An SoC can contain multiple power domains with individual or collection
of mesh partitions. This partition is called fabric cluster.
Certain type of meshes will need to run at the same frequency, they will
be placed in the same fabric cluster. Benefit of fabric cluster is that
it offers a scalable mechanism to deal with partitioned fabrics in a SoC.
The current sysfs interface supports control at package and die level.
This interface is not enough to support more granular control at
fabric cluster level.
SoCs with the support of TPMI (Topology Aware Register and PM Capsule
Interface), can have multiple power domains. Each power domain can
contain one or more fabric clusters.
To support such granular controls, enhance uncore common to optionally
create new directories to provide controls at fabric cluster level. It
is also important to have flexibility to change granularity for future
version of SoCs. If the directory name contains scope like:
"package_*_die_*_power_domain_*_cluster_*", then this is not expandable.
The cpufreq policies also have different scopes. There the scope of the
policy (affected_cpus) specified by attributes inside each policy.
So, follow the same model for uncore frequency scaling sysfs as:
"sys/devices/system/cpu/cpufreq/policy*"
Allow client drivers to optionally support granular control for each
fabric cluster. Here, the directory name will be "uncore" suffixed with
an unique instance number. For example: uncore00, uncore01 etc.
Attributes in the directory identify package id, power domain and
fabric cluster id. This interface is expandable even if some new level
of granularity is introduced. A new sysfs attribute can identify new
level.
For compatibility with the existing sysfs and provide easy way to set
limits for each fabric cluster in the package/die, the existing control
at package/die levels are still provided. For majority of users, this is
an easy approach.
For example: On a single package/die system, with three power domains
and one fabric cluster per power domain:
$tree -L 2 /sys/devices/system/cpu/intel_uncore_frequency/
/sys/devices/system/cpu/intel_uncore_frequency/
├── package_00_die_00
│ ├── current_freq_khz
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ └── min_freq_khz
├── uncore00
│ ├── current_freq_khz
│ ├── domain_id
│ ├── fabric_cluster_id
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ ├── min_freq_khz
│ └── package_id
├── uncore01
│ ├── current_freq_khz
│ ├── domain_id
│ ├── fabric_cluster_id
│ ├── initial_max_freq_khz
│ ├── initial_min_freq_khz
│ ├── max_freq_khz
│ ├── min_freq_khz
│ └── package_id
└── uncore02
├── current_freq_khz
├── domain_id
├── fabric_cluster_id
├── initial_max_freq_khz
├── initial_min_freq_khz
├── max_freq_khz
├── min_freq_khz
└── package_id
The attribute for cluster id is "fabric_cluster_id" instead of just
"cluster_id" is to avoid confusion with usage of term clusters in
other part of the Linux kernel.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
Link: https://lore.kernel.org/r/20230418171340.681662-3-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Implement support of uncore frequency control via TPMI (Topology Aware
Register and PM Capsule Interface). This driver provides the similar
functionality as the current uncore frequency driver using MSRs.
The hardware interface to read/write is basically substitution of MSR
0x620 and 0x621. There are specific MMIO offset and bits to get/set
minimum and maximum uncore ratio, similar to MSRs.
The scope of the uncore MSRs is package/die. But new generation of CPUs
have more granular control at a cluster level. Each package/die can have
multiple power domains, which further can have multiple clusters. The
TPMI interface allows control at cluster level.
The primary use case for uncore sysfs is to set maximum and minimum
uncore frequency to reduce power consumption or latency. The current
uncore sysfs control is per package/die. This is enough for the majority
of users as workload will move to different power domains as it moves
between different CPUs.
The current uncore sysfs provides controls at package/die level. When
user sets maximum/minimum limits, the driver sets the same limits to
each cluster.
Here number of power domains = number of resources in this aux device.
There are offsets and bits to discover number of clusters and offset for
each cluster level controls.
The TPMI documentation can be downloaded from:
https://github.com/intel/tpmi_power_management
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
Link: https://lore.kernel.org/r/20230420220514.747573-1-srinivas.pandruvada@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Synchronize the ABI documentation with the driver documentation
and direct users to the latter in case the search for more
detailed information.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Link: https://lore.kernel.org/r/20230508204241.11076-2-W_Armin@gmx.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The WMI interface used by the dell-wmi-ddv driver contains
many methods which are currently unused, making it difficult
to document these inside the drivers source code.
