Rob Herring [Tue, 18 Jul 2017 21:43:15 +0000 (16:43 -0500)]
misc: Convert to using %pOF instead of full_name
Now that we have a custom printf format specifier, convert users of
full_name to use %pOF instead. This is preparation to remove storing
of the full path string for each node.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
Arvind Yadav [Fri, 4 Aug 2017 06:37:57 +0000 (12:07 +0530)]
misc: ti-st: constify attribute_group structures.
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work with
const attribute_group. So mark the non-const structs as const.
Bhumika Goyal [Wed, 2 Aug 2017 09:10:06 +0000 (14:40 +0530)]
MISC: add const to bin_attribute structures
Add const to bin_attribute structures as they are only passed to the
functions sysfs_{remove/create}_bin_file. The arguments passed are of
type const, so declare the structures to be const.
pci_device_id are not supposed to change at runtime. All functions
working with pci_device_id provided by <linux/pci.h> work with
const pci_device_id. So mark the non-const structs as const.
pci_device_id are not supposed to change at runtime. All functions
working with pci_device_id provided by <linux/pci.h> work with
const pci_device_id. So mark the non-const structs as const.
pci_device_id are not supposed to change at runtime. All functions
working with pci_device_id provided by <linux/pci.h> work with
const pci_device_id. So mark the non-const structs as const.
checkpatch ERROR: space prohibited before open square bracket '['
pci_device_id are not supposed to change at runtime. All functions
working with pci_device_id provided by <linux/pci.h> work with
const pci_device_id. So mark the non-const structs as const.
misc: eeprom_93xx46: Include <linux/gpio/consumer.h>
eeprom_93xx46_platform_data struct has a 'struct gpio_desc'
type member, so it is better to include <linux/gpio/consumer.h>,
which provides 'struct gpio_desc' type.
misc: eeprom_93xx46: Simplify the usage of gpiod API
Commit 3ca9b1ac28398c ("misc: eeprom_93xx46: Add support for a GPIO
'select' line.") introduced the optional usage of 'select-gpios'
by using the gpiod API in a convoluted way.
Rewrite the gpiod handling to make the code simpler.
Sherry Yang [Wed, 23 Aug 2017 15:46:42 +0000 (08:46 -0700)]
android: binder: Add global lru shrinker to binder
Hold on to the pages allocated and mapped for transaction
buffers until the system is under memory pressure. When
that happens, use linux shrinker to free pages. Without
using shrinker, patch "android: binder: Move buffer out
of area shared with user space" will cause a significant
slow down for small transactions that fit into the first
page because free list buffer header used to be inlined
with buffer data.
In addition to prevent the performance regression for
small transactions, this patch improves the performance
for transactions that take up more than one page.
Modify alloc selftest to work with the shrinker change.
Test: Run memory intensive applications (Chrome and Camera)
to trigger shrinker callbacks. Binder frees memory as expected.
Test: Run binderThroughputTest with high memory pressure
option enabled.
Signed-off-by: Sherry Yang <sherryy@android.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sherry Yang [Wed, 23 Aug 2017 15:46:40 +0000 (08:46 -0700)]
android: binder: Add allocator selftest
binder_alloc_selftest tests that alloc_new_buf handles page allocation and
deallocation properly when allocate and free buffers. The test allocates 5
buffers of various sizes to cover all possible page alignment cases, and
frees the buffers using a list of exhaustive freeing order.
Test: boot the device with ANDROID_BINDER_IPC_SELFTEST config option
enabled. Allocator selftest passes.
Signed-off-by: Sherry Yang <sherryy@android.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Dan Carpenter [Wed, 16 Aug 2017 08:54:17 +0000 (11:54 +0300)]
thunderbolt: Fix reset response_type
There is a mistake here where we accidentally use sizeof(TB_CFG_PKG_RESET)
instead of just TB_CFG_PKG_RESET. The size of an int is 4 so it's the
same as TB_CFG_PKG_NOTIFY_ACK.
