[NET]: Fix locking in shaper driver.

 o use a semaphore instead of an opencoded and racy lock
 o move locking out of shaper_kick and into the callers - most just
   released the lock before calling shaper_kick
 o remove in_interrupt() tests.  from ->close we can always block, from
   ->hard_start_xmit and timer context never

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Kill export of fl6_sock_lookup.

There is no usage of this EXPORT_SYMBOL in the kernel.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[ATM]: [drivers] kill pointless NULL checks and casts before kfree()

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Chas Williams <chas@cmf.nrl.navy.mil>
Signed-off-by: David S. Miller <davem@davemloft.net>

[ATM]: fix ATM makefile for out-of-source-tree builds

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Chas Williams <chas@cmf.nrl.navy.mil>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPVS]: remove net/ipv4/ipvs/ip_vs_proto_icmp.c

ip_vs_proto_icmp.c was never finished.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] ppc64: Fix result code handling in prom_init

prom_init(), the trampoline code that "talks" to Open Firmware during
early boot, has various issues with managing OF result codes. Some of my
recent fixups in fact made the problem worse on some platforms.

This patch reworks it all. Tested on g5, Maple, POWER3 and POWER5.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge ... /pub/scm/linux/kernel/git/aegl/linux-2.6

Merge ... /linux/kernel/git/sfrench/cifs-2.6

[IA64] fix compilation warning in sys32_epoll_wait()

This gets rid of an unused variable `error' in sys_ia32.c:sys32_epoll_wait()

Getting rid of this one makes parsing the output of the kernecomp
autobuild easier --- searching for `Error' to find a problem kept
hitting this one, even though it's only a warning.

Signed-off-by: Tony Luck <tony.luck@intel.com>

[IA64] Cleanup compile warnings for ski config

The attached patch cleans up a compilation warning when ACPI
is turned off (i.e., when compiling for the Ski simulator).

Signed-off-by: Tony Luck <tony.luck@intel.com>

Merge ... /pub/scm/linux/kernel/git/davem/net-2.6

Merge ... /linux/kernel/git/davem/sparc-2.6

Merge ... /pub/scm/linux/kernel/git/gregkh/usb-2.6

Merge ... /pub/scm/linux/kernel/git/gregkh/i2c-2.6

Merge ... /pub/scm/linux/kernel/git/gregkh/pci-2.6

[PATCH] ppc32/ppc64: cleanup /proc/device-tree

This cleans up the /proc/device-tree representation of the Open Firmware
device-tree on ppc and ppc64.  It does the following things:

 - Workaround an issue in some Apple device-trees where a property may
   exist with the same name as a child node of the parent.  We now
   simply "drop" the property instead of creating duplicate entries in
   /proc with random result...

 - Do not try to chop off the "@0" at the end of a node name whose unit
   address is 0.  This is not useful, inconsistent, and the code was
   buggy and didn't always work anyway.

 - Do not create symlinks for the short name and unit address parts of a
   node.  These were never really used, bloated the memory footprint of
   the device-tree with useless struct proc_dir_entry and their matching
   dentry and inode cache bloat.

This results in smaller code, smaller memory footprint, and a more
accurate view of the tree presented to userland.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc32: Apple device-tree bug fix

This is the ppc32 patch equivalent to the just posted ppc64 one working
around a bug in Apple device-trees regarding the "cpus" nodes.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc64: Fix a device-tree bug on Apple's

Apple's Open Firmware has a funny bug when creating the /cpus nodes
where it leaves a dangling '\0' character in the CPU name which ends up
appearing in the full path of the node. This is bogus and
confuses /proc/device-tree badly.

This patch strips those bogus zero's from the node full path when
reading the device-tree from Open Firmware. The "name" property is not
modified and still contains the spurrious 0 (it basically contains 0
tailing 0 instead of one) but that shouldn't be a problem.

An equivalent patch for ppc32 will follow shortly

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[SPARC64]: Refine PCI strbuf ctx-based flush.

The initial peek read PIO of the match register is just a waste.
Just do the flush writes first, as that is more efficient.

Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] cpufreq-stats driver documentation

Documentation for cpufreq stats.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[PATCH] cpufreq-stats driver updates

Changes to the cpufreq stats driver:
* Changes the way P-state transition table looks in /sysfs providing more
  clear output
* Changes the time unit in the output from HZ to clock_t

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Typos.

cpfureq developers cant spel.

Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] longhaul - adjust transition latency.

