linux-2.6
15 years agonfsd: Revert "svcrpc: take advantage of tcp autotuning"
J. Bruce Fields [Wed, 27 May 2009 22:51:06 +0000 (18:51 -0400)] 
nfsd: Revert "svcrpc: take advantage of tcp autotuning"

This reverts commit 47a14ef1af48c696b214ac168f056ddc79793d0e "svcrpc:
take advantage of tcp autotuning", which uncovered some further problems
in the server rpc code, causing significant performance regressions in
common cases.

We will likely reinstate this patch after releasing 2.6.30 and applying
some work on the underlying fixes to the problem (developed by Trond).

Reported-by: Jeff Moyer <jmoyer@redhat.com>
Cc: Olga Kornievskaia <aglo@citi.umich.edu>
Cc: Jim Rees <rees@umich.edu>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
15 years agoPCI/ACPI: fix wrong ref count handling in acpi_pci_bind()
Kenji Kaneshige [Tue, 26 May 2009 00:08:03 +0000 (09:08 +0900)] 
PCI/ACPI: fix wrong ref count handling in acpi_pci_bind()

The 'dev' field of struct acpi_pci_data is having a pointer to struct
pci_dev without incrementing the reference counter. Because of this, I
got the following kernel oops when I was doing some pci hotplug
operations. This patch fixes this bug by replacing wrong hand-made
pci_find_slot() with pci_get_slot() in acpi_pci_bind().

 BUG: unable to handle kernel NULL pointer dereference at 00000000000000e8
 IP: [<ffffffff803f0e9b>] acpi_pci_unbind+0xb1/0xdd

 Call Trace:
  [<ffffffff803ecee4>] acpi_bus_remove+0x54/0x68
  [<ffffffff803ecf6d>] acpi_bus_trim+0x75/0xe3
  [<ffffffffa0345ddd>] acpiphp_disable_slot+0x16d/0x1e0 [acpiphp]
  [<ffffffffa03441f0>] disable_slot+0x20/0x60 [acpiphp]
  [<ffffffff803cfc18>] power_write_file+0xc8/0x110
  [<ffffffff803c6a54>] pci_slot_attr_store+0x24/0x30
  [<ffffffff803469ce>] sysfs_write_file+0xce/0x140
  [<ffffffff802e94e7>] vfs_write+0xc7/0x170
  [<ffffffff802e9aa0>] sys_write+0x50/0x90
  [<ffffffff8020bd6b>] system_call_fastpath+0x16/0x1b

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Reviewed-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Reviewed-by: Alex Chiang <achiang@hp.com>
Tested-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
15 years agonfsd: fix hung up of nfs client while sync write data to nfs server
Wei Yongjun [Tue, 19 May 2009 04:03:15 +0000 (12:03 +0800)] 
nfsd: fix hung up of nfs client while sync write data to nfs server

Commit 'Short write in nfsd becomes a full write to the client'
(31dec2538e45e9fff2007ea1f4c6bae9f78db724) broken the sync write.
With the following commands to reproduce:

  $ mount -t nfs -o sync 192.168.0.21:/nfsroot /mnt
  $ cd /mnt
  $ echo aaaa > temp.txt

Then nfs client is hung up.

In SYNC mode the server alaways return the write count 0 to the
client. This is because the value of host_err in nfsd_vfs_write()
will be overwrite in SYNC mode by 'host_err=nfsd_sync(file);',
and then we return host_err(which is now 0) as write count.

This patch fixed the problem.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
15 years agofsldma: Fix compile warnings
Kumar Gala [Wed, 13 May 2009 21:25:57 +0000 (16:25 -0500)] 
fsldma: Fix compile warnings

We we build with dma_addr_t as a 64-bit quantity we get:

drivers/dma/fsldma.c: In function 'fsl_chan_xfer_ld_queue':
drivers/dma/fsldma.c:625: warning: cast to pointer from integer of different size
drivers/dma/fsldma.c: In function 'fsl_dma_chan_do_interrupt':
drivers/dma/fsldma.c:737: warning: cast to pointer from integer of different size
drivers/dma/fsldma.c:737: warning: cast to pointer from integer of different size
drivers/dma/fsldma.c: In function 'of_fsl_dma_probe':
drivers/dma/fsldma.c:927: warning: cast to pointer from integer of different

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
15 years agoi915: Set object to gtt domain when faulting it back in
Kristian Høgsberg [Wed, 27 May 2009 18:37:28 +0000 (14:37 -0400)] 
i915: Set object to gtt domain when faulting it back in

When a GEM object is evicted from the GTT we set it to the CPU domain,
as it might get swapped in and out or ever mmapped regularly.  If the
object is mmapped through the GTT it can still get evicted in this way
by other objects requiring GTT space.  When the GTT mapping is touched
again we fault it back into the GTT, but fail to set it back to the
GTT domain.  This means we fail to flush any cached CPU writes to the
pages backing the object which will then happen "eventually", typically
after we write to the page through the uncached GTT mapping.

[anholt: Note that userland does do a set_domain(GTT, GTT) when starting
to access the GTT mapping.  That covers getting the existing mapping of the
object synchronized if it's bound to the GTT.  But set_domain(GTT, GTT)
doesn't do anything if the object is currently unbound.  This fix covers the
transition to being bound for GTT mapping.]

Fixes glyph and other pixmap corruption during swapping.  fd.o bug #21790

Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/vapier...
Linus Torvalds [Wed, 27 May 2009 17:58:49 +0000 (10:58 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/vapier/blackfin

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/vapier/blackfin:
  Blackfin: fix strncmp.o build error
  Blackfin: drop unneeded asm/.gitignore
  Blackfin: ignore generated vmlinux.lds
  MAINTAINERS: drop (subscribers-only) markings on Blackfin lists
  MAINTAINERS: update Blackfin items
  Blackfin: hook up preadv/pwritev syscalls

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 27 May 2009 17:58:23 +0000 (10:58 -0700)] 
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc: Fix up dma_alloc_coherent() on platforms without cache coherency.
  powerpc: Minor cleanups of kernel virt address space definitions
  powerpc: Move dma-noncoherent.c from arch/powerpc/lib to arch/powerpc/mm
  Revert "powerpc: Rework dma-noncoherent to use generic vmalloc layer"

15 years agoCacheFiles: Fixup renamed filenames in comments in internal.h
David Howells [Wed, 27 May 2009 14:46:55 +0000 (15:46 +0100)] 
CacheFiles: Fixup renamed filenames in comments in internal.h

Fix up renamed filenames in comments in fs/cachefiles/internal.h.

Originally, the files were all called cf-xxx.c, but they got renamed to
just xxx.c.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoFS-Cache: Fixup renamed filenames in comments in internal.h
David Howells [Wed, 27 May 2009 14:46:50 +0000 (15:46 +0100)] 
FS-Cache: Fixup renamed filenames in comments in internal.h

Fix up renamed filenames in comments in fs/fscache/internal.h.

Originally, the files were all called fsc-xxx.c, but they got renamed to
just xxx.c.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://neil.brown.name/md
Linus Torvalds [Wed, 27 May 2009 17:18:26 +0000 (10:18 -0700)] 
Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md: raid5: change incorrect usage of 'min' macro to 'min_t'

15 years agoMerge branch 'fix/pcm-jiffies-check' into for-linus
Takashi Iwai [Wed, 27 May 2009 14:51:27 +0000 (16:51 +0200)] 
Merge branch 'fix/pcm-jiffies-check' into for-linus

* fix/pcm-jiffies-check:
  ALSA: Enable PCM hw_ptr_jiffies check only in xrun_debug mode
  ALSA: Fix invalid jiffies check after pause

15 years agoMerge branch 'fix/misc' into for-linus
Takashi Iwai [Wed, 27 May 2009 14:51:15 +0000 (16:51 +0200)] 
Merge branch 'fix/misc' into for-linus

* fix/misc:
  sound: usb-audio: make the MotU Fastlane work again

15 years agonetfilter: nfnetlink_log: fix wrong skbuff size calculation
Pablo Neira Ayuso [Wed, 27 May 2009 13:49:11 +0000 (15:49 +0200)] 
netfilter: nfnetlink_log: fix wrong skbuff size calculation

This problem was introduced in 72961ecf84d67d6359a1b30f9b2a8427f13e1e71
since no space was reserved for the new attributes NFULA_HWTYPE,
NFULA_HWLEN and NFULA_HWHEADER.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
15 years agonetfilter: xt_hashlimit does a wrong SEQ_SKIP
Jesper Dangaard Brouer [Wed, 27 May 2009 13:45:34 +0000 (15:45 +0200)] 
netfilter: xt_hashlimit does a wrong SEQ_SKIP

The function dl_seq_show() returns 1 (equal to SEQ_SKIP) in case
a seq_printf() call return -1.  It should return -1.

