| Age | Commit message (Collapse) | Author |
|
Extract the GRF-level HPD interrupt mask/unmask register writes into
dedicated dw_hdmi_qp_rk35{88,76}_{mask,unmask}_hpd() helpers, and use
them from the .enable_hpd()/.disable_hpd() phy ops as well as the
hardirq, threaded re-arm and io_init() paths.
This is a preparatory refactor that separates the SoC-level HPD mask
from the phy op entry points, so a subsequent change can gate the
platform HPD IRQ line without affecting the debounce paths.
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-13-195157ba0b40@collabora.com
|
|
Implement .enable_hpd() and .disable_hpd() phy ops for RK3576 and RK3588
SoCs, used by the corresponding bridge callbacks for HPD activation
control.
Consolidate the interrupt clear-and-unmask sequence into enable_hpd()
and the mask-only operation into disable_hpd(), replacing the open-coded
register writes in the interrupt handlers and io_init().
The .setup_hpd() phy op, which was previously called from
dw_hdmi_qp_init_hw(), is no longer needed and its users are removed.
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-12-195157ba0b40@collabora.com
|
|
Start with HPD masked, matching RK3588, so the bridge framework controls
when interrupts are armed via hpd_enable().
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-11-195157ba0b40@collabora.com
|
|
The RK3576 and RK3588 HPD hardirq handlers read shared status registers
and currently wake the threaded handler whenever any bit in those
registers is set. This can wake the HDMI IRQ thread for unrelated
status bits, and on RK3588 can also wake the thread for the other HDMI
port.
Mask the status value with the SoC-specific HPD interrupt bit before
deciding whether to wake the threaded handler. On RK3588, select the
HDMI0 or HDMI1 HPD bit according to the instance port id.
No stable backport is needed. The issue may only cause spurious IRQ
thread wakeups and does not cause functional breakage.
Reported-by: Sashiko <sashiko-bot@kernel.org>
Closes: https://lore.kernel.org/all/20260702151346.5A4EC1F000E9@smtp.kernel.org/
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Link: https://lore.kernel.org/all/20260702151346.5A4EC1F000E9@smtp.kernel.org/
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-10-195157ba0b40@collabora.com
|
|
Replace indirect struct device accesses via hdmi->dev and pdev->dev with
the local dev parameter already available in dw_hdmi_qp_rockchip_bind(),
for consistency and readability.
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-9-195157ba0b40@collabora.com
|
|
Add the missing trailing newlines to a couple of dev_err_probe() calls
in dw_hdmi_qp_rockchip_bind().
Fixes: b6736a4ea3fa ("drm/rockchip: dw_hdmi_qp: Improve error handling with dev_err_probe()")
Fixes: e1f7b7cbd74c ("drm/rockchip: dw_hdmi_qp: Switch to drmm_encoder_init()")
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-8-195157ba0b40@collabora.com
|
|
The HPD threaded handler debounces hotplug events by delaying
hdmi->hpd_work for 150 ms. Nothing discards that work when the system
goes to sleep: drm_kms_helper_poll_disable() only disables the HPD IRQ
line via the bridge .hpd_disable() op, and disable_irq() merely waits
for the threaded handler, which has already returned after arming the
timer.
If the debounce expires during suspend, its non-freezable work may probe
the connector and read the EDID over DDC. Since genpd powers off the
HDMI controller's power domain in ->suspend_noirq(), this access may
hit unpowered hardware and hang the system.
Cancel the work in the suspend callback. The DRM device is registered
as a consumer of the HDMI device, so it always suspends first and the
HPD IRQ is already disabled by that point, meaning nothing can re-arm
the work behind our back. No counterpart is needed on resume, as the
work gets scheduled again on the next hotplug interrupt.