Fixes: d7f781bfdbf4 ("thunderbolt: Rework control channel to be more reliable") Reported-by: Colin King <colin.king@canonical.com> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: stable <stable@vger.kernel.org> # 4.13 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bernat, Yehezkel [Tue, 15 Aug 2017 05:19:20 +0000 (08:19 +0300)]
thunderbolt: Allow clearing the key
If secure authentication of a devices fails, either because the device
already has another key uploaded, or there is some other error sending
challenge to the device, and the user only wants to approve the device
just once (without a new key being uploaded to the device) the current
implementation does not allow this because the key cannot be cleared
once set even if we allow it to be changed.
Make this scenario possible and allow clearing the key by writing
empty string to the key sysfs file.
Stephen Boyd [Wed, 2 Aug 2017 16:22:20 +0000 (10:22 -0600)]
coresight: STM: Clean up __iomem type usage
The casting and other things here is odd, and causes sparse to
complain:
drivers/hwtracing/coresight/coresight-stm.c:279:35: warning: incorrect type in argument 1 (different address spaces)
drivers/hwtracing/coresight/coresight-stm.c:279:35: expected void [noderef] <asn:2>*addr
drivers/hwtracing/coresight/coresight-stm.c:279:35: got struct stm_drvdata *drvdata
drivers/hwtracing/coresight/coresight-stm.c:327:17: warning: incorrect type in argument 2 (different address spaces)
drivers/hwtracing/coresight/coresight-stm.c:327:17: expected void volatile [noderef] <asn:2>*addr
drivers/hwtracing/coresight/coresight-stm.c:327:17: got void *addr
drivers/hwtracing/coresight/coresight-stm.c:330:17: warning: incorrect type in argument 2 (different address spaces)
drivers/hwtracing/coresight/coresight-stm.c:330:17: expected void volatile [noderef] <asn:2>*addr
drivers/hwtracing/coresight/coresight-stm.c:330:17: got void *addr
drivers/hwtracing/coresight/coresight-stm.c:333:17: warning: incorrect type in argument 2 (different address spaces)
drivers/hwtracing/coresight/coresight-stm.c:333:17: expected void volatile [noderef] <asn:2>*addr
drivers/hwtracing/coresight/coresight-stm.c:333:17: got void *addr
>From what I can tell, we don't really need to treat ch_addr as
anything besides a pointer, and we can just do pointer math
instead of ORing in the bits of the offset and achieve the same
thing.
Also, we were passing a drvdata pointer to the
coresight_timeout() function, but we really wanted to pass the
address of the register base. Luckily the base is the first
member of the structure, so everything works out, but this is
quite unsafe if we ever change the structure layout. Clean this
all up so sparse stops complaining on this code.
coresight tmc: Add support for Coresight SoC 600 TMC
The coresight SoC 600 supports ETR save-restore which allows us
to restore a trace session by retaining the RRP/RWP/STS.Full values
when the TMC leaves the Disabled state. However, the TMC doesn't
have a scatter-gather unit in built.
Also, TMCs have different PIDs in different configurations (ETF,
ETB & ETR), unlike the previous generation.
While the DEVID exposes some of the features/changes in the TMC,
it doesn't explicitly advertises the new save-restore feature
as described above.
The Coresight SoC 600 TMC ETR supports save-restore feature,
where the values of the RRP/RWP and STS.Full are retained
when it leaves the Disabled state. Hence, we must program the
RRP/RWP and STS.Full to a proper value. For now, set the RRP/RWP
to the base address of the buffer and clear the STS.Full register.
This can be later exploited for proper save-restore of ETR
trace contexts (e.g, perf).
coresight tmc etr: Setup AXI cache encoding for read transfers
If the ETR supports split cache encoding (i.e, separate bits for
read and write transfers) unlike the older version (where read
and write transfers use the same encoding in AXICTL[2-5]).
This feature is not advertised and has to be described by the
static mask associated with the device id.
This patch cleans up how we setup the AXICTL register on
TMC ETR. At the moment we don't set the CacheCtrl bits, which
drives the arcache and awcache bits on AXI bus specifying the
cacheablitiy. Set this to Write-back Read and Write-allocate.
The SG unit in the TMC has been removed in Coresight SoC-600.
This is however advertised by DEVID:Bit 24 = 0b1. On the
previous generation, the bit is RES0, hence we can rely on the
DEVID to detect the support.
