Commit a76d7bd96d ("ARM: 7467/1: mutex: use generic xchg-based
implementation for ARMv6+") removed the barrier-less, ARM-specific
mutex implementation in favour of the generic xchg-based code.
Since then, a bug was uncovered in the xchg code when running on SMP
platforms, due to interactions between the locking paths and the
MUTEX_SPIN_ON_OWNER code. This was fixed in 0bce9c46bf ("mutex: place
lock in contended state after fastpath_lock failure"), however, the
atomic_dec-based mutex algorithm is now marginally more efficient for
ARM (~0.5% improvement in hackbench scores on dual A15).
This patch moves ARMv6+ platforms to the atomic_dec-based mutex code.
Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
ARM recently moved to asm-generic/mutex-xchg.h for its mutex
implementation after the previous implementation was found to be missing
some crucial memory barriers. However, this has revealed some problems
running hackbench on SMP platforms due to the way in which the
MUTEX_SPIN_ON_OWNER code operates.
The symptoms are that a bunch of hackbench tasks are left waiting on an
unlocked mutex and therefore never get woken up to claim it. This boils
down to the following sequence of events:
Task A Task B Task C Lock value
0 1
1 lock() 0
2 lock() 0
3 spin(A) 0
4 unlock() 1
5 lock() 0
6 cmpxchg(1,0) 0
7 contended() -1
8 lock() 0
9 spin(C) 0
10 unlock() 1
11 cmpxchg(1,0) 0
12 unlock() 1
At this point, the lock is unlocked, but Task B is in an uninterruptible
sleep with nobody to wake it up.
This patch fixes the problem by ensuring we put the lock into the
contended state if we fail to acquire it on the fastpath, ensuring that
any blocked waiters are woken up when the mutex is released.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Chris Mason <chris.mason@fusionio.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: <stable@vger.kernel.org>
Reviewed-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-6e9lrw2avczr0617fzl5vqb8@git.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The open-coded mutex implementation for ARMv6+ cores suffers from a
severe lack of barriers, so in the uncontended case we don't actually
protect any accesses performed during the critical section.
Furthermore, the code is largely a duplication of the ARMv6+ atomic_dec
code but optimised to remove a branch instruction, as the mutex fastpath
was previously inlined. Now that this is executed out-of-line, we can
reuse the atomic access code for the locking (in fact, we use the xchg
code as this produces shorter critical sections).
This patch uses the generic xchg based implementation for mutexes on
ARMv6+, which introduces barriers to the lock/unlock operations and also
has the benefit of removing a fair amount of inline assembly code.
Cc: <stable@vger.kernel.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Nicolas Pitre <nico@linaro.org>
Reported-by: Shan Kang <kangshan0910@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
If lock refcount is negative, lock will be always failed. To fix this
issue, do not set the lock refcount as negative value.
Change-Id: Ic89f393b60a0ff15313e91566a5055f620752452
Synopsys USB3 Controller (DWC3) has a restriction where size
of OUT requests (TRB) queued to the controller must be aligned
with the endpoint's max packet size.
Since, request buffer is already allocated with the size of
ADB_BULK_BUFFER_SIZE, hence, request this much data from the
controller.
Change-Id: Ibba1a627c9e34b92ddcb44ee2260a004a2b78dfc
Signed-off-by: Ido Shayevitz <idos@codeaurora.org>
Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
It is possible to use the wakelock before initializing wakelock.
In this case, wakelock error occurs. To avoid this issue, early initialize
the wakelock in init()
Change-Id: I8fb5450005efa6dbcee6c27b5a3fa551a0adb4dc
If you use the parameter 'maxcpus=n', restrict boot time cpus to n.
To initialize the ETM on cpu, the cpus should be online. So if you restrict
boot time cpus, you can not enable the ETM on cpu which is rectriced(not
enabled). Skip the initialization of the ETM.
Change-Id: Idccf537ff4dce0ae5d1b60ecc1631180ad323fb8
When being completely charged in the offmode charging and pulling out TA
from sleep condition, I gave 1 sec. delay until entering sleep condition again,
in order to prevent the phenomenon that a device doesn't turn off.
We need sometime for an userspace application to handle the usb disconnection
events. There is no effect of power consumption.