Fixes: 128a9bf8ace2 ("drm/rockchip: Add basic RK3588 HDMI output support")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Closes: https://lore.kernel.org/all/20260723015909.7F79A1F000E9@smtp.kernel.org/
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-7-195157ba0b40@collabora.com
|
|
The system-sleep PM callbacks fetch the driver state via
dev_get_drvdata() and dereference it unconditionally. However, the
platform device is bound by ->probe(), which merely registers a
component, hence the callbacks may well run when no valid state is
available:
- Before the aggregate driver calls ->bind(), e.g. when the system
suspends while probing is still deferred waiting for the VOP,
dev_get_drvdata() returns NULL.
- After ->unbind(), or when ->bind() fails past platform_set_drvdata(),
drvdata is left pointing to memory obtained via drmm_kzalloc(), which
is released together with the DRM device, while this platform device
remains bound and PM-capable.
Publish the driver data only after ->bind() completes successfully, and
clear it in ->unbind(). This ensures that a non-NULL drvdata always
represents live state, allowing each PM callback to rely on a single
NULL check.
Fixes: 128a9bf8ace2 ("drm/rockchip: Add basic RK3588 HDMI output support")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Closes: https://lore.kernel.org/all/20260702151550.8E14D1F000E9@smtp.kernel.org/
Closes: https://lore.kernel.org/all/20260723014811.1F9F01F000E9@smtp.kernel.org/
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-6-195157ba0b40@collabora.com
|
|
Export dw_hdmi_qp_hpd_notify() for platform drivers to report hot-plug
detection events. Unlike drm_helper_hpd_irq_event() which polls all
connectors, this helper targets only the affected connector and ensures
.detect_ctx() is invoked on reconnection events to trigger SCDC state
recovery.
Track the connector in use in curr_conn, set and cleared by the atomic
enable/disable hooks. Since the helper runs asynchronously from
IRQ/work context, hold a reference on that connector for as long as the
bridge is enabled, so that the pointer is always safe to dereference.
Otherwise validity would rest on the platform glue quiescing HPD before
the DRM-managed connector is freed during teardown, risking a
use-after-free on a late HPD event.
The helper must take its own temporary reference while running, because
the long-lived reference above only guarantees the pointer is valid at
the time it is read, not that the connector survives a concurrent
atomic_disable dropping that reference while the helper is still using
it. Moreover, reading the pointer and acquiring the temporary reference
must be atomic with respect to atomic_disable, otherwise a window exists
where drm_connector_get() is called on a connector whose refcount
already reached zero, hence guard curr_conn with a mutex for this
purpose. Holding the mutex across the whole helper instead is not an
option, as it takes modeset locks while atomic_disable is called with
them held.
Also note that this helper may sleep, so it must only be called from a
threaded HPD interrupt handler or a work function.
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-5-195157ba0b40@collabora.com
|
|
Add support for SCDC scrambling and the high TMDS clock ratio required
by HDMI 2.0 modes with TMDS character rates above 340 MHz, such as
4K@60Hz.
Hook the bridge into the DRM HDMI connector scrambling helpers so the
sink-side SCDC state and the controller scrambler are updated together
as part of the atomic enable and disable paths. This allows the driver
to advertise HDMI 2.0 support when HPD is available.
Keep no-HPD configurations limited to HDMI 1.4, since SCDC access
requires a connected HDMI sink.
Modes above the HDMI 2.0 600 MHz TMDS character-rate limit are now
implicitly rejected by the DRM HDMI connector/bridge infrastructure,
making the dw_hdmi_qp_bridge_tmds_char_rate_valid() callback redundant.
Drop it.
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-4-195157ba0b40@collabora.com
|
|
linux/workqueue.h is no longer needed, and SCRAMB_POLL_DELAY_MS is
unused.
Remove them.
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-3-195157ba0b40@collabora.com
|
|
Implement the .hpd_enable() and .hpd_disable() bridge callbacks and
extend dw_hdmi_qp_phy_ops with corresponding hooks.
This enables the DRM core to control when HPD interrupts are armed,
which is needed to prevent premature interrupt delivery before the
connector is fully initialized, and to properly quiesce HPD during
suspend.