With new version of TMC ETR, there are differing set of
features supported by the TMC. Add the capability of a
given TMC ETR for making safer decisions at runtime.
The device configuration register of the TMC (DEVID) lists
some of the capabilities. So, we can detect some of them at
probe. However, some of the features (or changes in behavior)
are not advertised and we have to depend on the PID to infer
the features. So we use a static description of the "unadvertised"
capabilities attached to the PID. Combining both, the static
and the dynamic capabilities, we maintain a bitmask of the
available features which can be later checked to take
appropriate actions.
Coresight SoC 600 defines a new configuration for TMC, Embedded Trace
Streamer (ETS), indicated by 0x3 in MODE:CONFIG_TYPE. This would break
the existing driver which will treat anything other than ETR/ETB as an
ETF. Fix the driver to check the configuration type properly and also
add a warning if we encounter an unsupported configuration (ETS).
The Linux coresight drivers define the programmable ATB replicator as
Qualcomm replicator, while this is designed by ARM. This can cause
confusion to a user selecting the driver. Cleanup all references to
make it explicitly clear. This patch :
1) Replace the compatible string for the replicator :
qcom,coresight-replicator1x => arm,coresight-dynamic-replicator
2) Changes the Kconfig symbol (since this is not part of any defconfigs)
CORESIGHT_QCOM_REPLICATOR => CORESIGHT_DYNAMIC_REPLICATOR
3) Improves the help message in the Kconfig.
4) Changes the name of the driver and the file :
coresight-replicator-qcom => coresight-dynamic-replicator
Cc: Pratik Patel <pratikp@codeaurora.org> Cc: Ivan T. Ivanov <ivan.ivanov@linaro.org> Cc: Mathieu Poirier <mathieu.poirier@linaro.org> Cc: devicetree@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Acked-by: Rob Herring <robh+dt@kernel.org> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com> Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Mike Leach [Wed, 2 Aug 2017 16:22:03 +0000 (10:22 -0600)]
coresight: etm4x: Adds trace return stack option programming for ETMv4.
Adds handling to program the return stack option into ETMv4 hardware if
specified in the perf command line.
If option is not supported by the hardware then it will be ignored.
This allows capture to move between core/ETM combinations that have the
hardware support to those that do not.
Mike Leach [Wed, 2 Aug 2017 16:22:02 +0000 (10:22 -0600)]
coresight: ptm: Adds trace return stack option programming for PTM.
Adds handling to program the return stack option into PTM hardware if
specified in the perf command line.
If option is not supported by the hardware then it will be ignored.
This allows capture to move between core/ETM combinations that have the
hardware support to those that do not.
attribute_groups are not supposed to change at runtime. All functions
working with attribute_groups provided by <linux/sysfs.h> work with const
attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
2573 288 296 3157 c55 coresight-etm-perf.o
File size After adding 'const':
text data bss dec hex filename
2613 224 296 3133 c3d coresight-etm-perf.o
Mathieu Poirier [Wed, 2 Aug 2017 16:21:59 +0000 (10:21 -0600)]
coresight: etm3x: Set synchronisation frequencty to TRM default
Register ETMSYNCFR holds the number of by that need to be generated before
periodic synchronisation packets are inserted in the trace stream. By
zeroing out the config structure, the current code effectively disable
periodic synchronization.
This patch simply initialise the recommended value for this register as
specified in the technical reference manual.
Mathieu Poirier [Wed, 2 Aug 2017 16:21:58 +0000 (10:21 -0600)]
coresight: etb10: Move etb_disable_hw() outside of lock
Function etb_disable_hw() is already taking care of unlocking and locking
the coresight access register and as such doesn't need to be placed
within the unlock/lock of function etb_update_buffer().
Mathieu Poirier [Wed, 2 Aug 2017 16:21:57 +0000 (10:21 -0600)]
coresight: Add barrier packet for synchronisation
When a buffer overflow happens the synchronisation patckets usually
present at the beginning of the buffer are lost, a situation that
prevents the decoder from knowing the context of the traces being
decoded.
This patch adds a barrier packet to be used by sink IPs when a buffer
overflow condition is detected. These barrier packets are then used
by the decoding library as markers to force re-synchronisation.
