Set MAC/BB/RF register according to specified channel. The function
rtw_set_channel_mac() is used to set MAC registers, but 8723D only need
some of them.
For channel 14, we need to set different CCK DFIR values, so restore the
values when channel 1 to 13 is selected.
Spur calibration is needed in channel 13 and 14, and we do notch if spur
is over threshold.
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200504105010.10780-4-yhchuang@realtek.com
This ops is used to parse RX descriptor to know the length of received
packet and containing PHY status. If PHY status is existing, the order is
RX descriptor, PHY status and then packet.
There are two types of PHY status, named CCK and OFDM. Their size are the
same, but formats are different.
struct ieee80211_rx_status is also filled depends on above information.
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200504105010.10780-3-yhchuang@realtek.com
Luke Nelson says:
====================
This patch series introduces a set of optimizations to the BPF JIT
on RV64. The optimizations are related to the verifier zero-extension
optimization and BPF_JMP BPF_K.
We tested the optimizations on a QEMU riscv64 virt machine, using
lib/test_bpf and test_verifier, and formally verified their correctness
using Serval.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch optimizes BPF_JSET BPF_K by using a RISC-V andi instruction
when the BPF immediate fits in 12 bits, instead of first loading the
immediate to a temporary register.
Examples of generated code with and without this optimization:
BPF_JMP_IMM(BPF_JSET, R1, 2, 1) without optimization:
20: li t1,2
24: and t1,a0,t1
28: bnez t1,0x30
BPF_JMP_IMM(BPF_JSET, R1, 2, 1) with optimization:
20: andi t1,a0,2
24: bnez t1,0x2c
BPF_JMP32_IMM(BPF_JSET, R1, 2, 1) without optimization:
20: li t1,2
24: mv t2,a0
28: slli t2,t2,0x20
2c: srli t2,t2,0x20
30: slli t1,t1,0x20
34: srli t1,t1,0x20
38: and t1,t2,t1
3c: bnez t1,0x44
BPF_JMP32_IMM(BPF_JSET, R1, 2, 1) with optimization:
20: andi t1,a0,2
24: bnez t1,0x2c
In these examples, because the upper 32 bits of the sign-extended
immediate are 0, BPF_JMP BPF_JSET and BPF_JMP32 BPF_JSET are equivalent
and therefore the JIT produces identical code for them.
Co-developed-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Luke Nelson <luke.r.nels@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Björn Töpel <bjorn.topel@gmail.com>
Acked-by: Björn Töpel <bjorn.topel@gmail.com>
Link: https://lore.kernel.org/bpf/20200506000320.28965-5-luke.r.nels@gmail.com
There are only two implementaions, one for ipv4 and one for ipv6.
Both are almost identical, they clear skb->cb[], set the TRANSFORMED flag
in IP(6)CB and then call the common xfrm_output() function.
By placing the IPCB handling into the common function, we avoid the need
for the output_finish indirection as the output functions can simply
use xfrm_output().
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Move this to xfrm_output.c. This avoids the state->extract_output
indirection.
This patch also removes the duplicated __xfrm6_extract_header helper
added in an earlier patch, we can now use the one from xfrm_inout.h .
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
so next patch can re-use it from net/xfrm/xfrm_output.c without
causing a linker error when IPV6 is a module.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
We cannot call this function from the core kernel unless we would force
CONFIG_IPV6=y.
Therefore expose this via ipv6_stubs so we can call it from net/xfrm
in the followup patch.
Since the call is expected to be unlikely, no extra code for the IPV6=y
case is added and we will always eat the indirection cost.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
The function only initializes the XFRM CB in the skb.
After previous patch xfrm4_extract_header is only called from
net/xfrm/xfrm_{input,output}.c.
Because of IPV6=m linker errors the ipv6 equivalent
(xfrm6_extract_header) was already placed in xfrm_inout.h because
we can't call functions residing in a module from the core.
So do the same for the ipv4 helper and place it next to the ipv6 one.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
In order to keep CONFIG_IPV6=m working, xfrm6_extract_header needs to be
duplicated. It will be removed again in a followup change when the
remaining caller is moved to net/xfrm as well.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
We can use a direct call for ipv4, so move the needed functions
to net/xfrm/xfrm_output.c and call them directly.