Tested-by: Maud Spierings <maud_spierings@hotmail.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS, Rock 5B
Tested-by: Robin Räber <robin.raebi@gmail.com>
Tested-by: Igor Paunovic <royalnet026@gmail.com> # RK3588 Orange Pi 5 Plus, dual HDMI
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-2-195157ba0b40@collabora.com
|
|
A NACK or timeout on the DDC bus is a normal condition when the sink is
absent or unpowered. Some displays keep HPD asserted and the EDID
EEPROM readable while in standby, but leave the SCDC slave unpowered.
This results in i2c-related error messages on every boot during
scrambling setup, even though the condition is handled gracefully and
the SCDC configuration is restored on the next hotplug event.
Log the i2c bus level failures with dev_dbg_ratelimited() instead,
matching the silent behaviour of the non-QP dw-hdmi i2c bus algorithm.
The DDC users already report failures at the appropriate level, e.g. the
SCDC helpers via drm_scdc_dbg(). This also makes the no-hpd special
casing of the read path messages redundant, so drop it.
Reported-by: Diederik de Haas <diederik@cknow-tech.com>
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Tested-by: Diederik de Haas <diederik@cknow-tech.com> # NanoPC-T6 LTS
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Link: https://patch.msgid.link/20261005-dw-hdmi-qp-scramb-v12-1-195157ba0b40@collabora.com
|
|
acpi_pci_link_remove() frees the link object without clearing the ACPI
device's driver_data pointer. The ACPI device survives hot-removal and
can later be retrieved by acpi_fetch_acpi_dev(), leaving callers with a
dangling pointer.
Clear driver_data before freeing the link.
Fixes: db89b4f0dbab ("ACPI: catch calls of acpi_driver_data on pointer of wrong type")
Signed-off-by: Jiale Yao <yaojiale02@163.com>
[ rjw: Subject tweak ]
Link: https://patch.msgid.link/20260925155846.2381956-1-yaojiale02@163.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Per the CXL 4.0 specification software must check the Command Effects
Log (CEL) for dynamic capacity command support.
Detect support for the DCD commands while reading the CEL, including:
Get DC Config
Get DC Extent List
Add DC Response
Release DC
Leave mds->dcd_supported false. Dynamic capacity is only usable once
the driver can handle Add and Release Capacity events. Detect and log
the capability here and let the commit which adds DCD event handling set
the flag.
Based on an original patch by Navneet Singh.
Signed-off-by: Ira Weiny <iweiny@kernel.org>
Signed-off-by: Anisa Su <anisa.su@samsung.com>
Tested-by: Wonjae Lee <wj28.lee@samsung.com>
Tested-by: Junhee Park <jh9934.park@samsung.com>
Tested-by: Heesoo Kim <habil.kim@samsung.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Jonathan Cameron <jonathan.cameron@oss.qualcomm.com>
Reviewed-by: Richard Cheng <icheng@nvidia.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Tested-by: Alison Schofield <alison.schofield@intel.com>
Link: https://patch.msgid.link/20261010225852.10727-2-anisa.su@samsung.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
ghes_do_proc() walks every error section of a CPER record, and
ghes_handle_arm_hw_error() walks every error-info entry of an ARM
processor section. Both assign to "queued" instead of OR-ing into it,
so only the last iteration counts: a later section or entry that queues
no memory-failure work clears the flag an earlier one set.
Instead of reassigning the variable at every time, just do an `OR`, so,
if there was a single match, it will return "queued" as set.
The comment above the check says "If no memory failure
work is queued", which is the semantics this patch implements. I.e, only
SIGBUS if there are queue was not handled at all. If we want to SIGBUS
if any event was not handled/queued at all, we will need something more
sophisticated.