Mathieu Poirier [Wed, 2 Aug 2017 16:21:56 +0000 (10:21 -0600)]
coresight: etb10: Remove useless conversion to LE
Internal CoreSight components are rendering trace data in little-endian
format. As such there is no need to convert the data once more, hence
removing the extra step.
Mathieu Poirier [Wed, 2 Aug 2017 16:21:55 +0000 (10:21 -0600)]
coresight: Correct buffer lost increment
Many conditions may cause synchronisation to be lost when updating
the perf ring buffer but the end result is still the same: synchronisation
is lost. As such there is no need to increment the lost count for each
condition, just once will suffice.
Kiran Gunda [Wed, 23 Aug 2017 12:46:26 +0000 (18:16 +0530)]
spmi: pmic-arb: Move the ownership check to irq_chip callback
Check the irq ownership in the irq_request_resources callback
instead of checking it during the irq mapping. This can prevent
installing the flow handler for the interrupt that is not owned by the EE
and allow the irq translation during the gpio driver probe.
Rob Herring [Tue, 18 Jul 2017 21:43:32 +0000 (16:43 -0500)]
spmi: Convert to using %pOF instead of full_name
Now that we have a custom printf format specifier, convert users of
full_name to use %pOF instead. This is preparation to remove storing
of the full path string for each node.
Signed-off-by: Rob Herring <robh@kernel.org> Reviewed-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
spmi: pmic-arb: Remove checking opc value not less than 0
The opc parameter in pmic_arb_write_cmd() function is defined with type
u8 and it's always greater than or equal to 0. Checking that it's not
less than 0 is redundant and it can cause a forbidden warning during
compilation. Remove the check.
David Collins [Fri, 28 Jul 2017 07:10:46 +0000 (12:40 +0530)]
spmi: pmic-arb: add support for HW version 5
Add support for version 5 of the SPMI PMIC arbiter. It utilizes
different offsets for registers than those found on version 3.
Also, the procedure to determine if writing and IRQ access is
allowed for a given PPID changes for version 5.
Signed-off-by: David Collins <collinsd@codeaurora.org> Signed-off-by: Kiran Gunda <kgunda@codeaurora.org> Reviewed-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
spmi: pmic-arb: return __iomem pointer instead of offset
Modify the pmic_arb version ops to return an __iomem pointer
to the address instead of an offset. That way we do not need to
care about the base address changes in the new HW version.
spmi: pmic-arb: use irq_chip callback to set spmi irq wakeup capability
Currently the driver sets the pmic arbiter core interrupt as wakeup capable
irrespective of the child irqs which causes the system to wakeup
unnecessarily. To fix this, set the core interrupt as wakeup capable
only if any of the child irqs request for it. Do this by marking it as
wakeup capable in the irq_set_wake callback.
spmi: pmic-arb: return the value instead of passing by pointer
Returning the output value from a function, when it is possible, is the
better and cleaner way than passing it by the pointer. Hence, modify
the ppid_to_apid mapping function to return apid instead of passing
it by a pointer. While at it, pass the ppid as function parameter to
ppid_to_apid mapping function instead of passing the sid and addr.
spmi: pmic-arb: replace the writel_relaxed with __raw_writel
Replace the writel_relaxed with __raw_writel to avoid byte swapping
in pmic_arb_write_data() function. That way the code is independent
of the CPU endianness.
Fixes: 111a10bf3e53 ("spmi: pmic-arb: rename spmi_pmic_arb_dev to
spmi_pmic_arb") Signed-off-by: Kiran Gunda <kgunda@codeaurora.org> Reviewed-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
spmi: pmic-arb: fix memory allocation for mapping_table
Allocate the correct memory size (max_pmic_peripherals) for the
mapping_table that holds the apid to ppid mapping. Also use a local
variable for mapping_table for better alignment of the code.
Fixes: 987a9f128b8a ("spmi: pmic-arb: Support more than 128 peripherals") Signed-off-by: Kiran Gunda <kgunda@codeaurora.org> Reviewed-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
spmi: pmic-arb: rename pa_xx to pmic_arb_xx and other cleanup
This patch cleans up the following.
