openwrt/staging/blogic.git
16 years agoMerge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Tue, 18 Mar 2008 14:32:23 +0000 (07:32 -0700)]
Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/ati_pcigart: fix the PCIGART to use drm_pci to allocate GART table.
  drm/radeon: fixup RV550 chip family
  drm/via: attempt again to stabilise the AGP DMA command submission.
  drm: Fix race that can lockup the kernel

16 years agopciehp: don't enable slot unless forced
Mark Lord [Mon, 17 Mar 2008 20:04:23 +0000 (16:04 -0400)]
pciehp: don't enable slot unless forced

This fixes a 2.6.25 regression reported by Alex Chiang.

Invoke pciehp_enable_slot() at startup only when pciehp_force=1.
Some HP equipment apparently cannot cope with it otherwise.

This restores the (previously working) 2.6.24 behaviour here,
while allowing machines that need a kick to use pciehp_force=1.

This was the original design back in October 2007,
but Kristen suggested we try without it first:

   Kristen Carlson Accardi wrote:
   >I think it would be ok to try allowing the slot to be enabled when not
   >using pciehp_force mode.  We can wrap it later if it proves to break things

This ended up breaking one of Alex's setups,
so it's time to put the wrapper back in now.

Signed-off-by: Mark Lord <mlord@pobox.com>
Acked-by: Alex Chiang <achiang@hp.com>
Acked-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Mon, 17 Mar 2008 16:52:24 +0000 (09:52 -0700)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ahci: Add Marvell 6121 SATA support
  pata_ali: use atapi_cmd_type() to determine cmd type instead of transfer size
  ahci: implement skip_host_reset parameter
  ahci: request all PCI BARs
  devres: implement pcim_iomap_regions_request_all()
  libata-acpi: improve dock event handling

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
Linus Torvalds [Mon, 17 Mar 2008 16:52:19 +0000 (09:52 -0700)]
Merge git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  virtio: fix race in enable_cb
  virtio: Enable netpoll interface for netconsole logging
  virtio: handle > 2 billion page balloon targets
  virtio: Fix sysfs bits to have proper block symlink
  virtio: Use spin_lock_irqsave/restore for virtio-pci

16 years agohfs_bnode_find() can fail, resulting in hfs_bnode_split() breakage
Al Viro [Sun, 16 Mar 2008 22:48:08 +0000 (22:48 +0000)]
hfs_bnode_find() can fail, resulting in hfs_bnode_split() breakage

oops and fs corruption; the latter can happen even on valid fs in case of oom.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoahci: Add Marvell 6121 SATA support
Jose Alberto Reguero [Thu, 13 Mar 2008 22:22:24 +0000 (23:22 +0100)]
ahci: Add Marvell 6121 SATA support

Signed-off-by: Jose Alberto Reguero <jareguero@telefonica.net>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agopata_ali: use atapi_cmd_type() to determine cmd type instead of transfer size
Tejun Heo [Tue, 11 Mar 2008 02:35:00 +0000 (11:35 +0900)]
pata_ali: use atapi_cmd_type() to determine cmd type instead of transfer size

pata_ali was using qc->nbytes to determine whether a command is
data transfer type or not.  As now qc->nbytes can be extended by
padding and draining buffers, these tests are not useful anymore.

Use atapi_cmd_type() instead.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agoahci: implement skip_host_reset parameter
Tejun Heo [Mon, 10 Mar 2008 01:25:25 +0000 (10:25 +0900)]
ahci: implement skip_host_reset parameter

Under certain circumstances (SSP turned off by the BIOS) and for
debugging purposes, skipping global controller reset is helpful.  Add
a kernel parameter for it.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agoahci: request all PCI BARs
Tejun Heo [Tue, 11 Mar 2008 10:52:31 +0000 (19:52 +0900)]
ahci: request all PCI BARs

ahci is often implemented with accompanying SFF compatible interface
and legacy IDE driver may attach to the legacy IO ports when the
controller is already claimed by ahci and vice-versa.  This patch
makes ahci use pcim_iomap_regions_request_all() so that all IO regions
are claimed on attach.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agodevres: implement pcim_iomap_regions_request_all()
Tejun Heo [Wed, 12 Mar 2008 06:26:34 +0000 (15:26 +0900)]
devres: implement pcim_iomap_regions_request_all()

Some drivers need to reserve all PCI BARs to prevent other drivers
misusing unoccupied BARs.  pcim_iomap_regions_request_all() requests
all BARs and iomap specified BARs.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agolibata-acpi: improve dock event handling
Tejun Heo [Wed, 12 Mar 2008 05:24:43 +0000 (14:24 +0900)]
libata-acpi: improve dock event handling

Improve ACPI hotplug handling such that dock event is handled properly.

* Register handlers for dock events.

* Directly detach device on EJECT_REQUEST instead of signaling hotplug
  event.  This prevents libata from accessing severed controller
  and/or device.

* While at it, use named constants for ACPI events and move uevent
  signaling inside host lock.

Original patch and testing by Holger Macht.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Holger Macht <hmacht@suse.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
16 years agovirtio: fix race in enable_cb
Christian Borntraeger [Fri, 14 Mar 2008 13:17:05 +0000 (14:17 +0100)]
virtio: fix race in enable_cb

There is a race in virtio_net, dealing with disabling/enabling the callback.
I saw the following oops:

kernel BUG at /space/kvm/drivers/virtio/virtio_ring.c:218!
illegal operation: 0001 [#1] SMP
Modules linked in: sunrpc dm_mod
CPU: 2 Not tainted 2.6.25-rc1zlive-host-10623-gd358142-dirty #99
Process swapper (pid: 0, task: 000000000f85a610, ksp: 000000000f873c60)
Krnl PSW : 0404300180000000 00000000002b81a6 (vring_disable_cb+0x16/0x20)
           R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:0 CC:3 PM:0 EA:3
Krnl GPRS: 0000000000000001 0000000000000001 0000000010005800 0000000000000001
           000000000f3a0900 000000000f85a610 0000000000000000 0000000000000000
           0000000000000000 000000000f870000 0000000000000000 0000000000001237
           000000000f3a0920 000000000010ff74 00000000002846f6 000000000fa0bcd8
Krnl Code: 00000000002b819aa7110001           tmll    %r1,1
           00000000002b819ea7840004           brc     8,2b81a6
           00000000002b81a2a7f40001           brc     15,2b81a4
          >00000000002b81a6a51b0001           oill    %r1,1
           00000000002b81aa40102000           sth     %r1,0(%r2)
           00000000002b81ae: 07fe               bcr     15,%r14
           00000000002b81b0eb7ff0380024       stmg    %r7,%r15,56(%r15)
           00000000002b81b6a7f13e00           tmll    %r15,15872
Call Trace:
([<000000000fa0bcd0>] 0xfa0bcd0)
 [<00000000002b8350>] vring_interrupt+0x5c/0x6c
 [<000000000010ab08>] do_extint+0xb8/0xf0
 [<0000000000110716>] ext_no_vtime+0x16/0x1a
 [<0000000000107e72>] cpu_idle+0x1c2/0x1e0

The problem can be triggered with a high amount of host->guest traffic.
I think its the following race:

poll says netif_rx_complete
poll calls enable_cb
enable_cb opens the interrupt mask
a new packet comes, an interrupt is triggered----\
enable_cb sees that there is more work           |
enable_cb disables the interrupt                 |
       .                                         V
       .                            interrupt is delivered
       .                            skb_recv_done does atomic napi test, ok
 some waiting                       disable_cb is called->check fails->bang!
       .
poll would do napi check
poll would do disable_cb

The fix is to let enable_cb not disable the interrupt again, but expect the
caller to do the cleanup if it returns false. In that case, the interrupt is
only disabled, if the napi test_set_bit was successful.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (cleaned up doco)
16 years agovirtio: Enable netpoll interface for netconsole logging
Amit Shah [Fri, 29 Feb 2008 10:54:50 +0000 (16:24 +0530)]
virtio: Enable netpoll interface for netconsole logging

Add a new poll_controller handler that the netpoll interface needs.

