Mark Brown [Fri, 27 Mar 2020 17:28:36 +0000 (17:28 +0000)]
Merge series "ASoC: Intel: add SoundWire machine driver" from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
To handle multiple hardware combinations, this patchset suggests a
single machine driver which will create and initialize dailinks
dynamically. This allows us to support new configurations easily, as
shown with the TigerLake rt5682 example.
Each configuration updates the card component string, and UCM can test
for the presence of components to configure them as needed.
Since we use a single the machine driver name, all previous ACPI
tables need to be updated. That should have no impact since the
machine drivers listed at the time were not upstreamed and are no
longer maintained.
Naveen Manohar (2):
ASoC: Intel: common: add match table for TGL RT5682 SoundWire driver
ASoC: Intel: sof_sdw: Add Volteer support with RT5682 SNDW helper
function
Dan Murphy [Fri, 27 Mar 2020 16:24:32 +0000 (11:24 -0500)]
ASoC: tlv320adcx140: Remove undocumented property
Remove undocumented and unneeded ti,use-internal-reg from the example as
it was an artifact from initial development. The code does not query
for this property and as the document indicates if areg-supply is
undefined then the internal regulator is used.
Fixes: 302c0b7490cd ("dt-bindings: sound: Add TLV320ADCx140 dt
bindings") Signed-off-by: Dan Murphy <dmurphy@ti.com> CC: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20200327162432.17067-1-dmurphy@ti.com Signed-off-by: Mark Brown <broonie@kernel.org>
Pierre-Louis Bossart [Wed, 25 Mar 2020 22:07:44 +0000 (17:07 -0500)]
ASoC: Intel: boards: add sof_sdw machine driver
This machine driver provides support for different configurations:
RT700, RT711, RT1308 (1x and 2x, I2S or SoundWire mode), and RT715
CometLake, Icelake, TigerLake.
PDM digital microphones
HDMI
To avoid introducing one driver per configuration, this common machine
driver relies on platform-specific information, tables and quirks to
dynamically create the relevant dailinks.
Unlike a lot of machine drivers, we use different DAI links for
SoundWire capture and playback since the Cadence PDIs can do capture
OR playback, not both simultaneously.
For each configuration, the card component string is updated so that UCM
can select the relevant parts.
Mark Brown [Fri, 27 Mar 2020 15:33:10 +0000 (15:33 +0000)]
Merge series "ASoC: remove rtd->cpu/codec_dai{s}" from Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>:
Hi Mark
Now, CPU/Codec DAI(s) were replaced by rtd->dais.
Thus, We don't need rtd->cpu/codec_dai{s} anymore.
This pathset replaces it by new macro.
Kuninori Morimoto (36):
ASoC: soc-core: add asoc_rtd_to_cpu/codec() macro
ASoC: amd: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: atmel: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: au1x: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: bcm: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: cirrus: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: dwc: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: fsl: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: generic: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: img: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: intel: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: kirkwood: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: mediatek: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: meson: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: mxs: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: pxa: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: qcom: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: rockchip: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: samsung: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: sh: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: sof: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: sprd: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: stm: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: sunxi: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: tegra: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: ti: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: txx9: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: uniphier: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: ux500: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: xtensa: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: arm: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: codecs: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: soc: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer
ASoC: soc-core: set rtd->num_cpu/codec at soc_new_pcm_runtime()
ASoC: soc-core: tidyup soc_new_pcm_runtime() rtd setups
ASoC: soc-core: remove cpu_dai/codec_dai/cpu_dais/codec_dais
Mark Brown [Fri, 27 Mar 2020 15:33:09 +0000 (15:33 +0000)]
Merge series "ASoC: SOF: Intel: add SoundWire support" from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
This patchset provides the support for SoundWire support on Intel
CometLake, IcelLake and TigerLake RVP platforms and form-factor
devices to be released 'soon'.
The bulk of the code is about detecting a valid SoundWire
configuration from ACPI, and implementing the interfaces suggested in
'[PATCH 0/8] soundwire: remove platform devices, add SOF interfaces'
for interrupts, PCI wakes and clock-stop configurations.