- Rename the "pa" to "pmic_arb".
- Rename the spmi_pmic_arb *dev to spmi_pmic_arb *pmic_arb.
- Rename the pa_{read,write}_data() functions to
pmic_arb_{read,write}_data().
- Rename channel to APID.
- Rename the HWIRQ_*() macros to hwirq_to_*().
spmi: pmic-arb: remove the read/write access checks
The access mode checks for peripheral ownership for read/write
permissions should not be required. Every peripheral enabled for
this master is expected to have a read/write permissions. If there
is any such invalid access due to wrong configuration in boot loader
or device tree files, then it should be fixed in those locations.
Hence, remove the access mode checks from the driver.
Jisheng Zhang [Fri, 25 Aug 2017 00:39:51 +0000 (08:39 +0800)]
Revert "staging: Fix build issues with new binder API"
This reverts commit d0bdff0db809 ("staging: Fix build issues with new
binder API"), because commit e38361d032f1 ("ARM: 8091/2: add get_user()
support for 8 byte types") has added the 64bit __get_user_asm_*
implementation.
Linus Torvalds [Mon, 28 Aug 2017 00:10:34 +0000 (17:10 -0700)]
Merge tag 'iommu-fixes-v4.13-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu
Pull IOMMU fix from Joerg Roedel:
"Another fix, this time in common IOMMU sysfs code.
In the conversion from the old iommu sysfs-code to the
iommu_device_register interface, I missed to update the release path
for the struct device associated with an IOMMU. It freed the 'struct
device', which was a pointer before, but is now embedded in another
struct.
Freeing from the middle of allocated memory had all kinds of nasty
side effects when an IOMMU was unplugged. Unfortunatly nobody
unplugged and IOMMU until now, so this was not discovered earlier. The
fix is to make the 'struct device' a pointer again"
* tag 'iommu-fixes-v4.13-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
iommu: Fix wrong freeing of iommu_device->dev
Linus Torvalds [Mon, 28 Aug 2017 00:08:37 +0000 (17:08 -0700)]
Merge tag 'char-misc-4.13-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc
Pull char/misc fix from Greg KH:
"Here is a single misc driver fix for 4.13-rc7. It resolves a reported
problem in the Android binder driver due to previous patches in
4.13-rc.
It's been in linux-next with no reported issues"
* tag 'char-misc-4.13-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
ANDROID: binder: fix proc->tsk check.
Linus Torvalds [Mon, 28 Aug 2017 00:03:33 +0000 (17:03 -0700)]
Merge tag 'staging-4.13-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
Pull staging/iio fixes from Greg KH:
"Here are few small staging driver fixes, and some more IIO driver
fixes for 4.13-rc7. Nothing major, just resolutions for some reported
problems.
All of these have been in linux-next with no reported problems"
* tag 'staging-4.13-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
iio: magnetometer: st_magn: remove ihl property for LSM303AGR
iio: magnetometer: st_magn: fix status register address for LSM303AGR
iio: hid-sensor-trigger: Fix the race with user space powering up sensors
iio: trigger: stm32-timer: fix get trigger mode
iio: imu: adis16480: Fix acceleration scale factor for adis16480
PATCH] iio: Fix some documentation warnings
staging: rtl8188eu: add RNX-N150NUB support
Revert "staging: fsl-mc: be consistent when checking strcmp() return"
iio: adc: stm32: fix common clock rate
iio: adc: ina219: Avoid underflow for sleeping time
iio: trigger: stm32-timer: add enable attribute
iio: trigger: stm32-timer: fix get/set down count direction
iio: trigger: stm32-timer: fix write_raw return value
iio: trigger: stm32-timer: fix quadrature mode get routine
iio: bmp280: properly initialize device for humidity reading
Linus Torvalds [Mon, 28 Aug 2017 00:01:54 +0000 (17:01 -0700)]
Merge tag 'ntb-4.13-bugfixes' of git://github.com/jonmason/ntb
Pull NTB fixes from Jon Mason:
"NTB bug fixes to address an incorrect ntb_mw_count reference in the
NTB transport, improperly bringing down the link if SPADs are
corrupted, and an out-of-order issue regarding link negotiation and
data passing"
* tag 'ntb-4.13-bugfixes' of git://github.com/jonmason/ntb:
ntb: ntb_test: ensure the link is up before trying to configure the mws
ntb: transport shouldn't disable link due to bogus values in SPADs
ntb: use correct mw_count function in ntb_tool and ntb_transport
Linus Torvalds [Sun, 27 Aug 2017 23:25:09 +0000 (16:25 -0700)]
Avoid page waitqueue race leaving possible page locker waiting
The "lock_page_killable()" function waits for exclusive access to the
page lock bit using the WQ_FLAG_EXCLUSIVE bit in the waitqueue entry
set.