This enables netconsole logging from a kvm guest over the virtio
net interface.

Signed-off-by: Amit Shah <amitshah@gmx.net>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agovirtio: handle > 2 billion page balloon targets
Rusty Russell [Tue, 18 Mar 2008 03:58:15 +0000 (22:58 -0500)]
virtio: handle > 2 billion page balloon targets

If the host asks for a huge target towards_target() can overflow, and
we up oops as we try to release more pages than we have.  The simple
fix is to use a 64-bit value.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agovirtio: Fix sysfs bits to have proper block symlink
Jeremy Katz [Sun, 2 Mar 2008 22:00:15 +0000 (17:00 -0500)]
virtio: Fix sysfs bits to have proper block symlink

Fix up so that the virtio_blk devices in sysfs link correctly to their
block device.  This then allows them to be detected by hal, etc

Signed-off-by: Jeremy Katz <katzj@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agovirtio: Use spin_lock_irqsave/restore for virtio-pci
Anthony Liguori [Sun, 2 Mar 2008 22:37:48 +0000 (16:37 -0600)]
virtio: Use spin_lock_irqsave/restore for virtio-pci

virtio-pci acquires its spin lock in an interrupt context so it's necessary
to use spin_lock_irqsave/restore variants.  This patch fixes guest SMP when
using virtio devices in KVM.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
16 years agodrm/ati_pcigart: fix the PCIGART to use drm_pci to allocate GART table.
Dave Airlie [Mon, 17 Mar 2008 00:24:24 +0000 (10:24 +1000)]
drm/ati_pcigart: fix the PCIGART to use drm_pci to allocate GART table.

This fixes a problem on 64-bit with 4GB with ATI RS690 chipsets. It
makes sure the pcigart table is allocated in coherent memory for DMA operations.

Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agodrm/radeon: fixup RV550 chip family
Alex Deucher [Mon, 17 Mar 2008 00:22:12 +0000 (10:22 +1000)]
drm/radeon: fixup RV550 chip family

This fixes up the RV550 chips which are based on RV515, not RV530.
It also adds another RS690 PCI ID.

Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agodrm/via: attempt again to stabilise the AGP DMA command submission.
Thomas Hellstrom [Mon, 17 Mar 2008 00:07:20 +0000 (10:07 +1000)]
drm/via: attempt again to stabilise the AGP DMA command submission.

It's worth remembering that all new bright ideas on how to make this command reader work properly and according to docs will probably fail :( Bring in some old code.

Also allow a larger SG-DMA download stride, and remove unnecessary waits for
command regulators pauses.

Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agodrm: Fix race that can lockup the kernel
Mike Isely [Thu, 13 Mar 2008 20:30:35 +0000 (15:30 -0500)]
drm: Fix race that can lockup the kernel

The i915_vblank_swap() function schedules an automatic buffer swap
upon receipt of the vertical sync interrupt.  Such an operation is
lengthy so it can't be allowed to happen in normal interrupt context,
thus the DRM implements this by scheduling the work in a kernel
softirq-scheduled tasklet.  In order for the buffer swap to work
safely, the DRM's central lock must be taken, via a call to
drm_lock_take() located in drivers/char/drm/drm_irq.c within the
function drm_locked_tasklet_func().  The lock-taking logic uses a
non-interrupt-blocking spinlock to implement the manipulations needed
to take the lock.  This semantic would be safe if all attempts to use
the spinlock only happen from process context.  However this buffer
swap happens from softirq context which is really a form of interrupt
context.  Thus we have an unsafe situation, in that
drm_locked_tasklet_func() can block on a spinlock already taken by a
thread in process context which will never get scheduled again because
of the blocked softirq tasklet.  This wedges the kernel hard.

To trigger this bug, run a dual-head cloned mode configuration which
uses the i915 drm, then execute an opengl application which
synchronizes buffer swaps against the vertical sync interrupt.  In my
testing, a lockup always results after running anywhere from 5 minutes
to an hour and a half.  I believe dual-head is needed to really
trigger the problem because then the vertical sync interrupt handling
is no longer predictable (due to being interrupt-sourced from two
different heads running at different speeds).  This raises the
probability of the tasklet trying to run while the userspace DRI is
doing things to the GPU (and manipulating the DRM lock).

The fix is to change the relevant spinlock semantics to be the
interrupt-blocking form.  After this change I am no longer able to
trigger the lockup; the longest test run so far was 20 hours (test
stopped after that point).

Note: I have examined the places where this spinlock is being
employed; all are reasonably short bounded sequences and should be
suitable for interrupts being blocked without impacting overall kernel
interrupt response latency.

Signed-off-by: Mike Isely <isely@pobox.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
16 years agoLinux 2.6.25-rc6
Linus Torvalds [Sun, 16 Mar 2008 23:32:14 +0000 (16:32 -0700)]
Linux 2.6.25-rc6

16 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6
Linus Torvalds [Sun, 16 Mar 2008 17:48:23 +0000 (10:48 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/kyle/parisc-2.6

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6:
  [PARISC] make ptr_to_pide() static
  [PARISC] head.S: section mismatch fixes
  [PARISC] add back Crestone Peak cpu
  [PARISC] futex: special case cmpxchg NULL in kernel space
  [PARISC] clean up show_stack
  [PARISC] add pa8900 CPUs to hardware inventory
  [PARISC] clean up include/asm-parisc/elf.h
  [PARISC] move defconfig to arch/parisc/configs/
  [PARISC] add back AD1889 MAINTAINERS entry
  [PARISC] pdc_console: fix bizarre panic on boot
  [PARISC] dump_stack in show_regs
  [PARISC] pdc_stable: fix compile errors
  [PARISC] remove unused pdc_iodc_printf function
  [PARISC] bump __NR_syscalls
  [PARISC] unbreak pgalloc.h
  [PARISC] move VMALLOC_* definitions to fixmap.h
  [PARISC] wire up timerfd syscalls
  [PARISC] remove old timerfd syscall

16 years ago[PARISC] make ptr_to_pide() static
FUJITA Tomonori [Mon, 10 Mar 2008 11:43:24 +0000 (20:43 +0900)]
[PARISC] make ptr_to_pide() static

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] head.S: section mismatch fixes
Helge Deller [Wed, 26 Dec 2007 17:07:01 +0000 (18:07 +0100)]
[PARISC] head.S: section mismatch fixes

- move boot_args[] into the init section
- move $global$ into the read_mostly section
- fix the following two section mismatches:
WARNING: vmlinux.o(.text+0x9c): Section mismatch: reference to .init.text:start_kernel (between '$pgt_fill_loop' and '$is_pa20')
WARNING: vmlinux.o(.text+0xa0): Section mismatch: reference to .init.text:start_kernel (between '$pgt_fill_loop' and '$is_pa20')

Signed-off-by: Helge Deller <deller@gmx.de>
SIgned-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] add back Crestone Peak cpu
Kyle McMartin [Sat, 1 Mar 2008 19:40:43 +0000 (11:40 -0800)]
[PARISC] add back Crestone Peak cpu

Crestone Peak Slow is the 800MHz PA-8800 cpu in the C8000.
0x88B is probably the Crestone Peak Fast.

Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] futex: special case cmpxchg NULL in kernel space
Kyle McMartin [Sat, 1 Mar 2008 18:25:52 +0000 (10:25 -0800)]
[PARISC] futex: special case cmpxchg NULL in kernel space

Commit a0c1e9073ef7428a14309cba010633a6cd6719ea added code to futex.c
to detect whether futex_atomic_cmpxchg_inatomic was implemented at run
time:

+       curval = cmpxchg_futex_value_locked(NULL, 0, 0);
+       if (curval == -EFAULT)
+               futex_cmpxchg_enabled = 1;

This is bogus on parisc, since page zero in kernel virtual space is the
gateway page for syscall entry, and should not be read from the kernel.
(That, and we really don't like the kernel faulting on its own address
 space...)

Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] clean up show_stack
Kyle McMartin [Sat, 1 Mar 2008 18:30:19 +0000 (10:30 -0800)]
[PARISC] clean up show_stack

When we show_regs, we obviously have a struct pt_regs of the calling
frame. Use these in show_stack so we don't have the entire bogus call trace
up to the show_stack call.

Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] add pa8900 CPUs to hardware inventory
James Bottomley [Wed, 20 Feb 2008 21:53:02 +0000 (15:53 -0600)]
[PARISC] add pa8900 CPUs to hardware inventory

This patch adds the known pa8900 CPUs to the inventory list and removes
the Crestone Peak one which apparently never escaped into the wild.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] clean up include/asm-parisc/elf.h
Randolph Chung [Sun, 24 Feb 2008 18:44:21 +0000 (10:44 -0800)]
[PARISC] clean up include/asm-parisc/elf.h

Cleanup some cruft. No functionality changes.

Signed-off-by: Randolph Chung <tausq@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] move defconfig to arch/parisc/configs/
Adrian Bunk [Tue, 26 Feb 2008 19:55:17 +0000 (21:55 +0200)]
[PARISC] move defconfig to arch/parisc/configs/

This patch moves the default parisc defconfig to
arch/parisc/configs/generic_defconfig where it belongs and selects it as
the default defconfig through KBUILD_DEFCONFIG.

Signed-off-by: Adrian Bunk <adrian.bunk@movial.fi>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] add back AD1889 MAINTAINERS entry
Thibaut VARENE [Wed, 20 Feb 2008 20:05:56 +0000 (21:05 +0100)]
[PARISC] add back AD1889 MAINTAINERS entry

Signed-off-by: Thibaut VARENE <T-Bone@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
16 years ago[PARISC] pdc_console: fix bizarre panic on boot
Kyle McMartin [Tue, 19 Feb 2008 07:34:34 +0000 (23:34 -0800)]
[PARISC] pdc_console: fix bizarre panic on boot

Commit 721fdf34167580ff98263c74cead8871d76936e6 introduced a subtle bug
by accidently removing the "static" from iodc_dbuf. This resulted in, what
appeared to be, a trap without *current set to a task. Probably the result of
a trap in real mode while calling firmware.

Also do other misc clean ups. Since the only input from firmware is non
blocking, share iodc_dbuf between input and output, and spinlock the
only callers.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] dump_stack in show_regs
Kyle McMartin [Tue, 19 Feb 2008 07:26:46 +0000 (23:26 -0800)]
[PARISC] dump_stack in show_regs

Originally, show_stack was used in BUG() output. However, a recent commit
changed it to print register state (no idea what that's supposed to help,
really...) and parisc was missing a backtrace because of it.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] pdc_stable: fix compile errors
Joel Soete [Tue, 19 Feb 2008 02:26:11 +0000 (18:26 -0800)]
[PARISC] pdc_stable: fix compile errors

Signed-off-by: Joel Soete <rubisher@scarlet.be>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] remove unused pdc_iodc_printf function
Kyle McMartin [Mon, 18 Feb 2008 22:26:41 +0000 (14:26 -0800)]
[PARISC] remove unused pdc_iodc_printf function

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] bump __NR_syscalls
Kyle McMartin [Mon, 18 Feb 2008 22:21:17 +0000 (14:21 -0800)]
[PARISC] bump __NR_syscalls

oops, forgot this in the previous commit.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] unbreak pgalloc.h
Kyle McMartin [Mon, 18 Feb 2008 22:16:26 +0000 (14:16 -0800)]
[PARISC] unbreak pgalloc.h

Commit 2f569afd9ced9ebec9a6eb3dbf6f83429be0a7b4 broke the compile
rather spectacularly. Fix code errors.

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] move VMALLOC_* definitions to fixmap.h
Kyle McMartin [Mon, 18 Feb 2008 22:13:43 +0000 (14:13 -0800)]
[PARISC] move VMALLOC_* definitions to fixmap.h

They make way more sense here, really...

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] wire up timerfd syscalls
Kyle McMartin [Mon, 18 Feb 2008 22:00:18 +0000 (14:00 -0800)]
[PARISC] wire up timerfd syscalls

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years ago[PARISC] remove old timerfd syscall
Kyle McMartin [Mon, 18 Feb 2008 21:57:26 +0000 (13:57 -0800)]
[PARISC] remove old timerfd syscall

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
16 years agoACPI: Remove ACPI_CUSTOM_DSDT_INITRD option
Linus Torvalds [Sat, 15 Mar 2008 18:53:32 +0000 (11:53 -0700)]
ACPI: Remove ACPI_CUSTOM_DSDT_INITRD option

This essentially reverts commit 71fc47a9adf8ee89e5c96a47222915c5485ac437
("ACPI: basic initramfs DSDT override support"), because the code simply
isn't ready.

It did ugly things to the init sequence to populate the rootfs image
early, but that just ended up showing other problems with the whole
approach.  The fact is, the VFS layer simply isn't initialized this
early, and the relevant ACPI code should either run much later, or this
shouldn't be done at all.

For 2.6.25, we'll just pick the latter option.  We can revisit this
concept later if necessary.

Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Tilman Schmidt <tilman@imap.cc>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Thomas Renninger <trenn@suse.de>
Cc: Eric Piel <eric.piel@tremplin-utc.net>
Cc: Len Brown <len.brown@intel.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Markus Gaugusch <dsdt@gaugusch.at>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotifm_sd: DATA_CARRY is not boolean in tifm_sd_transfer_data()
Roel Kluin [Sat, 15 Mar 2008 15:00:38 +0000 (16:00 +0100)]
tifm_sd: DATA_CARRY is not boolean in tifm_sd_transfer_data()

DATA_CARRY is not boolean

Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Sat, 15 Mar 2008 16:21:04 +0000 (09:21 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  [NET]: Fix tbench regression in 2.6.25-rc1

16 years agosched: simplify sched_slice()
Ingo Molnar [Fri, 14 Mar 2008 21:17:08 +0000 (22:17 +0100)]
sched: simplify sched_slice()

Use the existing calc_delta_mine() calculation for sched_slice(). This
saves a divide and simplifies the code because we share it with the
other /cfs_rq->load users.