This SEQ_SKIP behavior brakes processing the proc file e.g. via a
pipe or just through less.

Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: Patrick McHardy <kaber@trash.net>
15 years agoInput: usb1400_ts - fix access to "device data" in resume function
Manuel Traut [Wed, 27 May 2009 13:20:05 +0000 (06:20 -0700)] 
Input: usb1400_ts - fix access to "device data" in resume function

platform_data != driver_data

driver data is actually the "correct" place of the struct however it is
not placed there due to the need of the ac97 struct. This is broken since
d9105c2b01 aka "[ARM] 5184/1: Split ucb1400_ts into core and touchscreen"

Signed-off-by: Manuel Traut <manut@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
15 years agomd: raid5: change incorrect usage of 'min' macro to 'min_t'
NeilBrown [Wed, 27 May 2009 11:39:05 +0000 (21:39 +1000)] 
md: raid5: change incorrect usage of 'min' macro to 'min_t'

A recent patch to raid5.c use min on an int and a sector_t.
This isn't allowed.
So change it to min_t(sector_t,x,y).

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agosound: usb-audio: make the MotU Fastlane work again
Clemens Ladisch [Wed, 27 May 2009 08:49:30 +0000 (10:49 +0200)] 
sound: usb-audio: make the MotU Fastlane work again

Kernel 2.6.18 broke the MotU Fastlane, which uses duplicate endpoint
numbers in a manner that is not only illegal but also confuses the
kernel's endpoint descriptor caching mechanism.  To work around this, we
have to add a separate usb_set_interface() call to guide the USB core to
the correct descriptors.

Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Reported-and-tested-by: David Fries <david@fries.net>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agoALSA: Enable PCM hw_ptr_jiffies check only in xrun_debug mode
Takashi Iwai [Wed, 27 May 2009 08:53:33 +0000 (10:53 +0200)] 
ALSA: Enable PCM hw_ptr_jiffies check only in xrun_debug mode

The PCM hw_ptr jiffies check results sometimes in problems when a
hardware doesn't give smooth hw_ptr updates.  So far, au88x0 and some
other drivers appear not working due to this strict check.
However, this check is a nice debug tool, and the capability should be
still kept.

Hence, we disable this check now as default unless the user enables it
by setting the xrun_debug mode to the specific stream via a proc file.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agoPCI Hotplug: acpiphp: don't store a pci_dev in acpiphp_func
Alex Chiang [Thu, 21 May 2009 22:21:15 +0000 (16:21 -0600)] 
PCI Hotplug: acpiphp: don't store a pci_dev in acpiphp_func

An oops can occur if a user attempts to use both PCI logical
hotplug and the ACPI physical hotplug driver (acpiphp) in this
sequence, where $slot/address == $device.

In other words, if acpiphp has claimed a PCI device, and that
device is logically removed, then acpiphp may oops when it
attempts to access it again.

# echo 1 > /sys/bus/pci/devices/$device/remove
# echo 0 > /sys/bus/pci/slots/$slot/power

Unable to handle kernel NULL pointer dereference (address 0000000000000000)
Call Trace:
 [<a000000100016390>] show_stack+0x50/0xa0
 [<a000000100016c60>] show_regs+0x820/0x860
 [<a00000010003b390>] die+0x190/0x2a0
 [<a000000100066a40>] ia64_do_page_fault+0x8e0/0xa40
 [<a00000010000c7a0>] ia64_native_leave_kernel+0x0/0x270
 [<a0000001003b2660>] pci_remove_bus_device+0x120/0x260
 [<a0000002060549f0>] acpiphp_disable_slot+0x410/0x540 [acpiphp]
 [<a0000002060505c0>] disable_slot+0xc0/0x120 [acpiphp]
 [<a0000002040d21c0>] power_write_file+0x1e0/0x2a0 [pci_hotplug]
 [<a0000001003bb820>] pci_slot_attr_store+0x60/0xa0
 [<a000000100240f70>] sysfs_write_file+0x230/0x2c0
 [<a000000100195750>] vfs_write+0x190/0x2e0
 [<a0000001001961a0>] sys_write+0x80/0x100
 [<a00000010000c600>] ia64_ret_from_syscall+0x0/0x20
 [<a000000000010720>] __kernel_syscall_via_break+0x0/0x20

The root cause of this oops is that the logical remove ("echo 1 >
/sys/bus/pci/devices/$device/remove") destroyed the pci_dev. The
pci_dev struct itself wasn't deallocated because acpiphp kept a
reference, but some of its fields became invalid.

acpiphp doesn't have any real reason to keep a pointer to a
pci_dev around. It can always derive it using pci_get_slot().

If a logical remove destroys the pci_dev, acpiphp won't find it
and is thus prevented from causing mischief.

Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Acked-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoALSA: Fix invalid jiffies check after pause
Takashi Iwai [Wed, 27 May 2009 08:49:26 +0000 (10:49 +0200)] 
ALSA: Fix invalid jiffies check after pause

The hw_ptr_jiffies has to be reset properly to avoid the invalid
check of jiffies delta in snd_pcm_update_hw_ptr*() functions.
Especailly this patch fixes the bogus jiffies check after the puase
and resume.

This patch is a modified version of the original patch by Jaroslav.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agoBluetooth: Remove useless flush_work() causing lockdep warnings
Dave Young [Wed, 27 May 2009 07:10:28 +0000 (09:10 +0200)] 
Bluetooth: Remove useless flush_work() causing lockdep warnings

The calls to flush_work() are pointless in a single thread workqueue
and they are actually causing a lockdep warning.

=============================================
[ INFO: possible recursive locking detected ]
2.6.30-rc6-02911-gbb803cf #16
---------------------------------------------
bluetooth/2518 is trying to acquire lock:
 (bluetooth){+.+.+.}, at: [<c0130c14>] flush_work+0x28/0xb0

but task is already holding lock:
 (bluetooth){+.+.+.}, at: [<c0130424>] worker_thread+0x149/0x25e

other info that might help us debug this:
2 locks held by bluetooth/2518:
 #0:  (bluetooth){+.+.+.}, at: [<c0130424>] worker_thread+0x149/0x25e
 #1:  (&conn->work_del){+.+...}, at: [<c0130424>] worker_thread+0x149/0x25e

stack backtrace:
Pid: 2518, comm: bluetooth Not tainted 2.6.30-rc6-02911-gbb803cf #16
Call Trace:
 [<c03d64d9>] ? printk+0xf/0x11
 [<c0140d96>] __lock_acquire+0x7ce/0xb1b
 [<c0141173>] lock_acquire+0x90/0xad
 [<c0130c14>] ? flush_work+0x28/0xb0
 [<c0130c2e>] flush_work+0x42/0xb0
 [<c0130c14>] ? flush_work+0x28/0xb0
 [<f8b84966>] del_conn+0x1c/0x84 [bluetooth]
 [<c0130469>] worker_thread+0x18e/0x25e
 [<c0130424>] ? worker_thread+0x149/0x25e
 [<f8b8494a>] ? del_conn+0x0/0x84 [bluetooth]
 [<c0133843>] ? autoremove_wake_function+0x0/0x33
 [<c01302db>] ? worker_thread+0x0/0x25e
 [<c013355a>] kthread+0x45/0x6b
 [<c0133515>] ? kthread+0x0/0x6b
 [<c01034a7>] kernel_thread_helper+0x7/0x10

Based on a report by Oliver Hartkopp <oliver@hartkopp.net>

Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Tested-by: Oliver Hartkopp <oliver@hartkopp.net>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
15 years agopowerpc: Fix up dma_alloc_coherent() on platforms without cache coherency.
Benjamin Herrenschmidt [Wed, 27 May 2009 03:50:33 +0000 (13:50 +1000)] 
powerpc: Fix up dma_alloc_coherent() on platforms without cache coherency.

The implementation we just revived has issues, such as using a
Kconfig-defined virtual address area in kernel space that nothing
actually carves out (and thus will overlap whatever is there),
or having some dependencies on being self contained in a single
PTE page which adds unnecessary constraints on the kernel virtual
address space.