Create the necessary documentation based on current knowledge
so that all details of the WMI interface can be written down
for later use.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Link: https://lore.kernel.org/r/20230508204241.11076-1-W_Armin@gmx.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Add a place for device-specific documentation of WMI drivers.
The first entry is documentation for the wmi-bmof driver, with
additional documentation being expected to follow.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20230424222939.208137-5-W_Armin@gmx.de
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Add documentation for the WMI subsystem. The documentation describes
both the ACPI WMI interface and the driver API for interacting with
the WMI driver core. The information regarding the ACPI interface
was retrieved from the Ubuntu kernel references and the Windows driver
samples available on GitHub. The documentation is supposed to help
driver developers writing WMI drivers, as many modern machines designed
to run Windows provide an ACPI WMI interface.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20230424222939.208137-4-W_Armin@gmx.de
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The WMI driver core supports a more mordern bus-based interface for
interacting with WMI devices. The older GUID-based interface depends
on each WMI GUID and notification id being unique on a given system,
which turned out is not the case.
Mark the older interface as deprecated since new WMI drivers should
use the bus-based interface to avoid this issues.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20230424222939.208137-3-W_Armin@gmx.de
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Add kernel doc comments useful for documenting the functions/structs
used to interact with the WMI driver core.
Signed-off-by: Armin Wolf <W_Armin@gmx.de>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20230424222939.208137-2-W_Armin@gmx.de
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
This commit adds memory barrier for the 'vq' update in function
mlxbf_tmfifo_virtio_find_vqs() to avoid potential race due to
out-of-order memory write. It also adds barrier for the 'is_ready'
flag to make sure the initializations are visible before this flag
is checked.
Signed-off-by: Liming Sun <limings@nvidia.com>
Reviewed-by: Vadim Pasternak <vadimp@nvidia.com>
Link: https://lore.kernel.org/r/b98c0ab61d644ba38fa9b3fd1607b138b0dd820b.1682518748.git.limings@nvidia.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Perform 's@ \t@\t\t@g' so we wouldn't have spaces followed by tabs.
No functional change.
Picked from U-Boot commit 0cf207ec01c ("WS cleanup: remove SPACE(s) followed by TAB")
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20230507130120.7587-1-marek.vasut+renesas@mailbox.org
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
For development and testing purposes, the i915.force_probe module
parameter and DRM_I915_FORCE_PROBE kconfig option allow probing of
devices that aren't supported by the driver.
The i915.force_probe module parameter is "unsafe" and setting it taints
the kernel. However, using the kconfig option does not.
Always taint the kernel when force probing a device that is not
supported.
v2: Drop "depends on EXPERT" to avoid build breakage (kernel test robot)
Fixes: 7ef5ef5cde ("drm/i915: add force_probe module parameter to replace alpha_support")
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Dave Airlie <airlied@gmail.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230504103508.1818540-1-jani.nikula@intel.com
Merge series from Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>:
Many ASoC drivers are using dummy DAI.
I have 2 concern about it. 1st one is there is no guarantee that local
strings ("snd-soc-dummy-dai", "snd-soc-dummy") are kept until the card
was binded if it was added at subfunction.
2nd one is we can use common snd_soc_dai_link_component for it.
This patch-set adds common asoc_dummy_dlc, and use it.
If devm_reset_control_get_optional_exclusive() fails, some resources still
need to be released. So branch to the error handling path instead of
returning directly.
Fixes: aad53d4ee7 ("mmc: sdhci-cadence: Support mmc hardware reset")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Brad Larson <blarson@amd.com>
Link: https://lore.kernel.org/r/f61599a9ef23767c2d66e5af9c975f05ef1cec6b.1682430069.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Requests to the mmc layer usually come through a block device IO.
The exceptions are the ioctl interface, RPMB chardev ioctl
and debugfs, which issue their own blk_mq requests through
blk_execute_rq and do not query the BLK_STS error but the
mmcblk-internal drv_op_result. This patch ensures that drv_op_result
defaults to an error and has to be overwritten by the operation
to be considered successful.
The behavior leads to a bug where the request never propagates
the error, e.g. by directly erroring out at mmc_blk_mq_issue_rq if
mmc_blk_part_switch fails. The ioctl caller of the rpmb chardev then
can never see an error (BLK_STS_IOERR, but drv_op_result is unchanged)
and thus may assume that their call executed successfully when it did not.