It also improves code size:

      text    data     bss     dec     hex filename
     42659    2740     144   45543    b1e7 sched.o.before
     42093    2740     144   44977    afb1 sched.o.after

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agosched: fix fair sleepers
Ingo Molnar [Fri, 14 Mar 2008 21:16:08 +0000 (22:16 +0100)]
sched: fix fair sleepers

Fair sleepers need to scale their latency target down by runqueue
weight. Otherwise busy systems will gain ever larger sleep bonus.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agosched: fix overload performance: buddy wakeups
Peter Zijlstra [Fri, 14 Mar 2008 20:12:12 +0000 (21:12 +0100)]
sched: fix overload performance: buddy wakeups

Currently we schedule to the leftmost task in the runqueue. When the
runtimes are very short because of some server/client ping-pong,
especially in over-saturated workloads, this will cycle through all
tasks trashing the cache.

Reduce cache trashing by keeping dependent tasks together by running
newly woken tasks first. However, by not running the leftmost task first
we could starve tasks because the wakee can gain unlimited runtime.

Therefore we only run the wakee if its within a small
(wakeup_granularity) window of the leftmost task. This preserves
fairness, but does alternate server/client task groups.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: fix calc_delta_mine()
Ingo Molnar [Fri, 14 Mar 2008 21:20:01 +0000 (22:20 +0100)]
sched: fix calc_delta_mine()

lw->weight can be 0 for a short time during bootup.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agosched: fix update_load_add()/sub()
Ingo Molnar [Fri, 14 Mar 2008 22:48:28 +0000 (23:48 +0100)]
sched: fix update_load_add()/sub()

Clear the cached inverse value when updating load. This is needed for
calc_delta_mine() to work correctly when using the rq load.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
16 years agosched: min_vruntime fix
Peter Zijlstra [Fri, 14 Mar 2008 19:55:51 +0000 (20:55 +0100)]
sched: min_vruntime fix

Current min_vruntime tracking is incorrect and will cause serious
problems when we don't run the leftmost task for some reason.

min_vruntime does two things; 1) it's used to determine a forward
direction when the u64 vruntime wraps, 2) it's used to track the
leftmost vruntime to position newly enqueued tasks from.

The current logic advances min_vruntime whenever the current task's
vruntime advance. Because the current task may pass the leftmost task
still waiting we're failing the second goal. This causes new tasks to be
placed too far ahead and thus penalizes their runtime.

Fix this by making min_vruntime the min_vruntime of the waiting tasks by
tracking it in enqueue/dequeue, and compare against current's vruntime
to obtain the absolute minimum when placing new tasks.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agosched: fix race in schedule()
Hiroshi Shimamoto [Mon, 10 Mar 2008 18:01:20 +0000 (11:01 -0700)]
sched: fix race in schedule()

Fix a hard to trigger crash seen in the -rt kernel that also affects
the vanilla scheduler.

There is a race condition between schedule() and some dequeue/enqueue
functions; rt_mutex_setprio(), __setscheduler() and sched_move_task().

When scheduling to idle, idle_balance() is called to pull tasks from
other busy processor. It might drop the rq lock. It means that those 3
functions encounter on_rq=0 and running=1. The current task should be
put when running.

Here is a possible scenario:

   CPU0                               CPU1
    |                              schedule()
    |                              ->deactivate_task()
    |                              ->idle_balance()
    |                              -->load_balance_newidle()
rt_mutex_setprio()                     |
    |                              --->double_lock_balance()
    *get lock                          *rel lock
    * on_rq=0, ruuning=1               |
    * sched_class is changed           |
    *rel lock                          *get lock
    :                                  |
                                       :
                                   ->put_prev_task_rt()
                                   ->pick_next_task_fair()
                                       => panic

The current process of CPU1(P1) is scheduling. Deactivated P1, and the
scheduler looks for another process on other CPU's runqueue because CPU1
will be idle. idle_balance(), load_balance_newidle() and
double_lock_balance() are called and double_lock_balance() could drop
the rq lock. On the other hand, CPU0 is trying to boost the priority of
P1. The result of boosting only P1's prio and sched_class are changed to
RT. The sched entities of P1 and P1's group are never put. It makes
cfs_rq invalid, because the cfs_rq has curr and no leaf, but
pick_next_task_fair() is called, then the kernel panics.

Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394...
Linus Torvalds [Fri, 14 Mar 2008 23:49:41 +0000 (16:49 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ieee1394/linux1394-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
  firewire: fw-ohci: shut up false compiler warning on PPC32
  firewire: fw-ohci: use dma_alloc_coherent for ar_buffer
  ieee1394: sbp2: fix for SYM13FW500 bridge (Datafab disk)
  firewire: fw-sbp2: fix for SYM13FW500 bridge (Datafab disk)
  firewire: update Kconfig help text
  firewire: warn on fatal condition in topology code
  firewire: fw-sbp2: set single-phase retry_limit
  firewire: fw-ohci: Apple UniNorth 1st generation support
  firewire: fw-ohci: PPC PMac platform code
  firewire: endianess annotations
  firewire: endianess fix

16 years agonfsd: fix oops on access from high-numbered ports
J. Bruce Fields [Fri, 14 Mar 2008 23:37:11 +0000 (19:37 -0400)]
nfsd: fix oops on access from high-numbered ports

This bug was always here, but before my commit 6fa02839bf9412e18e77
("recheck for secure ports in fh_verify"), it could only be triggered by
failure of a kmalloc().  After that commit it could be triggered by a
client making a request from a non-reserved port for access to an export
marked "secure".  (Exports are "secure" by default.)

The result is a struct svc_export with a reference count one too low,
resulting in likely oopses next time the export is accessed.

The reference counting here is not straightforward; a later patch will
clean up fh_verify().

Thanks to Lukas Hejtmanek for the bug report and followup.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Cc: Lukas Hejtmanek <xhejtman@ics.muni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agostruct export_operations: adjust comments to match current members
Marc Dionne [Fri, 14 Mar 2008 13:11:29 +0000 (13:11 +0000)]
struct export_operations: adjust comments to match current members

The comments in the definition of struct export_operations don't match the
current members.

Add a comment for the 2 new functions and remove 2 comments for unused ones.

Signed-off-by: Marc Dionne <marc.c.dionne@gmail.com>
Acked-by: David Howells <dhowells@redhat.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agofirewire: fw-ohci: shut up false compiler warning on PPC32
Stefan Richter [Thu, 13 Mar 2008 23:27:49 +0000 (00:27 +0100)]
firewire: fw-ohci: shut up false compiler warning on PPC32

Shut up "may be used uninitialised in this function" warnings due to
PPC32's implementation of dma_alloc_coherent().

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agofirewire: fw-ohci: use dma_alloc_coherent for ar_buffer
Jarod Wilson [Wed, 12 Mar 2008 21:43:26 +0000 (17:43 -0400)]
firewire: fw-ohci: use dma_alloc_coherent for ar_buffer

Currently, we do nothing to guarantee we have a consistent DMA buffer for
asynchronous receive packets. Rather than doing several sync's following a
dma_map_single() to get consistent buffers, just switch to using
dma_alloc_coherent().