This fixes it by using more classic PTE accessors and automatically
locating the area for consistent memory, carving an appropriate hole
in the kernel virtual address space, leaving only the size of that
area as a Kconfig option. It also brings some dma-mask related fixes
from the ARM implementation which was almost identical initially but
grew its own fixes.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Minor cleanups of kernel virt address space definitions
Benjamin Herrenschmidt [Wed, 27 May 2009 03:44:50 +0000 (13:44 +1000)] 
powerpc: Minor cleanups of kernel virt address space definitions

Make FIXADDR_TOP a compile time constant and cleanup a
couple of definitions relative to the layout of the kernel
address space on ppc32. We also print out that layout at
boot time for debugging purposes.

This is a pre-requisite for properly fixing non-coherent
DMA allocactions.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Move dma-noncoherent.c from arch/powerpc/lib to arch/powerpc/mm
Benjamin Herrenschmidt [Wed, 27 May 2009 03:36:10 +0000 (13:36 +1000)] 
powerpc: Move dma-noncoherent.c from arch/powerpc/lib to arch/powerpc/mm

(pre-requisite to make the next patches more palatable)

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoBlackfin: fix strncmp.o build error
Mike Frysinger [Tue, 26 May 2009 09:03:52 +0000 (05:03 -0400)] 
Blackfin: fix strncmp.o build error

Fix some more fallout of the string changes:

  CC      arch/blackfin/lib/strncmp.o
In file included from include/linux/bitmap.h:9,
                 from include/linux/nodemask.h:90,
                 from include/linux/mmzone.h:17,
                 from include/linux/gfp.h:5,
                 from include/linux/kmod.h:23,
                 from include/linux/module.h:14,
                 from arch/blackfin/lib/strncmp.c:14:
include/linux/string.h: In function ‘strstarts’:
include/linux/string.h:132: error: implicit declaration of function ‘strncmp’
make[1]: *** [arch/blackfin/lib/strncmp.o] Error 1

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
CC: Rusty Russell <rusty@rustcorp.com.au>
15 years agoBlackfin: drop unneeded asm/.gitignore
Mike Frysinger [Wed, 20 May 2009 23:45:39 +0000 (19:45 -0400)] 
Blackfin: drop unneeded asm/.gitignore

We don't create a include/asm/mach/ symlink anymore, so we don't need the
.gitignore for it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
15 years agoBlackfin: ignore generated vmlinux.lds
Mike Frysinger [Tue, 19 May 2009 14:03:22 +0000 (10:03 -0400)] 
Blackfin: ignore generated vmlinux.lds

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
15 years agoMAINTAINERS: drop (subscribers-only) markings on Blackfin lists
Mike Frysinger [Mon, 18 May 2009 08:33:07 +0000 (04:33 -0400)] 
MAINTAINERS: drop (subscribers-only) markings on Blackfin lists

All of the Blackfin lists are transparently moderated for non-subscribers.
i.e. there are no annoying notices and people get whitelisted after first
their posting.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
15 years agoMAINTAINERS: update Blackfin items
Mike Frysinger [Sun, 24 May 2009 06:13:15 +0000 (02:13 -0400)] 
MAINTAINERS: update Blackfin items

With Bryan Wu having moved on to another job, push the slack onto some
other ADI lackeys.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
15 years agoBlackfin: hook up preadv/pwritev syscalls
Mike Frysinger [Tue, 26 May 2009 08:55:38 +0000 (04:55 -0400)] 
Blackfin: hook up preadv/pwritev syscalls

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
15 years agobfin_mac: fix build error due to net_device_ops convert
Mike Frysinger [Wed, 27 May 2009 03:55:33 +0000 (20:55 -0700)] 
bfin_mac: fix build error due to net_device_ops convert

The previous commit "convert to net_device_ops" broke the Blackfin MAC
driver as it declared the new structure before the function it used:
  CC      drivers/net/bfin_mac.o
drivers/net/bfin_mac.c:984: error: ‘bfin_mac_close’ undeclared here (not in a function)
make[1]: *** [drivers/net/bfin_mac.o] Error 1

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoatlx: move modinfo data from atlx.h to atl1.c
Alex Chiang [Wed, 27 May 2009 03:50:12 +0000 (20:50 -0700)] 
atlx: move modinfo data from atlx.h to atl1.c

Both atl1.c and atl2.c include atlx.h, which defines some modinfo
stuff. But atl2.c seems like it doesn't want the modinfo data
from atlx.h, as it defines its own.

Running modinfo on atl2.ko, we get conflicting information:

$ /sbin/modinfo drivers/net/atlx/atl2.ko | egrep "version|description|author"
version:        2.2.3
description:    Atheros Fast Ethernet Network Driver
author:         Atheros Corporation <xiong.huang@atheros.com>, Chris Snook <csnook@redhat.com>
version:        2.1.3
author:         Xiong Huang <xiong.huang@atheros.com>,  Chris Snook <csnook@redhat.com>, Jay Cliburn <jcliburn@gmail.com>

Move the modinfo data out of atlx.h and into atl1.c to eliminate
the confusion:

$ /sbin/modinfo drivers/net/atlx/atl1.ko | egrep "version|description|author"
version:        2.1.3
author:         Xiong Huang <xiong.huang@atheros.com>,  Chris Snook <csnook@redhat.com>, Jay Cliburn <jcliburn@gmail.com>
description:    Atheros L1 Gigabit Ethernet Driver

$ /sbin/modinfo drivers/net/atlx/atl2.ko | egrep "version|description|author"
version:        2.2.3
description:    Atheros Fast Ethernet Network Driver
author:         Atheros Corporation <xiong.huang@atheros.com>, Chris Snook <csnook@redhat.com>

Reported-by: Scott Scriven <scott.scriven@hp.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Acked-by: Jay Cliburn <jcliburn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agogianfar: fix babbling rx error event bug
Xiaotian Feng [Wed, 27 May 2009 03:48:04 +0000 (20:48 -0700)] 
gianfar: fix babbling rx error event bug

Gianfar interrupt handler uses IEVENT_ERR_MASK to check and handle errors.
Babbling RX error (IEVENT_BABR) should be included in IEVENT_ERROR_MASK.
Otherwise if BABR is raised, it never gets handled nor cleared, and an
interrupt storm results. This has been observed to happen on sending a
burst of ethernet frames to a gianfar based board.

Signed-off-by: Xiaotian Feng <xiaotian.feng@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agocls_cgroup: read classid atomically in classifier
Paul Menage [Wed, 27 May 2009 03:47:02 +0000 (20:47 -0700)] 
cls_cgroup: read classid atomically in classifier

Avoid reading the unsynchronized value cs->classid multiple times,
since it could change concurrently from non-zero to zero; this would
result in the classifier returning a positive result with a bogus
(zero) classid.

Signed-off-by: Paul Menage <menage@google.com>
Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agocpuidle: fix AMD C1E suspend hang
Shaohua Li [Tue, 19 May 2009 08:09:54 +0000 (16:09 +0800)] 
cpuidle: fix AMD C1E suspend hang

When AMD C1E is enabled, local APIC timer will stop even in C1. To avoid
suspend/resume hang, this patch removes C1 and replace it with a cpu_relax() in
suspend/resume path. This hasn't any impact in runtime path.

http://bugzilla.kernel.org/show_bug.cgi?id=13233

[ impact: avoid suspend/resume hang in AMD CPU with C1E enabled ]

Tested-by: Dmitry Lyzhyn <thisistempbox@yahoo.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
15 years agocpuidle: makes AMD C1E work in acpi_idle
Shaohua Li [Tue, 19 May 2009 08:09:42 +0000 (16:09 +0800)] 
cpuidle: makes AMD C1E work in acpi_idle

When AMD C1E is enabled, local APIC timer will stop even in C1.
This patch uses broadcast IPI to replace local APIC timer in C1.

http://bugzilla.kernel.org/show_bug.cgi?id=13233

[ impact: avoid boot hang in AMD CPU with C1E enabled ]

Tested-by: Dmitry Lyzhyn <thisistempbox@yahoo.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
15 years agoRevert "powerpc: Rework dma-noncoherent to use generic vmalloc layer"
Benjamin Herrenschmidt [Wed, 27 May 2009 03:33:14 +0000 (13:33 +1000)] 
Revert "powerpc: Rework dma-noncoherent to use generic vmalloc layer"

This reverts commit 33f00dcedb0e22cdb156a23632814fc580fcfcf8.

    While it was a good idea to try to use the mm/vmalloc.c allocator instead
    of our own (in fact, ours is itself a dup on an old variant of the vmalloc
    one), unfortunately, the approach is terminally busted since
    dma_alloc_coherent() can be called at interrupt time or in atomic contexts
    and there's little chances we'll make the code in mm/vmalloc.c cope with\       that :-(

    Until we can get the generic code to forbid that idiocy and fix all
    drivers abusing it, we pretty much have no choice but revert to
    our custom virtual space allocator.