That means that if it gets woken up, other waiters may have been
skipped.
That, in turn, means that if it sees the page being unlocked, it *must*
take that lock and return success, even if a lethal signal is also
pending.
So instead of checking for lethal signals first, we need to check for
them after we've checked the actual bit that we were waiting for. Even
if that might then delay the killing of the process.
This matches the order of the old "wait_on_bit_lock()" infrastructure
that the page locking used to use (and is still used in a few other
areas).
Note that if we still return an error after having unsuccessfully tried
to acquire the page lock, that is ok: that means that some other thread
was able to get ahead of us and lock the page, and when that other
thread then unlocks the page, the wakeup event will be repeated. So any
other pending waiters will now get properly woken up.
Fixes: 62906027091f ("mm: add PageWaiters indicating tasks are waiting for a page bit") Cc: Nick Piggin <npiggin@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Jan Kara <jack@suse.cz> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Andi Kleen <ak@linux.intel.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Sun, 27 Aug 2017 20:55:12 +0000 (13:55 -0700)]
Minor page waitqueue cleanups
Tim Chen and Kan Liang have been battling a customer load that shows
extremely long page wakeup lists. The cause seems to be constant NUMA
migration of a hot page that is shared across a lot of threads, but the
actual root cause for the exact behavior has not been found.
Tim has a patch that batches the wait list traversal at wakeup time, so
that we at least don't get long uninterruptible cases where we traverse
and wake up thousands of processes and get nasty latency spikes. That
is likely 4.14 material, but we're still discussing the page waitqueue
specific parts of it.
In the meantime, I've tried to look at making the page wait queues less
expensive, and failing miserably. If you have thousands of threads
waiting for the same page, it will be painful. We'll need to try to
figure out the NUMA balancing issue some day, in addition to avoiding
the excessive spinlock hold times.
That said, having tried to rewrite the page wait queues, I can at least
fix up some of the braindamage in the current situation. In particular:
(a) we don't want to continue walking the page wait list if the bit
we're waiting for already got set again (which seems to be one of
the patterns of the bad load). That makes no progress and just
causes pointless cache pollution chasing the pointers.
(b) we don't want to put the non-locking waiters always on the front of
the queue, and the locking waiters always on the back. Not only is
that unfair, it means that we wake up thousands of reading threads
that will just end up being blocked by the writer later anyway.
Also add a comment about the layout of 'struct wait_page_key' - there is
an external user of it in the cachefiles code that means that it has to
match the layout of 'struct wait_bit_key' in the two first members. It
so happens to match, because 'struct page *' and 'unsigned long *' end
up having the same values simply because the page flags are the first
member in struct page.
Cc: Tim Chen <tim.c.chen@linux.intel.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Christopher Lameter <cl@linux.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Sun, 27 Aug 2017 19:12:25 +0000 (12:12 -0700)]
Clarify (and fix) MAX_LFS_FILESIZE macros
We have a MAX_LFS_FILESIZE macro that is meant to be filled in by
filesystems (and other IO targets) that know they are 64-bit clean and
don't have any 32-bit limits in their IO path.
It turns out that our 32-bit value for that limit was bogus. On 32-bit,
the VM layer is limited by the page cache to only 32-bit index values,
but our logic for that was confusing and actually wrong. We used to
define that value to
(((loff_t)PAGE_SIZE << (BITS_PER_LONG-1))-1)
which is actually odd in several ways: it limits the index to 31 bits,
and then it limits files so that they can't have data in that last byte
of a page that has the highest 31-bit index (ie page index 0x7fffffff).