Resolves constant buffer failures on my own x86_64 laptop w/4GB of RAM and
likely to fix a number of other failures witnessed on x86_64 systems with
4GB of RAM or more.

Signed-off-by: Jarod Wilson <jwilson@redhat.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agoieee1394: sbp2: fix for SYM13FW500 bridge (Datafab disk)
Stefan Richter [Tue, 11 Mar 2008 21:32:52 +0000 (22:32 +0100)]
ieee1394: sbp2: fix for SYM13FW500 bridge (Datafab disk)

Fix I/O errors due to SYM13FW500's inability to handle larger request
sizes.  Reported by Piergiorgio Sartor <piergiorgio.sartor@nexgo.de> for
firewire-sbp2 in https://bugzilla.redhat.com/show_bug.cgi?id=436879

This fix is necessary because sbp2's default request size limit has been
lifted since 2.6.25-rc1.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Jarod Wilson <jwilson@redhat.com>
16 years agofirewire: fw-sbp2: fix for SYM13FW500 bridge (Datafab disk)
Stefan Richter [Tue, 11 Mar 2008 21:32:03 +0000 (22:32 +0100)]
firewire: fw-sbp2: fix for SYM13FW500 bridge (Datafab disk)

Fix I/O errors due to SYM13FW500's inability to handle larger request
sizes.  Reported by Piergiorgio Sartor <piergiorgio.sartor@nexgo.de> in
https://bugzilla.redhat.com/show_bug.cgi?id=436879

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Jarod Wilson <jwilson@redhat.com>
16 years agofirewire: update Kconfig help text
Stefan Richter [Sat, 8 Mar 2008 23:27:20 +0000 (00:27 +0100)]
firewire: update Kconfig help text

Remove some less necessary information, point out that video1394 and
dv1394 should be blacklisted along with ohci1394.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agofirewire: warn on fatal condition in topology code
Stefan Richter [Sat, 8 Mar 2008 21:38:16 +0000 (22:38 +0100)]
firewire: warn on fatal condition in topology code

If this ever happens to anybody, we want to have it in his log.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agofirewire: fw-sbp2: set single-phase retry_limit
Jarod Wilson [Fri, 7 Mar 2008 06:43:01 +0000 (01:43 -0500)]
firewire: fw-sbp2: set single-phase retry_limit

Per the SBP-2 specification, all SBP-2 target devices must have a BUSY_TIMEOUT
register. Per the 1394-1995 specification, the retry_limt portion of the
register should be set to 0x0 initially, and set on the target by a logged in
initiator (i.e., a Linux host w/firewire controller(s)).

Well, as it turns out, lots of devices these days have actually moved on to
starting to implement SBP-3 compliance, which says that retry_limit should
default to 0xf instead (yes, SBP-3 stomps directly on 1394-1995, oops).

Prior to this change, the firewire driver stack didn't touch retry_limit, and
any SBP-3 compliant device worked fine, while SBP-2 compliant ones were unable
to retransmit when the host returned an ack_busy_X, which resulted in stalled
out I/O, eventually causing the SCSI layer to give up and offline the device.

The simple fix is for us to set retry_limit to 0xf in the register for all
devices (which actually matches what the old ieee1394 stack did).

Prior to this change, a hard disk behind an SBP-2 Prolific PL-3507 bridge chip
would routinely encounter buffer I/O errors and wind up offlined by the SCSI
layer. With this change, I've encountered zero I/O failures moving tens of GB
of data around.

Signed-off-by: Jarod Wilson <jwilson@redhat.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agofirewire: fw-ohci: Apple UniNorth 1st generation support
Stefan Richter [Sat, 1 Mar 2008 01:47:15 +0000 (02:47 +0100)]
firewire: fw-ohci: Apple UniNorth 1st generation support

Mostly copied from ohci1394.c.  Necessary for some older Macs, e.g.
PowerBook G3 Pismo and early PowerBook G4 Titanium.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
16 years agofirewire: fw-ohci: PPC PMac platform code
Stefan Richter [Sat, 1 Mar 2008 01:42:56 +0000 (02:42 +0100)]
firewire: fw-ohci: PPC PMac platform code

Copied from ohci1394.c.  This code is necessary to prevent machine check
exceptions when reloading or resuming the driver.

Tested on a 1st generation PowerBook G4 Titanium, which also needs the
pci_probe() hunk.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
I was able to reproduce the system exception on resume with a 3rd-gen
Titanium PowerBook G4 667, and this patch does let the system resume
successfully now.

Not quite clear if there was possibly an updated version coming using
pci_enable_device() instead of the pair of pmac_call_feature() calls,
but either way, this is a definite must-have, at least for older ppc
macs -- my Aluminum PowerBook G4/1.67 suspends and resumes without this
patch just fine.

Signed-off-by: Jarod Wilson <jwilson@redhat.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
16 years agofirewire: endianess annotations
Stefan Richter [Sat, 23 Feb 2008 11:24:57 +0000 (12:24 +0100)]
firewire: endianess annotations

Kills warnings from 'make C=1 CHECKFLAGS="-D__CHECK_ENDIAN__" modules':

drivers/firewire/fw-transaction.c:771:10: warning: incorrect type in assignment (different base types)
drivers/firewire/fw-transaction.c:771:10:    expected unsigned int [unsigned] [usertype] <noident>
drivers/firewire/fw-transaction.c:771:10:    got restricted unsigned int [usertype] <noident>
drivers/firewire/fw-transaction.h:93:10: warning: incorrect type in assignment (different base types)
drivers/firewire/fw-transaction.h:93:10:    expected unsigned int [unsigned] [usertype] <noident>
drivers/firewire/fw-transaction.h:93:10:    got restricted unsigned int [usertype] <noident>
drivers/firewire/fw-ohci.c:1490:8: warning: restricted degrades to integer
drivers/firewire/fw-ohci.c:1490:35: warning: restricted degrades to integer
drivers/firewire/fw-ohci.c:1516:5: warning: cast to restricted type

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Jarod Wilson <jwilson@redhat.com>
16 years agofirewire: endianess fix
Stefan Richter [Sat, 23 Feb 2008 11:24:17 +0000 (12:24 +0100)]
firewire: endianess fix

The generation of incoming requests was filled in in wrong byte order on
machines with big endian CPU.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Jarod Wilson <jwilson@redhat.com>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Thu, 13 Mar 2008 20:25:32 +0000 (13:25 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IPoIB: Allocate priv->tx_ring with vmalloc()
  IPoIB/cm: Set tx_wr.num_sge in connected mode post_send()
  IPoIB: Don't drop multicast sends when they can be queued
  IB/ipath: Reset the retry counter for RDMA_READ_RESPONSE_MIDDLE packets
  IB/ipath: Fix error completion put on send CQ instead of recv CQ
  IB/ipath: Fix RC QP initialization
  IB/ipath: Fix potentially wrong RNR retry counter returned in ipath_query_qp()
  IB/ipath: Fix IB compliance problems with link state vs physical state

16 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx
Linus Torvalds [Thu, 13 Mar 2008 20:17:25 +0000 (13:17 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/djbw/async_tx

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
  async_tx: checkpatch says s/__FUNCTION__/__func__/g
  iop-adma.c: replace remaining __FUNCTION__ occurrences
  fsldma: Add a completed cookie updated action in DMA finish interrupt.
  fsldma: Add device_prep_dma_interrupt support to fsldma.c
  dmaengine: Fix a bug about BUG_ON() on DMA engine capability DMA_INTERRUPT.
  fsldma: Fix fsldma.c warning messages when it's compiled under PPC64.