Fixes: 7f17b4a121d0 ("ACPI: APEI: Kick the memory_failure() queue for synchronous errors")
Fixes: ccb5ecdc2dde ("ACPI: APEI: fix synchronous external aborts in user-mode")
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Shuai Xue <xueshuai@linux.alibaba.com>
Link: https://patch.msgid.link/20260925-ghes-queued-accumulate-v1-1-df50553957d4@debian.org
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Formerly, the reset sequence would race with panfrost_mmu_as_put()
when tearing down a perfcnt session. On top of that, poking GPU
registers to program a perfcnt session or obtaining a dump might lead to
undefined behaviour when done at the same time a reset was ongoing.
Use the reset r/w semaphore when disabling and re-enabling perfcnt
configuration, and also in the sections inside the 'enable'and 'dump'
ioctls where device registers are being accessed.
On top of that, expand the DRM uAPI for the perfcnt DUMP operation
so that user space can be made aware of a reset having happened,
whether it succeeded or failed to restore the original configuration.
UM needs to know about this condition because counter data is inaccurate
after a reset, so the best approach might be simply to try again.
Finally, update driver uAPI documentation to explain the meaning of the
new perfcnt dump ioctl's state flags, and bump DRM driver minor number
to reflect the new DUMP IOCTL req field.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Fixes: 73e467f60acd ("drm/panfrost: Consolidate reset handling")
Fixes: 7786fd108777 ("drm/panfrost: Expose performance counters through unstable ioctls")
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-15-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
In order to avoid accessing HW register while a reset is ongoing, a
read/write semaphore that envelopes the reset sequence will help driver
entry points avoid racing with it. For now, the only such racy entry
point is the ioctl that returns the current GPU timestamp.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-14-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
If the GPU and CPU are fully coherent, then there's no need to flush the
GPU caches after a perfcnt sample is made available.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-13-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
The GPU cache flush/invalidate operation is unnecessary. In the usual case,
before a perfcnt dump is requested, there will be no stale lines in GPU
caches for the sample BO's physmem pages, since their memory had not
recently been addressed by the GPU.
However, in the off-chance that the sample BO reuses memory that the GPU
had recently addressed and for which it contains stale lines in its caches,
there already was a FLUSH_PT command being issued as part of the buffer GPU
mapping procedure, so by the time panfrost_gem_open() comes back at perfcnt
session enable, stale GPU cache lines should've already been evicted.
In either case, forcing a GPU cache clean when enabling a new session makes
no difference, and removing it simplifies handling of the error path.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-12-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Just for the sake of avoiding repetition.
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-11-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
This will be of great help when testing potential races between the GPU
reset sequence and other parts of the code accessing HW registers.
We must also disable the reset work item rather than simply cancelling
it, to prevent the knob from triggering another reset when the device
is being removed.
At the moment, the only legitimate input value for the knob is '1'. This is
to allow for future inputs that would encode alternative reset scenarios
we'd want to trigger manually.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-10-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Rather than just failing silently, let's warn the user of device removal
not being able to take a PM reference, because it should be seen as fatal.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-9-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Ideally the probe() function will do as little as possible, and all device
initialisation and registration should happen inside the panfrost device
subsystem, just like it's done in Panthor. This also simplifies resource
unwinding in the error path.
Do the same thing for DRM driver remove, as in, sweep most of the action
into panfrost_device_fini(), just like we did for device probe.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-8-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Because the device must be in a position to accept jobs between the time
drm_dev_register() is called and autosuspend first kicks in, there's a very
narrow window inbetween during which jobs targeting the tiler buffer
object would time out, since the device's PM status is 'Active', but no MMU
interrupts were enabled at device initialisation time.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Fixes: 73e467f60acd ("drm/panfrost: Consolidate reset handling")
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-7-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
During device probe(), failure to do a PM get() will leave the usage_count
set to 0, which is the value assigned at device creation time. That means
when the autosuspend delay expires, runtime suspend callback won't be
invoked, so the device will remain powered on forever.
On top of that, failure to call PM put() during device unplug means
Panfrost device's PM usage_count increases monotonically for every new
module reload.
The outcome of both of the above meant that:
- Devfreq OPP transition notifications would be printed all the time,
even when no jobs are being submitted. This quickly fills the kernel
ring buffer with junk.