Since that SoundWire series will not be in 5.7, the build support for
SOF w/ SoundWire is not provided for now, and fall-back functions will
be used. This code is tested on a daily basis in the SOF tree and is
not expected to change in significant ways.
Changes since v2:
Corrected error in ACPI table (thanks Amadeusz)
Added patch 11 to add reset cycle required on some SoundWire platforms
If pci device is in D0, wakeen interrupt will be
aggregated at cAVS level as interrupt. This commit
check the wakeen status and process it in irq thread
Rander Wang [Wed, 25 Mar 2020 21:50:25 +0000 (16:50 -0500)]
ASoC: SOF: Intel: hda: add WAKEEN interrupt support for SoundWire
When a SoundWire link is in clock stop state, a Slave device may wake
up the Master for some events such as jack detection. The WAKEEN
interrupt will be triggered and processed by the audio pci device.
If audio device is in D3, the interrupt will be routed to PME, or
aggregated at cAVS level as interrupt when audio device is in D0. This
patch only supports D3 case, where the audio pci device will be
resumed by a PME event and the WAKEEN interrupt will be processed
after audio pci device is powered up and ROM is initialized
successfully.
The WAKEEN handling is only enabled after the first boot due to
dependencies on a shim_lock mutex being initialized.
We have a single irq handler for SOF interrupts. We can further merge
SoundWire ones to completely remove MSI interrupts handling issues
leading to timeouts.
For now we have a limited number of machine driver configurations, and
we can detect them based on the link configuration returned after
checking hardware and firmware (BIOS) configurations.
The link configuration is checked with a link_mask as well as a list
of _ADR descriptors for each link.
There is a chance that in extreme cases where the BIOS contains too
much information we would need to detect which Slave devices actually
report as 'attached'. This would be more accurate than static
table-based solutions, but it also introduces timing dependencies
since we don't know when those devices might become attached, so will
only be only be looked at if we see limitations with static methods
and the usual quirks based e.g. on DMI information.
Now that the SoundWire core supports the multi-step initialization,
call the relevant APIs.
The actual hardware enablement can be done in two places, ideally we'd
want to startup the SoundWire IP as soon as possible (while still
taking power rail dependencies into account)
However when suspend/resume is implemented, the DSP device will be
resumed first, and only when the DSP firmware is downloaded/booted
would the SoundWire child devices be resumed, so there are only
marginal benefits in starting the IP earlier for the first probe.
Pierre-Louis Bossart [Wed, 25 Mar 2020 21:50:17 +0000 (16:50 -0500)]
ASoC: soc-acpi: expand description of _ADR-based devices
For SoundWire, we need to know if endpoints needs to be 'aggregated'
(MIPI parlance, meaning logically grouped), e.g. when two speaker
amplifiers need to be handled as a single logical output.
We don't necessarily have the information at the firmware (BIOS)
level, so add a notion of endpoints and specify if a device/endpoint
is part of a group, with a position.
This may be expanded in future solutions, for now only provide a group
and position information.
Since we modify the header file, change all existing upstream tables
as well to avoid breaking compilation/bisect.
Mark Brown [Thu, 26 Mar 2020 19:04:33 +0000 (19:04 +0000)]
Merge series "ASoC: rt1308-sdw: configure amplifier with set_tdm_slot()" from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
When two (or more) amplifiers are on the same link, the integrator may:
a) assign dedicated slots for each of the amplifiers.
b) provide the same configuration to all amplifiers, and rely on
additional controls/processing in the amplifier to generate different
outputs.
case a) was the initial direction for SoundWire and is required for
amplifiers with limited capabilities, but to deal with orientation or
'posture' changes it's easier to implement case b) when the amplifier
can deal with multiple channels.
This patchset suggest the use of the set_tdm_slot() API to define
which of the channels will be consumed by what amplifiers. This maps
well with SoundWire's 'ChannelEnable' registers. The notion of
slot_width is however irrelevant here and ignored, and SoundWire ports
are typically single direction, so only one of the two masks shall be
used.