For ipv6 the indirection can be avoided as well but it will need
a bit more work -- to ease review it will be done in another patch.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Fix the following coccicheck warning:
include/sound/hdaudio.h:210:73-74: WARNING: return of 0/1 in function
'snd_hdac_is_in_pm' with return type bool
include/sound/hdaudio.h:211:76-77: WARNING: return of 0/1 in function
'snd_hdac_is_power_on' with return type bool
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Link: https://lore.kernel.org/r/20200506061716.19209-1-yanaijie@huawei.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
sizeof(flexible-array-member) triggers a warning because flexible array
members have incomplete type[1]. There are some instances of code in
which the sizeof operator is being incorrectly/erroneously applied to
zero-length arrays and the result is zero. Such instances may be hiding
some bugs. So, this work (flexible-array member conversions) will also
help to get completely rid of those sorts of issues.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200504201224.GA32282@embeddedor
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
sizeof(flexible-array-member) triggers a warning because flexible array
members have incomplete type[1]. There are some instances of code in
which the sizeof operator is being incorrectly/erroneously applied to
zero-length arrays and the result is zero. Such instances may be hiding
some bugs. So, this work (flexible-array member conversions) will also
help to get completely rid of those sorts of issues.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200504200838.GA31974@embeddedor
This merges the lockless page table walk rework series from Aneesh.
Because it touches powerpc KVM code we are sharing it with the kvm-ppc
tree in our topic/ppc-kvm branch.
This is the cover letter from Aneesh:
Avoid IPI while updating page table entries.
Problem Summary:
Slow termination of KVM guest with large guest RAM config due to a
large number of IPIs that were caused by clearing level 1 PTE
entries (THP) entries. This is shown in the stack trace below.
- qemu-system-ppc [kernel.vmlinux] [k] smp_call_function_many
- smp_call_function_many
- 36.09% smp_call_function_many
serialize_against_pte_lookup
radix__pmdp_huge_get_and_clear
zap_huge_pmd
unmap_page_range
unmap_vmas
unmap_region
__do_munmap
__vm_munmap
sys_munmap
system_call
__munmap
qemu_ram_munmap
qemu_anon_ram_free
reclaim_ramblock
call_rcu_thread
qemu_thread_start
start_thread
__clone
Why we need to do IPI when clearing PMD entries:
This was added as part of commit: 13bd817bb8 ("powerpc/thp: Serialize pmd clear against a linux page table walk")
serialize_against_pte_lookup makes sure that all parallel lockless
page table walk completes before we convert a PMD pte entry to regular
pmd entry. We end up doing that conversion in the below scenarios
1) __split_huge_zero_page_pmd
2) do_huge_pmd_wp_page_fallback
3) MADV_DONTNEED running parallel to page faults.
local_irq_disable and lockless page table walk:
The lockless page table walk work with the assumption that we can
dereference the page table contents without holding a lock. For this
to work, we need to make sure we read the page table contents
atomically and page table pages are not going to be freed/released
while we are walking the table pages. We can achieve by using a rcu
based freeing for page table pages or if the architecture implements
broadcast tlbie, we can block the IPI as we walk the page table pages.
To support both the above framework, lockless page table walk is done
with irq disabled instead of rcu_read_lock()
We do have two interface for lockless page table walk, gup fast and
__find_linux_pte. This patch series makes __find_linux_pte table walk
safe against the conversion of PMD PTE to regular PMD.
gup fast:
gup fast is already safe against THP split because kernel now
differentiate between a pmd split and a compound page split. gup fast
can run parallel to a pmd split and we prevent a parallel gup fast to
a hugepage split, by freezing the page refcount and failing the
speculative page ref increment.
Similar to how gup is safe against parallel pmd split, this patch
series updates the __find_linux_pte callers to be safe against a
parallel pmd split. We do that by enforcing the following rules.
1) Don't reload the pte value, because that can be updated in
parallel.
2) Code should be able to work with a stale PTE value and not the
recent one. ie, the pte value that we are looking at may not be the
latest value in the page table.
3) Before looking at pte value check for _PAGE_PTE bit. We now do this
as part of pte_present() check.
Performance:
This speeds up Qemu guest RAM del/unplug time as below
128 core, 496GB guest:
Without patch:
munmap start: timer = 13162 ms, PID=7684
munmap finish: timer = 95312 ms, PID=7684 - delta = 82150 ms
With patch (upto removing IPI)
munmap start: timer = 196449 ms, PID=6681
munmap finish: timer = 196488 ms, PID=6681 - delta = 39ms
With patch (with adding the tlb invalidate in pmdp_huge_get_and_clear_full)
munmap start: timer = 196345 ms, PID=6879
munmap finish: timer = 196714 ms, PID=6879 - delta = 369ms
Link: https://lore.kernel.org/r/20200505071729.54912-1-aneesh.kumar@linux.ibm.com
On some X86 devices we do not register an input-device, because the
power-button is also handled by the soc_button_array (GPIO) input driver,
and we want to avoid reporting power-button presses to userspace twice.
