This moves the VME core, VME board drivers, and VME bridge drivers out
of the drivers/staging/vme/ area to drivers/vme/.
The VME device drivers have not moved out yet due to some API questions
they are still working through, that should happen soon, hopefully.
Cc: Martyn Welch <martyn.welch@ge.com>
Cc: Manohar Vanga <manohar.vanga@cern.ch>
Cc: Vincent Bossier <vincent.bossier@gmail.com>
Cc: "Emilio G. Cota" <cota@braap.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some on SoC DSP HW is very tightly coupled with DMA and DAI drivers. It's
necessary to allow some flexability wrt to PCM operations here so that we
can define a bespoke DPCM trigger() PCM operation for such HW.
A bespoke DPCM trigger() allows exact ordering and timing of component
triggering by allowing a component driver to manage the final enable
and disable configurations without adding extra complexity to other
component drivers. e.g. The McPDM DAI and ABE are tightly coupled on
OMAP4 so we have a bespoke trigger to manage the trigger to improve
performance and reduce complexity when triggering new McPDM BEs.
Signed-off-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Some component drivers will need to be able to look up their
DAI link substream and RTD data. Provide a mechanism for this.
Signed-off-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch allows DPCM to dynamically alter the FE to BE PCM links
at runtime based on mixer setting updates. DAPM is looked up after
every mixer update and we perform a DPCM runtime update if the
mixer has a change of value.
This patchs adds/changes the following :-
o Adds DPCM runtime update core.
o Changes soc_dapm_mixer_update_power() and soc_dapm_mux_update_power()
to return if a change has occured rather than 0. No other users check
atm.
Signed-off-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The Dynamic PCM core allows digital audio data to be dynamically
routed between different ALSA PCMs and DAI links on SoC CPUs with
on chip DSP devices. e.g. audio data could be played on pcm:0,0 and
routed to any (or all) SoC DAI links.
Dynamic PCM introduces the concept of Front End (FE) PCMs and Back
End (BE) PCMs. The FE PCMs are normal ALSA PCM devices except that
they can dynamically route digital audio data to any supported BE
PCM. A BE PCM has no ALSA device, but represents a DAI link and it's
substream and audio HW parameters.
e.g. pcm:0,0 routing digital data to 2 external codecs.
FE pcm:0,0 ----> BE (McBSP.0) ----> CODEC 0
+--> BE (McPDM.0) ----> CODEC 1
e.g. pcm:0,0 and pcm:0,1 routing digital data to 1 external codec.
FE pcm:0,0 ---
+--> BE (McBSP.0) ----> CODEC
FE pcm:0,1 ---
The digital audio routing is controlled by the usual ALSA method
of mixer kcontrols. Dynamic PCM uses a DAPM graph to work out the
routing based upon the mixer settings and configures the BE PCMs
based on routing and the FE HW params.
DPCM is designed so that most ASoC component drivers will need no
modification at all. It's intended that existing CODEC, DAI and
platform drivers can be used in DPCM based audio devices without
any changes. However, there will be some cases where minor changes
are required (e.g. for very tightly coupled HW) and there are
helpers to support this too.
Somethimes the HW params of a FE and BE do not match or are
incompatible, so in these cases the machine driver can reconfigure
any hw_params and make any DSP perform sample rate / format conversion.
This patch adds the core DPCM code and contains :-
o The FE and BE PCM operations.
o FE and BE DAI link support.
o FE and BE PCM creation.
o BE support API.
o BE and FE link management.
Signed-off-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
If multiple clients are registered on a single camera host interface,
the user-space hot-plug software can try to access the one, that probed
first, before probing of the second one has completed. This can be
handled by individual host drivers, but it is even better to hold back
the user-space until all the probing on this host has completed. This
fixes a race on ecovec with two clients registered on the CEU1 host, which
otherwise triggers a BUG() in sh_mobile_ceu_remove_device().
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Don't pick __u8/__u16 values directly from raw pointers, but instead use
an array of structures of code:value pairs. This is OK, since the buffer
we take options from is not an skb memory, but a user-to-kernel one.