While always checking the blk_execute_rq return value would be
advised, let's eliminate the error by always setting
drv_op_result as -EIO to be overwritten on success (or other error)
Fixes: 614f0388f5 ("mmc: block: move single ioctl() commands to block requests")
Signed-off-by: Christian Loehle <cloehle@hyperstone.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/59c17ada35664b818b7bd83752119b2d@hyperstone.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Add return value for dim_calc_stats. This is an indication for the
caller if curr_stats was assigned by the function. Avoid using
curr_stats uninitialized over {rdma/net}_dim, when no time delta between
samples. Coverity reported this potential use of an uninitialized
variable.
Fixes: 4c4dbb4a73 ("net/mlx5e: Move dynamic interrupt coalescing code to include/linux")
Fixes: cb3c7fd4f8 ("net/mlx5e: Support adaptive RX coalescing")
Signed-off-by: Roy Novich <royno@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Link: https://lore.kernel.org/r/20230507135743.138993-1-tariqt@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add documentation for the DRM_MODE_TV_MODE_MAX enumerator to fix the
kernel-doc warning:
include/drm/drm_connector.h:204: warning: Enum value 'DRM_MODE_TV_MODE_MAX' not described in enum 'drm_connector_tv_mode'
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Link: https://patchwork.freedesktop.org/patch/msgid/20230504123444.1843795-1-jani.nikula@intel.com
There has been a lot of confusion around which platform profiles are
supported on various platforms and it would be useful to have a debug
method to be able to override the profile mode that is selected.
I don't expect this to be used in anything other than debugging in
conjunction with Lenovo engineers - but it does give a way to get a
system working whilst we wait for either FW fixes, or a driver fix
to land upstream, if something is wonky in the mode detection logic
Signed-off-by: Mark Pearson <mpearson-lenovo@squebb.ca>
Link: https://lore.kernel.org/r/20230505132523.214338-2-mpearson-lenovo@squebb.ca
Cc: stable@vger.kernel.org
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
I had incorrectly thought that PSC profiles were not usable on Intel
platforms so had blocked them in the driver initialistion. This broke
platform profiles on the T490.
After discussion with the FW team PSC does work on Intel platforms and
should be allowed.
Note - it's possible this may impact other platforms where it is advertised
but special driver support that only Windows has is needed. But if it does
then they will need fixing via quirks. Please report any issues to me so I
can get them addressed - but I haven't found any problems in testing...yet
Fixes: bce6243f76 ("platform/x86: thinkpad_acpi: do not use PSC mode on Intel platforms")
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2177962
Cc: stable@vger.kernel.org
Signed-off-by: Mark Pearson <mpearson-lenovo@squebb.ca>
Link: https://lore.kernel.org/r/20230505132523.214338-1-mpearson-lenovo@squebb.ca
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Currently when the uncore_write() returns error, it is silently
ignored. Return error to user space when uncore_write() fails.
Fixes: 49a474c7ba ("platform/x86: Add support for Uncore frequency control")
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20230418153230.679094-1-srinivas.pandruvada@linux.intel.com
Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
As all level 2 and level 3 caches are unified, add required
cache-unified properties to fix warnings like:
meson-a1-ad401.dtb: l2-cache0: 'cache-unified' is a required property
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Link: https://lore.kernel.org/r/20230421223211.115612-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
With the current device tree for meson8b, uarts B (e.g. available on pins
8/10 on Odroid-C1) and C (pins 3/5 on Odroid-C1) do not work, because they
are relying on incorrect clocks. Change the references of pclk to the
correct CLKID, to allow use of the two uarts.
Fixes: 3375aa7713 ("ARM: dts: meson8b: Fix the UART device-tree schema validation")
Signed-off-by: Hans-Frieder Vogt <hfdevel@gmx.net>
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Link: https://lore.kernel.org/r/trinity-bf20bcb9-790b-4ab9-99e3-0831ef8257f4-1680878185420@3c-app-gmx-bap55
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Document the new C3 SoC/board device tree bindings.
C3 is an advanced edge AI processor designed for smart IP camera
applications, which does not belong to Meson series.
Therefore, modify the title field accordingly.
Signed-off-by: Kelvin Zhang <kelvin.zhang@amlogic.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230407102704.1055152-1-kelvin.zhang@amlogic.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
When `QUIRK_AUTO_CLEAR_INT` isn't set, interrupt masking should be
cleared by writing to Interrupt Mask Clear (IMR) and interrupt
status should be cleared properly at shutdown/init.
This fixes an error where interrupts are left enabled during resume
from hibernation with `CONFIG_USB4=y`.