Change-Id: I0851cc2b47c3780fd68d3e8965fd0d6f5b7e8312
In off-mode charging, hsic host driver never release wake_lock.
Due to this, system cannot enter the suspend during off-mode
charging. This patch disable the hsic host driver and mdm
driver in off-mode charging to reduce power consumption.
Change-Id: I8580af1d5a9d754982b198c9bd47a98e4b69c725
The interrupts added to the ringbuffer on PTFLUSH and TLBUPDATE
were causing a major increase in the number of interrupts from the GPU.
This was leading to increase in power and loss of performance. Add a check
to turn off IOMMU clocks when going to SLEEP.
Change-Id: I41617dd3b7b3f7d9622523f2a1407b912dbd989e
Signed-off-by: Shubhraprakash Das <sadas@codeaurora.org>
Signed-off-by: Jeff Boody <jboody@codeaurora.org>
Remove ts_notifier_list from the driver since it is not being
used and is causing extra work to be done in the interrupt
handlers for A2XX, A3XX and Z180.
Change-Id: I5512e36f1e807f3a3e62aeac54cfd3075d4cf7a4
Signed-off-by: Carter Cooper <ccooper@codeaurora.org>
Signed-off-by: Jeff Boody <jboody@codeaurora.org>
This reverts commit de86e6482897db0006151ae9154f181e75f5ab2c.
Commit 7c8527ce ("PM: Prevent runtime suspend during system resume")
fixes the general issue of a parent device being prematurely
runtime suspended before a child device is resumed. It does so
by moving the pm_runtime_put_sync() call from the device_resume()
stage to device_complete(), which happens to act on the devices
in reverse (children before parent) order.
Thus when usb_resume() is called by device_resume(), we know its
parent will be runtime PM enabled and its usage counter is still
non-zero. This obviates the need to do an additional
pm_runtime_get_sync() before calling usb_resume_both() as the
device and its interfaces now have the opportunity to increment
the parent's power.child_count if necessary without fear of
having a runtime suspended parent.
CRs-fixed: 396041
Change-Id: I793ddfbb9bd88634f45f0c1e570a5d616248e3a8
Signed-off-by: Jack Pham <jackp@codeaurora.org>
git://codeaurora.org/external/wlan/prima.git
8cfd26d wlan: Release 3.2.1.13
148c518 Update csrRoamInitGlobals() to support static linking
fff59a1 Send EDCA Parameters After Reassoc
c7f1b8f Populate Correct Scan Mode for Roam
c9990e6 Fix driver hang in csrRoamRoamingStateDisassocRspProcessor
d1ef85b Remove call to palStopTimer for preauth
a31968f Insure that LFR BG scanning is not performed for any non-Infra association
96a9106 Only flush non-P2P results from scan cache prior to LFR scan
ac37d9d De-register lookup UP indication during pre-auth
def1bff Improve convergence of LFR roam candidate scan algorithm
3774f3b Dump neighbor state info when scan is not allowed
524cf61 wlan: Release 3.2.1.12
70e619c Conditionally invoke driver restart API
556271d Notify the Kernel suspend to WCNSS
8aef3d0 Update restart driver API to support statically linked WLAN driver
f9f6457 wlan: Release 3.2.1.11i
d5b0e74 Post proper response to SME when reassoc is rejected
e6c5a60 wlan: Release 3.2.1.11i
bd2182f Post proper response to SME when reassoc is rejected
Signed-off-by: Iliyan Malchev <malchev@google.com>
commit 9ab4233dd0 upstream.
Otherwise the code races with munmap (causing a use-after-free
of the vma) or with close (causing a use-after-free of the struct
file).
The bug was introduced by commit 90ed52ebe4 ("[PATCH] holepunch: fix
mmap_sem i_mutex deadlock")
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Miklos Szeredi <mszeredi@suse.cz>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bwh: Backported to 3.2:
- Adjust context
- madvise_remove() calls vmtruncate_range(), not do_fallocate()]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the normal case, when the device is turned off, lcd off function doesn't
work. However, we still need to turn off the panel and DSV convertor. This
patch adds an LCD turn-off function and calls it on device shut down.
