page_frag_free() calls __free_pages_ok() to free the page back to Buddy.
This is OK for high order page, but for order-0 pages, it misses the
optimization opportunity of using Per-Cpu-Pages and can cause zone lock
contention when called frequently.
Pawel Staszewski recently shared his result of 'how Linux kernel handles
normal traffic'[1] and from perf data, Jesper Dangaard Brouer found the
lock contention comes from page allocator:
mlx5e_poll_tx_cq
|
--16.34%--napi_consume_skb
|
|--12.65%--__free_pages_ok
| |
| --11.86%--free_one_page
| |
| |--10.10%--queued_spin_lock_slowpath
| |
| --0.65%--_raw_spin_lock
|
|--1.55%--page_frag_free
|
--1.44%--skb_release_data
Jesper explained how it happened: mlx5 driver RX-page recycle mechanism is
not effective in this workload and pages have to go through the page
allocator. The lock contention happens during mlx5 DMA TX completion
cycle. And the page allocator cannot keep up at these speeds.[2]
I thought that __free_pages_ok() are mostly freeing high order pages and
thought this is an lock contention for high order pages but Jesper
explained in detail that __free_pages_ok() here are actually freeing
order-0 pages because mlx5 is using order-0 pages to satisfy its page pool
allocation request.[3]
The free path as pointed out by Jesper is:
skb_free_head()
-> skb_free_frag()
-> page_frag_free()
And the pages being freed on this path are order-0 pages.
Fix this by doing similar things as in __page_frag_cache_drain() - send
the being freed page to PCP if it's an order-0 page, or directly to Buddy
if it is a high order page.
With this change, Paweł hasn't noticed lock contention yet in his
workload and Jesper has noticed a 7% performance improvement using a micro
benchmark and lock contention is gone. Ilias' test on a 'low' speed 1Gbit
interface on an cortex-a53 shows ~11% performance boost testing with
64byte packets and __free_pages_ok() disappeared from perf top.
[1]: https://www.spinics.net/lists/netdev/msg531362.html
[2]: https://www.spinics.net/lists/netdev/msg531421.html
[3]: https://www.spinics.net/lists/netdev/msg531556.html
[akpm@linux-foundation.org: add comment]
Link: http://lkml.kernel.org/r/20181120014544.GB10657@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reported-by: Pawel Staszewski <pstaszewski@itcare.pl>
Analysed-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Tested-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Acked-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Pankaj gupta <pagupta@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since a2468cc9bf ("swap: choose swap device according to numa node"),
avail_lists field of swap_info_struct is changed to an array with
MAX_NUMNODES elements. This made swap_info_struct size increased to 40KiB
and needs an order-4 page to hold it.
This is not optimal in that:
1 Most systems have way less than MAX_NUMNODES(1024) nodes so it
is a waste of memory;
2 It could cause swapon failure if the swap device is swapped on
after system has been running for a while, due to no order-4
page is available as pointed out by Vasily Averin.
Solve the above two issues by using nr_node_ids(which is the actual
possible node number the running system has) for avail_lists instead of
MAX_NUMNODES.
nr_node_ids is unknown at compile time so can't be directly used when
declaring this array. What I did here is to declare avail_lists as zero
element array and allocate space for it when allocating space for
swap_info_struct. The reason why keep using array but not pointer is
plist_for_each_entry needs the field to be part of the struct, so pointer
will not work.
This patch is on top of Vasily Averin's fix commit. I think the use of
kvzalloc for swap_info_struct is still needed in case nr_node_ids is
really big on some systems.
Link: http://lkml.kernel.org/r/20181115083847.GA11129@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vasily Averin <vvs@virtuozzo.com>
Cc: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The LKP robot found a 27% will-it-scale/page_fault3 performance
regression regarding commit e27be240df53("mm: memcg: make sure
memory.events is uptodate when waking pollers").
What the test does is:
1 mkstemp() a 128M file on a tmpfs;
2 start $nr_cpu processes, each to loop the following:
2.1 mmap() this file in shared write mode;
2.2 write 0 to this file in a PAGE_SIZE step till the end of the file;
2.3 unmap() this file and repeat this process.
3 After 5 minutes, check how many loops they managed to complete, the
higher the better.
The commit itself looks innocent enough as it merely changed some event
counting mechanism and this test didn't trigger those events at all.
Perf shows increased cycles spent on accessing root_mem_cgroup->stat_cpu
in count_memcg_event_mm()(called by handle_mm_fault()) and in
__mod_memcg_state() called by page_add_file_rmap(). So it's likely due
to the changed layout of 'struct mem_cgroup' that either make stat_cpu
falling into a constantly modifying cacheline or some hot fields stop
being in the same cacheline.
I verified this by moving memory_events[] back to where it was:
: --- a/include/linux/memcontrol.h
: +++ b/include/linux/memcontrol.h
: @@ -205,7 +205,6 @@ struct mem_cgroup {
: int oom_kill_disable;
:
: /* memory.events */
: - atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
: struct cgroup_file events_file;
:
: /* protect arrays of thresholds */
: @@ -238,6 +237,7 @@ struct mem_cgroup {
: struct mem_cgroup_stat_cpu __percpu *stat_cpu;
: atomic_long_t stat[MEMCG_NR_STAT];
: atomic_long_t events[NR_VM_EVENT_ITEMS];
: + atomic_long_t memory_events[MEMCG_NR_MEMORY_EVENTS];
:
: unsigned long socket_pressure;
And performance restored.
Later investigation found that as long as the following 3 fields
moving_account, move_lock_task and stat_cpu are in the same cacheline,
performance will be good. To avoid future performance surprise by other
commits changing the layout of 'struct mem_cgroup', this patch makes
sure the 3 fields stay in the same cacheline.
One concern of this approach is, moving_account and move_lock_task could
be modified when a process changes memory cgroup while stat_cpu is a
always read field, it might hurt to place them in the same cacheline. I
assume it is rare for a process to change memory cgroup so this should
be OK.
Link: https://lkml.kernel.org/r/20180528114019.GF9904@yexl-desktop
Link: http://lkml.kernel.org/r/20180601071115.GA27302@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reported-by: kernel test robot <xiaolong.ye@intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a page is freed back to the global pool, its buddy will be checked
to see if it's possible to do a merge. This requires accessing buddy's
page structure and that access could take a long time if it's cache
cold.
This patch adds a prefetch to the to-be-freed page's buddy outside of
zone->lock in hope of accessing buddy's page structure later under
zone->lock will be faster. Since we *always* do buddy merging and check
an order-0 page's buddy to try to merge it when it goes into the main
allocator, the cacheline will always come in, i.e. the prefetched data
will never be unused.
Normally, the number of prefetch will be pcp->batch(default=31 and has
an upper limit of (PAGE_SHIFT * 8)=96 on x86_64) but in the case of
pcp's pages get all drained, it will be pcp->count which has an upper
limit of pcp->high. pcp->high, although has a default value of 186
(pcp->batch=31 * 6), can be changed by user through
/proc/sys/vm/percpu_pagelist_fraction and there is no software upper
limit so could be large, like several thousand. For this reason, only
the first pcp->batch number of page's buddy structure is prefetched to
avoid excessive prefetching.