For those options which don't require any value now, require this to be
zero (for potential future extension of this API).
v2: Changed tcp_repair_opt to use two __u32-s as spotted by David Laight.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- Convert the old uid mapping functions into compatibility wrappers
- Add a uid/gid mapping layer from user space uid and gids to kernel
internal uids and gids that is extent based for simplicty and speed.
* Working with number space after mapping uids/gids into their kernel
internal version adds only mapping complexity over what we have today,
leaving the kernel code easy to understand and test.
- Add proc files /proc/self/uid_map /proc/self/gid_map
These files display the mapping and allow a mapping to be added
if a mapping does not exist.
- Allow entering the user namespace without a uid or gid mapping.
Since we are starting with an existing user our uids and gids
still have global mappings so are still valid and useful they just don't
have local mappings. The requirement for things to work are global uid
and gid so it is odd but perfectly fine not to have a local uid
and gid mapping.
Not requiring global uid and gid mappings greatly simplifies
the logic of setting up the uid and gid mappings by allowing
the mappings to be set after the namespace is created which makes the
slight weirdness worth it.
- Make the mappings in the initial user namespace to the global
uid/gid space explicit. Today it is an identity mapping
but in the future we may want to twist this for debugging, similar
to what we do with jiffies.
- Document the memory ordering requirements of setting the uid and
gid mappings. We only allow the mappings to be set once
and there are no pointers involved so the requirments are
trivial but a little atypical.
Performance:
In this scheme for the permission checks the performance is expected to
stay the same as the actuall machine instructions should remain the same.
The worst case I could think of is ls -l on a large directory where
all of the stat results need to be translated with from kuids and
kgids to uids and gids. So I benchmarked that case on my laptop
with a dual core hyperthread Intel i5-2520M cpu with 3M of cpu cache.
My benchmark consisted of going to single user mode where nothing else
was running. On an ext4 filesystem opening 1,000,000 files and looping
through all of the files 1000 times and calling fstat on the
individuals files. This was to ensure I was benchmarking stat times
where the inodes were in the kernels cache, but the inode values were
not in the processors cache. My results:
v3.4-rc1: ~= 156ns (unmodified v3.4-rc1 with user namespace support disabled)
v3.4-rc1-userns-: ~= 155ns (v3.4-rc1 with my user namespace patches and user namespace support disabled)
v3.4-rc1-userns+: ~= 164ns (v3.4-rc1 with my user namespace patches and user namespace support enabled)
All of the configurations ran in roughly 120ns when I performed tests
that ran in the cpu cache.
So in summary the performance impact is:
1ns improvement in the worst case with user namespace support compiled out.
8ns aka 5% slowdown in the worst case with user namespace support compiled in.
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
- Transform userns->creator from a user_struct reference to a simple
kuid_t, kgid_t pair.
In cap_capable this allows the check to see if we are the creator of
a namespace to become the classic suser style euid permission check.
This allows us to remove the need for a struct cred in the mapping
functions and still be able to dispaly the user namespace creators
uid and gid as 0.
- Remove the now unnecessary delayed_work in free_user_ns.
All that is left for free_user_ns to do is to call kmem_cache_free
and put_user_ns. Those functions can be called in any context
so call them directly from free_user_ns removing the need for delayed work.
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Pull NFS client bugfixes from Trond Myklebust:
- Fix NFSv4 infinite loops on open(O_TRUNC)
- Fix an Oops and an infinite loop in the NFSv4 flock code
- Don't register the PipeFS filesystem until it has been set up
- Fix an Oops in nfs_try_to_update_request
- Don't reuse NFSv4 open owners: fixes a bad sequence id storm.
* tag 'nfs-for-3.4-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
NFSv4: Keep dropped state owners on the LRU list for a while
NFSv4: Ensure that we don't drop a state owner more than once
NFSv4: Ensure we do not reuse open owner names
nfs: Enclose hostname in brackets when needed in nfs_do_root_mount
NFS: put open context on error in nfs_flush_multi
NFS: put open context on error in nfs_pagein_multi
NFSv4: Fix open(O_TRUNC) and ftruncate() error handling
NFSv4: Ensure that we check lock exclusive/shared type against open modes
NFSv4: Ensure that the LOCK code sets exception->inode
NFS: check for req==NULL in nfs_try_to_update_request cleanup
SUNRPC: register PipeFS file system after pernet sybsystem
Pull x86 fixes from H. Peter Anvin.