    There's also a problem with SMP safety since freeing such mapping
    would require an IPI which cannot be done at interrupt time.

    However, right now, I don't think we support any platform that is
    both SMP and has non-coherent DMA (don't laugh, I know such things
    do exist !) so we can sort that out later.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agodrm/i915: Apply a big hammer to 865 GEM object CPU cache flushing.
Eric Anholt [Wed, 27 May 2009 01:46:16 +0000 (18:46 -0700)] 
drm/i915: Apply a big hammer to 865 GEM object CPU cache flushing.

On the 865, but not the 855, the clflush we do appears to not actually make
it out to the hardware all the time.  An easy way to safely reproduce was
X -retro, which would show that some of the blits involved in drawing the
lovely root weave didn't make it out to the hardware.  Those blits are 32
bytes each, and 1-2 would be missing at various points around the screen.
Other experimentation (doing more clflush, doing more AGP chipset flush,
poking at some more device registers to maybe trigger more flushing) didn't
help.  krh came up with the wbinvd as a way to successfully get all those
blits to appear.

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Fix tiling pitch handling on 8xx.
Eric Anholt [Wed, 27 May 2009 00:44:56 +0000 (17:44 -0700)] 
drm/i915: Fix tiling pitch handling on 8xx.

The pitch field is an exponent on pre-965, so we were rejecting buffers
on 8xx that we shouldn't have.  915 got lucky in that the largest legal
value happened to match (8KB / 512 = 0x10), but 8xx has a smaller tile width.
Additionally, we programmed that bad value into the register on 8xx, so the
only pitch that would work correctly was 4096 (512-1023 pixels), while others
would probably give bad rendering or hangs.

Signed-off-by: Eric Anholt <eric@anholt.net>
fd.o bug #20473.

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Wed, 27 May 2009 01:03:40 +0000 (18:03 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6:
  tomoyo: add missing call to cap_bprm_set_creds

15 years agotomoyo: add missing call to cap_bprm_set_creds
Herton Ronaldo Krzesinski [Tue, 26 May 2009 03:15:53 +0000 (12:15 +0900)] 
tomoyo: add missing call to cap_bprm_set_creds

cap_bprm_set_creds() has to be called from security_bprm_set_creds().
TOMOYO forgot to call cap_bprm_set_creds() from tomoyo_bprm_set_creds()
and suid executables were not being working.

Make sure we call cap_bprm_set_creds() with TOMOYO, to set credentials
properly inside tomoyo_bprm_set_creds().

Signed-off-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Acked-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <jmorris@namei.org>
15 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 26 May 2009 22:06:12 +0000 (15:06 -0700)] 
Merge branch 'x86-fixes-for-linus' of git://git./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: avoid back to back on_each_cpu in cpa_flush_array
  x86, relocs: ignore R_386_NONE in kernel relocation entries

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt...
Linus Torvalds [Tue, 26 May 2009 21:48:36 +0000 (14:48 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/anholt/drm-intel

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel:
  drm/i915: Add support for VGA load detection (pre-945).
  drm/i915: Use an I2C algo to do the flip to SDVO DDC bus.
  drm/i915: Determine type before initialising connector
  drm/i915: Return SDVO LVDS VBT mode if no EDID modes are detected.
  drm/i915: Fetch SDVO LVDS mode lines from VBT, then reserve them
  i915: support 8xx desktop cursors
  drm/i915: allocate large pointer arrays with vmalloc

15 years agoMerge branch 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 26 May 2009 21:48:03 +0000 (14:48 -0700)] 
Merge branch 'core-fixes-for-linus' of git://git./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:
  oprofile: fix cpu buffer size

15 years agox86: avoid back to back on_each_cpu in cpa_flush_array
Pallipadi, Venkatesh [Tue, 26 May 2009 17:33:35 +0000 (10:33 -0700)] 
x86: avoid back to back on_each_cpu in cpa_flush_array

Cleanup cpa_flush_array() to avoid back to back on_each_cpu() calls.

[ Impact: optimizes fix 0af48f42df15b97080b450d24219dd95db7b929a ]

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
15 years agoMerge branch 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
Linus Torvalds [Tue, 26 May 2009 19:15:35 +0000 (12:15 -0700)] 
Merge branch 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6

* 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
  NFSv4: Fix the case where NFSv4 renewal fails
  nfs: fix build error in nfsroot with initconst
  XPRTRDMA: fix client rpcrdma FRMR registration on mlx4 devices

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Tue, 26 May 2009 19:14:46 +0000 (12:14 -0700)] 
Merge branch 'for-linus' of git://git./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 - Add missing check of pin vref 50 and others in Realtek codecs
  ALSA: hda - Add 5stack-no-fp model for STAC927x
  ALSA: hda - Add forced codec-slots for ASUS W5Fm

15 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Tue, 26 May 2009 19:13:33 +0000 (12:13 -0700)] 
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] powernow-k8: determine exact CPU frequency for HW Pstates
  [CPUFREQ] powernow-k8 cleanup msg if BIOS does not export ACPI _PSS cpufreq data
  [CPUFREQ] fix timer teardown in ondemand governor
  [CPUFREQ] fix timer teardown in conservative governor
  [CPUFREQ] remove rwsem lock from CPUFREQ_GOV_STOP call
  [CPUFREQ] powernow-k7 build fix when ACPI=n
  [CPUFREQ] add atom family to p4-clockmod

15 years agolguest: fix on Intel when KVM loaded (unhandled trap 13)
Rusty Russell [Tue, 26 May 2009 11:24:41 +0000 (20:54 +0930)] 
lguest: fix on Intel when KVM loaded (unhandled trap 13)

When KVM is loaded, and hence VT set up, the vmcall instruction in an
lguest guest causes a #GP, not #UD.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agokmod: Release sub_info on cred allocation failure.
Tetsuo Handa [Tue, 26 May 2009 08:11:03 +0000 (09:11 +0100)] 
kmod: Release sub_info on cred allocation failure.

call_usermodehelper_setup() forgot to kfree(sub_info)
when prepare_usermodehelper_creds() failed.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Tue, 26 May 2009 19:10:19 +0000 (12:10 -0700)] 
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (21 commits)
  r8169: avoid losing MSI interrupts
  tcp: tcp_vegas ssthresh bugfix
  mac8390: fix regression caused during net_device_ops conversion
  gianfar: fix BUG under load after introduction of skb recycling
  wimax/i2400m: usb: fix device reset on autosuspend while not yet idle
  RxRPC: Error handling for rxrpc_alloc_connection()
  ipv4: Fix oops with FIB_TRIE
  pktgen: do not access flows[] beyond its length
  gigaset: beyond ARRAY_SIZE of iwb->data
  IPv6: set RTPROT_KERNEL to initial route
  net: fix rtable leak in net/ipv4/route.c
  net: fix length computation in rt_check_expire()
  wireless: beyond ARRAY_SIZE of intf->crypto_stats
  iwlwifi: update 5000 ucode support to version 2 of API
  cfg80211: fix race between core hint and driver's custom apply
  airo: fix airo_get_encode{,ext} buffer overflow like I mean it...
  ath5k: fix interpolation with equal power levels
  iwlwifi: do not cancel delayed work inside spin_lock_irqsave
  ath5k: fix exp off-by-one when computing OFDM delta slope
  wext: verify buffer size for SIOCSIWENCODEEXT
  ...

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Tue, 26 May 2009 19:09:32 +0000 (12:09 -0700)] 
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/mm: Fix broken MMU PID stealing on !SMP

15 years agoMerge branch 'for-linus' of git://neil.brown.name/md
Linus Torvalds [Tue, 26 May 2009 19:09:03 +0000 (12:09 -0700)] 
Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md: don't use locked_ioctl.
  md: don't update curr_resync_completed without also updating reshape_position.
  md: raid5: avoid sector values going negative when testing reshape progress.
  md: export 'frozen' resync state through sysfs
  md: bitmap: improve bitmap maintenance code.
  md: improve errno return when setting array_size
  md: always update level / chunk_size / layout when writing v1.x metadata.

15 years agoNFSv4: Fix the case where NFSv4 renewal fails
Trond Myklebust [Tue, 26 May 2009 18:51:00 +0000 (14:51 -0400)] 
NFSv4: Fix the case where NFSv4 renewal fails

If the asynchronous lease renewal fails (usually due to a soft timeout),
then we _must_ schedule state recovery in order to ensure that we don't
lose the lease unnecessarily or, if the lease is already lost, that we
recover the locking state promptly...

