This is the main workhorse that deals with semop user calls such that
the waitforzero or semval update operations, on the set, can complete on
not as the sma currently stands. Currently, the set is iterated twice
(setting semval, then backwards for the sempid value). Slowpaths, and
particularly SEM_UNDO calls, must undo any altered sem when it is
detected that the caller must block or has errored-out.
With larger sets, there can occur situations where this involves a lot
of cycles and can obviously be a suboptimal use of cached resources in
shared memory. Ie, discarding CPU caches that are also calling semop
and have the sembuf cached (and can complete), while the current lock
holder doing the semop will block, error, or does a waitforzero
operation.
This patch proposes still iterating the set twice, but the first scan is
read-only, and we perform the actual updates afterward, once we know
that the call will succeed. In order to not suffer from the overhead of
dealing with sops that act on the same sem_num, such (rare) cases use
perform_atomic_semop_slow(), which is exactly what we have now.
Duplicates are detected before grabbing sem_lock, and uses simple a
32/64-bit hash array variable to based on the sem_num we are working on.
In addition add some comments to when we expect to the caller to block.
[akpm@linux-foundation.org: coding-style fixes]
[colin.king@canonical.com: ensure we left shift a ULL rather than a 32 bit integer]
Link: http://lkml.kernel.org/r/20161028181129.7311-1-colin.king@canonical.com
Link: http://lkml.kernel.org/r/20160921194603.GB21438@linux-80c1.suse
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Our sysv sems have been using the notion of lockless wakeups for a
while, ever since commit 0a2b9d4c79 ("ipc/sem.c: move wake_up_process
out of the spinlock section"), in order to reduce the sem_lock hold
times. This in-house pending queue can be replaced by wake_q (just like
all the rest of ipc now), in that it provides the following advantages:
o Simplifies and gets rid of unnecessary code.
o We get rid of the IN_WAKEUP complexities. Given that wake_q_add()
grabs reference to the task, if awoken due to an unrelated event,
between the wake_q_add() and wake_up_q() window, we cannot race with
sys_exit and the imminent call to wake_up_process().
o By not spinning IN_WAKEUP, we no longer need to disable preemption.
In consequence, the wakeup paths (after schedule(), that is) must
acknowledge an external signal/event, as well spurious wakeup occurring
during the pending wakeup window. Obviously no changes in semantics
that could be visible to the user. The fastpath is _only_ for when we
know for sure that we were awoken due to a the waker's successful semop
call (queue.status is not -EINTR).
On a 48-core Haswell, running the ipcscale 'waitforzero' test, the
following is seen with increasing thread counts:
v4.8-rc5 v4.8-rc5
semopv2
Hmean sembench-sem-2 574733.00 ( 0.00%) 578322.00 ( 0.62%)
Hmean sembench-sem-8 811708.00 ( 0.00%) 824689.00 ( 1.59%)
Hmean sembench-sem-12 842448.00 ( 0.00%) 845409.00 ( 0.35%)
Hmean sembench-sem-21 933003.00 ( 0.00%) 977748.00 ( 4.80%)
Hmean sembench-sem-48 935910.00 ( 0.00%) 1004759.00 ( 7.36%)
Hmean sembench-sem-79 937186.00 ( 0.00%) 983976.00 ( 4.99%)
Hmean sembench-sem-234 974256.00 ( 0.00%) 1060294.00 ( 8.83%)
Hmean sembench-sem-265 975468.00 ( 0.00%) 1016243.00 ( 4.18%)
Hmean sembench-sem-296 991280.00 ( 0.00%) 1042659.00 ( 5.18%)
Hmean sembench-sem-327 975415.00 ( 0.00%) 1029977.00 ( 5.59%)
Hmean sembench-sem-358 1014286.00 ( 0.00%) 1049624.00 ( 3.48%)
Hmean sembench-sem-389 972939.00 ( 0.00%) 1043127.00 ( 7.21%)
Hmean sembench-sem-420 981909.00 ( 0.00%) 1056747.00 ( 7.62%)
Hmean sembench-sem-451 990139.00 ( 0.00%) 1051609.00 ( 6.21%)
Hmean sembench-sem-482 965735.00 ( 0.00%) 1040313.00 ( 7.72%)
[akpm@linux-foundation.org: coding-style fixes]
[sfr@canb.auug.org.au: merge fix for WAKE_Q to DEFINE_WAKE_Q rename]
Link: http://lkml.kernel.org/r/20161122210410.5eca9fc2@canb.auug.org.au
Link: http://lkml.kernel.org/r/1474225896-10066-3-git-send-email-dave@stgolabs.net
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The OpenRISC compiler (so far) fails to optimize away a large portion of
code containing a reference to posix_timer_event in alarmtimer.c when
CONFIG_POSIX_TIMERS is unset. Let's give it a direct clue to let the
build succeed.