- Because MMU interrupts are only enabled when the device is reset,
the very first job targeting the tiler heap BO after device probe()
would always time out, since the driver's PM runtime resume callback
would not be invoked.
To fix the above:
- Manually adjust the PM refcnt at device probe and removal time.
- Ensure pm_runtime_dont_use_autosuspend is called in the wind-down path.
- Call pm_runtime_put_autosuspend() when device is ready to accept jobs
Fixes: 635430797d3f ("drm/panfrost: Rework runtime PM initialization")
Fixes: 876b15d2c88d ("drm/panfrost: Fix module unload")
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-6-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Gather all clock enables and disables into a single function to avoid
repetitions between driver init/fini and device resume/suspend, since
these clocks are always handled in bulk.
Also do clk (un)prepares and dis/enables at the same time, since the
clk_prepare_* family of functions can simply increase the refcnt of
an already prepared clock.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-5-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
The clk_* API can deal with NULL clock objects, so we can do away with NULL
checks in client code.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-4-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Because debugfs knobs are a device interface, and also the different knobs
deal with specific subsystems, it's best to move their initialisations away
from panfrost_drv.c and into the subsystem they provide information about.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-3-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
The device init function feels a bit cluttered with things that should be
done in the specific subsystems where those locks and lists are being
used. Same goes for module parameters, it's best to limit them to the
subsystems where they're relevant.
Then we can also discard panfrost_drv.h since the transparent_huge_page
doesn't need to be accessed elsewhere. We can also simplify the
creation of a mount point, because drm_gem_get_huge_mnt() would never
return NULL if drm_gem_huge_mnt_create() returns 0 when kernel is built
with support for transparent huge pages.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-2-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
Since the moment we call drm_dev_register() the device should be in a
position to accept jobs, so it's best if the shrinker is already
initialized by then.
On top of that, make shrinker functions take an panfrost_device pointer
like other functions in the same sequence and rename them accordingly.
Essentially mimic the init/fini behaviour in Panthor.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Link: https://patch.msgid.link/20261012-claude-fixes-v15-1-bc0a6c498553@collabora.com
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
|
|
The mlx5 DMA pool structs (mlx5_dma_pool, mlx5_dma_pool_page and its
block bitmap, and mlx5_frag_buf_node_pools) are allocated without a
node hint, so they land on the node of whichever CPU happens to create
or fill the pool, while the DMA pages they describe are allocated on
the pool's node.
The pool, page and bitmap are dereferenced on every block allocation
and free. Allocate them on the pool's NUMA node as well, so that all
of a pool's state lives on one node.
These allocations happen on the control path, so the performance gain
is expected to be small; the change is mainly for consistency.
Signed-off-by: Seongjun Hong <hsj0512@snu.ac.kr>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20261006-net-mlx5-numa-allocate-pool-page-v2-1-0dc4c2eb1376@snu.ac.kr
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
On the Microsoft Surface Pro 11 and Surface Laptop 7 (X1E80100), the
WCN7850 hardware rfkill input reads as "radio off" and the OS has no
way to control it. ath12k enables hardware rfkill handling, so Wi-Fi
stays hard-blocked:
iwd: Error bringing interface 4 up: Operation not possible due to RF-kill
ACPI platforms can opt out through ath12k_acpi_get_disable_rfkill(),
but DT platforms cannot. A DT property was proposed for this and
rejected, because it describes what the OS should do rather than the
hardware. The suggested alternative was to handle these machines in
the driver until a connector binding can describe a broken rfkill line.
Add a table of board compatibles and hardware revisions on which
hardware rfkill is known to be unusable, and skip the rfkill setup
there. Other affected machines then only need a new table entry.
Tested on a Surface Pro 11 (microsoft,denali) with next-20260929,
booted with the unmodified mainline DTB: the interface is no longer
blocked and connects. The Surface Laptop 7 entries (microsoft,romulus13
and microsoft,romulus15) are based on reports from Jérôme de Bretagne
and Dale Whinham. I have not tested them.