From patch by: Ken Staton <ken_staton@agilent.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Longhaul: Magic timer frobbing.

As mandated by the spec, disable timer around transitions.

From code by : Ken Staton <ken_staton@agilent.com
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] longhaul - disable PCI mastering around transition.

The spec states that we have to do this, which is *horrid*.

Based on code from: Ken Staton <ken_staton@agilent.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand governor default sampling downfactor as 1

[PATCH] [5/5] ondemand governor default sampling downfactor as 1

Make default sampling downfactor 1.
This works better with earlier auto downscaling change in ondemand governor.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand governor automatic downscaling

[PATCH] [4/5] ondemand governor automatic downscaling

Here is a change of policy for the ondemand governor. The modification
concerns the frequency downscaling. Instead of decreasing to a lower
frequency when the CPU usage is under 20%, this new policy automatically
scales to the optimal frequency. The optimal frequency being the lowest
frequency which provides enough power to not trigger the upscaling policy.

Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand,conservative governor idle_tick clean-up

[PATCH] [3/5] ondemand,conservative governor idle_tick clean-up

Ondemand and conservative governor clean-up, it factorises the idle ticks
measurement.

Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand,conservative governor store the idle ticks for all cpus

[PATCH] [2/5] ondemand,conservative governor store the idle ticks for all cpus

Ondemand, conservative governor did not store prev_cpu_idle_up into
prev_cpu_idle_down for other CPUs than the current CPU.

Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand,conservative minor bug-fix and cleanup

[PATCH] [1/5] ondemand,conservative minor bug-fix and cleanup

Attached patch fixes some minor issues with Alexander's patch and related
cleanup in both ondemand and conservative governor.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Allow ondemand stepping to be changed by user.

Adds support so that the cpufreq change stepping is no longer fixed at 5% and
can be changed dynamically by the user

Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Prevents un-necessary cpufreq changes if we are already at min/max

Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Add support to cpufreq_ondemand to ignore 'nice' cpu time

Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Conservative cpufreq governer

A new cpufreq module, based on the ondemand one with my additional patches
just posted.  This one is more suitable for battery environments where its
probably more appealing to have the cpu freq gracefully increase and decrease
rather than flip between the min and max freq's.

N.B. Bruno Ducrot pointed out that the amd64's "do have unacceptable latency
between min and max freq transition, due to the step-by-step requirements
(200MHz IIRC)"; so AMD64 users would probably benefit from this too.

Signed-off-by: Alexander Clouter <alex-kernel@digriz.org.uk>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] fix up comment in cpufreq.h

Fix up comment in cpufreq.h stating transition latency should be passed
in microseconds -- it was decided long ago to switch to nanoseconds.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] dual-core powernow-k8

With the release of the dual-core AMD Opterons last week,
it's high time that cpufreq supported them.  The attached
patch applies cleanly to 2.6.12-rc3 and updates powernow-k8
to support the latest Athlon 64 and Opteron processors.

Update the driver to version 1.40.0 and provide support
for dual-core processors.

Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] make cpufreq_gov_dbs static

This patch makes a needlessly global and EXPORT_SYMBOL'ed struct static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Recalibrate cpu_khz [2/2]

Some cpufreq drivers (at that time, only powernow-k7) need to recalibrate the
cpu_khz at runtime.

Signed-off-by: Bruno Ducrot <ducrot@poupinou.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Recalibrate cpu_khz [1/2]

We have to recalibrate cpu_khz in order to use the current FID instead the max
FID since some BIOS do not put the processor at maximum frequency at POST.
Also, some BIOS will change the processor frequency at our back after cpu_khz
was calibrate.  Finally, this will fix a long standing bug when we do
something like this:

# rmmod powernow-k7
# modprobe powernow-k7

Signed-off-by: Bruno Ducrot <ducrot@poupinou.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] AMD Elan SC520 cpufreq driver.

From: Sean Young <sean@mess.org>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] Add warning comment about default governors.

This comes up time and time again. Until its fixed, place this
comment in the Kconfig which should stem the flow of resubmissions.