This fixes
[linux-next:master 6682/7183] alarmtimer.c:undefined reference to `posix_timer_event'
reported by kbuild test robot.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When LONG_MIN is passed to msgrcv, one would expect to recieve any
message. But convert_mode does *msgtyp = -*msgtyp and -LONG_MIN is
undefined. In particular, with my gcc -LONG_MIN produces -LONG_MIN
again.
So handle this case properly by assigning LONG_MAX to *msgtyp if
LONG_MIN was specified as msgtyp to msgrcv.
This code:
long msg[] = { 100, 200 };
int m = msgget(IPC_PRIVATE, IPC_CREAT | 0644);
msgsnd(m, &msg, sizeof(msg), 0);
msgrcv(m, &msg, sizeof(msg), LONG_MIN, 0);
produces currently nothing:
msgget(IPC_PRIVATE, IPC_CREAT|0644) = 65538
msgsnd(65538, {100, "\310\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"}, 16, 0) = 0
msgrcv(65538, ...
Except a UBSAN warning:
UBSAN: Undefined behaviour in ipc/msg.c:745:13
negation of -9223372036854775808 cannot be represented in type 'long int':
With the patch, I see what I expect:
msgget(IPC_PRIVATE, IPC_CREAT|0644) = 0
msgsnd(0, {100, "\310\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"}, 16, 0) = 0
msgrcv(0, {100, "\310\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"}, 16, -9223372036854775808, 0) = 16
Link: http://lkml.kernel.org/r/20161024082633.10148-1-jslaby@suse.cz
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Choosing the appropriate compression option when using an embedded
initramfs can result in significant size differences in the resulting
data.
This is caused by avoiding double compression of the initramfs contents.
For example on my tests, choosing CONFIG_INITRAMFS_COMPRESSION_NONE when
compressing the kernel using XZ) results in up to 500KiB differences
(9MiB to 8.5MiB) in the kernel size as the dictionary will not get
polluted with uncomprensible data and may reuse kernel data too.
Despite embedding an uncompressed initramfs, a user may want to allow
for a compressed extra initramfs to be passed using the rd system, for
example to boot a recovery system. 9ba4bcb645 ("initramfs: read
CONFIG_RD_ variables for initramfs compression") broke that behavior by
making the choice based on CONFIG_RD_* instead of adding
CONFIG_INITRAMFS_COMPRESSION_LZ4. Saddly, CONFIG_RD_* is also used to
choose the supported RD compression algorithms by the kernel and a user
may want to support more than one.
This patch also reverts commit 3e4e0f0a87 ("initramfs: remove
"compression mode" choice") restoring back the "compression mode" choice
and includes the CONFIG_INITRAMFS_COMPRESSION_LZ4 option which was never
added.
As a result the following options are added or readed affecting the embedded
initramfs compression:
INITRAMFS_COMPRESSION_NONE Do no compression
INITRAMFS_COMPRESSION_GZIP Compress using gzip
INITRAMFS_COMPRESSION_BZIP2 Compress using bzip2
INITRAMFS_COMPRESSION_LZMA Compress using lzma
INITRAMFS_COMPRESSION_XZ Compress using xz
INITRAMFS_COMPRESSION_LZO Compress using lzo
INITRAMFS_COMPRESSION_LZ4 Compress using lz4
These depend on the corresponding CONFIG_RD_* option being set (except
NONE which has no dependencies).