Sofar when we did this we also did not register our interrupt handlers,
since those were only necessary to report input events.
But on at least 2 device models the Medion Akoya E1239T and the GPD win,
the GPIO pin used by the soc_button_array driver for the power-button
cannot wakeup the system from suspend. Why this does not work is not clear,
I've tried comparing the value of all relevant registers on the Cherry
Trail SoC, with those from models where this does work. I've checked:
PMC registers: FUNC_DIS, FUNC_DIS2, SOIX_WAKE_EN, D3_STS_0, D3_STS_1,
D3_STDBY_STS_0, D3_STDBY_STS_1; PMC ACPI I/O regs: PM1_STS_EN, GPE0a_EN
and they all have identical contents in the working and non working cases.
I suspect that the firmware either sets some unknown register to a value
causing this, or that it turns off a power-plane which is necessary for
GPIO wakeups to work during suspend.
What does work on the Medion Akoya E1239T is letting the AXP288 wakeup
the system on a power-button press (the GPD win has a different PMIC).
Move the registering of the power-button press/release interrupt-handler
from axp20x_pek_probe_input_device() to axp20x_pek_probe() so that the
PMIC will wakeup the system on a power-button press, even if we do not
register an input device.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Link: https://lore.kernel.org/r/20200426155757.297087-1-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
imx7ulp uses different USB PHY with imx7d (MXS PHY vs PICO PHY), so the
features are supported by non-core register are a little different.
For example, autoresume feature is supported by all controllers for
imx7ulp, but for imx7d, it is only supported by non-HSIC controller.
Besides, these two platforms use different HSIC controller, imx7ulp
needs software operation, but imx7d doesn't.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
The bootloader may use device mode, and keep dp up. We need dp
to be pulled down before possbile charger detection operation.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
imx7d (and imx8mm, imx8mn) uses Samsung PHY and USB generic PHY driver.
The USB generic PHY driver is impossible to have a charger detection
for every user, so we implement USB charger detection routine at glue
layer. After the detection has finished, it will notify USB PHY
charger framework, and the uevents will be triggered.
Signed-off-by: Jun Li <jun.li@nxp.com>
Signed-off-by: Peter Chen <peter.chen@nxp.com>
Some vendors glue layer need to handle some events for vbus, eg,
some i.mx platforms (imx7d, imx8mm, imx8mn, etc) needs vbus event
to handle charger detection, its charger detection is finished at
glue layer code, but not at USB PHY driver.
Signed-off-by: Peter Chen <peter.chen@nxp.com>
This patch removes the unused functions set_kernel_text_rw/ro.
Currently, it is not being invoked from anywhere and no other architecture
(except arm) uses this code. Even in ARM, these functions are not invoked
from anywhere currently.
Fixes: d27c3c9081 ("riscv: add STRICT_KERNEL_RWX support")
Signed-off-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Zong Li <zong.li@sifive.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Pull x86 platform driver fixes from Andy Shevchenko:
- Avoid loading asus-nb-wmi module on selected laptop models
- Fix S0ix debug support for Jasper Lake PMC
- Few fixes which have been reported by Hulk bot and others
* tag 'platform-drivers-x86-v5.7-2' of git://git.infradead.org/linux-platform-drivers-x86:
platform/x86: thinkpad_acpi: Remove always false 'value < 0' statement
platform/x86: intel_pmc_core: avoid unused-function warnings
platform/x86: asus-nb-wmi: Do not load on Asus T100TA and T200TA
platform/x86: intel_pmc_core: Change Jasper Lake S0ix debug reg map back to ICL
platform/x86/intel-uncore-freq: make uncore_root_kobj static
platform/x86: wmi: Make two functions static
platform/x86: surface3_power: Fix a NULL vs IS_ERR() check in probe
In bpf_tcp_ingress we used apply_bytes to subtract bytes from sg.size
which is used to track total bytes in a message. But this is not
correct because apply_bytes is itself modified in the main loop doing
the mem_charge.
Then at the end of this we have sg.size incorrectly set and out of
sync with actual sk values. Then we can get a splat if we try to
cork the data later and again try to redirect the msg to ingress. To
fix instead of trying to track msg.size do the easy thing and include
it as part of the sk_msg_xfer logic so that when the msg is moved the
sg.size is always correct.
To reproduce the below users will need ingress + cork and hit an
error path that will then try to 'free' the skmsg.