Fixes: 468c49f447 ("thunderbolt: Disable interrupt auto clear for rings")
Cc: stable@vger.kernel.org # v6.3
Reported-by: Takashi Iwai <tiwai@suse.de>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217343
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Merge series from AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>:
This series performs some cleanups for mainly MT8195 and switches
both MT8195 and MT8186's SOF driver to the snd_sof_ipc_process_reply()
helper.
Merge series from Mårten Lindahl <marten.lindahl@axis.com>:
This series adds basic support for TI's TPS62870/TPS62871/TPS62872/
TPS62873 high-frequency single-channel step-down converters with an
I2C interface.
The devices can operate in power save mode for maximum efficiency, or
forced-PWM mode for best transient performance and lowest output
voltage ripple. All chip variants have four output voltage ranges and
the driver changes active range depending on the requested voltage
setting.
There are differences in the electrical characteristics and packaging
between the variants, but the register interfaces are identical.
assigned-clocks are a dependency of clocks, however the dtschema has
limitation and expects clocks to be present in the binding using
assigned-clocks, not in other referenced bindings. The clocks were
defined in common fsl,imx6q-pcie-common.yaml, which is referenced by
fsl,imx6q-pcie-ep.yaml. The fsl,imx6q-pcie-ep.yaml used assigned-clocks
thus leading to warnings:
Documentation/devicetree/bindings/pci/fsl,imx6q-pcie-ep.example.dtb: pcie-ep@33800000:
Unevaluated properties are not allowed ('assigned-clock-parents', 'assigned-clock-rates', 'assigned-clocks' were unexpected)
From schema: Documentation/devicetree/bindings/pci/fsl,imx6q-pcie-ep.yaml
Fix this by moving clocks to each specific schema from the common one
and narrowing them to strictly match what is expected for given device.
Fixes: b10f82380e ("dt-bindings: imx6q-pcie: Restruct i.MX PCIe schema")
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Richard Zhu <hongxing.zhu@nxp.com>
Link: https://lore.kernel.org/r/20230508071837.68552-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
rasize (ra_pages) should be set higher than read size by default
to allow parallel reads when reading large files in order to
improve performance (otherwise there is much dead time on the
network when doing readahead of large files). Default rasize
to twice readsize.
Signed-off-by: Steve French <stfrench@microsoft.com>
The driver's probe() first registers regulators in a loop and then in a
second loop passes them as irq data to the interrupt handlers. However
the function to get the regulator for given name
tps65219_get_rdev_by_name() was a no-op due to argument passed by value,
not pointer, thus the second loop assigned always same value - from
previous loop. The interrupts, when fired, where executed with wrong
data. Compiler also noticed it:
drivers/regulator/tps65219-regulator.c: In function ‘tps65219_get_rdev_by_name’:
drivers/regulator/tps65219-regulator.c:292:60: error: parameter ‘dev’ set but not used [-Werror=unused-but-set-parameter]
Fixes: c12ac5fc3e ("regulator: drivers: Add TI TPS65219 PMIC regulators support")
Cc: <stable@vger.kernel.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org
Reviewed-by: Markus Schneider-Pargmann <msp@baylibre.com
Link: https://lore.kernel.org/r/20230507144656.192800-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org
Add support for Texas Instruments TPS6287x, single-channel
synchronous step-down converters with four output voltage ranges.
Signed-off-by: Mårten Lindahl <marten.lindahl@axis.com
Link: https://lore.kernel.org/r/20230502-tps6287x-driver-v3-2-e25140a023f5@axis.com
Signed-off-by: Mark Brown <broonie@kernel.org
Add bindings for the TPS62870/TPS62871/TPS62872/TPS62873 voltage
regulators.