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x32, siginfo: Provide proper overrides for x32 siginfo_t
asm-generic: Allow overriding clock_t and add attributes to siginfo_t
x32: Check __ILP32__ instead of __LP64__ for x32
x86, acpi: Call acpi_enter_sleep_state via an asmlinkage C function from assembler
ACPI: Convert wake_sleep_flags to a value instead of function
x86, apic: APIC code touches invalid MSR on P5 class machines
i387: ptrace breaks the lazy-fpu-restore logic
x86/platform: Remove incorrect error message in x86_default_fixup_cpu_id()
x86, efi: Add dedicated EFI stub entry point
x86/amd: Remove broken links from comment and kernel message
x86, microcode: Ensure that module is only loaded on supported AMD CPUs
x86, microcode: Fix sysfs warning during module unload on unsupported CPUs
The "pgsteal" stat is confusing because it counts both direct reclaim as
well as background reclaim. However, we have "kswapd_steal" which also
counts background reclaim value.
This patch fixes it and also makes it match the existng "pgscan_" stats.
Test:
pgsteal_kswapd_dma32 447623
pgsteal_kswapd_normal 42272677
pgsteal_kswapd_movable 0
pgsteal_direct_dma32 2801
pgsteal_direct_normal 44353270
pgsteal_direct_movable 0
Signed-off-by: Ying Han <yinghan@google.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dan Magenheimer <dan.magenheimer@oracle.com>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I thought this had been removed years ago. All in-kernel users of this
call have now been cleaned up and converted over to use dev_err()
instead, which is the correct thing to do. Now that there are no users,
the macro can be removed so no one else accidentally starts to use it.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reorder structure writeback_control to remove 8 bytes of padding on 64
bit builds, this shrinks its size from 48 to 40 bytes.
This structure is always on the stack and uses C99 named initialisation,
so should be safe and have a small impact on stack usage.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
fallocate filesystem operation preallocates media space for the given file.
If fallocate returns success then any subsequent write to the given range
never fails with 'not enough space' error.
Signed-off-by: Anatol Pomozov <anatol.pomozov@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
A small L3 cache index disable fix from Srivatsa Bhat which unifies the
way the code checks for already disabled indices.
( Pulling it into v3.4 despite the v3.5 tag - the fix is small and we better
keep the same code across kernel versions for such user facing interfaces. )
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Traces of rcu_prep_idle events can be confusing because
rcu_cleanup_after_idle() does no tracing. This commit therefore adds
this tracing.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The rcu_blocking_is_gp() function tests to see if there is only one
online CPU, and if so, synchronize_sched() and friends become no-ops.
However, for larger systems, num_online_cpus() scans a large vector,
and might be preempted while doing so. While preempted, any number
of CPUs might come online and go offline, potentially resulting in
num_online_cpus() returning 1 when there never had only been one
CPU online. This could result in a too-short RCU grace period, which
could in turn result in total failure, except that the only way that
the grace period is too short is if there is an RCU read-side critical
section spanning it. For RCU-sched and RCU-bh (which are the only
cases using rcu_blocking_is_gp()), RCU read-side critical sections
have either preemption or bh disabled, which prevents CPUs from going
offline. This in turn prevents actual failures from occurring.