Neither of those limitations make sense. The index is actually the full
32 bit unsigned value, and we can use that whole full page. So the
maximum size of the file would logically be "PAGE_SIZE << BITS_PER_LONG".
However, we do wan tto avoid the maximum index, because we have code
that iterates over the page indexes, and we don't want that code to
overflow. So the maximum size of a file on a 32-bit host should
actually be one page less than the full 32-bit index.
So the actual limit is ULONG_MAX << PAGE_SHIFT. That means that we will
not actually be using the page of that last index (ULONG_MAX), but we
can grow a file up to that limit.
The wrong value of MAX_LFS_FILESIZE actually caused problems for Doug
Nazar, who was still using a 32-bit host, but with a 9.7TB 2 x RAID5
volume. It turns out that our old MAX_LFS_FILESIZE was 8TiB (well, one
byte less), but the actual true VM limit is one page less than 16TiB.
This was invisible until commit c2a9737f45e2 ("vfs,mm: fix a dead loop
in truncate_inode_pages_range()"), which started applying that
MAX_LFS_FILESIZE limit to block devices too.
NOTE! On 64-bit, the page index isn't a limiter at all, and the limit is
actually just the offset type itself (loff_t), which is signed. But for
clarity, on 64-bit, just use the maximum signed value, and don't make
people have to count the number of 'f' characters in the hex constant.
So just use LLONG_MAX for the 64-bit case. That was what the value had
been before too, just written out as a hex constant.
Fixes: c2a9737f45e2 ("vfs,mm: fix a dead loop in truncate_inode_pages_range()") Reported-and-tested-by: Doug Nazar <nazard@nazar.ca> Cc: Andreas Dilger <adilger@dilger.ca> Cc: Mark Fasheh <mfasheh@versity.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Sat, 26 Aug 2017 19:48:29 +0000 (12:48 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Pull input fixes from Dmitry Torokhov:
- a tweak to the IBM Trackpoint driver that helps recognizing
trackpoints on never Lenovo Carbons
- a fix to the ALPS driver solving scroll issues on some Dells
- yet another ACPI ID has been added to Elan I2C toucpad driver
- quieted diagnostic message in soc_button_array driver
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: ALPS - fix two-finger scroll breakage in right side on ALPS touchpad
Input: soc_button_array - silence -ENOENT error on Dell XPS13 9365
Input: trackpoint - add new trackpoint firmware ID
Input: elan_i2c - add ELAN0602 ACPI ID to support Lenovo Yoga310
Linus Torvalds [Sat, 26 Aug 2017 19:46:14 +0000 (12:46 -0700)]
Merge tag 'pci-v4.13-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Pull PCI fix from Bjorn Helgaas:
"Remove needlessly alarming MSI affinity warning (this is not actually
a bug fix, but the warning prompts unnecessary bug reports)"
* tag 'pci-v4.13-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
PCI/MSI: Don't warn when irq_create_affinity_masks() returns NULL
Linus Torvalds [Sat, 26 Aug 2017 16:06:28 +0000 (09:06 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Ingo Molnar:
"Two fixes: one for an ldt_struct handling bug and a cherry-picked
objtool fix"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mm: Fix use-after-free of ldt_struct
objtool: Fix '-mtune=atom' decoding support in objtool 2.0
Linus Torvalds [Sat, 26 Aug 2017 16:02:18 +0000 (09:02 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timer fix from Ingo Molnar:
"Fix a timer granularity handling race+bug, which would manifest itself
by spuriously increasing timeouts of some timers (from 1 jiffy to ~500
jiffies in the worst case measured) in certain nohz states"
* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
timers: Fix excessive granularity of new timers after a nohz idle
Linus Torvalds [Sat, 26 Aug 2017 01:02:27 +0000 (18:02 -0700)]
Merge branch 'akpm' (patches from Andrew)
Merge misc fixes from Andrew Morton:
"6 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