In the meantime, there are two concerns:
1. the prefetch could potentially evict existing cachelines, especially
for L1D cache since it is not huge
2. there is some additional instruction overhead, namely calculating
buddy pfn twice
For 1, it's hard to say, this microbenchmark though shows good result
but the actual benefit of this patch will be workload/CPU dependant;
For 2, since the calculation is a XOR on two local variables, it's
expected in many cases that cycles spent will be offset by reduced
memory latency later. This is especially true for NUMA machines where
multiple CPUs are contending on zone->lock and the most time consuming
part under zone->lock is the wait of 'struct page' cacheline of the
to-be-freed pages and their buddies.
Test with will-it-scale/page_fault1 full load:
kernel Broadwell(2S) Skylake(2S) Broadwell(4S) Skylake(4S)
v4.16-rc2+ 9034215 7971818 13667135 15677465
patch2/3 9536374 +5.6% 8314710 +4.3% 14070408 +3.0% 16675866 +6.4%
this patch 10180856 +6.8% 8506369 +2.3% 14756865 +4.9% 17325324 +3.9%
Note: this patch's performance improvement percent is against patch2/3.
(Changelog stolen from Dave Hansen and Mel Gorman's comments at
http://lkml.kernel.org/r/148a42d8-8306-2f2f-7f7c-86bc118f8ccd@intel.com)
[aaron.lu@intel.com: use helper function, avoid disordering pages]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180320113146.GB24737@intel.com
[aaron.lu@intel.com: v4]
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Link: http://lkml.kernel.org/r/20180309082431.GB30868@intel.com
Link: http://lkml.kernel.org/r/20180301062845.26038-4-aaron.lu@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Suggested-by: Ying Huang <ying.huang@intel.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__rmqueue(), __rmqueue_fallback(), __rmqueue_smallest() and
__rmqueue_cma_fallback() are all in page allocator's hot path and better
be finished as soon as possible. One way to make them faster is by making
them inline. But as Andrew Morton and Andi Kleen pointed out:
https://lkml.org/lkml/2017/10/10/1252https://lkml.org/lkml/2017/10/10/1279
To make sure they are inlined, we should use __always_inline for them.
With the will-it-scale/page_fault1/process benchmark, when using nr_cpu
processes to stress buddy, the results for will-it-scale.processes with
and without the patch are:
On a 2-sockets Intel-Skylake machine:
compiler base head
gcc-4.4.7 6496131 6911823 +6.4%
gcc-4.9.4 7225110 7731072 +7.0%
gcc-5.4.1 7054224 7688146 +9.0%
gcc-6.2.0 7059794 7651675 +8.4%
On a 4-sockets Intel-Skylake machine:
compiler base head
gcc-4.4.7 13162890 13508193 +2.6%
gcc-4.9.4 14997463 15484353 +3.2%
gcc-5.4.1 14708711 15449805 +5.0%
gcc-6.2.0 14574099 15349204 +5.3%
The above 4 compilers are used because I've done the tests through
Intel's Linux Kernel Performance(LKP) infrastructure and they are the
available compilers there.
The benefit being less on 4 sockets machine is due to the lock
contention there(perf-profile/native_queued_spin_lock_slowpath=81%) is
less severe than on the 2 sockets machine(85%).
What the benchmark does is: it forks nr_cpu processes and then each
process does the following:
1 mmap() 128M anonymous space;
2 writes to each page there to trigger actual page allocation;
3 munmap() it.
in a loop.
https://github.com/antonblanchard/will-it-scale/blob/master/tests/page_fault1.c
Binary size wise, I have locally built them with different compilers:
[aaron@aaronlu obj]$ size */*/mm/page_alloc.o
text data bss dec hex filename
37409 9904 8524 55837 da1d gcc-4.9.4/base/mm/page_alloc.o
38273 9904 8524 56701 dd7d gcc-4.9.4/head/mm/page_alloc.o
37465 9840 8428 55733 d9b5 gcc-5.5.0/base/mm/page_alloc.o
38169 9840 8428 56437 dc75 gcc-5.5.0/head/mm/page_alloc.o
37573 9840 8428 55841 da21 gcc-6.4.0/base/mm/page_alloc.o
38261 9840 8428 56529 dcd1 gcc-6.4.0/head/mm/page_alloc.o
36863 9840 8428 55131 d75b gcc-7.2.0/base/mm/page_alloc.o
37711 9840 8428 55979 daab gcc-7.2.0/head/mm/page_alloc.o
Text size increased about 800 bytes for mm/page_alloc.o.
[aaron@aaronlu obj]$ size */*/vmlinux
text data bss dec hex filename
10342757 5903208 17723392 33969357 20654cd gcc-4.9.4/base/vmlinux
10342757 5903208 17723392 33969357 20654cd gcc-4.9.4/head/vmlinux
10332448 5836608 17715200 33884256 2050860 gcc-5.5.0/base/vmlinux
10332448 5836608 17715200 33884256 2050860 gcc-5.5.0/head/vmlinux
10094546 5836696 17715200 33646442 201676a gcc-6.4.0/base/vmlinux
10094546 5836696 17715200 33646442 201676a gcc-6.4.0/head/vmlinux
10018775 5828732 17715200 33562707 2002053 gcc-7.2.0/base/vmlinux
10018775 5828732 17715200 33562707 2002053 gcc-7.2.0/head/vmlinux
Text size for vmlinux has no change though, probably due to function
alignment.
Link: http://lkml.kernel.org/r/20171013063111.GA26032@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the system has more than one swap device and swap device has the node
information, we can make use of this information to decide which swap
device to use in get_swap_pages() to get better performance.
The current code uses a priority based list, swap_avail_list, to decide
which swap device to use and if multiple swap devices share the same
priority, they are used round robin. This patch changes the previous
single global swap_avail_list into a per-numa-node list, i.e. for each
numa node, it sees its own priority based list of available swap
devices. Swap device's priority can be promoted on its matching node's
swap_avail_list.
The current swap device's priority is set as: user can set a >=0 value,
or the system will pick one starting from -1 then downwards. The
priority value in the swap_avail_list is the negated value of the swap
device's due to plist being sorted from low to high. The new policy
doesn't change the semantics for priority >=0 cases, the previous
starting from -1 then downwards now becomes starting from -2 then
downwards and -1 is reserved as the promoted value.
Take 4-node EX machine as an example, suppose 4 swap devices are
available, each sit on a different node:
swapA on node 0
swapB on node 1
swapC on node 2
swapD on node 3
After they are all swapped on in the sequence of ABCD.