Signed-off-by: Rob Weryk <rjweryk@uwo.ca>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] speedstep-smi: it works on at least one P4M

The speedstep-smi driver actually works on >=1 notebook with a
Pentium 4-M CPU where all other cpufreq drivers fail. Therefore,
allow speedstep-smi on P4Ms again, but warn users of likely failure

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] ondemand: trivial clean-ups

Trivial ondemand governor clean-ups:
- change from sampling_rate_in_HZ() to the official function
usecs_to_jiffies().
- use for_each_online_cpu() to instead of using "if (cpu_online(i))"

Signed-off-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] speedstep-centrino: Pentium 4 - M (HT) support

The Pentium 4 - Ms (HT) with CPUID 0xF34 and 0xF41 seem to support
centrino-like enhanced speedstep; however, no "table" support is possible.
Therefore, put NULL entries into speedstep-centrino.c

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] cpufreq-core: reduce warning messages.

cpufreq core is printing out messages at KERN_WARNING level that the core
recovers from without intervention, and that the system administrator can
do nothing about.  Patch below reduces the severity of these messages to
debug.

Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Jones <davej@redhat.com>

[CPUFREQ] powernow-k7: don't print khz element of FSB.

Signed-off-by: Dave Jones <davej@redhat.com>

[IPSEC]: Fix esp_decap_data size verification in esp4.

Signed-off-by: Edgar E Iglesias <edgar@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Fix streaming buffer flushing on PCI and SBUS.

Firstly, if the direction is TODEVICE, then dirty data in the
streaming cache is impossible so we can elide the flush-flag
synchronization in that case.

Next, the context allocator is broken.  It is highly likely
that contexts get used multiple times for different dma
mappings, which confuses the strbuf flushing code and makes
it run inefficiently.

Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED]: Disable dsmark debugging messages by default

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED]: make dsmark try using pfifo instead of noop while grafting

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED]: Fix dsmark to count ignored indices while walking

Unused indices which are ignored while walking must still
be counted to avoid dumping the same index twice.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] UDF filesystem: array '__mon_yday' declared as not static

in fs/udf/udftime.c the global array '__mon_yday' is not static, and it
conflicts with the glibc one when the kernel is compiled as user mode.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] pcdp.c build fix

In file included from drivers/firmware/pcdp.c:18:
drivers/firmware/pcdp.h:48: error: field `addr' has incomplete type
drivers/firmware/pcdp.c: In function `setup_serial_console':
drivers/firmware/pcdp.c:27: error:  `ACPI_ADR_SPACE_SYSTEM_MEMORY' undeclared (first use in this  function)

Cc: <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] flush icache in correct context

flush_icache_range() is used in two different situation - in binfmt_elf.c &
co for user space mappings and module.c for kernel modules.  On m68k
flush_icache_range() doesn't know which data to flush, as it has separate
address spaces and the pointer argument can be valid in either address
space.

First I considered splitting flush_icache_range(), but this patch is
simpler.  Setting the correct context gives flush_icache_range() enough
information to flush the correct data.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc64: allow timer based profiling on iseries

We used to have an iseries specific profiler that used /proc/profile.  Now
thats gone we can use the generic timer based stuff.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] h8300 sleep problem

h8300 appears to sleep (halt) when need_resched IS set.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] acpi build fix: x86 setup.c

This is a neverending story

linux/acpi.h contains empty declarations for acpi_boot_init() &
acpi_boot_table_init() but they are nested inside #ifdef CONFIG_ACPI.

So we'll have to #ifdef in arch/i386/kernel/setup.c: setup_arch()

Signed-off-by: Alexander Nyberg <alexn@telia.com>
Cc: "Brown, Len" <len.brown@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64 CONFIG_ACPI=n build fix

Make CONFIG_X86_PM_TIMER dependent on CONFIG_ACPI

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64: More fixes for compilation without CONFIG_ACPI

Suggested by Alexander Nyberg

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] PCI Hotplug: more CPCI updates

Here is my third attempt at a patch to further update the CompactPCI
hotplug driver infrastructure to address the pci_enable_device issue
discussed on the list as well as a few other issues I discovered during
some more testing.  This version addresses a few more issues pointed out
by Prarit Bhargava.  Changes include:
- cpci_enable_device and its recursive calling of pci_enable_device on
  new devices removed.
- Use list_rwsem to avoid slot status change races between disable_slot
  and check_slots.
- Fixed oopsing in cpci_hp_unregister_bus caused by calling list_del on
  a slot after calling pci_hp_deregister.
- Removed kfree calls in cleanup_slots since release_slot will have
  done it already.
- Reworked init_slots a bit to fix latch and adapter file updating on
  subsequent calls to cpci_hp_start.
- Improved sanity checking in cpci_hp_register_controller.
- Now shut things down correctly in cpci_hotplug_exit.
- Switch to pci_get_slot instead of deprecated pci_find_slot.
- A bunch of CodingStyle fixes.