This patch depends on the previous one (the previous version didn't) to
simplify the way in which the algorithm is chosen and keep backwards
compatibility with the behaviour introduced by 9ba4bcb645
("initramfs: read CONFIG_RD_ variables for initramfs compression").
Link: http://lkml.kernel.org/r/57EAD77B.7090607@klondike.es
Signed-off-by: Francisco Blas Izquierdo Riera (klondike) <klondike@klondike.es>
Cc: P J P <ppandit@redhat.com>
Cc: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kdb_trap_printk allows to pass normal printk() messages to kdb via
vkdb_printk(). For example, it is used to get backtrace using the
classic show_stack(), see kdb_show_stack().
vkdb_printf() tries to avoid a potential infinite loop by disabling the
trap. But this approach is racy, for example:
CPU1 CPU2
vkdb_printf()
// assume that kdb_trap_printk == 0
saved_trap_printk = kdb_trap_printk;
kdb_trap_printk = 0;
kdb_show_stack()
kdb_trap_printk++;
Problem1: Now, a nested printk() on CPU0 calls vkdb_printf()
even when it should have been disabled. It will not
cause a deadlock but...
// using the outdated saved value: 0
kdb_trap_printk = saved_trap_printk;
kdb_trap_printk--;
Problem2: Now, kdb_trap_printk == -1 and will stay like this.
It means that all messages will get passed to kdb from
now on.
This patch removes the racy saved_trap_printk handling. Instead, the
recursion is prevented by a check for the locked CPU.
The solution is still kind of racy. A non-related printk(), from
another process, might get trapped by vkdb_printf(). And the wanted
printk() might not get trapped because kdb_printf_cpu is assigned. But
this problem existed even with the original code.
A proper solution would be to get_cpu() before setting kdb_trap_printk
and trap messages only from this CPU. I am not sure if it is worth the
effort, though.
In fact, the race is very theoretical. When kdb is running any of the
commands that use kdb_trap_printk there is a single active CPU and the
other CPUs should be in a holding pen inside kgdb_cpu_enter().
The only time this is violated is when there is a timeout waiting for
the other CPUs to report to the holding pen.
Finally, note that the situation is a bit schizophrenic. vkdb_printf()
explicitly allows recursion but only from KDB code that calls
kdb_printf() directly. On the other hand, the generic printk()
recursion is not allowed because it might cause an infinite loop. This
is why we could not hide the decision inside vkdb_printf() easily.
Link: http://lkml.kernel.org/r/1480412276-16690-4-git-send-email-pmladek@suse.com
Signed-off-by: Petr Mladek <pmladek@suse.com>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kdb_printf_lock does not prevent other CPUs from entering the critical
section because it is ignored when KDB_STATE_PRINTF_LOCK is set.
The problematic situation might look like:
CPU0 CPU1
vkdb_printf()
if (!KDB_STATE(PRINTF_LOCK))
KDB_STATE_SET(PRINTF_LOCK);
spin_lock_irqsave(&kdb_printf_lock, flags);
vkdb_printf()
if (!KDB_STATE(PRINTF_LOCK))
BANG: The PRINTF_LOCK state is set and CPU1 is entering the critical
section without spinning on the lock.
The problem is that the code tries to implement locking using two state
variables that are not handled atomically. Well, we need a custom
locking because we want to allow reentering the critical section on the
very same CPU.
Let's use solution from Petr Zijlstra that was proposed for a similar
scenario, see
https://lkml.kernel.org/r/20161018171513.734367391@infradead.org
This patch uses the same trick with cmpxchg(). The only difference is
that we want to handle only recursion from the same context and
therefore we disable interrupts.
In addition, KDB_STATE_PRINTF_LOCK is removed. In fact, we are not able
to set it a non-racy way.