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
15 years agonfs: fix build error in nfsroot with initconst
Sam Ravnborg [Tue, 26 May 2009 18:51:00 +0000 (14:51 -0400)] 
nfs: fix build error in nfsroot with initconst

fix build error with latest kbuild adjustments to initconst.

The commit a447c0932445f92ce6f4c1bd020f62c5097a7842 ("vfs: Use
const for kernel parser table") changed:

    static match_table_t __initdata tokens = {
to
    static match_table_t __initconst tokens = {

But the missing const causes popwerpc to fail with latest
updates to __initconst like this:

fs/nfs/nfsroot.c:400: error: __setup_str_nfs_root_setup causes a section type conflict
fs/nfs/nfsroot.c:400: error: __setup_str_nfs_root_setup causes a section type conflict

The bug is only present with kbuild-next.
Following patch has been build tested.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
15 years agoXPRTRDMA: fix client rpcrdma FRMR registration on mlx4 devices
Vu Pham [Tue, 26 May 2009 18:51:00 +0000 (14:51 -0400)] 
XPRTRDMA: fix client rpcrdma FRMR registration on mlx4 devices

mlx4/connectX FRMR requires local write enable together with remote
rdma write enable. This fixes NFS/RDMA operation over the ConnectX
Infiniband HCA in the default memreg mode.

Signed-off-by: Vu Pham <vu@mellanox.com>
Signed-off-by: Tom Talpey <tmtalpey@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
15 years agodrm/i915: Add support for VGA load detection (pre-945).
Ma Ling [Tue, 26 May 2009 03:31:00 +0000 (11:31 +0800)] 
drm/i915: Add support for VGA load detection (pre-945).

Two approaches for VGA detections: hot plug detection for 945G onwards
and load pipe detection for Pre-945G.  Load pipe detection will get one free
pipe, set border color as red and blue, then check CRT status by
swf register.  This is a sync-up with the 2D driver.

Signed-off-by: Ma Ling <ling.ma@intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years ago[CPUFREQ] powernow-k8: determine exact CPU frequency for HW Pstates
Andreas Herrmann [Wed, 22 Apr 2009 11:48:33 +0000 (13:48 +0200)] 
[CPUFREQ] powernow-k8: determine exact CPU frequency for HW Pstates

Slightly modified by trenn@suse.de -> only do this on fam 10h and fam 11h.

Currently powernow-k8 determines CPU frequency from ACPI PSS objects, but
according to AMD family 11h BKDG this frequency is just a rounded value:

  "CoreFreq (MHz) = The CPU COF specified by MSRC001_00[6B:64][CpuFid]
  rounded to the nearest 100 Mhz."

As a consequnce powernow-k8 reports wrong CPU frequency on some systems,
e.g. on Turion X2 Ultra:

  powernow-k8: Found 1 AMD Turion(tm)X2 Ultra DualCore Mobile ZM-82
               processors (2 cpu cores) (version 2.20.00)
  powernow-k8:    0 : pstate 0 (2200 MHz)
  powernow-k8:    1 : pstate 1 (1100 MHz)
  powernow-k8:    2 : pstate 2 (600 MHz)

But this is wrong as frequency for Pstate2 is 550 MHz. x86info reports it
correctly:

  #x86info -a |grep Pstate
  ...
  Pstate-0: fid=e, did=0, vid=24 (2200MHz)
  Pstate-1: fid=e, did=1, vid=30 (1100MHz)
  Pstate-2: fid=e, did=2, vid=3c (550MHz) (current)

Solution is to determine the frequency directly from Pstate MSRs instead
of using rounded values from ACPI table.

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] powernow-k8 cleanup msg if BIOS does not export ACPI _PSS cpufreq data
Thomas Renninger [Wed, 22 Apr 2009 11:48:32 +0000 (13:48 +0200)] 
[CPUFREQ] powernow-k8 cleanup msg if BIOS does not export ACPI _PSS cpufreq data

- Make the message shorter and easier to grep for
- Use printk_once instead of WARN_ONCE (functionality of these was mixed)

Signed-off-by: Thomas Renninger <trenn@suse.de>
Cc: Langsdorf, Mark <mark.langsdorf@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] fix timer teardown in ondemand governor
Mathieu Desnoyers [Sun, 17 May 2009 14:30:45 +0000 (10:30 -0400)] 
[CPUFREQ] fix timer teardown in ondemand governor

* Rafael J. Wysocki (rjw@sisk.pl) wrote:
> This message has been generated automatically as a part of a report
> of regressions introduced between 2.6.28 and 2.6.29.
>
> The following bug entry is on the current list of known regressions
> introduced between 2.6.28 and 2.6.29.  Please verify if it still should
> be listed and let me know (either way).
>
>
> Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=13186
> Subject : cpufreq timer teardown problem
> Submitter : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Date : 2009-04-23 14:00 (24 days old)
> References : http://marc.info/?l=linux-kernel&m=124049523515036&w=4
> Handled-By : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Patch : http://patchwork.kernel.org/patch/19754/
>    http://patchwork.kernel.org/patch/19753/
>

(updated changelog)

cpufreq fix timer teardown in ondemand governor

The problem is that dbs_timer_exit() uses cancel_delayed_work() when it should
use cancel_delayed_work_sync(). cancel_delayed_work() does not wait for the
workqueue handler to exit.

The ondemand governor does not seem to be affected because the
"if (!dbs_info->enable)" check at the beginning of the workqueue handler returns
immediately without rescheduling the work. The conservative governor in
2.6.30-rc has the same check as the ondemand governor, which makes things
usually run smoothly. However, if the governor is quickly stopped and then
started, this could lead to the following race :

dbs_enable could be reenabled and multiple do_dbs_timer handlers would run.
This is why a synchronized teardown is required.

The following patch applies to, at least, 2.6.28.x, 2.6.29.1, 2.6.30-rc2.

Depends on patch
cpufreq: remove rwsem lock from CPUFREQ_GOV_STOP call

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: gregkh@suse.de
CC: stable@kernel.org
CC: cpufreq@vger.kernel.org
CC: Ingo Molnar <mingo@elte.hu>
CC: rjw@sisk.pl
CC: Ben Slusky <sluskyb@paranoiacs.org>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] fix timer teardown in conservative governor
Mathieu Desnoyers [Sun, 17 May 2009 14:29:33 +0000 (10:29 -0400)] 
[CPUFREQ] fix timer teardown in conservative governor

* Rafael J. Wysocki (rjw@sisk.pl) wrote:
> This message has been generated automatically as a part of a report
> of regressions introduced between 2.6.28 and 2.6.29.
>
> The following bug entry is on the current list of known regressions
> introduced between 2.6.28 and 2.6.29.  Please verify if it still should
> be listed and let me know (either way).
>
>
> Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=13186
> Subject : cpufreq timer teardown problem
> Submitter : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Date : 2009-04-23 14:00 (24 days old)
> References : http://marc.info/?l=linux-kernel&m=124049523515036&w=4
> Handled-By : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Patch : http://patchwork.kernel.org/patch/19754/
>    http://patchwork.kernel.org/patch/19753/
>

(re-send with updated changelog)

cpufreq fix timer teardown in conservative governor

The problem is that dbs_timer_exit() uses cancel_delayed_work() when it should
use cancel_delayed_work_sync(). cancel_delayed_work() does not wait for the
workqueue handler to exit.

The ondemand governor does not seem to be affected because the
"if (!dbs_info->enable)" check at the beginning of the workqueue handler returns
immediately without rescheduling the work. The conservative governor in
2.6.30-rc has the same check as the ondemand governor, which makes things
usually run smoothly. However, if the governor is quickly stopped and then
started, this could lead to the following race :

dbs_enable could be reenabled and multiple do_dbs_timer handlers would run.
This is why a synchronized teardown is required.

Depends on patch
cpufreq: remove rwsem lock from CPUFREQ_GOV_STOP call

The following patch applies to 2.6.30-rc2. Stable kernels have a similar
issue which should also be fixed, but the code changed between 2.6.29
and 2.6.30, so this patch only applies to 2.6.30-rc.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: gregkh@suse.de
CC: stable@kernel.org
CC: cpufreq@vger.kernel.org
CC: Ingo Molnar <mingo@elte.hu>
CC: rjw@sisk.pl
CC: Ben Slusky <sluskyb@paranoiacs.org>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] remove rwsem lock from CPUFREQ_GOV_STOP call
Mathieu Desnoyers [Sun, 17 May 2009 14:23:52 +0000 (10:23 -0400)] 
[CPUFREQ] remove rwsem lock from CPUFREQ_GOV_STOP call

* Rafael J. Wysocki (rjw@sisk.pl) wrote:
> This message has been generated automatically as a part of a report
> of regressions introduced between 2.6.28 and 2.6.29.
>
> The following bug entry is on the current list of known regressions
> introduced between 2.6.28 and 2.6.29.  Please verify if it still should
> be listed and let me know (either way).
>
>
> Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=13186
> Subject : cpufreq timer teardown problem
> Submitter : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Date : 2009-04-23 14:00 (24 days old)
> References : http://marc.info/?l=linux-kernel&m=124049523515036&w=4
> Handled-By : Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> Patch : http://patchwork.kernel.org/patch/19754/
>    http://patchwork.kernel.org/patch/19753/

The patches linked above depend on the following patch to remove
circular locking dependency :

cpufreq: remove rwsem lock from CPUFREQ_GOV_STOP call

(the following issue was faced when using cancel_delayed_work_sync() in the
timer teardown (which fixes a race).