Pierre-Louis Bossart (2):
ASoC: rt1308-sdw: add set_tdm_slot() support
ASoC: rt1308-sdw: use slot and rx_mask to configure stream
Mark Brown [Thu, 26 Mar 2020 18:01:16 +0000 (18:01 +0000)]
ASoC: pxa: Enable AC'97 bus support for PXA machines
The AC'97 based PXA machines currently don't build reliably as they don't
ensure that an AC'97 bus is built, causing at least eseries_pxa_defconfig
to fail to build. Add selects to fix this.
Mark Brown [Thu, 26 Mar 2020 15:10:53 +0000 (15:10 +0000)]
ASoC: pxa: Select regmap from AC'97 machines
regmap needs to be selected by users which for machine drivers that select
AC'97 CODEC drivers means that we need to also select regmap to ensure that
the CODEC driver will build if nothing else enables regmap as is likely for
such systems.
Pierre-Louis Bossart [Wed, 25 Mar 2020 21:29:05 +0000 (16:29 -0500)]
ASoC: rt1308-sdw: use slot and rx_mask to configure stream
If the DAI was configured with a set_tdm_slots() call, use the information.
A platform or machine driver may configure each amplifier to extract
different bitSlots from the frame, or extract the same data and use
processing to generate the relevant output. The latter case is easier
to handle in case of orientation changes.
Guennadi Liakhovetski [Wed, 25 Mar 2020 21:12:31 +0000 (16:12 -0500)]
ASoC: SOF: fix uninitialised "work" with VirtIO
In the VirtIO case the sof_pcm_open() function isn't called on the
host during guest streaming, which then leaves "work" structures
uninitialised. However it is then used to handle position update
messages from the DSP. Move their initialisation to immediately after
allocation of the containing structure.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com> Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com> Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com> Link: https://lore.kernel.org/r/20200325211233.27394-4-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown <broonie@kernel.org>
Sathyanarayana Nujella [Wed, 25 Mar 2020 21:32:44 +0000 (16:32 -0500)]
ASoC: Intel: sof_rt5682: Add support for tgl-max98373-rt5682
This patch does the below:
1. Adds the driver data and updates quirk info for TGL
with Max98373 speaker amp and ALC5682 headset codec.
2. Added max98373 speaker related code to common file for re-use.
Add "Spk Switch" and associated widget, route to max98360a
speaker amp for power saving, also remove the speaker_amp_init()
callback with complete separated tables for max98373 and max98360a.
Curtis Malainey [Wed, 25 Mar 2020 21:32:42 +0000 (16:32 -0500)]
ASoC: Intel: Make glk+rt5682 echo ref dynamic
Without the dynamic flag to allow runtime routing, the card cannot
probe on chromebooks because SOF is constantly waiting for the link.
Adding flag back to allow upstream kernels to work on rt5682 based
chromebooks since SOF can now ignore the hard coded front end.
Cezary Rojewski [Wed, 25 Mar 2020 13:16:11 +0000 (14:16 +0100)]
ASoC: Intel: bdw-rt5650: Revert SSP0 link to use dummy components
Recent series of patches targeting broadwell boards, while enabling
SOF, changed behavior for non-SOF solutions. In essence replacing
platform 'dummy' with actual 'platform' causes redundant stream
initialization to occur during audio start. hw_params for haswell-pcm
destroys initial stream right after its creation - only to recreate it
again from proceed from there.
While harmless so far, this flow isn't right and should be corrected.
The actual need for dummy components for SSP0 link is questionable but
that issue is subject for another series.
Fixes: a40acc6bfceb ("ASoC: Intel: bdw-rt5650: change cpu_dai and platform components for SOF") Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com> Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20200325131611.545-4-cezary.rojewski@intel.com Signed-off-by: Mark Brown <broonie@kernel.org>
Cezary Rojewski [Wed, 25 Mar 2020 13:16:10 +0000 (14:16 +0100)]
ASoC: Intel: bdw-rt5677: Revert SSP0 link to use dummy components
Recent series of patches targeting broadwell boards, while enabling
SOF, changed behavior for non-SOF solutions. In essence replacing
platform 'dummy' with actual 'platform' causes redundant stream
initialization to occur during audio start. hw_params for haswell-pcm
destroys initial stream right after its creation - only to recreate it
again from proceed from there.