This commit therefore adds a large block comment to rcu_blocking_is_gp()
documenting why it is safe. This commit also moves rcu_blocking_is_gp()
into kernel/rcutree.c, which should help prevent unwary developers from
mistaking it for a generally useful function.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Currently, __kfree_rcu() is implemented as an inline function, and
contains a BUILD_BUG_ON() that malfunctions if __kfree_rcu() is compiled
as an out-of-line function. Unfortunately, there are compiler settings
(e.g., -O0) that can result in __kfree_rcu() being compiled out of line,
resulting in annoying build breakage. This commit therefore converts
both __kfree_rcu() and __is_kfree_rcu_offset() from inline functions to
macros to prevent such misbehavior on the part of the compiler.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The list_first_entry_rcu() macro is inherently unsafe because it cannot
be applied to an empty list. But because RCU readers do not exclude
updaters, a list might become empty between the time that list_empty()
claimed it was non-empty and the time that list_first_entry_rcu() is
invoked. Therefore, the list_empty() test cannot be separated from the
list_first_entry_rcu() call. This commit therefore combines these to
macros to create a new list_first_or_null_rcu() macro that replaces
the old (and unsafe) list_first_entry_rcu() macro.
This patch incorporates Paul's review comments on the previous version of
this patch available here:
https://lkml.org/lkml/2012/4/2/536
This patch cannot break any upstream code because list_first_entry_rcu()
is not being used anywhere in the kernel (tested with grep(1)), and any
external code using it is probably broken as a result of using it.
Signed-off-by: Michel Machado <michel@digirati.com.br>
CC: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
CC: Dipankar Sarma <dipankar@in.ibm.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
* Make __list_add_rcu check the next->prev and prev->next pointers
just like __list_add does.
* Make list_del_rcu use __list_del_entry, which does the same checking
at deletion time.
Has been running for a week here without anything being tripped up,
but it seems worth adding for completeness just in case something
ever does corrupt those lists.
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
All macros used for creating different kind of clocks have similar code for
initializing struct clk. This patch removes those redundant lines and create
another macro DEFINE_CLK.
Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
CLK_MUX_INDEX_BIT is mistakenly written as CLK_MUX_INDEX_BITWISE in comment. Fix
it.
CLK_GATE_SET_TO_DISABLE is mistakenly written as CLK_GATE_SET_DISABLE in
comment. Fix it.
Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
For most of .set_rate implementation, parent_rate will be used, so just
like passing parent_rate into .recalc_rate, let's pass parent_rate into
.set_rate too.
It also updates the kernel doc for .set_rate ops.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
This patch is the basic clk version of 'clk: core: copy parent_names &
return error codes'.
The registration functions are changed to allow the core code to copy
the array of strings and allow platforms to declare those arrays as
__initdata.
This patch also converts all of the basic clk registration functions to
return error codes which better aligns them with the existing clk.h api.
Signed-off-by: Mike Turquette <mturquette@linaro.org>
This patch cleans up clk_register and solves a few bugs by teaching
clk_register and __clk_init to return error codes (instead of just NULL)
to better align with the existing clk.h api.
Along with that change this patch also introduces a new behavior whereby
clk_register copies the parent_names array, thus allowing platforms to
declare their parent_names arrays as __initdata.
Signed-off-by: Mike Turquette <mturquette@linaro.org>
Drivers should be able to declare their arrays of parent names as const
so the APIs need to accept const arguments.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
[mturquette@linaro.org: constified gate]
Signed-off-by: Mike Turquette <mturquette@linaro.org>
The comment is inaccurate (it actually ends the CONFIG_COMMON_CLK
section, there's no else) and given that we've just got a single level
of ifdef isn't really needed anyway.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
Besides the static initialization, the clk_ops of basic clks could
also be used by particular clk type being subclass of the basic clks.
For example, clk_busy_divider has the same clk_ops as clk_divider,
except it has to wait for a busy bit before return success with
.set_rate. clk_busy_divider will somehow reuse clk_ops of clk_divider.
Since clk-provider.h is included by clk-private.h, it's safe to move
those clk_ops declaration of basic clks form clk-private.h into
clk-provider.h, so that implementation of clks like clk_busy_divider
above do not need to include clk-private.h to access those clk_ops.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
The clk_ops of basic clks should have "const" to match the definition
in "struct clk" and clk_register prototype.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Mike Turquette <mturquette@linaro.org>
When we do this it becomes clear the lock we should be holding is the vc
lock, and in fact many of our other helpers are properly invoked this way.