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/pci-2.6
Linus Torvalds [Thu, 13 Mar 2008 20:16:22 +0000 (13:16 -0700)]
Merge git://git./linux/kernel/git/gregkh/pci-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/pci-2.6:
  PCI: fix issue with busses registering multiple times in sysfs

16 years agoavoid endless loops in lib/swiotlb.c
Jan Beulich [Thu, 13 Mar 2008 09:13:30 +0000 (09:13 +0000)]
avoid endless loops in lib/swiotlb.c

Commit 681cc5cd3efbeafca6386114070e0bfb5012e249 ("iommu sg merging:
swiotlb: respect the segment boundary limits") introduced two
possibilities for entering an endless loop in lib/swiotlb.c:

 - if max_slots is zero (possible if mask is ~0UL)
 - if the number of slots requested fits into a swiotlb segment, but is
   too large for the part of a segment which remains after considering
   offset_slots

This fixes them

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Thu, 13 Mar 2008 20:13:47 +0000 (13:13 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6: (32 commits)
  ACPI: thermal: show temperature in millidegree Celsius
  thermal: fix generic thermal I/F for hwmon
  acer-wmi: build depends on i8042
  documentation:  Move power-related files to Documentation/power/
  ACPI: buffer array too short in drivers/acpi/system.c
  acer-wmi: Add DMI quirk for mail LED support on Acer Aspire 3610/ 5610
  acer-wmi: Fix DSDT path in documentation
  acer-wmi: Make device detection error messages more descriptive
  laptops: move laptop-mode.txt to Documentation/laptops/
  ACPICA: Warn if packages with invalid references are evaluated
  ACPI: add _PRT quirks to work around broken firmware
  Hibernation: Fix mark_nosave_pages()
  ACPI: Ignore _BQC object when registering backlight device
  ACPI: WMI: Clean up handling of spec violating data blocks
  acer-wmi: Don't warn if mail LED cannot be detected
  acer-wmi: Rename mail LED correctly & remove hardcoded colour
  ACPI: use ACPI_DEBUG_PRINT instead of printk in acpi_processor_hotplug_notify()
  ACPI: button: make real parent for input devices in device tree
  toshiba_acpi: Enable autoloading
  ACPI: EC: Handle IRQ storm on Acer laptops
  ...

16 years agoblock: floppy: fix rmmod lockup
Jiri Slaby [Thu, 13 Mar 2008 19:32:43 +0000 (12:32 -0700)]
block: floppy: fix rmmod lockup

Floppy rmmod locks up when no such hardware was initialized, since there is
nobody to wake the remove code up.  Remove the completion, because release is
called during platform_unregister anyway.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agokernel-doc: set verbose mode via environment
Randy Dunlap [Thu, 13 Mar 2008 19:32:43 +0000 (12:32 -0700)]
kernel-doc: set verbose mode via environment

Honor the environment variable "KBUILD_VERBOSE=1" (as set by make V=1) to
enable verbose mode in scripts/kernel-doc.  Useful for getting more info and
warnings from kernel-doc.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoalpha: use iommu_is_span_boundary helper function
FUJITA Tomonori [Thu, 13 Mar 2008 19:32:40 +0000 (12:32 -0700)]
alpha: use iommu_is_span_boundary helper function

iommu_is_span_boundary in lib/iommu-helper.c was exported for PARISC IOMMUs
(commit 3715863aa142c4f4c5208f5f3e5e9bac06006d2f).  Alpha's IOMMU can use it.

This removes the check on the boundary size alignment because
iommu_is_span_boundary does.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Acked-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agochar: riscom, fix rc_board indexing
Jiri Slaby [Thu, 13 Mar 2008 19:32:39 +0000 (12:32 -0700)]
char: riscom, fix rc_board indexing

In riscom8_init_module, rc_board should be indexed by i, not by 0, otherwise
the loop is useless.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi_bitbang: short transfer status fix
Jan Nikitenko [Thu, 13 Mar 2008 19:32:39 +0000 (12:32 -0700)]
spi_bitbang: short transfer status fix

SPI controller drivers return number of bytes actually transfered from
bitbang->txrx_bufs() method.  This updates handling of short transfers (where
the transfer size is less than requested):

 - Even zero byte short transfers should report errors;
 - Include short transfers in the total of transferred bytes;
 - Use EREMOTEIO (like USB) not EMSGSIZE to report short transfers

Short transfers don't normally mean invalid message sizes, but if the
underlying controller driver needs to use EMSGSIZE it can still do so.

[db: fix two more minor issues]
Signed-off-by: Jan Nikitenko <jan.nikitenko@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoau1550_spi: fix prototype of irq handler
Jan Nikitenko [Thu, 13 Mar 2008 19:32:38 +0000 (12:32 -0700)]
au1550_spi: fix prototype of irq handler

Remove struct pt_regs *regs from prototype of au1550_spi_irq handler to avoid
warning in request_irq().  Also fix comment type leadin.

Signed-off-by: Jan Nikitenko <jan.nikitenko@gmail.com>
Signed-off-by: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoh8300: fix recent uaccess breakage
Yoshinori Sato [Thu, 13 Mar 2008 19:32:37 +0000 (12:32 -0700)]
h8300: fix recent uaccess breakage

Al Viro wrote:
>
>  After that commit in asm-h8300/uaccess.h we have
>
> #define get_user(x, ptr)                                        \
> ({                                                              \
>     int __gu_err = 0;                                           \
>     uint32_t __gu_val = 0;                              \
>     ^^^^^^^^^^^^^^^^^
>     switch (sizeof(*(ptr))) {                                   \
>     case 1:                                                     \
>     case 2:                                                     \
>     case 4:                                                     \
>         __gu_val = *(ptr);                                      \
>         break;                                                  \
>     case 8:                                                     \
>         memcpy(&__gu_val, ptr, sizeof (*(ptr)));                \
>                                ^^^^^^^^^^^^^^^^
>
> which, of course, is FUBAR whenever we actually hit that case - memcpy of
> 8 bytes into uint32_t is obviously wrong.  Why don't we simply do

Cc: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopagemap: proper read error handling
Marcelo Tosatti [Thu, 13 Mar 2008 19:32:35 +0000 (12:32 -0700)]
pagemap: proper read error handling

Fix pagemap_read() error handling by releasing acquired resources and checking
for get_user_pages() partial failure.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agosmackfs: do not trust `count' in inodes write()s
Ahmed S. Darwish [Thu, 13 Mar 2008 19:32:34 +0000 (12:32 -0700)]
smackfs: do not trust `count' in inodes write()s

Smackfs write() implementation does not put a higher bound on the number of
bytes to copy from user-space.  This may lead to a DOS attack if a malicious
`count' field is given.

Assure that given `count' is exactly the length needed for a /smack/load rule.
 In case of /smack/cipso where the length is relative, assure that `count'
does not exceed the size needed for a buffer representing maximum possible
number of CIPSO 2.2 categories.

Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
Acked-by: Casey Schaufler <casey@schaufler-ca.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoasync_tx: checkpatch says s/__FUNCTION__/__func__/g
Dan Williams [Fri, 14 Mar 2008 00:45:28 +0000 (17:45 -0700)]
async_tx: checkpatch says s/__FUNCTION__/__func__/g

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agoiop-adma.c: replace remaining __FUNCTION__ occurrences
Harvey Harrison [Fri, 14 Mar 2008 00:45:28 +0000 (17:45 -0700)]
iop-adma.c: replace remaining __FUNCTION__ occurrences

__FUNCTION__ is gcc-specific, use __func__

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agofsldma: Add a completed cookie updated action in DMA finish interrupt.
Zhang Wei [Fri, 14 Mar 2008 00:45:28 +0000 (17:45 -0700)]
fsldma: Add a completed cookie updated action in DMA finish interrupt.

The patch 'fsldma: do not cleanup descriptors in hardirq context'
(commit 222ccf9ab838a1ca7163969fabd2cddc10403fb5) removed descriptors
cleanup function to tasklet but the completed cookie do not updated.
Thus, the DMA controller will get lots of duplicated transfer
interrupts. Just make a completed cookie update in interrupt handler.
And keep other cleanup jobs in tasklet function.

Tested-by: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agofsldma: Add device_prep_dma_interrupt support to fsldma.c
Zhang Wei [Fri, 14 Mar 2008 00:45:28 +0000 (17:45 -0700)]
fsldma: Add device_prep_dma_interrupt support to fsldma.c

This is a bug that I assigned DMA_INTERRUPT capability to fsldma
but missing device_prep_dma_interrupt function. For a bug in
dmaengine.c the driver passed BUG_ON() checking. The patch fixes it.

Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agodmaengine: Fix a bug about BUG_ON() on DMA engine capability DMA_INTERRUPT.
Zhang Wei [Fri, 14 Mar 2008 00:45:28 +0000 (17:45 -0700)]
dmaengine: Fix a bug about BUG_ON() on DMA engine capability DMA_INTERRUPT.

The device->device_prep_dma_interrupt function is used by
DMA_INTERRUPT capability, not DMA_ZERO_SUM.

Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agofsldma: Fix fsldma.c warning messages when it's compiled under PPC64.
Zhang Wei [Thu, 13 Mar 2008 17:45:27 +0000 (10:45 -0700)]
fsldma: Fix fsldma.c warning messages when it's compiled under PPC64.

There are warning messages reported by Stephen Rothwell with
ARCH=powerpc allmodconfig build:

drivers/dma/fsldma.c: In function 'fsl_dma_prep_memcpy':
drivers/dma/fsldma.c:439: warning: comparison of distinct pointer types
lacks a cast
drivers/dma/fsldma.c: In function 'fsl_chan_xfer_ld_queue':
drivers/dma/fsldma.c:584: warning: format '%016llx' expects type 'long long
unsigned int', but argument 4 has type 'dma_addr_t'
drivers/dma/fsldma.c: In function 'fsl_dma_chan_do_interrupt':
drivers/dma/fsldma.c:668: warning: format '%x' expects type 'unsigned int',
but argument 5 has type 'dma_addr_t'
drivers/dma/fsldma.c:684: warning: format '%016llx' expects type 'long long
unsigned int', but argument 4 has type 'dma_addr_t'
drivers/dma/fsldma.c:684: warning: format '%016llx' expects type 'long long
unsigned int', but argument 5 has type 'dma_addr_t'
drivers/dma/fsldma.c:701: warning: format '%02x' expects type 'unsigned
int', but argument 4 has type 'dma_addr_t'
drivers/dma/fsldma.c: In function 'fsl_dma_self_test':
drivers/dma/fsldma.c:840: warning: format '%d' expects type 'int', but
argument 5 has type 'size_t'
drivers/dma/fsldma.c: In function 'of_fsl_dma_probe':
drivers/dma/fsldma.c:1010: warning: format '%08x' expects type 'unsigned
int', but argument 5 has type 'resource_size_t'

This patch fixed the above warning messages.

Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
16 years agoPCI: fix issue with busses registering multiple times in sysfs
Greg Kroah-Hartman [Thu, 13 Mar 2008 04:48:03 +0000 (21:48 -0700)]
PCI: fix issue with busses registering multiple times in sysfs

PCI busses can be registered multiple times, so we need to detect if we
have registered our bus structure in sysfs already.  If so, don't do it
again.

Thanks to Guennadi Liakhovetski <g.liakhovetski@gmx.de> for reporting
the problem, and to Linus for poking me to get me to believe that it was
a real problem.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
16 years agoMerge branches 'release' and 'doc' into release
Len Brown [Thu, 13 Mar 2008 05:59:53 +0000 (01:59 -0400)]
Merge branches 'release' and 'doc' into release

16 years agoMerge branches 'release', 'ejd', 'sony' and 'wmi' into release
Len Brown [Thu, 13 Mar 2008 05:59:52 +0000 (01:59 -0400)]
Merge branches 'release', 'ejd', 'sony' and 'wmi' into release

16 years agoMerge branches 'release', 'button-sysfs', 'misc', 'mismatch', 'randconfig' and 'toshi...
Len Brown [Thu, 13 Mar 2008 05:59:49 +0000 (01:59 -0400)]
Merge branches 'release', 'button-sysfs', 'misc', 'mismatch', 'randconfig' and 'toshiba' into release

16 years agoMerge branches 'release' and 'thermal' into release
Len Brown [Thu, 13 Mar 2008 05:59:46 +0000 (01:59 -0400)]
Merge branches 'release' and 'thermal' into release

16 years agoMerge branches 'release', 'bugzilla-8570', 'bugzilla-9966', 'bugzilla-9998', 'bugzill...
Len Brown [Thu, 13 Mar 2008 05:59:45 +0000 (01:59 -0400)]
Merge branches 'release', 'bugzilla-8570', 'bugzilla-9966', 'bugzilla-9998', 'bugzilla-10100', 'bugzilla-10132', 'bugzilla-10138' and 'bugzilla-10206' into release

16 years ago[NET]: Fix tbench regression in 2.6.25-rc1
Zhang Yanmin [Thu, 13 Mar 2008 05:52:37 +0000 (22:52 -0700)]
[NET]: Fix tbench regression in 2.6.25-rc1

Comparing with kernel 2.6.24, tbench result has regression with
2.6.25-rc1.

1) On 2 quad-core processor stoakley: 4%.
2) On 4 quad-core processor tigerton: more than 30%.

bisect located below patch.

b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b is first bad commit
commit b4ce92775c2e7ff9cf79cca4e0a19c8c5fd6287b
Author: Herbert Xu <herbert@gondor.apana.org.au>
Date:   Tue Nov 13 21:33:32 2007 -0800

    [IPV6]: Move nfheader_len into rt6_info

    The dst member nfheader_len is only used by IPv6.  It's also currently
    creating a rather ugly alignment hole in struct dst.  Therefore this patch
    moves it from there into struct rt6_info.

Above patch changes the cache line alignment, especially member
__refcnt. I did a testing by adding 2 unsigned long pading before
lastuse, so the 3 members, lastuse/__refcnt/__use, are moved to next
cache line. The performance is recovered.

I created a patch to rearrange the members in struct dst_entry.

With Eric and Valdis Kletnieks's suggestion, I made finer arrangement.