Current behaviour:
their priorities will be:
swapA: -1
swapB: -2
swapC: -3
swapD: -4
And their position in the global swap_avail_list will be:
swapA -> swapB -> swapC -> swapD
prio:1 prio:2 prio:3 prio:4
New behaviour:
their priorities will be(note that -1 is skipped):
swapA: -2
swapB: -3
swapC: -4
swapD: -5
And their positions in the 4 swap_avail_lists[nid] will be:
swap_avail_lists[0]: /* node 0's available swap device list */
swapA -> swapB -> swapC -> swapD
prio:1 prio:3 prio:4 prio:5
swap_avali_lists[1]: /* node 1's available swap device list */
swapB -> swapA -> swapC -> swapD
prio:1 prio:2 prio:4 prio:5
swap_avail_lists[2]: /* node 2's available swap device list */
swapC -> swapA -> swapB -> swapD
prio:1 prio:2 prio:3 prio:5
swap_avail_lists[3]: /* node 3's available swap device list */
swapD -> swapA -> swapB -> swapC
prio:1 prio:2 prio:3 prio:4
To see the effect of the patch, a test that starts N process, each mmap
a region of anonymous memory and then continually write to it at random
position to trigger both swap in and out is used.
On a 2 node Skylake EP machine with 64GiB memory, two 170GB SSD drives
are used as swap devices with each attached to a different node, the
result is:
runtime=30m/processes=32/total test size=128G/each process mmap region=4G
kernel throughput
vanilla 13306
auto-binding 15169 +14%
runtime=30m/processes=64/total test size=128G/each process mmap region=2G
kernel throughput
vanilla 11885
auto-binding 14879 +25%
[aaron.lu@intel.com: v2]
Link: http://lkml.kernel.org/r/20170814053130.GD2369@aaronlu.sh.intel.com
Link: http://lkml.kernel.org/r/20170816024439.GA10925@aaronlu.sh.intel.com
[akpm@linux-foundation.org: use kmalloc_array()]
Link: http://lkml.kernel.org/r/20170814053130.GD2369@aaronlu.sh.intel.com
Link: http://lkml.kernel.org/r/20170816024439.GA10925@aaronlu.sh.intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Cc: "Chen, Tim C" <tim.c.chen@intel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Inter-Processor-Interrupt(IPI) is needed when a page is unmapped and the
process' mm_cpumask() shows the process has ever run on other CPUs. page
migration, page reclaim all need IPIs. The number of IPI needed to send
to different CPUs is especially large for multi-threaded workload since
mm_cpumask() is per process.
For smp_call_function_many(), whenever a CPU queues a CSD to a target
CPU, it will send an IPI to let the target CPU to handle the work.
This isn't necessary - we need only send IPI when queueing a CSD
to an empty call_single_queue.
The reason:
flush_smp_call_function_queue() that is called upon a CPU receiving an
IPI will empty the queue and then handle all of the CSDs there. So if
the target CPU's call_single_queue is not empty, we know that:
i. An IPI for the target CPU has already been sent by 'previous queuers';
ii. flush_smp_call_function_queue() hasn't emptied that CPU's queue yet.
Thus, it's safe for us to just queue our CSD there without sending an
addtional IPI. And for the 'previous queuers', we can limit it to the
first queuer.
To demonstrate the effect of this patch, a multi-thread workload that
spawns 80 threads to equally consume 100G memory is used. This is tested
on a 2 node broadwell-EP which has 44cores/88threads and 32G memory. So
after 32G memory is used up, page reclaiming starts to happen a lot.
With this patch, IPI number dropped 88% and throughput increased about
15% for the above workload.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20170519075331.GE2084@aaronlu.sh.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Linus found there still is a race in mremap after commit 5d1904204c
("mremap: fix race between mremap() and page cleanning").
As described by Linus:
"the issue is that another thread might make the pte be dirty (in the
hardware walker, so no locking of ours will make any difference)
*after* we checked whether it was dirty, but *before* we removed it
from the page tables"
Fix it by moving the check after we removed it from the page table.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Prior to 3.15, there was a race between zap_pte_range() and
page_mkclean() where writes to a page could be lost. Dave Hansen
discovered by inspection that there is a similar race between
move_ptes() and page_mkclean().
We've been able to reproduce the issue by enlarging the race window with
a msleep(), but have not been able to hit it without modifying the code.
So, we think it's a real issue, but is difficult or impossible to hit in
practice.
The zap_pte_range() issue is fixed by commit 1cf35d47712d("mm: split
'tlb_flush_mmu()' into tlb flushing and memory freeing parts"). And
this patch is to fix the race between page_mkclean() and mremap().
Here is one possible way to hit the race: suppose a process mmapped a
file with READ | WRITE and SHARED, it has two threads and they are bound
to 2 different CPUs, e.g. CPU1 and CPU2. mmap returned X, then thread
1 did a write to addr X so that CPU1 now has a writable TLB for addr X
on it. Thread 2 starts mremaping from addr X to Y while thread 1
cleaned the page and then did another write to the old addr X again.
The 2nd write from thread 1 could succeed but the value will get lost.
thread 1 thread 2
(bound to CPU1) (bound to CPU2)
1: write 1 to addr X to get a
writeable TLB on this CPU
2: mremap starts
3: move_ptes emptied PTE for addr X
and setup new PTE for addr Y and
then dropped PTL for X and Y
4: page laundering for N by doing
fadvise FADV_DONTNEED. When done,
pageframe N is deemed clean.
5: *write 2 to addr X
6: tlb flush for addr X
7: munmap (Y, pagesize) to make the
page unmapped
8: fadvise with FADV_DONTNEED again
to kick the page off the pagecache
9: pread the page from file to verify
the value. If 1 is there, it means
we have lost the written 2.
*the write may or may not cause segmentation fault, it depends on
if the TLB is still on the CPU.
Please note that this is only one specific way of how the race could
occur, it didn't mean that the race could only occur in exact the above
config, e.g. more than 2 threads could be involved and fadvise() could
be done in another thread, etc.
For anonymous pages, they could race between mremap() and page reclaim:
THP: a huge PMD is moved by mremap to a new huge PMD, then the new huge
PMD gets unmapped/splitted/pagedout before the flush tlb happened for
the old huge PMD in move_page_tables() and we could still write data to
it. The normal anonymous page has similar situation.
To fix this, check for any dirty PTE in move_ptes()/move_huge_pmd() and
if any, did the flush before dropping the PTL. If we did the flush for
every move_ptes()/move_huge_pmd() call then we do not need to do the
flush in move_pages_tables() for the whole range. But if we didn't, we
still need to do the whole range flush.
Alternatively, we can track which part of the range is flushed in
move_ptes()/move_huge_pmd() and which didn't to avoid flushing the whole
range in move_page_tables(). But that would require multiple tlb
flushes for the different sub-ranges and should be less efficient than
the single whole range flush.
KBuild test on my Sandybridge desktop doesn't show any noticeable change.
v4.9-rc4:
real 5m14.048s
user 32m19.800s
sys 4m50.320s
With this commit:
real 5m13.888s
user 32m19.330s
sys 4m51.200s
Reported-by: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The global zero page is used to satisfy an anonymous read fault. If
THP(Transparent HugePage) is enabled then the global huge zero page is
used. The global huge zero page uses an atomic counter for reference
counting and is allocated/freed dynamically according to its counter
value.
CPU time spent on that counter will greatly increase if there are a lot
of processes doing anonymous read faults. This patch proposes a way to
reduce the access to the global counter so that the CPU load can be
reduced accordingly.