Change-Id: I5526f0f08567190050527121d7e0890c05607a68
Signed-off-by: Iliyan Malchev <malchev@google.com>
commit 8a5bdaf3 never allow the runtime suspend if prepare is failed.
The complete phase is invoked when calling __device_suspend.
It allows the runtime suspend again. But if prepare is failed,
__device_suspend won't be called and the complete phase won't also
called. Thus this runtime suspend will be blocked forever.
If prepare is failed, you should allow the runtime suspend again.
Change-Id: I57c7c04609299244a042f2524e1c65dbbdd91ae3
mako does not need loadable kernel modules; disabling support for them reduces
the attack surface on the kernel
Signed-off-by: Iliyan Malchev <malchev@google.com>
In the case which is below,
1. acquire slab for cpu partial list
2. free object to it by remote cpu
3. page->freelist = t
then memory leak is occurred.
Change acquire_slab() not to zap freelist when it works for cpu partial list.
I think it is a sufficient solution for fixing a memory leak.
Below is output of 'slabinfo -r kmalloc-256'
when './perf stat -r 30 hackbench 50 process 4000 > /dev/null' is done.
***Vanilla***
Sizes (bytes) Slabs Debug Memory
------------------------------------------------------------------------
Object : 256 Total : 468 Sanity Checks : Off Total: 3833856
SlabObj: 256 Full : 111 Redzoning : Off Used : 2004992
SlabSiz: 8192 Partial: 302 Poisoning : Off Loss : 1828864
Loss : 0 CpuSlab: 55 Tracking : Off Lalig: 0
Align : 8 Objects: 32 Tracing : Off Lpadd: 0
***Patched***
Sizes (bytes) Slabs Debug Memory
------------------------------------------------------------------------
Object : 256 Total : 300 Sanity Checks : Off Total: 2457600
SlabObj: 256 Full : 204 Redzoning : Off Used : 2348800
SlabSiz: 8192 Partial: 33 Poisoning : Off Loss : 108800
Loss : 0 CpuSlab: 63 Tracking : Off Lalig: 0
Align : 8 Objects: 32 Tracing : Off Lpadd: 0
Total and loss number is the impact of this patch.
Cc: <stable@vger.kernel.org>
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Memory allocated by kstrdup should be freed,
when kmalloc(kmem_size, GFP_KERNEL) is failed.
Acked-by: Christoph Lameter <cl@linux.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
SDR104 (supported by SD3.0 spec compliant SD cards) bus speed mode
requires DLL (Delay Locked Loop for sampling clock generation) HW
block to be tuned if clock rate is >100MHz. To help the sampling clock
tuning SD3.0 specification has added CMD19.
During card initialization, driver uses these tuning commands to search
for optimal sampling point and then programs the DLL HW with this optimal
sampling point.
In addition, our SDCC controller's DLL HW has a mechansim of data
tracking (CDR) which can be enabled during read transaction so the host
will not lose the sampling point as a result of voltage and, especially,
temperature variations. Tuning pattern commands (CMD19) can be sent
automatically by HW (by using AUTO_CMD19 bit) before any RX transaction
for helping the CDR to track after the correct sampling point.
Although sending CMD19 automatically before every read commands is
overhead and reduces read performance by ~1 MB/s. SD spec does
not mandate to send these commands before every read operations.
This change keeps AUTO_CMD19 disabled by default but provides sysfs
attribute to enable or disable AUTO_CMD19 as per need.
Change-Id: I9ef99a8dde9b9733ea99639cdab2b6714ac50d76
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
IBM reported a deadlock in select_parent(). This was found to be caused
by taking rename_lock when already locked when restarting the tree
traversal.
There are two cases when the traversal needs to be restarted:
1) concurrent d_move(); this can only happen when not already locked,
since taking rename_lock protects against concurrent d_move().
2) racing with final d_put() on child just at the moment of ascending
to parent; rename_lock doesn't protect against this rare race, so it
can happen when already locked.
Because of case 2, we need to be able to handle restarting the traversal
when rename_lock is already held. This patch fixes all three callers of
try_to_ascend().
IBM reported that the deadlock is gone with this patch.