While harmless so far, this flow isn't right and should be corrected.
The actual need for dummy components for SSP0 link is questionable but
that issue is subject for another series.
Fixes: 4865bde187b2 ("ASoC: Intel: bdw-rt5677: change cpu_dai and platform components for SOF") Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com> Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20200325131611.545-3-cezary.rojewski@intel.com Signed-off-by: Mark Brown <broonie@kernel.org>
Cezary Rojewski [Wed, 25 Mar 2020 13:16:09 +0000 (14:16 +0100)]
ASoC: Intel: broadwell: Revert back SSP0 link to use dummy components
Recent series of patches targeting broadwell boards, while enabling
SOF, changed behavior for non-SOF solutions. In essence replacing
platform 'dummy' with actual 'platform' causes redundant stream
initialization to occur during audio start. hw_params for haswell-pcm
destroys initial stream right after its creation - only to recreate it
again from proceed from there.
While harmless so far, this flow isn't correct and should be corrected.
The actual need for dummy components for SSP0 link is questionable but
that issue is subject for another series.
Link to first message in conversation:
https://lkml.org/lkml/2020/3/18/54
Fixes: 64df6afa0dab ("ASoC: Intel: broadwell: change cpu_dai and platform components for SOF") Reported-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com> Acked-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20200325131611.545-2-cezary.rojewski@intel.com Signed-off-by: Mark Brown <broonie@kernel.org>
Paul Cercueil [Fri, 6 Mar 2020 22:29:26 +0000 (23:29 +0100)]
ASoC: Convert jz4740-i2s doc to YAML
Convert the textual binding documentation for the AIC (AC97/I2S
Controller) of Ingenic SoCs to a YAML schema, and add the new compatible
strings in the process.
YueHaibing [Tue, 24 Mar 2020 07:06:15 +0000 (15:06 +0800)]
ASoC: wm8974: remove unused variables
sound/soc/codecs/wm8974.c:200:38: warning:
wm8974_aux_boost_controls defined but not used [-Wunused-const-variable=]
sound/soc/codecs/wm8974.c:204:38: warning:
wm8974_mic_boost_controls defined but not used [-Wunused-const-variable=]
Jonghwan Choi [Thu, 19 Mar 2020 14:00:44 +0000 (23:00 +0900)]
ASoC: tas2562: Fixed incorrect amp_level setting.
According to the tas2562 datasheet,the bits[5:1] represents the amp_level value.
So to set the amp_level value correctly,the shift value should be set to 1.
Mark Brown [Mon, 23 Mar 2020 18:17:26 +0000 (18:17 +0000)]
Merge series "Support built-in Mic on Tegra boards that use WM8903" from Dmitry Osipenko <digetx@gmail.com>:
Hello,
This small series adds audio route for built-in microphone on NVIDIA Tegra
boards that use WM8903 CODEC. In particular this is needed in order to unmute
internal microphone on Acer A500 tablet device. I'm planning to send out the
device tree for the A500 for 5.8, so will be nice to get the microphone
sorted out. Please review and apply, thanks in advance.
Dmitry Osipenko (2):
dt-bindings: sound: tegra-wm8903: Document built-in microphone audio
source
ASoC: tegra: tegra_wm8903: Support DAPM events for built-in microphone
The internal microphone source is needed in order to be able to describe
the hardware audio routing for devices that have the built-in microphone
in addition to the external Mic Jack.
Linus Torvalds [Sun, 22 Mar 2020 18:35:33 +0000 (11:35 -0700)]
Merge tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs fixes from David Sterba:
"Two fixes.