Link: http://lkml.kernel.org/r/1480412276-16690-3-git-send-email-pmladek@suse.com
Signed-off-by: Petr Mladek <pmladek@suse.com>
Reviewed-by: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We've got a delay loop waiting for secondary CPUs. That loop uses
loops_per_jiffy. However, loops_per_jiffy doesn't actually mean how
many tight loops make up a jiffy on all architectures. It is quite
common to see things like this in the boot log:
Calibrating delay loop (skipped), value calculated using timer
frequency.. 48.00 BogoMIPS (lpj=24000)
In my case I was seeing lots of cases where other CPUs timed out
entering the debugger only to print their stack crawls shortly after the
kdb> prompt was written.
Elsewhere in kgdb we already use udelay(), so that should be safe enough
to use to implement our timeout. We'll delay 1 ms for 1000 times, which
should give us a full second of delay (just like the old code wanted)
but allow us to notice that we're done every 1 ms.
[akpm@linux-foundation.org: simplifications, per Daniel]
Link: http://lkml.kernel.org/r/1477091361-2039-1-git-send-email-dianders@chromium.org
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Brian Norris <briannorris@chromium.org>
Cc: <stable@vger.kernel.org> [4.0+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add support for mapping and unmapping a page with attributes.
The primary use for this is currently to allow for us to pass the
DMA_ATTR_SKIP_CPU_SYNC attribute when mapping and unmapping a page. On
some architectures such as ARM the synchronization has significant
overhead and if we are already taking care of the sync_for_cpu and
sync_for_device from the driver there isn't much need to handle this in
the map/unmap calls as well.
Link: http://lkml.kernel.org/r/20161110113601.76501.46095.stgit@ahduyck-blue-test.jf.intel.com
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "Add support for DMA writable pages being writable by the
network stack", v3.
The first 19 patches in the set add support for the DMA attribute
DMA_ATTR_SKIP_CPU_SYNC on multiple platforms/architectures. This is
needed so that we can flag the calls to dma_map/unmap_page so that we do
not invalidate cache lines that do not currently belong to the device.
Instead we have to take care of this in the driver via a call to
sync_single_range_for_cpu prior to freeing the Rx page.
Patch 20 adds support for dma_map_page_attrs and dma_unmap_page_attrs so
that we can unmap and map a page using the DMA_ATTR_SKIP_CPU_SYNC
attribute.
Patch 21 adds support for freeing a page that has multiple references
being held by a single caller. This way we can free page fragments that
were allocated by a given driver.
The last 2 patches use these updates in the igb driver, and lay the
groundwork to allow for us to reimplement the use of build_skb.
This patch (of 23):
This change allows us to pass DMA_ATTR_SKIP_CPU_SYNC which allows us to
avoid invoking cache line invalidation if the driver will just handle it
later via a sync_for_cpu or sync_for_device call.
Link: http://lkml.kernel.org/r/20161110113419.76501.38491.stgit@ahduyck-blue-test.jf.intel.com
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A soft lookup will occur when I run trinity in syscall kexec_load. the
corresponding stack information is as follows.
BUG: soft lockup - CPU#6 stuck for 22s! [trinity-c6:13859]
Kernel panic - not syncing: softlockup: hung tasks
CPU: 6 PID: 13859 Comm: trinity-c6 Tainted: G O L ----V------- 3.10.0-327.28.3.35.zhongjiang.x86_64 #1
Hardware name: Huawei Technologies Co., Ltd. Tecal BH622 V2/BC01SRSA0, BIOS RMIBV386 06/30/2014
Call Trace:
<IRQ> dump_stack+0x19/0x1b
panic+0xd8/0x214
watchdog_timer_fn+0x1cc/0x1e0
__hrtimer_run_queues+0xd2/0x260
hrtimer_interrupt+0xb0/0x1e0
? call_softirq+0x1c/0x30
local_apic_timer_interrupt+0x37/0x60
smp_apic_timer_interrupt+0x3f/0x60
apic_timer_interrupt+0x6d/0x80
<EOI> ? kimage_alloc_control_pages+0x80/0x270
? kmem_cache_alloc_trace+0x1ce/0x1f0
? do_kimage_alloc_init+0x1f/0x90
kimage_alloc_init+0x12a/0x180
SyS_kexec_load+0x20a/0x260
system_call_fastpath+0x16/0x1b
the first time allocation of control pages may take too much time
because crash_res.end can be set to a higher value. we need to add
cond_resched to avoid the issue.