Signed-off-by: Scott Murray <scottm@somanetworks.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] PCI: amd74xx patch for new NVIDIA device IDs

Here's the 2.6 amd74xx patch for NVIDIA MCP51.

Signed-off-by: Andy Currid <acurrid@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] PCI Hotplug: SHPCHP driver doesn't enable PERR and SERR properly

Current shpchp driver doesn't seem to program command register to
enable PERR and SERR properly. The following patch fixes this issue.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] PCI Hotplug: shpchp driver doesn't program _HPP values properly

Current shpchp driver doesn't seem to program _HPP values
properly. The following patch fixes this issue.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] USB: fix usb-serial generic initialization

At module load time, if a generic device is found, the tty information
for the device is not set up properly (as the tty structures aren't initialized
yet.)  This can cause big problems for things like udev.  This patch fixes this.

Thanks to Kay Sievers for the original patch for this problem.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] USB: ftdi_sio: new PID for ELV UM100

ftdi_sio: Add PID for "ELV USB Module UM100".
PID sent by Armin Laugher.

Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] USB: add sl811_cs support

This adds support for a CF-card USB Host adapter, the Ratoc REX-CFU1U, by
wrapping a PCMCIA driver around the existing "sl811-hcd" platform driver.

This CF card is especially useful for PDAs, which currently tend to have
no other solution for USB host capability.

From: Botond Botyanszki <boti@rocketmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] USB: sl811-hcd fixes

Various fixes to the sl811-hcd driver:

  * Fix small glitches that crept in during recent evolution of usbcore's hcd
    glue layer, coupling endpoint state records to usbcore and active urbs.
    (As noted by folk whose boards weren't stuck on 2.6.9 kernels...)

  * Cope with various system-specific issues:
      - Some configurations (e.g. a CF-card uses this chip) have iospace
        addresses for the two registers, rather than memory mapped ones.
      - Some configurations do interesting things with IRQs; maybe the
        line is shared, or it doesn't support level triggering.
      - Not all boards can drive the chip reset line in software.

  * Address a potential race during unlinking.

  * Tweak probe/remove section info to handle the case where this segment
    of a platform bus is hotpluggable (e.g. CF card).  (The basic problem
    is that CONFIG_HOTPLUG is global, which is wrong since not all busses
    can hotplug even on hotplug-friendly systems...)  Also export the
    driver, so that the CF driver can depend on it.

Also removed some annoying end-of-line whitespace.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

[PATCH] I2C: ALI1563 SMBus driver fix

This patch fixes "grave" bugs in i2c-ali1563 driver. It seems on recent
chipset revisions the HSTS_DONE is set only for block transfers, so we
must detect the end of ordinary transaction other way. Also due to missing
and mask, setting other transfer modes was not possible. Moreover the
continous byte mode transfer uses DAT0 for command rather than CMD command.
All those changes were tested with help of Chunhao Huang from Winbond.

I'm willing to maintain the driver. Second patch adds me as maintainer
if this is neccessary.

Signed-Off-By: Rudolf Marek <r.marek@sh.cvut.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

Merge ... /pub/scm/linux/kernel/git/aegl/linux-2.6

[IA64] Use "PER_CPU" form of EXPORT macro

I was gently reminded that there are per-cpu forms of the EXPORT_SYMBOL macros.

Signed-off-by: Tony Luck <tony.luck@intel.com>

[PATCH] Relax idecd dma alignment check

Only the address needs alignment of mask bits, length should work with
a relaxed alignment check.

Signed-off-by: Jens Axboe <axboe@suse.de>
[ This is take 2: make the length check be for 16-byte alignment, not
  just word alignment.  That should hopefully keep everybody happy,
  while still allowing CD writing with DMA ]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc64: actually call prom_send_capabilities

When I sent in the patch adding the code for the kernel to tell the
firmware about its capabilities on pSeries machines, I included the
function to give the capabilities to firmware but somehow forgot the
hunk that adds the call to the new function.  This patch adds the
call.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[IPV4]: Fix BUG() in 2.6.x, udp_poll(), fragments + CONFIG_HIGHMEM

Steven Hand <Steven.Hand@cl.cam.ac.uk> wrote:
>
> Reconstructed forward trace:
>
>   net/ipv4/udp.c:1334   spin_lock_irq()
>   net/ipv4/udp.c:1336   udp_checksum_complete()
> net/core/skbuff.c:1069   skb_shinfo(skb)->nr_frags > 1
> net/core/skbuff.c:1086   kunmap_skb_frag()
> net/core/skbuff.h:1087   local_bh_enable()
> kernel/softirq.c:0140   WARN_ON(irqs_disabled());

The receive queue lock is never taken in IRQs (and should never be) so
we can simply substitute bh for irq.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Fix deadlock with ip_queue and tcp local input path.