* KOSAKI Motohiro (kosaki.motohiro@jp.fujitsu.com) wrote:
> Hi
>
> my box output following warnings.
> it seems regression by commit 7ccc7608b836e58fbacf65ee4f8eefa288e86fac.
>
> A: work -> do_dbs_timer()  -> cpu_policy_rwsem
> B: store() -> cpu_policy_rwsem -> cpufreq_governor_dbs() -> work
>
>

Hrm, I think it must be due to my attempt to fix the timer teardown race
in ondemand governor mixed with new locking behavior in 2.6.30-rc.

The rwlock seems to be taken around the whole call to
cpufreq_governor_dbs(), when it should be only taken around accesses to
the locked data, and especially *not* around the call to
dbs_timer_exit().

Reverting my fix attempt would put the teardown race back in place
(replacing the cancel_delayed_work_sync by cancel_delayed_work).
Instead, a proper fix would imply modifying this critical section :

cpufreq.c: __cpufreq_remove_dev()
...
        if (cpufreq_driver->target)
                __cpufreq_governor(data, CPUFREQ_GOV_STOP);

        unlock_policy_rwsem_write(cpu);

To make sure the __cpufreq_governor() callback is not called with rwsem
held. This would allow execution of cancel_delayed_work_sync() without
being nested within the rwsem.

Applies on top of the 2.6.30-rc5 tree.

Required to remove circular dep in teardown of both conservative and
ondemande governors so they can use cancel_delayed_work_sync().
CPUFREQ_GOV_STOP does not modify the policy, therefore this locking seemed
unneeded.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
CC: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Greg KH <greg@kroah.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: "Rafael J. Wysocki" <rjw@sisk.pl>
CC: Ben Slusky <sluskyb@paranoiacs.org>
CC: Chris Wright <chrisw@sous-sol.org>
CC: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] powernow-k7 build fix when ACPI=n
Dave Jones [Thu, 23 Apr 2009 17:36:12 +0000 (13:36 -0400)] 
[CPUFREQ] powernow-k7 build fix when ACPI=n

arch/x86/kernel/cpu/cpufreq/powernow-k7.c:172: warning: 'invalidate_entry' defined but not used

Reported-by: Toralf Förster <toralf.foerster@gmx.de>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years ago[CPUFREQ] add atom family to p4-clockmod
Jarod Wilson [Fri, 6 Mar 2009 20:24:57 +0000 (20:24 +0000)] 
[CPUFREQ] add atom family to p4-clockmod

Some atom procs don't do freq scaling (such as the atom 330 on my own
littlefalls2 board). By adding the atom family here, we at least get
the benefit of passive cooling in a thermal emergency. Not sure how
to see that its actually helping any, but the driver does bind and
claim its functioning on my atom 330.

Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: Dave Jones <davej@redhat.com>
15 years agoMerge branch 'linux-2.6.30.y' of git://git.kernel.org/pub/scm/linux/kernel/git/inaky...
David S. Miller [Tue, 26 May 2009 05:56:11 +0000 (22:56 -0700)] 
Merge branch 'linux-2.6.30.y' of git://git./linux/kernel/git/inaky/wimax

15 years agor8169: avoid losing MSI interrupts
David Dillow [Fri, 22 May 2009 15:29:34 +0000 (15:29 +0000)] 
r8169: avoid losing MSI interrupts

The 8169 chip only generates MSI interrupts when all enabled event
sources are quiescent and one or more sources transition to active. If
not all of the active events are acknowledged, or a new event becomes
active while the existing ones are cleared in the handler, we will not
see a new interrupt.

The current interrupt handler masks off the Rx and Tx events once the
NAPI handler has been scheduled, which opens a race window in which we
can get another Rx or Tx event and never ACK'ing it, stopping all
activity until the link is reset (ifconfig down/up). Fix this by always
ACK'ing all event sources, and loop in the handler until we have all
sources quiescent.

Signed-off-by: David Dillow <dave@thedillows.org>
Tested-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agox86, relocs: ignore R_386_NONE in kernel relocation entries
Tejun Heo [Tue, 26 May 2009 05:42:40 +0000 (14:42 +0900)] 
x86, relocs: ignore R_386_NONE in kernel relocation entries

For relocatable 32bit kernels, boot/compressed/relocs.c processes
relocation entries in the kernel image and appends it to the kernel
image such that boot/compressed/head_32.S can relocate the kernel.
The kernel image is one statically linked object and only uses two
relocation types - R_386_PC32 and R_386_32, of the two only the latter
needs massaging during kernel relocation and thus handled by relocs.
R_386_PC32 is ignored and all other relocation types are considered
error.

When the target of a relocation resides in a discarded section,
binutils doesn't throw away the relocation record but nullifies it by
changing it to R_386_NONE, which unfortunately makes relocs fail.

The problem was triggered by yet out-of-tree x86 stack unwind patches
but given the binutils behavior, ignoring R_386_NONE is the right
thing to do.

The problem has been tracked down to binutils behavior by Jan Beulich.

[ Impact: fix build with certain binutils by ignoring R_386_NONE ]

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jan Beulich <JBeulich@novell.com>
Cc: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <4A1B8150.40702@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
15 years agotcp: tcp_vegas ssthresh bugfix
Doug Leith [Tue, 26 May 2009 05:44:59 +0000 (22:44 -0700)] 
tcp: tcp_vegas ssthresh bugfix

This patch fixes ssthresh accounting issues in tcp_vegas when cwnd decreases

Signed-off-by: Doug Leith <doug.leith@nuim.ie>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomac8390: fix regression caused during net_device_ops conversion
Finn Thain [Tue, 26 May 2009 05:43:49 +0000 (22:43 -0700)] 
mac8390: fix regression caused during net_device_ops conversion

Changeset ca17584bf2ad1b1e37a5c0e4386728cc5fc9dabc ("mac8390: update
to net_device_ops") broke mac8390 by adding 8390.o to the link. That
meant that lib8390.c was included twice, once in mac8390.c and once in
8390.c, subject to different macros. This patch reverts that by
avoiding the wrappers in 8390.c. They seem to be of no value since
COMPAT_NET_DEV_OPS is going away soon.

Tested with a Kinetics EtherPort card.

Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopowerpc/mm: Fix broken MMU PID stealing on !SMP
Hideo Saito [Sun, 24 May 2009 15:33:34 +0000 (15:33 +0000)] 
powerpc/mm: Fix broken MMU PID stealing on !SMP

The recent rework of the MMU PID handling for non-hash CPUs has a
subtle bug in the !SMP "optimized" variant of the PID stealing
function.  It clears the PID in the mm context before it calls
local_flush_tlb_mm(). However, the later will not flush anything
if the PID in the context is clear...

Signed-off-by: Hideo Saito <hsaito.ppc@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agomd: don't use locked_ioctl.
NeilBrown [Tue, 26 May 2009 02:57:36 +0000 (12:57 +1000)] 
md: don't use locked_ioctl.

md has no need for the BKL - it does its own locking.
So md_ioctl doesn't need to be a locked_ioctl.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: don't update curr_resync_completed without also updating reshape_position.
NeilBrown [Tue, 26 May 2009 02:57:21 +0000 (12:57 +1000)] 
md: don't update curr_resync_completed without also updating reshape_position.

In order for the metadata to always be consistent, we mustn't updated
curr_resync_completed without also updating reshape_position.

The reshape code updates both at the same time.  However since
commit 97e4f42d62badb0f9fbc27c013e89bc1336a03bc
the common md_do_sync will sometimes update curr_resync_completed
but is not in a position to update reshape_position.
So if MD_RECOVERY_RESHAPE is set (indicating that a reshape is
happening, so reshape_position might change), don't update
curr_resync_completed in md_do_sync, leave it to the per-personality
reshape code.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: raid5: avoid sector values going negative when testing reshape progress.
NeilBrown [Tue, 26 May 2009 02:41:08 +0000 (12:41 +1000)] 
md: raid5: avoid sector values going negative when testing reshape progress.