1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So
   sizeof(dst_entry)=200 no matter if CONFIG_NET_CLS_ROUTE=y/n. I
   tested many patches on my 16-core tigerton by moving tclassid to
   different place. It looks like tclassid could also have impact on
   performance.  If moving tclassid before metrics, or just don't move
   tclassid, the performance isn't good. So I move it behind metrics.

2) Add comments before __refcnt.

On 16-core tigerton:

If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18%
better than the one without the patch;

If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30%
better than the one without the patch.

With 32bit 2.6.25-rc1 on 8-core stoakley, the new patch doesn't
introduce regression.

Thank Eric, Valdis, and David!

Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoACPI: thermal: show temperature in millidegree Celsius
Zhang, Rui [Wed, 27 Feb 2008 23:51:30 +0000 (07:51 +0800)]
ACPI: thermal: show temperature in millidegree Celsius

as now required by the generic thermal I/F

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Len Brown <len.brown@intel.com>
16 years agothermal: fix generic thermal I/F for hwmon
Zhang, Rui [Wed, 27 Feb 2008 00:37:50 +0000 (08:37 +0800)]
thermal: fix generic thermal I/F for hwmon

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
16 years agoacer-wmi: build depends on i8042
Jiri Kosina [Mon, 10 Mar 2008 21:29:37 +0000 (22:29 +0100)]
acer-wmi: build depends on i8042

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Carlos Corbacho <carlos@strangeworlds.co.uk>
Signed-off-by: Len Brown <len.brown@intel.com>
16 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 13 Mar 2008 00:56:37 +0000 (17:56 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm: (26 commits)
  [ARM] 4856/1: Orion: initialise the sixth PCIe MBUS mapping window as well
  [ARM] 4855/1: Orion: use correct ethernet unit address range
  [ARM] 4853/1: include uImage target in make help
  [ARM] 4851/1: ns9xxx: fix size of gpiores
  [ARM] AT91: correct at91sam9263ek LCD power gpio pin
  [ARM] replace remaining __FUNCTION__ occurrences
  [ARM] 4850/1: include generic pgtable.h for !CONFIG_MMU case
  [ARM] 4849/1: move ATAGS asm definitions
  [ARM] 4848/1: at91: remove false lockdep warnings
  [ARM] 4847/1: kprobes: fix compilation with CONFIG_DEBUG_FS=y
  [ARM] include/asm-arm - use angle brackets for includes
  [ARM] 4845/1: Orion: Ignore memory tags with invalid data
  ARM: OMAP2: Register the L4 io bus to boot OMAP2
  ARM: OMAP1: Compile in other 16xx boards to OSK defconfig
  ARM: OMAP1: Refresh H2 defconfig
  ARM: OMAP1: Refresh OSK defconfig
  ARM: OMAP: gpio lockdep updates
  ARM: OMAP1: omap1/pm.c build fix
  ARM: OMAP1: omap h2 regression fix
  ARM: OMAP1: Fix compile for boards depending on old gpio expander
  ...

16 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Thu, 13 Mar 2008 00:00:35 +0000 (17:00 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/paulus/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc: (22 commits)
  [POWERPC] Fix large hash table allocation on Cell blades
  [POWERPC] Export empty_zero_page
  [POWERPC] Fix viodasd driver with scatterlist debug
  [POWERPC] Fix arch/powerpc/platforms/powermac/pic.c when !CONFIG_ADB_PMU
  [POWERPC] Fix drivers/macintosh/mediabay.c when !CONFIG_ADB_PMU
  [POWERPC] Fix undefined pmu_sys_suspended compilation error
  [POWERPC] Fix build of modular drivers/macintosh/apm_emu.c
  [POWERPC] Fix sleep on some powerbooks
  [POWERPC] Fix bogus test for unassigned PCI resources
  [POWERPC] Fix zImage-dtb.initrd build error
  [POWERPC] Add __ucmpdi2 for 64-bit comparisons in 32-bit kernels
  [POWERPC] spufs: fix rescheduling of non-runnable contexts
  [POWERPC] spufs: don't (ab)use SCHED_IDLE
  [POWERPC] QE: Make qe_get_firmware_info reentrant
  [POWERPC] 83xx: Make 83xx perfmon support selectable
  [PPC] 8xx: swap bug-fix
  [POWERPC] 85xx: sbc8548 - Fix incorrect PCI-X and PCI interrupt map
  [POWERPC] QE: Fix QE firmware uploading limit
  [POWERPC] 8xx: Fix wrapper platform for adder875, and combine defconfigs.
  [POWERPC] 8xx: fix swap
  ...

16 years ago[POWERPC] Fix large hash table allocation on Cell blades
Michael Ellerman [Wed, 12 Mar 2008 07:03:24 +0000 (18:03 +1100)]
[POWERPC] Fix large hash table allocation on Cell blades

My recent hack to allocate the hash table under 1GB on cell was poorly
tested, *cough*. It turns out on blades with large amounts of memory we
fail to allocate the hash table at all. This is because RTAS has been
instantiated just below 768MB, and 0-x MB are used by the kernel,
leaving no areas that are both large enough and also naturally-aligned.

For the cell IOMMU hack the page tables must be under 2GB, so use that
as the limit instead. This has been tested on real hardware and boots
happily.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years ago[POWERPC] Export empty_zero_page
Theodore Ts'o [Wed, 12 Mar 2008 20:24:31 +0000 (07:24 +1100)]
[POWERPC] Export empty_zero_page

Once again, this time with feeling....

- Ted

>From c91cfaabc17f8a53807a2f31f067a732e34a1550 Mon Sep 17 00:00:00 2001
From: Theodore Ts'o <tytso@mit.edu>
Date: Wed, 12 Mar 2008 11:50:39 -0400
Subject: Export empty_zero_page

The empty_zero_page symbol is exported by most other architectures
(s390, ia64, x86, um), and an upcoming ext4 patch needs it because
ZERO_PAGE() references empty_zero_page, and we need it to zero out an
unitialized extents in ext4 files.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years ago[POWERPC] Fix viodasd driver with scatterlist debug
Benjamin Herrenschmidt [Wed, 12 Mar 2008 06:23:56 +0000 (17:23 +1100)]
[POWERPC] Fix viodasd driver with scatterlist debug

The iSeries viodasd drivers does some very strange things with
scatterlists, one of these causing a BUG_ON to trigger when
scatterlist debugging is enabled due to initializing the
scatterlist with memset instead of sg_init_table().

This fixes it by using sg_init_table().  The rest of the stuff
it does to that poor list is still pretty awful but it will work.

I may look into fixing things in a nicer way some other time.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
16 years ago[POWERPC] Fix arch/powerpc/platforms/powermac/pic.c when !CONFIG_ADB_PMU
Tony Breeds [Tue, 11 Mar 2008 23:48:48 +0000 (10:48 +1100)]
[POWERPC] Fix arch/powerpc/platforms/powermac/pic.c when !CONFIG_ADB_PMU

When building arch/powerpc/platforms/powermac/pic.c when !CONFIG_ADB_PMU
we get the following warnings:

arch/powerpc/platforms/powermac/pic.c: In function 'pmacpic_find_viaint':
arch/powerpc/platforms/powermac/pic.c:623: warning: label 'not_found' defined but not used

This fixes it.

Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>