instead, as the edge IRQ handler allows for
one edge IRQ event to be queued as the IRQ is actually masked only when
that second IRQ is received, at which point the hardware can queue an
additional IRQ event, making edge triggered interrupts a bit more

Tested-by: Martin Michlmayr <>
Signed-off-by: Nicolas Pitre <>
Signed-off-by: Russell King <>
14 years agoMerge branch 'sched-fixes-for-linus' of git://
Linus Torvalds [Tue, 17 Feb 2009 22:30:06 +0000 (14:30 -0800)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://
  sched: cpu hotplug fix

14 years agoMerge branch 'timers-fixes-for-linus' of git://
Linus Torvalds [Tue, 17 Feb 2009 22:29:42 +0000 (14:29 -0800)]
Merge branch 'timers-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'timers-fixes-for-linus' of git://
  timers: more consistently use clock vs timer

14 years agoMerge branch 'tracing-fixes-for-linus' of git://
Linus Torvalds [Tue, 17 Feb 2009 22:29:15 +0000 (14:29 -0800)]
Merge branch 'tracing-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://
  doc: mmiotrace.txt, buffer size control change
  trace: mmiotrace to the tracer menu in Kconfig
  mmiotrace: count events lost due to not recording