As sector_t in unsigned, we cannot afford to let 'safepos' etc go
negative.
So replace
   a -= b;
by
   a -= min(b,a);

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: export 'frozen' resync state through sysfs
NeilBrown [Mon, 25 May 2009 23:41:17 +0000 (09:41 +1000)] 
md: export 'frozen' resync state through sysfs

The md resync engine has a 'frozen' state which ensures that
no resync/recovery.  This is used to avoid races.

Export this state through the 'sync_action' sysfs attribute
so that user-space can benefit and also avoid some races.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: bitmap: improve bitmap maintenance code.
NeilBrown [Mon, 25 May 2009 23:41:17 +0000 (09:41 +1000)] 
md: bitmap: improve bitmap maintenance code.

The code for checking which bits in the bitmap can be cleared
has 2 problems:
 1/ it repeatedly takes and drops a spinlock, where it would make
    more sense to just hold on to it most of the time.
 2/ it doesn't make use of some opportunities to skip large sections
    of the bitmap

This patch fixes those.  It will only affect CPU consumption, not
correctness.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: improve errno return when setting array_size
NeilBrown [Mon, 25 May 2009 23:41:17 +0000 (09:41 +1000)] 
md: improve errno return when setting array_size

Instead of always returns EINVAL if anything goes wrong
when setting the array size, add the option of
  E2BIG
if the size requested is too large.  This makes it easier
for user-space to be sure what went wrong.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agomd: always update level / chunk_size / layout when writing v1.x metadata.
NeilBrown [Mon, 25 May 2009 23:40:59 +0000 (09:40 +1000)] 
md: always update level / chunk_size / layout when writing v1.x metadata.

We previously didn't update these fields when writing the metadata
because they could never change.  They can now, so we better write
them.
v0.90 metadata always updated these fields.

Signed-off-by: NeilBrown <neilb@suse.de>
15 years agoMerge branch 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Mon, 25 May 2009 22:51:27 +0000 (15:51 -0700)] 
Merge branch 'kvm-updates/2.6.30' of git://git./virt/kvm/kvm

* 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Fix PDPTR reloading on CR4 writes
  KVM: Make paravirt tlb flush also reload the PAE PDPTRs

15 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 25 May 2009 22:50:32 +0000 (15:50 -0700)] 
Merge branch 'x86-fixes-for-linus' of git://git./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: Remove remap percpu allocator for the time being
  x86: cpa_flush_array wbinvd should be done on all CPUs
  x86: bugfix wbinvd() model check instead of family check
  x86: introduce noxsave boot parameter
  x86, setup: revert ACPI 3 E820 extended attributes support
  x86: DMI match for the Sony VGN-Z540N as it needs BIOS reboot

15 years agoKVM: Fix PDPTR reloading on CR4 writes
Avi Kivity [Sun, 24 May 2009 19:19:00 +0000 (22:19 +0300)] 
KVM: Fix PDPTR reloading on CR4 writes

The processor is documented to reload the PDPTRs while in PAE mode if any
of the CR4 bits PSE, PGE, or PAE change.  Linux relies on this
behaviour when zapping the low mappings of PAE kernels during boot.

The code already handled changes to CR4.PAE; augment it to also notice changes
to PSE and PGE.

This triggered while booting an F11 PAE kernel; the futex initialization code
runs before any CR3 reloads and writes to a NULL pointer; the futex subsystem
ended up uninitialized, killing PI futexes and pulseaudio which uses them.

Cc: stable@kernel.org
Signed-off-by: Avi Kivity <avi@redhat.com>
15 years agoKVM: Make paravirt tlb flush also reload the PAE PDPTRs
Avi Kivity [Sun, 24 May 2009 19:15:25 +0000 (22:15 +0300)] 
KVM: Make paravirt tlb flush also reload the PAE PDPTRs

The paravirt tlb flush may be used not only to flush TLBs, but also
to reload the four page-directory-pointer-table entries, as it is used
as a replacement for reloading CR3.  Change the code to do the entire
CR3 reloading dance instead of simply flushing the TLB.

Cc: stable@kernel.org
Signed-off-by: Avi Kivity <avi@redhat.com>
15 years agonetfilter: nf_ct_dccp: add missing DCCP protocol changes in event cache
Pablo Neira Ayuso [Mon, 25 May 2009 15:29:43 +0000 (17:29 +0200)] 
netfilter: nf_ct_dccp: add missing DCCP protocol changes in event cache

This patch adds the missing protocol state-change event reporting
for DCCP.

$ sudo conntrack -E
    [NEW] dccp     33 240 src=192.168.0.2 dst=192.168.1.2 sport=57040 dport=5001 [UNREPLIED] src=192.168.1.2 dst=192.168.1.100 sport=5001 dport=57040

With this patch:

$ sudo conntrack -E
    [NEW] dccp     33 240 REQUEST src=192.168.0.2 dst=192.168.1.2 sport=57040 dport=5001 [UNREPLIED] src=192.168.1.2 dst=192.168.1.100 sport=5001 dport=57040

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
15 years agonetfilter: nf_ct_tcp: fix accepting invalid RST segments
Jozsef Kadlecsik [Mon, 25 May 2009 15:23:15 +0000 (17:23 +0200)] 
netfilter: nf_ct_tcp: fix accepting invalid RST segments

Robert L Mathews discovered that some clients send evil TCP RST segments,
which are accepted by netfilter conntrack but discarded by the
destination. Thus the conntrack entry is destroyed but the destination
retransmits data until timeout.

The same technique, i.e. sending properly crafted RST segments, can easily
be used to bypass connlimit/connbytes based restrictions (the sample
script written by Robert can be found in the netfilter mailing list
archives).

The patch below adds a new flag and new field to struct ip_ct_tcp_state so
that checking RST segments can be made more strict and thus TCP conntrack
can catch the invalid ones: the RST segment is accepted only if its
sequence number higher than or equal to the highest ack we seen from the
other direction. (The last_ack field cannot be reused because it is used
to catch resent packets.)

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
15 years agogianfar: fix BUG under load after introduction of skb recycling
Lennert Buytenhek [Mon, 25 May 2009 07:42:34 +0000 (00:42 -0700)] 
gianfar: fix BUG under load after introduction of skb recycling

Since commit 0fd56bb5be6455d0d42241e65aed057244665e5e ("gianfar:
Add support for skb recycling"), gianfar puts skbuffs that are in
the rx ring back onto the recycle list as-is in case there was a
receive error, but this breaks the following invariant: that all
skbuffs on the recycle list have skb->data = skb->head + NET_SKB_PAD.

The RXBUF_ALIGNMENT realignment done in gfar_new_skb() will be done
twice on skbuffs recycled in this way, causing there not to be enough
room in the skb anymore to receive a full packet, eventually leading
to an skb_over_panic from gfar_clean_rx_ring() -> skb_put().

Resetting the skb->data pointer to skb->head + NET_SKB_PAD before
putting the skb back onto the recycle list restores the mentioned
invariant, and should fix this issue.

Reported-by: Michael Guntsche <mike@it-loops.com>
Tested-by: Michael Guntsche <mike@it-loops.com>
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoALSA: hda - Add missing check of pin vref 50 and others in Realtek codecs
Takashi Iwai [Mon, 25 May 2009 06:06:02 +0000 (08:06 +0200)] 
ALSA: hda - Add missing check of pin vref 50 and others in Realtek codecs

Some Realtek codecs like ALC861 seem to support only VREF50 while the
current driver assumes it's only VREF80.  Check other VREF bits to set
the correct value.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agox86: Remove remap percpu allocator for the time being
Tejun Heo [Mon, 25 May 2009 03:01:59 +0000 (12:01 +0900)] 
x86: Remove remap percpu allocator for the time being

Remap percpu allocator has subtle bug when combined with page
attribute changing.  Remap percpu allocator aliases PMD pages for the
first chunk and as pageattr doesn't know about the alias it ends up
updating page attributes of the original mapping thus leaving the
alises in inconsistent state which might lead to subtle data
corruption.  Please read the following threads for more information:

  http://thread.gmane.org/gmane.linux.kernel/835783

The following is the proposed fix which teaches pageattr about percpu
aliases.

  http://thread.gmane.org/gmane.linux.kernel/837157

However, the above changes are deemed too pervasive for upstream
inclusion for 2.6.30 release, so this patch essentially disables
the remap allocator for the time being.