[ I rewrote the patch to be smaller and just do the "goto again" if the
lock was already held, but credit goes to Miklos for the real work.
- Linus ]
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
git://codeaurora.org/external/wlan/prima.git
b4798fe Merge branch 'jb-mr1-dev' of git.codeaurora.org:/external/wlan/prima into HEAD
f9f6457 wlan: Release 3.2.1.11i
d5b0e74 Post proper response to SME when reassoc is rejected
e6c5a60 wlan: Release 3.2.1.11i
bd2182f Post proper response to SME when reassoc is rejected
Signed-off-by: Iliyan Malchev <malchev@google.com>
Expose the packing control test.
Test the packing control feature under these scenarios:
- Packing expected: entering write packing state,
staying in write packing state.
- Packing not expected: not entering write packing state
when not supposed to.
- Mixed states: test the ability to shift from packing
to no-packing and back, and the opposite.
(cherry picked from commit 202befb5a5a344f88a0b4fd1ed07b9a2c7ce4e3d)
(cherry picked from commit 91e1e32573321986b141f153ffdb5bcc8affe95b)
Change-Id: I06638dc19aa7703d0571d9699b24ce9977b1bcb8
Signed-off-by: Lee Susman <lsusman@codeaurora.org>
Signed-off-by: Tatyana Brokhman <tlinder@codeaurora.org>
(cherry picked from commit 273c29cec62e3f771bc726cf226c66234ceb80cb)
When hitting a stop potential packing event, we should zero
num_of_potential_packed_wr_reqs, so that the packing won't be
enabled too early.
(cherry picked from commit 285ee17e20580759ef349b43e48565cfad8adb10)
Change-Id: I384d74dc6eece67358a51fb0add5d988aee44d5d
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Tatyana Brokhman <tlinder@codeaurora.org>
(cherry picked from commit 659fc08c2d5f7b46c3543feb618f4708dc78b77b)
Since clk_prepare_enable is being used when enabling clocks,
clk_disable_prepare should be used to disable them.
CRs-fixed: 453579
Change-Id: Ic15c99e1fa82e45a3b9165af60c104a6bb736f92
Signed-off-by: Amy Maloche <amaloche@codeaurora.org>
Add device tree (DT) support for isa1200. Extract device
properties from DT and store it in platform data structure.
Change-Id: I9ed9cc36cb07db00634b9e73a8052e62d313d484
Signed-off-by: Amy Maloche <amaloche@codeaurora.org>
Make sure secure map and unmap get called in each mfd and balanced.
Otherwise, secure contents can not be played for both primary and
hdmi.
CRs-fixed: 450427
Change-Id: I5399be64fbb104dc3cbe2c5c65b3f4406f47473c
Signed-off-by: Huaibin Yang <huaibiny@codeaurora.org>
NV bin is downloaded when wlan driver is loaded. This change
adds support for NV bin download during coldboot of WCNSS and
thus much before wlan driver is loaded. NV bin id not downloaded
by platform driver during SSR.
Change-Id: I070aae4270c54a3af4d4c0a24897e8c0c9cbfaa5
Acked-by: Rajesh Chauhan <rajeshc@qca.qualcomm.com>
Signed-off-by: Sameer Thalappil <sameert@codeaurora.org>
The current fence wait times out too early in cases where the GPU
is under a heavy workload. Change this behaviour to warn and wait
for a longer time for GPU to complete it's work.
Change-Id: I8b20c22733a77c1aaea4a1fd1e4daaa14b5ec2fd
Signed-off-by: Naseer Ahmed <naseer@codeaurora.org>
Updating the sample rate for the compressed stream playback of the HDMI
interface through the mixer control.
Change-Id: Ib7b01f490f5e209512433a82faf96af0a3a810bd
Signed-off-by: Aviral Gupta <aviralg@codeaurora.org>
For the use case of Playback over HDMI device, the number of
channels supported varies with the sink capabilities.
Add support to fix the channels based on the sink capabilites
configured by userspace for msm8960
Change-Id: Iba4f1fead17832d7832fabdcba02da671e5fd005
Signed-off-by: Subhash Chandra Bose Naripeddy <snariped@codeaurora.org>
The print messages are helpful in debugging adb stability issues.