mm/memblock.c: reversed logic in memblock_discard()
fork: fix incorrect fput of ->exe_file causing use-after-free
mm/madvise.c: fix freeing of locked page with MADV_FREE
dax: fix deadlock due to misaligned PMD faults
mm, shmem: fix handling /sys/kernel/mm/transparent_hugepage/shmem_enabled
PM/hibernate: touch NMI watchdog when creating snapshot
Linus Torvalds [Sat, 26 Aug 2017 00:46:23 +0000 (17:46 -0700)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull Paolo Bonzini:
"Bugfixes for x86, PPC and s390"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: PPC: Book3S: Fix race and leak in kvm_vm_ioctl_create_spapr_tce()
KVM, pkeys: do not use PKRU value in vcpu->arch.guest_fpu.state
KVM: x86: simplify handling of PKRU
KVM: x86: block guest protection keys unless the host has them enabled
KVM: PPC: Book3S HV: Add missing barriers to XIVE code and document them
KVM: PPC: Book3S HV: Workaround POWER9 DD1.0 bug causing IPB bit loss
KVM: PPC: Book3S HV: Use msgsync with hypervisor doorbells on POWER9
KVM: s390: sthyi: fix specification exception detection
KVM: s390: sthyi: fix sthyi inline assembly
Linus Torvalds [Sat, 26 Aug 2017 00:40:03 +0000 (17:40 -0700)]
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio fixes from Michael Tsirkin:
"Fixes two obvious bugs in virtio pci"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
virtio_pci: fix cpu affinity support
virtio_blk: fix incorrect message when disk is resized
Linus Torvalds [Sat, 26 Aug 2017 00:32:35 +0000 (17:32 -0700)]
Merge tag 'powerpc-4.13-8' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fix from Michael Ellerman:
"Just one fix, to add a barrier in the switch_mm() code to make sure
the mm cpumask update is ordered vs the MMU starting to load
translations. As far as we know no one's actually hit the bug, but
that's just luck.
Thanks to Benjamin Herrenschmidt, Nicholas Piggin"
* tag 'powerpc-4.13-8' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/mm: Ensure cpumask update is ordered
Linus Torvalds [Sat, 26 Aug 2017 00:27:26 +0000 (17:27 -0700)]
Merge tag 'nfsd-4.13-2' of git://linux-nfs.org/~bfields/linux
Pull nfsd fixes from Bruce Fields:
"Two nfsd bugfixes, neither 4.13 regressions, but both potentially
serious"
* tag 'nfsd-4.13-2' of git://linux-nfs.org/~bfields/linux:
net: sunrpc: svcsock: fix NULL-pointer exception
nfsd: Limit end of page list when decoding NFSv4 WRITE
Linus Torvalds [Sat, 26 Aug 2017 00:22:33 +0000 (17:22 -0700)]
Merge tag 'cifs-fixes-for-4.13-rc6-and-stable' of git://git.samba.org/sfrench/cifs-2.6
Pull cifs fixes from Steve French:
"Some bug fixes for stable for cifs"
* tag 'cifs-fixes-for-4.13-rc6-and-stable' of git://git.samba.org/sfrench/cifs-2.6:
cifs: return ENAMETOOLONG for overlong names in cifs_open()/cifs_lookup()
cifs: Fix df output for users with quota limits
Linus Torvalds [Sat, 26 Aug 2017 00:09:19 +0000 (17:09 -0700)]
Merge tag 'for-linus-20170825' of git://git.infradead.org/linux-mtd
Pull MTD fixes from Brian Norris:
"Two fixes - one for a 4.13 regression, and the other for an older one:
- Atmel NAND: since we started utilizing ONFI timings, we found that
we were being too restrict at rejecting them, partly due to
discrepancies in ONFI 4.0 and earlier versions. Relax the
restriction to keep these platforms booting. This is a 4.13-rc1
regression.
- nandsim: repeated probe/removal may not work after a failed init,
because we didn't free up our debugfs files properly on the failure
path. This has been around since 3.8, but it's nice to get this
fixed now in a nice easy patch that can target -stable, since
there's already refactoring work (that also fixes the issue)
targeted for the next merge window"
* tag 'for-linus-20170825' of git://git.infradead.org/linux-mtd:
mtd: nand: atmel: Relax tADL_min constraint
mtd: nandsim: remove debugfs entries in error path
projects / mirror_ubuntu-kernels.git / log