To do this, a new flag of the mm_struct is introduced:
MMF_USED_HUGE_ZERO_PAGE. With this flag, the process only need to touch
the global counter in two cases:
1 The first time it uses the global huge zero page;
2 The time when mm_user of its mm_struct reaches zero.
Note that right now, the huge zero page is eligible to be freed as soon
as its last use goes away. With this patch, the page will not be
eligible to be freed until the exit of the last process from which it
was ever used.
And with the use of mm_user, the kthread is not eligible to use huge
zero page either. Since no kthread is using huge zero page today, there
is no difference after applying this patch. But if that is not desired,
I can change it to when mm_count reaches zero.
Case used for test on Haswell EP:
usemem -n 72 --readonly -j 0x200000 100G
Which spawns 72 processes and each will mmap 100G anonymous space and
then do read only access to that space sequentially with a step of 2MB.
CPU cycles from perf report for base commit:
54.03% usemem [kernel.kallsyms] [k] get_huge_zero_page
CPU cycles from perf report for this commit:
0.11% usemem [kernel.kallsyms] [k] mm_get_huge_zero_page
Performance(throughput) of the workload for base commit: 1784430792
Performance(throughput) of the workload for this commit: 4726928591
164% increase.
Runtime of the workload for base commit: 707592 us
Runtime of the workload for this commit: 303970 us
50% drop.
Link: http://lkml.kernel.org/r/fe51a88f-446a-4622-1363-ad1282d71385@intel.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jerome Marchand <jmarchan@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Ebru Akagunduz <ebru.akagunduz@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since commit:
52aec3308d ("x86/tlb: replace INVALIDATE_TLB_VECTOR by CALL_FUNCTION_VECTOR")
the TLB remote shootdown is done through call function vector. That
commit didn't take care of irq_tlb_count, which a later commit:
fd0f586972 ("x86: Distinguish TLB shootdown interrupts from other functions call interrupts")
... tried to fix.
The fix assumes every increase of irq_tlb_count has a corresponding
increase of irq_call_count. So the irq_call_count is always bigger than
irq_tlb_count and we could substract irq_tlb_count from irq_call_count.
Unfortunately this is not true for the smp_call_function_single() case.
The IPI is only sent if the target CPU's call_single_queue is empty when
adding a csd into it in generic_exec_single. That means if two threads
are both adding flush tlb csds to the same CPU's call_single_queue, only
one IPI is sent. In other words, the irq_call_count is incremented by 1
but irq_tlb_count is incremented by 2. Over time, irq_tlb_count will be
bigger than irq_call_count and the substract will produce a very large
irq_call_count value due to overflow.
Considering that:
1) it's not worth to send more IPIs for the sake of accurate counting of
irq_call_count in generic_exec_single();
2) it's not easy to tell if the call function interrupt is for TLB
shootdown in __smp_call_function_single_interrupt().
Not to exclude TLB shootdown from call function count seems to be the
simplest fix and this patch just does that.
This bug was found by LKP's cyclic performance regression tracking recently
with the vm-scalability test suite. I have bisected to commit:
3dec0ba0be ("mm/rmap: share the i_mmap_rwsem")
This commit didn't do anything wrong but revealed the irq_call_count
problem. IIUC, the commit makes rwc->remap_one in rmap_walk_file
concurrent with multiple threads. When remap_one is try_to_unmap_one(),
then multiple threads could queue flush TLB to the same CPU but only
one IPI will be sent.
Since the commit was added in Linux v3.19, the counting problem only
shows up from v3.19 onwards.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Cc: Alex Shi <alex.shi@linaro.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com>
Link: http://lkml.kernel.org/r/20160811074430.GA18163@aaronlu.sh.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
commit 059500940d (ACPI/video: export acpi_video_get_levels)
mistakenly dropped the correct value of max_level and that caused the
set_level function following failed and the acpi_video backlight interface
didn't get created. Fix this by passing back the correct max_level value.
While at it, also fix the param used in acpi_video_device_lcd_query_levels
where acpi_handle is expected but acpi_video_device is passed.
Fixes: 059500940d (ACPI/video: export acpi_video_get_levels)
Reported-and-tested-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
INT3406 ACPI device object resembles an ACPI video output device, but its
_BCM is said to be deprecated and should not be used. So we will make
use of the raw interface to do the actual cooling.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The acpi_video_get_levels is useful for other drivers, i.e. the
to-be-added int3406 thermal driver, so export it.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since we will need the backlight_device_get_by_type API, we can use it
instead of the backlight_device_registered API whenever necessary so
remove the backlight_device_registered API.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
It is useful to get the backlight device's pointer and use it to set
backlight in some cases(the following patch will make use of it) so add
the two APIs and export them.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Some ACPI node's _STA will touch operation region field, since the
evaluation of _STA in acpi_bus_type_and_status is very early, the
operation region handler is not ready yet. Instead of fail that function
and not creating the acpi_device node consequently, set status to 0 so
that later when the driver for that device is probing, it can find
the acpi_device node and proceed normally. And at that time, the
handler for the operation region is ready and its _STA evaluation will
succeed, its present status can be checked there.
Even there will be no driver using this node later, it doesn't seem
hurt to have one more acpi_device node created with status set to 0.
This happens on Microsoft Surface 3, where the SPI device node NTRG's
_STA touches GPIO fields and the SPI core driver will only enumerate SPI
devices from ACPI if the acpi_device node is 1: created; 2: _STA
indicates it's present.
Note that due to another problem in SPI driver, for NTRG to be actually
enumerated, some changes have to be made in the SPI layer, which is
addressed by Mika(not send out yet):
https://bugzilla.kernel.org/show_bug.cgi?id=104291#c23
Link: https://bugzilla.kernel.org/show_bug.cgi?id=104291
Reported-by: Bastien Nocera <bugzilla@hadess.net>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The firmware of ESPRIMO Mobile M9410 has two video output devices that
have _BCM control method, one is the type of "External Digital Monitor"
while the other is the type of "Internal/Integrated Digital Flat Panel".
Only the 2nd video output device's _BCM control method works, but
since we have created two and the 1st one got picked up by user space,
the backlight functionality is broken. To solve this problem, only
register backlight interface for "Internal/Integrated Digital Flat Panel"
type video output device on this laptop.
Another problem of this laptop is that the IDs listed by the _DOD method
doesn't have bit 31 set, which means it doesn't follow the format
specified by ACPI spec. But the value indicates that it actually follows
that format so I've added a DMI quirk and a module level parameter to
force use the device_id_scheme so that we can get the video output
device's type to do the decision if we should register backlight
interface.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=104121
Suggested-by: Hans de Goede <hdegoede@redhat.com>
Reported-and-tested-by: Christian Scharl <zahlsum-kernelbugs@yahoo.de>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The CrystalCove GPIO irqchip doesn't have irq_set_wake callback defined
so we should set IRQCHIP_SKIP_SET_WAKE for it or it would cause an irq
desc's wake_depth unbalanced warning during system resume phase from the
gpio_keys driver, which is the driver for the power button of the ASUS
T100 laptop.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The PCI "ACPI additions for FW latency optimizations" ECN (link below)
defines two functions in the PCI _DSM:
Function 8, "Reset Delay," applies to the entire hierarchy below a PCI
host bridge. If it returns one, the OS may assume that all devices in
the hierarchy have already completed power-on reset delays.