Signed-off-by: Mårten Lindahl <marten.lindahl@axis.com
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org
Link: https://lore.kernel.org/r/20230502-tps6287x-driver-v3-1-e25140a023f5@axis.com
Signed-off-by: Mark Brown <broonie@kernel.org
trans_pcie->rba.alloc_wq only hosts a single work item and thus doesn't need
explicit concurrency limit. Let's use the default @max_active. This doesn't
cost anything and clearly expresses that @max_active doesn't matter.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Gregory Greenman <gregory.greenman@intel.com>
Cc: Avraham Stern <avraham.stern@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mordechay Goodstein <mordechay.goodstein@intel.com>
Cc: "Haim, Dreyfuss" <haim.dreyfuss@intel.com>
Cc: linux-wireless@vger.kernel.org
Cc: netdev@vger.kernel.org
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
Cc: xen-devel@lists.xenproject.org
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Fei Li <fei1.li@intel.com>
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Sunil Goutham <sgoutham@marvell.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Ratheesh Kannoth <rkannoth@marvell.com>
Cc: Srujana Challa <schalla@marvell.com>
Cc: Geetha sowjanya <gakula@marvell.com>
Cc: netdev@vger.kernel.org
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Sunil Goutham <sgoutham@marvell.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: netdev@vger.kernel.org
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Johan Hovold <johan@kernel.org>
Acked-by: Alex Elder <elder@kernel.org>
Cc: greybus-dev@lists.linaro.org
BACKGROUND
==========
When multiple work items are queued to a workqueue, their execution order
doesn't match the queueing order. They may get executed in any order and
simultaneously. When fully serialized execution - one by one in the queueing
order - is needed, an ordered workqueue should be used which can be created
with alloc_ordered_workqueue().
However, alloc_ordered_workqueue() was a later addition. Before it, an
ordered workqueue could be obtained by creating an UNBOUND workqueue with
@max_active==1. This originally was an implementation side-effect which was
broken by 4c16bd327c ("workqueue: restore WQ_UNBOUND/max_active==1 to be
ordered"). Because there were users that depended on the ordered execution,
5c0338c687 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
made workqueue allocation path to implicitly promote UNBOUND workqueues w/
@max_active==1 to ordered workqueues.
While this has worked okay, overloading the UNBOUND allocation interface
this way creates other issues. It's difficult to tell whether a given
workqueue actually needs to be ordered and users that legitimately want a
min concurrency level wq unexpectedly gets an ordered one instead. With
planned UNBOUND workqueue updates to improve execution locality and more
prevalence of chiplet designs which can benefit from such improvements, this
isn't a state we wanna be in forever.
This patch series audits all callsites that create an UNBOUND workqueue w/
@max_active==1 and converts them to alloc_ordered_workqueue() as necessary.
WHAT TO LOOK FOR
================
The conversions are from
alloc_workqueue(WQ_UNBOUND | flags, 1, args..)
to
alloc_ordered_workqueue(flags, args...)
which don't cause any functional changes. If you know that fully ordered
execution is not ncessary, please let me know. I'll drop the conversion and
instead add a comment noting the fact to reduce confusion while conversion
is in progress.
If you aren't fully sure, it's completely fine to let the conversion
through. The behavior will stay exactly the same and we can always
reconsider later.
As there are follow-up workqueue core changes, I'd really appreciate if the
patch can be routed through the workqueue tree w/ your acks. Thanks.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Nathan Lynch <nathanl@linux.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
cpuset_can_attach() can fail. Postpone DL BW allocation until all tasks
have been checked. DL BW is not allocated per-task but as a sum over
all DL tasks migrating.
If multiple controllers are attached to the cgroup next to the cpuset
controller a non-cpuset can_attach() can fail. In this case free DL BW
in cpuset_cancel_attach().
Finally, update cpuset DL task count (nr_deadline_tasks) only in
cpuset_attach().
Suggested-by: Waiman Long <longman@redhat.com>
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
While moving a set of tasks between exclusive cpusets,
cpuset_can_attach() -> task_can_attach() calls dl_cpu_busy(..., p) for
DL BW overflow checking and per-task DL BW allocation on the destination
root_domain for the DL tasks in this set.
This approach has the issue of not freeing already allocated DL BW in
the following error cases:
(1) The set of tasks includes multiple DL tasks and DL BW overflow
checking fails for one of the subsequent DL tasks.
(2) Another controller next to the cpuset controller which is attached
to the same cgroup fails in its can_attach().
To address this problem rework dl_cpu_busy():
(1) Split it into dl_bw_check_overflow() & dl_bw_alloc() and add a
dedicated dl_bw_free().
(2) dl_bw_alloc() & dl_bw_free() take a `u64 dl_bw` parameter instead of
a `struct task_struct *p` used in dl_cpu_busy(). This allows to
allocate DL BW for a set of tasks too rather than only for a single
task.
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
update_tasks_root_domain currently iterates over all tasks even if no
DEADLINE task is present on the cpuset/root domain for which bandwidth
accounting is being rebuilt. This has been reported to introduce 10+ ms
delays on suspend-resume operations.
Skip the costly iteration for cpusets that don't contain DEADLINE tasks.