The first is a regression: when dropping some incompat bits the
conditions were reversed. The other is a fix for rename whiteout
potentially leaving stack memory linked to a list"
* tag 'for-5.6-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: fix removal of raid[56|1c34} incompat flags after removing block group
btrfs: fix log context list corruption after rename whiteout error
Linus Torvalds [Sun, 22 Mar 2020 17:46:50 +0000 (10:46 -0700)]
Merge branch 'akpm' (patches from Andrew)
Merge misc fixes from Andrew Morton:
"10 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
x86/mm: split vmalloc_sync_all()
mm, slub: prevent kmalloc_node crashes and memory leaks
mm/mmu_notifier: silence PROVE_RCU_LIST warnings
epoll: fix possible lost wakeup on epoll_ctl() path
mm: do not allow MADV_PAGEOUT for CoW pages
mm, memcg: throttle allocators based on ancestral memory.high
mm, memcg: fix corruption on 64-bit divisor in memory.high throttling
page-flags: fix a crash at SetPageError(THP_SWAP)
mm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case
memcg: fix NULL pointer dereference in __mem_cgroup_usage_unregister_event
Joerg Roedel [Sun, 22 Mar 2020 01:22:41 +0000 (18:22 -0700)]
x86/mm: split vmalloc_sync_all()
Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in
__purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in
the vunmap() code-path. While this change was necessary to maintain
correctness on x86-32-pae kernels, it also adds additional cycles for
architectures that don't need it.
Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported
severe performance regressions in micro-benchmarks because it now also
calls the x86-64 implementation of vmalloc_sync_all() on vunmap(). But
the vmalloc_sync_all() implementation on x86-64 is only needed for newly
created mappings.
To avoid the unnecessary work on x86-64 and to gain the performance
back, split up vmalloc_sync_all() into two functions:
* vmalloc_sync_mappings(), and
* vmalloc_sync_unmappings()
Most call-sites to vmalloc_sync_all() only care about new mappings being
synchronized. The only exception is the new call-site added in the
above mentioned commit.
Shile Zhang directed us to a report of an 80% regression in reaim
throughput.
Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") Reported-by: kernel test robot <oliver.sang@intel.com> Reported-by: Shile Zhang <shile.zhang@linux.alibaba.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Borislav Petkov <bp@suse.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [GHES] Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: <stable@vger.kernel.org> Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/ Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This only happens with a mmotm patch "mm/memcontrol.c: allocate
shrinker_map on appropriate NUMA node" [2] which effectively calls
kmalloc_node for each possible node. SLUB however only allocates
kmem_cache_node on online N_NORMAL_MEMORY nodes, and relies on
node_to_mem_node to return such valid node for other nodes since commit a561ce00b09e ("slub: fall back to node_to_mem_node() node if allocating
on memoryless node"). This is however not true in this configuration
where the _node_numa_mem_ array is not initialized for nodes 0 and 2-31,
thus it contains zeroes and get_partial() ends up accessing
non-allocated kmem_cache_node.
A related issue was reported by Bharata (originally by Ramachandran) [3]
where a similar PowerPC configuration, but with mainline kernel without
patch [2] ends up allocating large amounts of pages by kmalloc-1k
kmalloc-512. This seems to have the same underlying issue with
node_to_mem_node() not behaving as expected, and might probably also
lead to an infinite loop with CONFIG_SLUB_CPU_PARTIAL [4].
This patch should fix both issues by not relying on node_to_mem_node()
anymore and instead simply falling back to NUMA_NO_NODE, when
kmalloc_node(node) is attempted for a node that's not online, or has no
usable memory. The "usable memory" condition is also changed from
node_present_pages() to N_NORMAL_MEMORY node state, as that is exactly
the condition that SLUB uses to allocate kmem_cache_node structures.
The check in get_partial() is removed completely, as the checks in
___slab_alloc() are now sufficient to prevent get_partial() being
reached with an invalid node.
Qian Cai [Sun, 22 Mar 2020 01:22:34 +0000 (18:22 -0700)]
mm/mmu_notifier: silence PROVE_RCU_LIST warnings
It is safe to traverse mm->notifier_subscriptions->list either under
SRCU read lock or mm->notifier_subscriptions->lock using
hlist_for_each_entry_rcu(). Silence the PROVE_RCU_LIST false positives,
for example,
Signed-off-by: Qian Cai <cai@lca.pw> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Jason Gunthorpe <jgg@mellanox.com> Link: http://lkml.kernel.org/r/20200317175640.2047-1-cai@lca.pw Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Roman Penyaev [Sun, 22 Mar 2020 01:22:30 +0000 (18:22 -0700)]
epoll: fix possible lost wakeup on epoll_ctl() path
This fixes possible lost wakeup introduced by commit a218cc491420.