[ 173.699981] BUG: KASAN: null-ptr-deref in sk_msg_free_elem+0xdd/0x120
[ 173.699987] Read of size 8 at addr 0000000000000008 by task test_sockmap/5317
[ 173.700000] CPU: 2 PID: 5317 Comm: test_sockmap Tainted: G I 5.7.0-rc1+ #43
[ 173.700005] Hardware name: Dell Inc. Precision 5820 Tower/002KVM, BIOS 1.9.2 01/24/2019
[ 173.700009] Call Trace:
[ 173.700021] dump_stack+0x8e/0xcb
[ 173.700029] ? sk_msg_free_elem+0xdd/0x120
[ 173.700034] ? sk_msg_free_elem+0xdd/0x120
[ 173.700042] __kasan_report+0x102/0x15f
[ 173.700052] ? sk_msg_free_elem+0xdd/0x120
[ 173.700060] kasan_report+0x32/0x50
[ 173.700070] sk_msg_free_elem+0xdd/0x120
[ 173.700080] __sk_msg_free+0x87/0x150
[ 173.700094] tcp_bpf_send_verdict+0x179/0x4f0
[ 173.700109] tcp_bpf_sendpage+0x3ce/0x5d0
Fixes: 604326b41a ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/158861290407.14306.5327773422227552482.stgit@john-Precision-5820-Tower
When sk_msg_pop() is called where the pop operation is working on
the end of a sge element and there is no additional trailing data
and there _is_ data in front of pop, like the following case,
|____________a_____________|__pop__|
We have out of order operations where we incorrectly set the pop
variable so that instead of zero'ing pop we incorrectly leave it
untouched, effectively. This can cause later logic to shift the
buffers around believing it should pop extra space. The result is
we have 'popped' more data then we expected potentially breaking
program logic.
It took us a while to hit this case because typically we pop headers
which seem to rarely be at the end of a scatterlist elements but
we can't rely on this.
Fixes: 7246d8ed4d ("bpf: helper to pop data from messages")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/158861288359.14306.7654891716919968144.stgit@john-Precision-5820-Tower
There is no logic in elf_fdpic_core_dump itself or in the various arch
helpers called from it which use uaccess routines on kernel pointers
except for the file writes thate are nicely encapsulated by using
__kernel_write in dump_emit.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
There is no logic in elf_core_dump itself or in the various arch helpers
called from it which use uaccess routines on kernel pointers except for
the file writes thate are nicely encapsulated by using __kernel_write in
dump_emit.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
The code in binfmt_elf.c is differnt from the rest of the code that
processes siginfo, as it sends siginfo from a kernel buffer to a file
rather than from kernel memory to userspace buffers. To remove it's
use of set_fs the code needs some different siginfo helpers.
Add the helper copy_siginfo_to_external to copy from the kernel's
internal siginfo layout to a buffer in the siginfo layout that
userspace expects.
Modify fill_siginfo_note to use copy_siginfo_to_external instead of
set_fs and copy_siginfo_to_user.
Update compat_binfmt_elf.c to use the previously added
copy_siginfo_to_external32 to handle the compat case.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Factor out a copy_siginfo_to_external32 helper from
copy_siginfo_to_user32 that fills out the compat_siginfo, but does so
on a kernel space data structure. With that we can let architectures
override copy_siginfo_to_user32 with their own implementations using
copy_siginfo_to_external32. That allows moving the x32 SIGCHLD purely
to x86 architecture code.
As a nice side effect copy_siginfo_to_external32 also comes in handy
for avoiding a set_fs() call in the coredump code later on.
Contains improvements from Eric W. Biederman <ebiederm@xmission.com>
and Arnd Bergmann <arnd@arndb.de>.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Replace the coredump ->read method with a ->dump method that must call
dump_emit itself. That way we avoid a buffer allocation an messing with
set_fs() to call into code that is intended to deal with user buffers.
For the ->get case we can now use a small on-stack buffer and avoid
memory allocations as well.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Just use the proper non __-prefixed get/put_user variants where that is
not done yet.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Currently, we may perform a copy_to_user (through
simple_read_from_buffer()) while holding a context's register_lock,
while accessing the context save area.
This change uses a temporary buffer for the context save area data,
which we then pass to simple_read_from_buffer.
Includes changes from Christoph Hellwig <hch@lst.de>.
Fixes: bf1ab978be ("[POWERPC] coredump: Add SPU elf notes to coredump.")
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
[hch: renamed to function to avoid ___-prefixes]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
When a new neighbor entry has been added, event is generated but it does not
include protocol, because its value is assigned after the event notification
routine has run, so move protocol assignment code earlier.
Fixes: df9b0e30d4 ("neighbor: Add protocol attribute")
Cc: David Ahern <dsahern@gmail.com>
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