Function 9, "Device Readiness Durations," applies only to the object
where it is located. It returns delay durations required after various
events if the device requires less time than the spec requires. Delays
from this function take precedence over the Reset Delay function.
Add support for Reset Delay and part of Device Readiness Durations.
[bhelgaas: changelog, comments]
Link: https://www.pcisig.com/specifications/conventional/pci_firmware/ECN_fw_latency_optimization_final.pdf
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The find_pci_host_bridge() function can be useful to other PCI code so
export it. Change its name to pci_find_host_bridge().
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The PCI Firmware Specification, r3.0, sec 4.6.4.1.3, defines a single UUID
for an ACPI _DSM method to provide device-specific control functions. This
_DSM method support several functions, including PCI Express Slot
Information, PCI Express Slot Number, PCI Bus Capabilities, etc.
Move the UUID definition from pci/pci-label.c, where it could be used only
for one function, to pci/pci-acpi.c where it can be shared for all these
functions.
[bhelgaas: changelog]
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The native backlight behavior (so not registering both the acpi-video
and the vendor backlight driver) can be useful on some non win8 machines
too, so change the behavior of the video.use_native_backlight=1 or 0
kernel cmdline option to be: if user has set video.use_native_backlight=1
or 0, use that no matter if it is a win8 system or not. Also, we will
put some known systems into the DMI table to make them either use native
backlight interface or not, and the use_native_backlight_dmi is used to
reflect that.
Original-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The fixed event handler should return a value that is either 0 or 1
meanning if the event is handled or not, instead of an acpi_status to
mean if the handler runs well or not.
Suggested-by: Lv Zheng <lv.zheng@intel.com>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If the firmware has declared more than 8 video output devices, and the
one that control the internal panel's backlight is listed after the
first 8 output devices, the _DOD will not include it due to the current
i915 operation region implementation. As a result, we will not create a
backlight device for it while we should. Solve this problem by special
case the firmware that has 8+ output devices in that if we see such a
firmware, we do not test if the device is in _DOD list. The creation of
the backlight device will also enable the firmware to emit events on
backlight hotkey press when the acpi_osi= cmdline option is specified on
those affected ASUS laptops.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=70241
Reported-and-tested-by: Oleksij Rempel <linux@rempel-privat.de>
Reported-and-tested-by: Dmitry Tunin <hanipouspilot@gmail.com>
Reported-and-tested-by: Jimbo <jaime.91@hotmail.es>
Cc: 3.18+ <stable@vger.kernel.org> # 3.18+
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 0b8db271f1 ("ACPI / video: check _DOD list when creating
backlight devices") checks if the video device is in the bind devices
list to decide if we should create backlight device for it, that causes
problem for one Dell Latitude E6410, where none of the video output
devices are properly bound due to the way how we did the comparing
between its _ADR and the _DOD's values. Solve this problem by comparing
the lower 12 bits of both the device's _ADR and the _DOD's values instead
of relying on bind result.
Fixes: 0b8db271f1 ("ACPI / video: check _DOD list when creating backlight devices")
Reported-and-tested-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The same virtual GPIO strategy is also used for the AXP288 PMIC in that
various control methods that are used to do power rail handling and
sensor reading/setting will touch GPIO fields defined under the PMIC
device. The GPIO fileds are only defined by the ACPI code while the
actual hardware doesn't really have a GPIO controller, but to make those
control method execution succeed, we have to install a GPIO handler for
the PMIC device handle. Since we do not need the virtual GPIO strategy,
we can simply do nothing in that handler.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The Baytrail-T-CR platform firmware has defined two customized operation
regions for PMIC chip Dollar Cove XPower - one is for power resource
handling and one is for thermal just like the CrystalCove one. This patch
adds support for them on top of the common PMIC opregion region code.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Lee Jones <lee.jones@linaro.org> for the MFD part
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The Baytrail-T platform firmware has defined two customized operation
regions for PMIC chip Crystal Cove - one is for power resource handling
and one is for thermal: sensor temperature reporting, trip point setting,
etc. This patch adds support for them on top of the existing Crystal Cove
PMIC driver.
The reason to split code into a separate file intel_pmic.c is that there
are more PMIC drivers with ACPI operation region support coming and we can
re-use those code. The intel_pmic_opregion_data structure is created also
for this purpose: when we need to support a new PMIC's operation region,
we just need to fill those callbacks and the two register mapping tables.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Lee Jones <lee.jones@linaro.org> for the MFD part
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The ASUS K53SM's ACPI table queries _OSI("Windows 2012") in the video
output device's _BCL and _BCM control method instead of the usual _INI
functions of the _SB or PCI host bridge PCI0 devices. This made our
video module thought this is a pre-Win8 system when deciding if we
should register a backlight interface for it and the end result is that
a non-working acpi_video interface is registered and user is unable to
control backlight from GUI. Solve this problem by evaluating _BCL control
method before doing the decision.
Note that for some Thinkpad systems, the _BCL is also required to be
evaluated for the hotkey event to be generated no matter if we will
register an ACPI video backlight interface for it or not. Since the
thinkpad_acpi module will do this anyway we didn't add such a thing in
the video module previously. But now with this change here, the
thinkpad_acpi module is no more necessary for those systems regarding
backlight functionality.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=85051
Reported-and-tested-by: Ralf Jung <post+kernel@ralfj.de>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
ACPI INT3402 device object could report temperature for the memory module.
To expose such information to user space, a thermal zone device is registered
for it so that the thermal sysfs interface can expose such information for
userspace to use.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
This macro can be used by other component so move it to a common header,
but in a slightly different way: define two macros, one macro with an
offset and the other doesn't.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
INT3404 ACPI object follows the ACPI 5.0 fan object definition as
described in section 11.3 of the ACPI 5.0 Specification.
Thus we can reuse the ACPI fan driver for INT3404 ACPI object.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
This patch adds support for ACPI 4.0 style fan, lacking part is: no
support for 'Low Speed Notification Support', 'Fine Grain Control' is
not used yet.
It's not clear what to do on suspend/resume callback for 4.0 style ACPI
fan, so it does nothing for now.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Convert ACPI fan driver to a platform driver for the purpose of phasing
out ACPI bus.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
When we have the acpi_device pointer, there is no need to pass the
device's handle to the acpi_bus_xxx_power functions to get/set/update
the device's power state, instead, use the acpi_device_xxx_power
functions directly.