The patch have been tested and above issue is not appear.
Link: http://lkml.kernel.org/r/1481164674-42775-1-git-send-email-zhongjiang@huawei.com
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Xunlei Pang <xpang@redhat.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently in x86_64, the symbol address of phys_base is exported to
vmcoreinfo. Dave Anderson complained this is really useless for his
Crash implementation. Because in user-space utility Crash and
Makedumpfile which exported vmcore information is mainly used for, value
of phys_base is needed to covert virtual address of exported kernel
symbol to physical address. Especially init_level4_pgt, if we want to
access and go over the page table to look up a PA corresponding to VA,
firstly we need calculate
page_dir = SYMBOL(init_level4_pgt) - __START_KERNEL_map + phys_base;
Now in Crash and Makedumpfile, we have to analyze the vmcore elf program
header to get value of phys_base. As Dave said, it would be preferable
if it were readily availabl in vmcoreinfo rather than depending upon the
PT_LOAD semantics.
Hence in this patch change to export the value of phys_base instead of
its virtual address.
And people also complained that KERNEL_IMAGE_SIZE exporting is x86_64
only, should be moved into arch dependent function
arch_crash_save_vmcoreinfo. Do the moving in this patch.
Link: http://lkml.kernel.org/r/1478568596-30060-2-git-send-email-bhe@redhat.com
Signed-off-by: Baoquan He <bhe@redhat.com>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H . Peter Anvin" <hpa@zytor.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Xunlei Pang <xlpang@redhat.com>
Cc: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Eugene Surovegin <surovegin@google.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Atsushi Kumagai <ats-kumagai@wm.jp.nec.com>
Cc: Dave Anderson <anderson@redhat.com>
Cc: Pratyush Anand <panand@redhat.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I was amused to find "unsafe core_pattern" warning having these lines in
/etc/sysctl.conf:
fs.suid_dumpable=2
kernel.core_pattern=/core/core-%e-%p-%E
kernel.core_uses_pid=0
Turns out kernel is formally right. Default core_pattern is just "core",
which doesn't qualify for secure path while setting suid.dumpable.
Hint admins about solution, clarify sysctl names, delete unnecessary '\'
characters (string literals are concatenated regardless) and reformat for
easier grepping.
Link: http://lkml.kernel.org/r/20161029152124.GA1258@avx2
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When running certain database workload on a high-end system with many
CPUs, it was found that spinlock contention in the sigprocmask syscalls
became a significant portion of the overall CPU cycles as shown below.
9.30% 9.30% 905387 dataserver /proc/kcore 0x7fff8163f4d2
[k] _raw_spin_lock_irq
|
---_raw_spin_lock_irq
|
|--99.34%-- __set_current_blocked
| sigprocmask
| sys_rt_sigprocmask
| system_call_fastpath
| |
| |--50.63%-- __swapcontext
| | |
| | |--99.91%-- upsleepgeneric
| |
| |--49.36%-- __setcontext
| | ktskRun
Looking further into the swapcontext function in glibc, it was found that
the function always call sigprocmask() without checking if there are
changes in the signal mask.
A check was added to the __set_current_blocked() function to avoid taking
the sighand->siglock spinlock if there is no change in the signal mask.
This will prevent unneeded spinlock contention when many threads are
trying to call sigprocmask().
With this patch applied, the spinlock contention in sigprocmask() was
gone.
Link: http://lkml.kernel.org/r/1474979209-11867-1-git-send-email-Waiman.Long@hpe.com
Signed-off-by: Waiman Long <Waiman.Long@hpe.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Stas Sergeev <stsp@list.ru>
Cc: Scott J Norton <scott.norton@hpe.com>
Cc: Douglas Hatch <doug.hatch@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