However they get flooded if adb daemon is restarted continuously.
If serial console is enabled this is leading to watchdog bark.
CRs-Fixed: 456662
Change-Id: I1381d44b7a5081ee04053c739b9c83111b191125
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
An error message is printed when user space tries to enable/disable
android configuration when it is already enabled/disabled. This
message is helpful in debugging USB composition issues. However
this message is getting flooded in an error condition. If serial
console is enabled, it is leading to watchdog bark.
CRs-Fixed: 457725
Change-Id: I2d61ac46ba3c8464bada23a64ad0d8027a4adfe9
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
All debugging outputs from KPSS pass 3 are divided by four in
hardware. Multiply the measured clock rate to compensate.
Change-Id: Id0e0e0116bb469b690b4d1b2a46e8a2b3b164a86
Signed-off-by: Patrick Daly <pdaly@codeaurora.org>
(cherry picked from commit 73db3ad247a9ab74c8edeca59b961594d1cdb13d)
msm_fb_wait_for_fence() will put fence fd list if sync_fence_wait()
failed. After that same fd list wil be put one more time. This
cause a fd put twice and cause system crash at the time fd
was closed.
Change-Id: If6a358e051c98841e34c23d81098243003ef0836
Signed-off-by: Kuogee Hsieh <khsieh@codeaurora.org>
Fence sync seems to react badly when it gets a unexpected timestamp
on a timeline. When a event expires always send the expected timestamp
even if the event expired because the context was destroyed.
Change-Id: Ic0dedbad770363a00257d62c4940f7743bc33060
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
There are some testing showing that L2 at 1188MHz may cause
some stability issue. Lower the max speed to 1134Mhz until
the root cause is found.
Change-Id: I327c33c72ea0327e269d7839fadfb25ac715408b
Signed-off-by: Tianyi Gou <tgou@codeaurora.org>
It is possible that pipe_commit still be issued by
frame work after suspend had been started. Therefore
mutex lock within video_off is necessary to prevent
race condition from happening between these two threads.
Change-Id: I57f51744c1654d6def8664c492035631a5b7081e
Signed-off-by: Kuogee Hsieh <khsieh@codeaurora.org>
There is a missing step to clear OTHR_RESET_CTR before power down.
It could cause a glitch and power hit without it. It has to be reset.
original gerrit is https://review-android.quicinc.com/#/c/191312/
This is cherry-pick gerrit by developer to jb_2.5
Change-Id: Id96070352480be6ea412844d4f71765e2df76afe
CRs-fixed: 420142
Signed-off-by: Fred Oh <fred@codeaurora.org>
The current code does not abort suspend when there are pending URBs.
They are simply killed during suspend. This may lead to QMI stall.
Add proper checks to ensure that there is no pending Tx/Rx data
during suspend. Clear the "SUSPENDED" flag in resume only after
taking out all the URBs from "tx_deferred" anchor.
CRs-Fixed: 451414
Change-Id: Icc48afc5eeedcc126d0ab1bec951bde4a53ac165
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Synchronous runtime PM API is used to increment the PM usage counter
in Tx path. It means Tx data path will not race with interface suspend.
Remove "delayed" anchor which is meant for Tx requests, that are arrived
during suspend. There is no point in distinguishing auto suspend and
system wide suspend. Abort suspend when we have pending Tx URBs.
CRs-Fixed: 451414
Change-Id: I302b120ccecfe30db1ebd2b1fcd9945a3675cf83
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
To deduce whether secure demux is available, an attempt is made to
load the secure demux appliction during module initialization.
The loading process retrieves the application from filesystem.
When kernel configuration is set so that the module is built-in
as part of the kernel image, filesystem is not yet available
and loading fails.
This change moves the loading attempt from initialization
to first usage of status query, which occurs after device
is opened by user-space and filesystem is sure to be available.
Change-Id: I22d24054196f6068bf28e2ba6f2a4da6a9340730
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
The platform data was requesting the wrong number,
it should request the first perf irq item.
Change-Id: I4a25b4704ed9e76172c6b0d4ca4b28a3286ab2ad
Signed-off-by: Neil Leeder <nleeder@codeaurora.org>