To make this happen for fan module, export acpi_device_update_power.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
The device pointer will not be NULL in the PM callback and ACPI driver's
add/remove callback, so checking NULL for them isn't necessary.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
The _COMPONENT, ACPI_MODULE_NAME(name) and ACPI_FAN_FILE_STATE are not
used anywhere so remove them.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
The _DOD method lists which video output device is currently attached so
we should only care about them and ignore others. An user recently
reported that there are two acpi_video interfaces appeared on his system
and one of them doesn't work. From the acpidump, it is found that there
are more than one video output devices that have _BCM control method but
the _DOD lists only one of them. So this patch checks if the video output
device is in the _DOD list and will not create backlight device if it is
not in the list. Also, we consider the broken _DOD case(reflected by the
video->attached_count is 0) and do not change behaviour for those broken
_DOD systems.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=84111
Reported-and-tested-by: ntrrgc@gmail.com
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Some of the Thinkpads' firmware will issue a backlight change request
through i915 operation region unconditionally on AC plug/unplug, the
backlight level used is arbitrary and thus should be ignored. This is
handled by commit 0b9f7d93ca (ACPI / i915: ignore firmware requests
for backlight change). Then there is a Dell laptop whose vendor backlight
interface also makes use of operation region to change backlight level
and with the above commit, that interface no long works. The condition
used to ignore the backlight change request from firmware is thus
changed to: if the vendor backlight interface is not in use and the ACPI
backlight interface is broken, we ignore the requests; oterwise, we keep
processing them.
Fixes: 0b9f7d93ca (ACPI / i915: ignore firmware requests for backlight change)
Link: https://lkml.org/lkml/2014/9/23/854
Reported-and-tested-by: Pali Rohár <pali.rohar@gmail.com>
Cc: 3.16+ <stable@vger.kernel.org> # 3.16+
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The virtual GPIO introduced in ACPI table of Baytrail-T based system is
used to solve a problem under Windows. We do not have such problems
under Linux so we do not actually need them. But we have to tell GPIO
library that the Crystal Cove GPIO chip has this many GPIO pins or the
common GPIO handler will refuse any access to those high number GPIO
pins, which will resulted in a failure evaluation of every ACPI control
method that is used to turn on/off power resource and/or report sensor
temperatures.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
[changed vgpio number from 0x5e to 94]
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
On 05/21/2014 04:22 PM, Aaron Lu wrote:
> On 05/21/2014 01:57 PM, Kui Zhang wrote:
>> Hello,
>>
>> I get following error when rmmod thermal.
>>
>> rmmod thermal
>> Killed
While dealing with this problem, I found another problem that also
results in a kernel crash on thermal module removal:
From: Aaron Lu <aaron.lu@intel.com>
Date: Wed, 21 May 2014 16:05:38 +0800
Subject: [PATCH] thermal: hwmon: Make the check for critical temp valid consistent
We used the tz->ops->get_crit_temp && !tz->ops->get_crit_temp(tz, temp)
to decide if we need to create the temp_crit attribute file but we just
check if tz->ops->get_crit_temp exists to decide if we need to remove
that attribute file. Some ACPI thermal zone doesn't have a valid critical
trip point and that would result in removing a non-existent device file
on thermal module unload.
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
The GPIO operation region handler should be called where sleep is
allowed, so we should use the *_cansleep version of gpiod_get/set APIs
or we will get a warning message complaining invalid context if the GPIO
chip has the cansleep flag set.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
When we set backlight on behalf of ACPI opregion, we will convert the
backlight value in the 0-255 range defined in opregion to the actual
hardware level. Commit 22505b82a2 (drm/i915: avoid brightness overflow
when doing scale) is meant to fix the overflow problem when doing the
conversion, but it also caused a problem that the converted hardware
level doesn't quite represent the intended value: say user wants maximum
backlight level(255 in opregion's range), then we will calculate the
actual hardware level to be: level = freq / max * level, where freq is
the hardware's max backlight level(937 on an user's box), and max and
level are all 255. The converted value should be 937 but the above
calculation will yield 765.
To fix this issue, just use 64 bits to do the calculation to keep the
precision and avoid overflow at the same time.
Buglink: https://bugzilla.kernel.org/show_bug.cgi?id=72491
Reported-by: Nico Schottelius <nico-bugzilla.kernel.org@schottelius.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Fixed hardware does not exist on HW-reduced ACPI platforms since the
programming interface for them is not implemented on them, so no need
to scan that hardware on them.
This patch avoids creating the fixed power button ACPI device and
eliminates a probe error message from ACPI button driver on ASUS T100.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Now that we can directly get the ACPI device conterpart of the physical
ATA transport device, the odd_can_poweroff can be eliminated.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
If ACPI handle for an ATA device is NULL, we shouldn't call
ata_dev_get_GTF as that function will use handle to do some ACPI
evaluation.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
ZPODD is built on top of runtime PM functionality, it doesn't make sense
to have it in a kernel that doesn't have CONFIG_PM_RUNTIME set.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
On resume we do not need to wait for acpi_thermal_check to finish.
Instead, we can run it asynchronously and not block the whole system
resume. Also, we make sure when we are suspending again, previously
queued work for acpi_thermal_check is done.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Migrate sr to make use of block layer runtime PM. Accordingly, the
SCSI bus layer runtime PM callback is simplified as all SCSI drivers
implementing runtime PM now use the block layer's request-based
mechanism.
Note that due to the device will be polled by kernel at a constant
interval, if the autosuspend delay is set longer than the polling
interval then the device will never suspend.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
With commit 2c62333a40 "ACPI / video: Quirk initial backlight level 0"
we do not need to have the following systems in DMI table, so remove them.
HP Pavilion m4, HP 1000 Notebook PC, HP Pavilion g6 Notebook PC,
HP Pavilion dm4, Fujitsu E753, HP Folio 13-2000.
With this change, the use_bios_initial_backlight module parameter is no
longer needed and thus removed.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Tested-by: Alex Hung <alex.hung@canonical.com> # for HP 1000 Notebook PC
Tested-by: Gustavo Maciel Dias Vieira <gustavo@sagui.org> # for HP Pavilion dm4
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When testing if the firmware's initial value is valid, we should use
the corrected level value instead of the raw value returned from
firmware.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When system has a lot of highmem (e.g. 16GiB using a 32 bits kernel),
the code to calculate how much memory we need to preallocate in
normal zone may cause overflow. As Leon has analysed:
It looks that during computing 'alloc' variable there is overflow:
alloc = (3943404 - 1970542) - 1978280 = -5418 (signed)
And this function goes to err_out.
Fix this by avoiding that overflow.
References: https://bugzilla.kernel.org/show_bug.cgi?id=60817
Reported-and-tested-by: Leon Drugi <eyak@wp.pl>
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Sujit has found a race condition that would make q->nr_pending
unbalanced, it occurs as Sujit explained:
"
sd_probe_async() ->
add_disk() ->
disk_add_event() ->
schedule(disk_events_workfn)
sd_revalidate_disk()
blk_pm_runtime_init()
return;
Let's say the disk_events_workfn() calls sd_check_events() which tries
to send test_unit_ready() and because of sd_revalidate_disk() trying to
send another commands the test_unit_ready() might be re-queued as the
tagged command queuing is disabled.
So the race condition is -
Thread 1 | Thread 2
sd_revalidate_disk() | sd_check_events()
...nr_pending = 0 as q->dev = NULL| scsi_queue_insert()
blk_runtime_pm_init() | blk_pm_requeue_request() ->
| nr_pending = -1 since
| q->dev != NULL
"
The problem is, the test_unit_ready request doesn't get counted the
first time it is queued, so the later decrement of q->nr_pending in
blk_pm_requeue_request makes it unbalanced.