Reported-by: Qais Yousef <qyousef@layalina.io>
Link: https://lore.kernel.org/lkml/20230206221428.2125324-1-qyousef@layalina.io/
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Qais reported that iterating over all tasks when rebuilding root domains
for finding out which ones are DEADLINE and need their bandwidth
correctly restored on such root domains can be a costly operation (10+
ms delays on suspend-resume).
To fix the problem keep track of the number of DEADLINE tasks belonging
to each cpuset and then use this information (followup patch) to only
perform the above iteration if DEADLINE tasks are actually present in
the cpuset for which a corresponding root domain is being rebuilt.
Reported-by: Qais Yousef <qyousef@layalina.io>
Link: https://lore.kernel.org/lkml/20230206221428.2125324-1-qyousef@layalina.io/
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Turns out percpu_cpuset_rwsem - commit 1243dc518c ("cgroup/cpuset:
Convert cpuset_mutex to percpu_rwsem") - wasn't such a brilliant idea,
as it has been reported to cause slowdowns in workloads that need to
change cpuset configuration frequently and it is also not implementing
priority inheritance (which causes troubles with realtime workloads).
Convert percpu_cpuset_rwsem back to regular cpuset_mutex. Also grab it
only for SCHED_DEADLINE tasks (other policies don't care about stable
cpusets anyway).
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
rebuild_root_domains() and update_tasks_root_domain() have neutral
names, but actually deal with DEADLINE bandwidth accounting.
Rename them to use 'dl_' prefix so that intent is more clear.
No functional change.
Suggested-by: Qais Yousef <qyousef@layalina.io>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
When a tick broadcast clockevent device is initialized for one shot mode
then tick_broadcast_setup_oneshot() OR's the periodic broadcast mode
cpumask into the oneshot broadcast cpumask.
This is required when switching from periodic broadcast mode to oneshot
broadcast mode to ensure that CPUs which are waiting for periodic
broadcast are woken up on the next tick.
But it is subtly broken, when an active broadcast device is replaced and
the system is already in oneshot (NOHZ/HIGHRES) mode. Victor observed
this and debugged the issue.
Then the OR of the periodic broadcast CPU mask is wrong as the periodic
cpumask bits are sticky after tick_broadcast_enable() set it for a CPU
unless explicitly cleared via tick_broadcast_disable().
That means that this sets all other CPUs which have tick broadcasting
enabled at that point unconditionally in the oneshot broadcast mask.
If the affected CPUs were already idle and had their bits set in the
oneshot broadcast mask then this does no harm. But for non idle CPUs
which were not set this corrupts their state.
On their next invocation of tick_broadcast_enable() they observe the bit
set, which indicates that the broadcast for the CPU is already set up.
As a consequence they fail to update the broadcast event even if their
earliest expiring timer is before the actually programmed broadcast
event.
If the programmed broadcast event is far in the future, then this can
cause stalls or trigger the hung task detector.
Avoid this by telling tick_broadcast_setup_oneshot() explicitly whether
this is the initial switch over from periodic to oneshot broadcast which
must take the periodic broadcast mask into account. In the case of
initialization of a replacement device this prevents that the broadcast
oneshot mask is modified.
There is a second problem with broadcast device replacement in this
function. The broadcast device is only armed when the previous state of
the device was periodic.
That is correct for the switch from periodic broadcast mode to oneshot
broadcast mode as the underlying broadcast device could operate in
oneshot state already due to lack of periodic state in hardware. In that
case it is already armed to expire at the next tick.
For the replacement case this is wrong as the device is in shutdown
state. That means that any already pending broadcast event will not be
armed.
This went unnoticed because any CPU which goes idle will observe that
the broadcast device has an expiry time of KTIME_MAX and therefore any
CPUs next timer event will be earlier and cause a reprogramming of the
broadcast device. But that does not guarantee that the events of the
CPUs which were already in idle are delivered on time.
Fix this by arming the newly installed device for an immediate event
which will reevaluate the per CPU expiry times and reprogram the
broadcast device accordingly. This is simpler than caching the last
expiry time in yet another place or saving it before the device exchange
and handing it down to the setup function. Replacement of broadcast
devices is not a frequent operation and usually happens once somewhere
late in the boot process.
Fixes: 9c336c9935 ("tick/broadcast: Allow late registered device to enter oneshot mode")
Reported-by: Victor Hassan <victor@allwinnertech.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/87pm7d2z1i.ffs@tglx