Originally modifications to ep->wq were serialized by ep->wq.lock, but
in commit a218cc491420 ("epoll: use rwlock in order to reduce
ep_poll_callback() contention") a new rw lock was introduced in order to
relax fd event path, i.e. callers of ep_poll_callback() function.
After the change ep_modify and ep_insert (both are called on epoll_ctl()
path) were switched to ep->lock, but ep_poll (epoll_wait) was using
ep->wq.lock on wqueue list modification.
The bug doesn't lead to any wqueue list corruptions, because wake up
path and list modifications were serialized by ep->wq.lock internally,
but actual waitqueue_active() check prior wake_up() call can be
reordered with modifications of ep ready list, thus wake up can be lost.
And yes, can be healed by explicit smp_mb():
list_add_tail(&epi->rdlink, &ep->rdllist);
smp_mb();
if (waitqueue_active(&ep->wq))
wake_up(&ep->wp);
But let's make it simple, thus current patch replaces ep->wq.lock with
the ep->lock for wqueue modifications, thus wake up path always observes
activeness of the wqueue correcty.
Fixes: a218cc491420 ("epoll: use rwlock in order to reduce ep_poll_callback() contention") Reported-by: Max Neunhoeffer <max@arangodb.com> Signed-off-by: Roman Penyaev <rpenyaev@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Max Neunhoeffer <max@arangodb.com> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Christopher Kohlhoff <chris.kohlhoff@clearpool.io> Cc: Davidlohr Bueso <dbueso@suse.de> Cc: Jason Baron <jbaron@akamai.com> Cc: Jes Sorensen <jes.sorensen@gmail.com> Cc: <stable@vger.kernel.org> [5.1+] Link: http://lkml.kernel.org/r/20200214170211.561524-1-rpenyaev@suse.de
References: https://bugzilla.kernel.org/show_bug.cgi?id=205933 Bisected-by: Max Neunhoeffer <max@arangodb.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Michal Hocko [Sun, 22 Mar 2020 01:22:26 +0000 (18:22 -0700)]
mm: do not allow MADV_PAGEOUT for CoW pages
Jann has brought up a very interesting point [1]. While shared pages
are excluded from MADV_PAGEOUT normally, CoW pages can be easily
reclaimed that way. This can lead to all sorts of hard to debug
problems. E.g. performance problems outlined by Daniel [2].
There are runtime environments where there is a substantial memory
shared among security domains via CoW memory and a easy to reclaim way
of that memory, which MADV_{COLD,PAGEOUT} offers, can lead to either
performance degradation in for the parent process which might be more
privileged or even open side channel attacks.
The feasibility of the latter is not really clear to me TBH but there is
no real reason for exposure at this stage. It seems there is no real
use case to depend on reclaiming CoW memory via madvise at this stage so
it is much easier to simply disallow it and this is what this patch
does. Put it simply MADV_{PAGEOUT,COLD} can operate only on the
exclusively owned memory which is a straightforward semantic.
Chris Down [Sun, 22 Mar 2020 01:22:23 +0000 (18:22 -0700)]
mm, memcg: throttle allocators based on ancestral memory.high
Prior to this commit, we only directly check the affected cgroup's
memory.high against its usage. However, it's possible that we are being
reclaimed as a result of hitting an ancestor memory.high and should be
penalised based on that, instead.
This patch changes memory.high overage throttling to use the largest
overage in its ancestors when considering how many penalty jiffies to
charge. This makes sure that we penalise poorly behaving cgroups in the
same way regardless of at what level of the hierarchy memory.high was
breached.
Fixes: 0e4b01df8659 ("mm, memcg: throttle allocators when failing reclaim over memory.high") Reported-by: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Chris Down <chris@chrisdown.name> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Tejun Heo <tj@kernel.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Nathan Chancellor <natechancellor@gmail.com> Cc: Roman Gushchin <guro@fb.com> Cc: <stable@vger.kernel.org> [5.4.x+] Link: http://lkml.kernel.org/r/8cd132f84bd7e16cdb8fde3378cdbf05ba00d387.1584036142.git.chris@chrisdown.name Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
projects / users / jedix / linux-maple.git / log