Fix this by calling blk_pm_runtime_init before add_disk so that all
requests initiated there will all be counted.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reported-and-tested-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
Cc: stable@vger.kernel.org
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Previously, we wanted SCSI devices corrsponding to ATA devices to
be runtime resumed when the power resource for those ATA device was
turned on by some other device, so we added the SCSI device to the
dependent device list of the ATA device's ACPI node. However, this
code has no effect after commit 41863fc (ACPI / power: Drop automaitc
resume of power resource dependent devices) and the mechanism it was
supposed to implement is regarded as a bad idea now, so drop it.
[rjw: Changelog]
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The tpacpi_acpi_handle_locate function makes use of acpi_get_devices to
locate handle for ACPI video by HID, the problem is, ACPI video node
doesn't really have HID defined(i.e. no _HID control method is defined
for video device), so.. that function would fail. This can be solved by
enhancing the callback function for acpi_get_devices, where we can use
acpi_device_hid function to check if the ACPI node corresponds to a
video controller.
In addition to that, the _BCL control method only exists under a video
output device node, not a video controller device node. So to evaluate
_BCL, we need the handle of a video output device node, which is child
of the located video controller node from tpacpi_acpi_handle_locate.
The two fix are necessary for some Thinkpad models to emit notification
on backlight hotkey press as a result of evaluation of _BCL.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Tested-by: Igor Gnatenko <i.gnatenko.brain@gmail.com>
Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
According to Matthew Garrett, "Windows 8 leaves backlight control up
to individual graphics drivers rather than making ACPI calls itself.
There's plenty of evidence to suggest that the Intel driver for
Windows [8] doesn't use the ACPI interface, including the fact that
it's broken on a bunch of machines when the OS claims to support
Windows 8. The simplest thing to do appears to be to disable the
ACPI backlight interface on these systems".
So for Win8 systems, if there is native backlight control interface
registered by GPU driver, ACPI video does not need to register its own.
Since there are systems that don't work well with this approach, a
parameter for video module named use_native_backlight is introduced and
has the value of false by default. For users who have a broken ACPI
video backlight interface, video.use_native_backlight=1 is needed in
kernel cmdline.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The backlight control and event delivery functionality provided by ACPI
video module is mixed together and registered all during video device
enumeration time. As a result, the two functionality are also removed
together on module unload time or by the acpi_video_unregister function.
The two functionalities are actually independent and one may be useful
while the other one may be broken, so it is desirable to seperate the
two functionalities such that it is clear and easy to disable one
functionality without affecting the other one.
APIs to selectively remove backlight control interface and/or event
delivery functionality can be easily added once needed.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Introduce a new API for modules to query if a specific type of backlight
device has been registered. This is useful for some backlight device
provider module (e.g. ACPI video) to know if a native control
interface(e.g. the interface created by i915) is available and then do
things accordingly (e.g. avoid registering its own on Win8 systems).
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The freeze and restore functions defined in virtio drivers are used
for suspend and hibernate, so CONFIG_PM_SLEEP is more appropriate than
CONFIG_PM. This patch replace all CONFIG_PM with CONFIG_PM_SLEEP for
virtio drivers that implement freeze and restore callbacks.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The virtio_pci_freeze/restore are defined under CONFIG_PM but is used
by SET_SYSTEM_SLEEP_PM_OPS macro, which is defined under
CONFIG_PM_SLEEP. So if CONFIG_PM_SLEEP is not cofigured but
CONFIG_PM_RUNTIME is, the following warning message appeared:
drivers/virtio/virtio_pci.c:770:12: warning: ‘virtio_pci_freeze’ defined but not used [-Wunused-function]
static int virtio_pci_freeze(struct device *dev)
^
drivers/virtio/virtio_pci.c:790:12: warning: ‘virtio_pci_restore’ defined but not used [-Wunused-function]
static int virtio_pci_restore(struct device *dev)
^
Fix it by changing CONFIG_PM to CONFIG_PM_SLEEP.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The ata_dev_acpi_handle is defined in libata-acpi.c and the only
external user is libata-zpodd.c, which is built when CONFIG_ATA_ACPI
is set, so there is no need to make an empty stub function for
ONFIG_ATA_ACPI case in libata.h. It also causes compile errors due to
acpi_handle is not defined when !CONFIG_ACPI. This patch fixes this
problem by removing the empty stub.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Binding ACPI handle to SCSI device has several drawbacks, namely:
1 During ATA device initialization time, ACPI handle will be needed
while SCSI devices are not created yet. So each time ACPI handle is
needed, instead of retrieving the handle by ACPI_HANDLE macro,
a namespace scan is performed to find the handle for the corresponding
ATA device. This is inefficient, and also expose a restriction on
calling path not holding any lock.
2 The binding to SCSI device tree makes code complex, while at the same
time doesn't bring us any benefit. All ACPI handlings are still done
in ATA module, not in SCSI.
Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
devices(ATA port and ATA device). The binding needs to be done only once,
since the ATA transport devices do not go away with hotplug. And due to
this, the flush_work call in hotplug handler for ATA bay is no longer
needed.
Tested on an Intel test platform for binding and runtime power off for
ODD(ZPODD) and hard disk; on an ASUS S400C for binding and normal boot
and S3, where its SATA port node has _SDD and _GTF control methods when
configured as an AHCI controller and its PATA device node has _GTF
control method when configured as an IDE controller. SATA PMP binding
and ATA hotplug is not tested.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Tested-by: Dirk Griesbach <spamthis@freenet.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Some card's max brightness level is pretty large, e.g. on Acer Aspire
4732Z, the max level is 989910. If user space set a large enough level
then the current scale done in intel_panel_set_backlight will cause an
integer overflow and the scaled level will be mistakenly small, leaving
user with an almost black screen. This patch fixes this problem.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
[danvet: Add a comment to explain what's going on.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The state information can be useful to know what the problem is when
an error message about a device can not being set to a higher power
state than its parent appeared, so this patch adds such state
information for both the target state of the device and the current
state of its parent.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Now that acpi_device_set_power() checks whether or not the given
device is power manageable, it is not necessary to do this check in
acpi_bus_set_power() any more, so remove it.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 7381fe737 "libata-acpi: remove redundent code for power resource
handling" removed ata_acpi_(un)bind but left their prototypes in libata.h,
so remove them.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Expose acpi_gbl_osi_data so that code outside of ACPICA can check
the value of the last successfull _OSI call. The definitions for
OSI versions are moved to actypes.h so that other components can
access them too.
Based on a patch from Matthew Garrett which in turn was based on
an earlier patch from Seth Forshee.
[rjw: Changelog]
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 30dcf76acc "libata: migrate ACPI code over to new bindings"
mistakenly dropped the code to register hotplug notificaion handler
for ATA port/devices, causing regression for people using ATA bay,
as kernel bug #59871 shows.