We don't at this point guarantee not to race the keyboard code but the results
of that appear harmless and that was true before we started as well.
We now have no users of tty_lock in the console driver...
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch (as1545) fixes a problem affecting several ASUS computers:
The machine crashes or corrupts memory when going into suspend if the
ehci-hcd driver is bound to any controllers. Users have been forced
to unbind or unload ehci-hcd before putting their systems to sleep.
After extensive testing, it was determined that the machines don't
like going into suspend when any EHCI controllers are in the PCI D3
power state. Presumably this is a firmware bug, but there's nothing
we can do about it except to avoid putting the controllers in D3
during system sleep.
The patch adds a new flag to indicate whether the problem is present,
and avoids changing the controller's power state if the flag is set.
Runtime suspend is unaffected; this matters only for system suspend.
However as a side effect, the controller will not respond to remote
wakeup requests while the system is asleep. Hence USB wakeup is not
functional -- but of course, this is already true in the current state
of affairs.
This fixes Bugzilla #42728.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Steven Rostedt <rostedt@goodmis.org>
Tested-by: Andrey Rahmatullin <wrar@wrar.name>
Tested-by: Oleksij Rempel (fishor) <bug-track@fisher-privat.net>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull HSI fixes and ABI documentation from Carlos Chinea
* tag 'hsi_fixes_for_3.4' of git://gitorious.org/kernel-hsi/kernel-hsi:
HSI: Add HSI ABI documentation
HSI: hsi_char: Remove max_data_size from sysfs
HSI: hsi: Rework hsi_event interface
HSI: hsi: Remove controllers and ports from the bus
HSI: hsi: Fix error path cleanup on client registration
HSI: hsi: Rework hsi_controller release
Currently, MSI messages can only be injected to in-kernel irqchips by
defining a corresponding IRQ route for each message. This is not only
unhandy if the MSI messages are generated "on the fly" by user space,
IRQ routes are a limited resource that user space has to manage
carefully.
By providing a direct injection path, we can both avoid using up limited
resources and simplify the necessary steps for user land.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
The default VGA device is a somewhat fluid concept on platforms with
multiple GPUs. Add support for setting it so switching code can update
things appropriately, and make sure that the sysfs code returns the right
device if it's changed.
v2: Updated to fix builds when __ARCH_HAS_VGA_DEFAULT_DEVICE is false.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Acked-by: benh@kernel.crashing.org
Cc: airlied@redhat.com
Signed-off-by: Dave Airlie <airlied@redhat.com>
bridge: set fake_rtable's dst to NULL to avoid kernel Oops
when bridge is deleted before tap/vif device's delete, kernel may
encounter an oops because of NULL reference to fake_rtable's dst.
Set fake_rtable's dst to NULL before sending packets out can solve
this problem.
v4 reformat, change br_drop_fake_rtable(skb) to {}
v3 enrich commit header
v2 introducing new flag DST_FAKE_RTABLE to dst_entry struct.
[ Use "do { } while (0)" for nop br_drop_fake_rtable()
implementation -DaveM ]
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Peter Huang <peter.huangpeng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix merge between commit 3adadc08cc ("net ax25: Reorder ax25_exit to
remove races") and commit 0ca7a4c87d ("net ax25: Simplify and
cleanup the ax25 sysctl handling")
The former moved around the sysctl register/unregister calls, the
later simply removed them.
With help from Stephen Rothwell.
Signed-off-by: David S. Miller <davem@davemloft.net>
Randy Dunlap <rdunlap@xenotime.net> reported:
> On 04/23/2012 12:07 AM, Stephen Rothwell wrote:
>
>> Hi all,
>>
>> Changes since 20120420:
>
>
> include/net/ax25.h:447:75: error: expected ';' before '}' token
>
> static inline int ax25_register_dev_sysctl(ax25_dev *ax25_dev) { return 0 };
> static inline void ax25_unregister_dev_sysctl(ax25_dev *ax25_dev) {};
>
> First function: move ';' inside braces.
> Second function: drop the ';'.
Put the semicolons where it makes sense.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