When we have ip_queue being used from LOCAL_IN, then we end up with a
situation where the verdicts coming back from userspace traverse the TCP
input path from syscall context.  While this seems to work most of the
time, there's an ugly deadlock:

syscall context is interrupted by the timer interrupt.  When the timer
interrupt leaves, the timer softirq get's scheduled and calls
tcp_delack_timer() and alike.  They themselves do bh_lock_sock(sk),
which is already held from somewhere else -> boom.

I've now tested the suggested solution by Patrick McHardy and Herbert Xu to
simply use local_bh_{en,dis}able().

Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge ... /pub/scm/linux/kernel/git/davem/tg3-2.6

Merge of /home/davem/src/GIT/tg3-2.6/

[NET]: Use %lx for netdev->features sysfs formatting.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Clear up user copy warning in flowlabel code.

We are intentionally ignoring the copy_to_user() value,
make it clear to the compiler too.

Noted by Jeff Garzik.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Add ethtool support for NETIF_F_HW_CSUM.

Signed-off-by: Jon Mason <jdmason@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Kill MULTIPATHHOLDROUTE flag.

It cannot work properly, so just ignore it in drr
and rr multipath algorithms just like the random
multipath algorithm does.

Suggested by Herbert Xu.

Signed-off by: Pravin B. Shelar <pravins@calsoftinc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Add is_multicast_ether_addr() in include/linux/etherdevice.h

This patch adds is_multicast_ether_addr() to go along with
is_valid_ether_addr() and friends. It then changes
is_valid_ether_addr() to use the new macro, and fixes up the comment
on that function to move implementation details out of the API doco.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Primary and secondary addresses

Add an option to make secondary IP addresses get promoted
when primary IP addresses are removed from the device.
It defaults to off to preserve existing behavior.

Signed-off-by: Harald Welte <laforge@gnumonks.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Automatic merge of 'misc-fixes' branch from

rsync://rsync.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev

[TG3]: Fix bug in tg3_load_firmware_cpu

Add tg3_nvram_lock() and tg3_nvram_unlock() calls around tg3_halt_cpu().
It is possible that the bootcode may be loading code from nvram during
this call and stopping the cpu without getting the lock may cause
uncompleted nvram data to be left in the nvram data register. Subsequent
calls to read/write nvram data will fail.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add interrupt test

This test uses the previously added tg3_test_interrupt() to perform the
test.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add loopback test

The test will loopback one packet in MAC loopback mode and verify the
packet data.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add memory test

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add register test

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add parameter to tg3_halt

Add a reset kind parameter to tg3_halt() so that the RESET_KIND_SUSPEND
parameter can be passed to tg3_halt() before doing offline tests.

All other calls to tg3_halt() will use the RESET_KIND_SHUTDOWN
parameter.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add link test

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add nvram test

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Add basic selftest infrastructure

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: receive path optimization

This improves the bridge local receive path by avoiding going
through another softirq.  The bridge receive path is already being called
from a netif_receive_skb() there is no point in going through another
receiveq round trip.

Recursion is limited because bridge can never be a port of a bridge
so handle_bridge() always returns.

Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: prevent bad forwarding table updates

Avoid poisoning of the bridge forwarding table by frames that have been
dropped by filtering. This prevents spoofed source addresses on hostile
side of bridge from causing packet leakage, a small but possible security
risk.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: set features based on enslaved devices

Make features of the bridge pseudo-device be a subset of the underlying
devices.  Motivated by Xen and others who use bridging to do failover.

Signed-off-by: Catalin BOIE <catab at umrella.ro>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: make dev->features unsigned

The features field in netdevice is really a bitmask, and bitmask's should
be unsigned.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: features change notification

Resend of earlier patch (no changes) from Catalin used to provide
device feature change notification.

Signed-off-by: Catalin BOIE <catab at umbrella.ro>
Acked-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

libata: bump version

Merge ... /spare/repo/netdev-2.6 branch use-after-unmap

Automatic merge of 'for-linus' branch from

rsync://rsync.kernel.org/pub/scm/linux/kernel/git/dtor/input