Fix this by adding back the hotplug notification handler registration
code. Since this code has to be run once and notification needs to
be installed on every ATA port/devices handle no matter if there is
actual device attached, we can't do this in binding time for ATA
device ACPI handle, as the binding only occurs when a SCSI device is
created, i.e. there is device attached. So introduce the
ata_acpi_hotplug_init() function to loop scan all ATA ACPI handles
and if it is available, install the notificaion handler for it during
ATA init time.
With the ATA ACPI handle binding to SCSI device tree, it is possible
now that when the SCSI hotplug work removes the SCSI device, the ACPI
unbind function will find that the corresponding ACPI device has
already been deleted by dock driver, causing a scaring message like:
[ 128.263966] scsi 4:0:0:0: Oops, 'acpi_handle' corrupt
Fix this by waiting for SCSI hotplug task finish in our notificaion
handler, so that the removal of ACPI device done in ACPI unbind
function triggered by the removal of SCSI device is run earlier when
ACPI device is still available.
[rjw: Rebased]
References: https://bugzilla.kernel.org/show_bug.cgi?id=59871
Reported-bisected-and-tested-by: Dirk Griesbach <spamthis@freenet.de>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: 3.6+ <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The ACPI video driver has changed a lot, and it doesn't export
interfaces in /proc any more, so the documentation for it should
be updated.
This update focuses on ACPI video driver's backlight control.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
ACPI video driver is written according to ACPI spec, appendix B: Video
Extensions. So it better be put under the acpi directory instead of the
power directory. This patch moves the file there without any other
change.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add description for video module's parameter brightness_switch_enabled
into kernel-parameters.txt.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
With the introduction of ACPI scan handlers, an ACPI device object
with an ACPI scan handler attached to it must not be bound to an ACPI
driver any more. Therefore it doesn't make sense to match those
ACPI device objects against a newly registered ACPI driver in
acpi_bus_match(), so make that function return 0 if the device
object passed to it has an ACPI scan handler attached.
This also addresses a regression related to a broken ACPI table in
the BIOS, where it has defined a _ROM method under the PCI root
bridge object. This causes the video module to treat that object
as a display controller device (since only display devices are
supposed to have a _ROM method defined according to the ACPI spec).
As a result, the ACPI video driver binds to the PCI root bridge
object and overwrites the previously assigned driver_data field of
it, causing subsequent calls to acpi_get_pci_dev() to fail.
[rjw: Subject and changelog]
References: https://bugzilla.kernel.org/show_bug.cgi?id=58091
Reported-by: Jason Cassell <bluesloth600@gmail.com>
Reported-and-bisected-by: Dmitry S. Demin <dmitryy.demin@gmail.com>
Cc: 3.9+ <stable@kernel.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In blk_post_runtime_resume, an autosuspend request will be initiated for
the device. Since we are holding the queue lock, we can't sleep and thus
we should use the async version to initiate an autosuspend, i.e.
pm_request_suspend instead of pm_runtime_suspend, which might sleep.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Since commit 30dcf76acc, ata_ap_acpi_handle will always do a namespace
walk, which requires acquiring an acpi namespace mutex. This made it
impossible to be used when calling path has held a spinlock.
For example, it can occur in the following code path for pata_acpi:
ata_scsi_queuecmd (ap->lock is acquired)
__ata_scsi_queuecmd
ata_scsi_translate
ata_qc_issue
pacpi_qc_issue
ata_acpi_stm
ata_ap_acpi_handle
acpi_get_child
acpi_walk_namespace
acpi_ut_acquire_mutex (acquire mutex while holding lock)
This caused scheduling while atomic bug, as reported in bug #56781.
Actually, ata_ap_acpi_handle doesn't have to walk the namespace every
time it is called, it can simply return the bound acpi handle on the
corresponding SCSI host. The reason previously it is not done this way
is, ata_ap_acpi_handle is used in the binding function
ata_acpi_bind_host by ata_acpi_gtm when the handle is not bound to the
SCSI host yet. Since we already have the ATA port's handle in its
binding function, we can simply use it instead of calling
ata_ap_acpi_handle there. So introduce a new function __ata_acpi_gtm,
where it will receive an acpi handle param in addition to the ATA port
which is solely used for debug statement. With this change, we can make
ata_ap_acpi_handle simply return the bound handle for SCSI host instead
of walking the acpi namespace now.
Buglink: https://bugzilla.kernel.org/show_bug.cgi?id=56781
Reported-and-tested-by: <kenzopl@o2.pl>
Cc: <stable@vger.kernel.org>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
acpi_video_bus_get_devices() may fail due to some video output device
doesn't have the _ADR method, and in this case, the error processing
is to simply free the video structure in acpi_video_bus_add(), while
leaving those already registered video output devices in the wild,
which means for some video output device, we have already registered
a backlight interface and installed a notification handler for it.
So it can happen when user is using this system, on hotkey pressing,
the notification handler will send a keycode through a non-existing
input device, causing kernel freeze.
To solve this problem, free all those already registered video output
devices once something goes wrong in acpi_video_bus_get_devices(), so
that no wild backlight interfaces and notification handlers exist.
References: https://bugzilla.kernel.org/show_bug.cgi?id=51731
Reported-and-tested-by: <i-tek@web.de>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Currently we decide if the _BQC is using index by first setting the
level to maximum, and then check if _BQC returned maximum; if not, we
say it is using index.
This is not true for some buggy systems, where the _BQC method will
always return a constant value(e.g. 0 or 100 for the two broken system)
and thus break the current logic. So this patch tries to enhance the
quirk detect logic for _BQC: we do this by picking a test_level, it can
be the maximum level or the mininum one based on some condition. And we
don't make the assumption that if _BQC returned a value that is not what
we just set, it must be using an index. Instead, we will compare the
value returned from _BQC and if it doesn't match, see if the returned
value is an index. And if still no, clear the capability of _BQC.
References: https://bugzilla.kernel.org/show_bug.cgi?id=42861
References: https://bugzilla.kernel.org/show_bug.cgi?id=56011
Reported-and-tested-by: Artem Savkov <artem.savkov@gmail.com>
Reported-by: Luis Medinas <lmedinas@gmail.com>
Reported-by: Cheppes <cheppes@mailinator.com>
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Acked-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
With commit:
bc9b6407bd
ACPI / PM: Rework the handling of devices depending on power resources
The ACPI core now takes care of the power resources an acpi device
depends on in that when the power resources are turned on, any devices
that are bound to or in the dependent list of this acpi device will be
runtime resumed. So there is no need for ata acpi code to duplicate this
effort, and thus, the ata_acpi_(un)register_power_resource functions are
no longer needed.
The above commit thinks the scsi device is not bound to the acpi device,
so needs to be added to the dependent list. But actually, it is. So
there is no need to add it to the dependent list, or it will be runtime
resumed twice(though this wouldn't cause any problem).
This patch fixes it, and as a result, the
ata_acpi_(un)register_power_resource and ata_acpi_(un)bind functions
are removed.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
ACPI spec 5 defined the _ADR encoding for sdio bus as:
High word - slot number (0 based)
Low word - function number
This patch adds support for binding sdio function device with acpi node,
and if successful, involve acpi into its power management.
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
Aaron Lu