Signed-off-by: Tejun Heo <tj@kernel.org>
LKML-Reference: <4A1A0A27.4050301@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Mon, 25 May 2009 02:38:25 +0000 (19:38 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/rafael/suspend-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspend-2.6:
  PM: Do not hold dpm_list_mtx while disabling/enabling nonboot CPUs

15 years agoasync: make sure independent async domains can't accidentally entangle
James Bottomley [Sun, 24 May 2009 20:03:43 +0000 (13:03 -0700)] 
async: make sure independent async domains can't accidentally entangle

The problem occurs when async_synchronize_full_domain() is called when
the async_pending list is not empty.  This will cause lowest_running()
to return the cookie of the first entry on the async_pending list, which
might be nothing at all to do with the domain being asked for and thus
cause the domain synchronization to wait for an unrelated domain.   This
can cause a deadlock if domain synchronization is used from one domain
to wait for another.

Fix by running over the async_pending list to see if any pending items
actually belong to our domain (and return their cookies if they do).

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoPM: Do not hold dpm_list_mtx while disabling/enabling nonboot CPUs
Rafael J. Wysocki [Sun, 24 May 2009 19:15:07 +0000 (21:15 +0200)] 
PM: Do not hold dpm_list_mtx while disabling/enabling nonboot CPUs

We shouldn't hold dpm_list_mtx while executing
[disable|enable]_nonboot_cpus(), because theoretically this may lead
to a deadlock as shown by the following example (provided by Johannes
Berg):

CPU 3       CPU 2                     CPU 1
                                      suspend/hibernate
            something:
            rtnl_lock()               device_pm_lock()
                                       -> mutex_lock(&dpm_list_mtx)

            mutex_lock(&dpm_list_mtx)

linkwatch_work
 -> rtnl_lock()
                                      disable_nonboot_cpus()
                                       -> flush CPU 3 workqueue

Fortunately, device drivers are supposed to stop any activities that
might lead to the registration of new device objects way before
disable_nonboot_cpus() is called, so it shouldn't be necessary to
hold dpm_list_mtx over the entire late part of device suspend and
early part of device resume.

Thus, during the late suspend and the early resume of devices acquire
dpm_list_mtx only when dpm_list is going to be traversed and release
it right after that.

This patch is reported to fix the regressions tracked as
http://bugzilla.kernel.org/show_bug.cgi?id=13245.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Reported-by: Miles Lane <miles.lane@gmail.com>
Tested-by: Ming Lei <tom.leiming@gmail.com>
15 years agoUse a format for linux_banner
Alex Riesen [Sun, 24 May 2009 13:30:48 +0000 (15:30 +0200)] 
Use a format for linux_banner

There is no format specifiers left in the linux_banner, and gcc-4.3
complains seeing the printk.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoALSA: hda - Add 5stack-no-fp model for STAC927x
Takashi Iwai [Sun, 24 May 2009 17:00:08 +0000 (19:00 +0200)] 
ALSA: hda - Add 5stack-no-fp model for STAC927x

The recent fix for the headphone volume control on IDT/STAC codecs
resulted in the removal of invalid "Side" volume eventually.  But,
if the front panel doesn't exist, this setup could be regarded as a
sort of regression, as reported in kernel bug #13250.

Now as a workaround, a new model 5stack-no-fp is added so that the user
without the front panel can choose this one explicitly.

Reference: bko#13250
http://bugzilla.kernel.org/show_bug.cgi?id=13250

Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agoALSA: hda - Add forced codec-slots for ASUS W5Fm
Ozan Çağlayan [Sat, 23 May 2009 12:00:04 +0000 (15:00 +0300)] 
ALSA: hda - Add forced codec-slots for ASUS W5Fm

ASUS W5Fm needs the fixed codec-slots to probe to override the BIOS
problem like W5F.

Tested-by: Alp Kılıç <kilic.alp@gmail.com>
Signed-off-by: Ozan Çağlayan <ozan@pardus.org.tr>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
15 years agoLinux 2.6.30-rc7 v2.6.30-rc7
Linus Torvalds [Sat, 23 May 2009 21:47:00 +0000 (14:47 -0700)] 
Linux 2.6.30-rc7

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Sat, 23 May 2009 20:44:00 +0000 (13:44 -0700)] 
Merge git://git./linux/kernel/git/jejb/scsi-rc-fixes-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
  [SCSI] mpt2sas: fix driver version inconsistency
  [SCSI] 3w-xxxx: scsi_dma_unmap fix
  [SCSI] 3w-9xxx: scsi_dma_unmap fix
  [SCSI] ses: fix problems caused by empty SES provided name
  [SCSI] fc-transport: Close state transition-window during rport deletion.
  [SCSI] initialize max_target_blocked in scsi_alloc_target
  [SCSI] fnic: Add new Cisco PCI-Express FCoE HBA

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Sat, 23 May 2009 20:42:53 +0000 (13:42 -0700)] 
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] Avoid open on possible directories since Samba now rejects them

15 years ago[CIFS] Avoid open on possible directories since Samba now rejects them
Steve French [Sat, 23 May 2009 18:57:25 +0000 (18:57 +0000)] 
[CIFS] Avoid open on possible directories since Samba now rejects them

Small change (mostly formatting) to limit lookup based open calls to
file create only.

After discussion yesteday on samba-technical about the posix lookup
regression,  and looking at a problem with cifs posix open to one
particular Samba version, Jeff and JRA realized that Samba server's
behavior changed in this area (posix open behavior on files vs.
directories).   To make this behavior consistent, JRA just made a
fix to Samba server to alter how it handles open of directories (now
returning the equivalent of EISDIR instead of success). Since we don't
know at lookup time whether the inode is a directory or file (and
thus whether posix open will succeed with most current Samba server),
this change avoids the posix open code on lookup open (just issues
posix open on creates).    This gets the semantic benefits we want
(atomicity, posix byte range locks, improved write semantics on newly
created files) and file create still is fast, and we avoid the problem
that Jeff noticed yesterday with "openat" (and some open directory
calls) of non-cached directories to one version of Samba server, and
will work with future Samba versions (which include the fix jra just
pushed into Samba server).  I confirmed this approach with jra
yesterday and with Shirish today.

Posix open is only called (at lookup time) for file create now.
For opens (rather than creates), because we do not know if it
is a file or directory yet, and current Samba no longer allows
us to do posix open on dirs, we could end up wasting an open call
on what turns out to be a dir. For file opens, we wait to call posix
open till cifs_open.  It could be added here (lookup) in the future
but the performance tradeoff of the extra network request when EISDIR
or EACCES is returned would have to be weighed against the 50%
reduction in network traffic in the other paths.

Reviewed-by: Shirish Pargaonkar <shirishp@us.ibm.com>
Tested-by: Jeff Layton <jlayton@redhat.com>
CC: Jeremy Allison <jra@samba.org>
Signed-off-by: Steve French <sfrench@us.ibm.com>
15 years agoInput: multitouch - augment event semantics documentation
Henrik Rydberg [Sat, 23 May 2009 16:51:21 +0000 (09:51 -0700)] 
Input: multitouch - augment event semantics documentation

Through the collaboration to adapt the N-trig and Stantum HID
drivers to the MT protocol, some semantic clarifications to the
protocol have been made. This patch adds them to the MT documentation.

Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Tested-by: Stéphane Chatty <chatty@enac.fr>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
15 years agoInput: multitouch - add tracking ID to the protocol
Henrik Rydberg [Sat, 23 May 2009 16:51:20 +0000 (09:51 -0700)] 
Input: multitouch - add tracking ID to the protocol

There are a few multi-touch devices that support finger tracking
well in hardware, Stantum being the prime example. By exposing the
tracking ID in the MT protocol, evdev bandwidth and cpu usage in
user space can be reduced.

This patch adds the ABS_MT_TRACKING_ID to the MT protocol.

Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Tested-by: Stéphane Chatty <chatty@enac.fr>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
15 years agoMerge branch 'for-rmk' of git://git.marvell.com/orion
Russell King [Sat, 23 May 2009 16:24:51 +0000 (17:24 +0100)] 
Merge branch 'for-rmk' of git://git.marvell.com/orion

15 years ago[ARM] disable NX support for OABI-supporting kernels
Russell King [Sat, 23 May 2009 10:36:20 +0000 (11:36 +0100)] 
[ARM] disable NX support for OABI-supporting kernels

Our signal syscall restart handling for these kernels still uses
the userspace stack to build code for restarting the syscall.
Unfortunately, fixing this is non-trivial, and so for the time
being, we resolve the problem by disabling NX support.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>