Link: https://lore.kernel.org/all/2ixwny6hzumlqmok5cjwutvtk7nzmx4t3dhxc44ao4l2u755pj@n6m7jyxf2o62/
Suggested-by: Manivannan Sadhasivam <mani@kernel.org>
Assisted-by: LLM
Signed-off-by: François Roux <franzelfranzel@gmail.com>
Tested-by: Dale Whinham <daleyo@gmail.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20261006062728.45007-1-franzelfranzel@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
The GPIO callbacks take a mutex and issue USB transfers, so this GPIO
chip must not be used in atomic context. Change the can_sleep switch in
struct gpio_chip to true.
Fixes: 611a995e8ae1 ("gpio: Add Nuvoton NCT6694 GPIO support")
Signed-off-by: Ming Yu <a0282524688@gmail.com>
Link: https://patch.msgid.link/20261007092102.3768818-2-a0282524688@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
|
|
The WLAN enable GPIO is requested with GPIOD_ASIS and then kept at its
current level, so that a WLAN module left powered on by the firmware is
not switched off. The FIXME explains why: toggling it would take the
PCIe link down, and the controller driver could not recover from that.
That reasoning held while the PCIe link was trained before the
sequencer probed. Since commit b921aa3f8dec ("PCI/pwrctrl: Switch to
pwrctrl create, power on/off, destroy APIs"), qcom_pcie_host_init()
powers the endpoint through pwrctrl before it starts link training, and
defers until the pwrctrl driver is bound. That driver cannot bind
before this sequencer has probed. Power-on therefore always comes
before link training, and keeping the firmware state is now harmful.
On the Microsoft Surface Pro 11 (X1E80100), the WCN7850 sits on a
PCIe port without a PERST# GPIO. While pci-pwrctrl-pwrseq is not yet
bound, the host init is deferred about ten times, and each attempt
powers the PHY and controller back down. The later power-on only sets
a GPIO that is already high, so the chip is never reset and the link
never comes up:
qcom-pcie 1c08000.pcie: Device found, but not active
The endpoint is not enumerated and ath12k never probes. Bluetooth on
the same chip fails too ("QCA Failed to send TLV segment (-110)").
Request the GPIO as GPIOD_OUT_LOW. The chip is then off when the
sequencer probes, and pwrctrl powers it up cleanly right before link
training. Drop the FIXME and the code that preserved the firmware
state.
Tested on a Surface Pro 11:
- b921aa3f8dec plus a one-line version of this change: endpoint
enumerated at 2.4 s, ath12k and Bluetooth working.
- next-20260929 plus this patch: "PCIe Gen.3 x2 link up" at 3.4 s,
ath12k and Bluetooth working, Wi-Fi connected.
Without the change, neither kernel enumerates the endpoint.
The culprit was found by bisecting v6.17..v7.1, with the DTB and
.config held constant.
Fixes: b921aa3f8dec ("PCI/pwrctrl: Switch to pwrctrl create, power on/off, destroy APIs")
Closes: https://lore.kernel.org/all/CAPjyS8dY0Q_o3XmFjuKZZsjhzhoSNL+ZGJtw5cQt8HiV3sUtxA@mail.gmail.com/
Cc: stable@vger.kernel.org
Assisted-by: LLM
Signed-off-by: François Roux <franzelfranzel@gmail.com>
Tested-by: Dale Whinham <daleyo@gmail.com>
Link: https://patch.msgid.link/20261004165124.4277-1-franzelfranzel@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
|
|
ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/superm1/linux
Pull amd-pstate content for 7.4 (10/4/26) from Mario Limonciello:
"* Fixups for a few bugs identified by Sashiko
* Initial support for maximum frequency capacity handling on Zen6 / client
* Updated EPP tunings for Zen6 / client"
|
|
Building arm64 allmodconfig with clang fails:
drivers/resctrl/mpam_internal.h:207:7: error: ignoring return value
of function declared with 'warn_unused_result' attribute
[-Werror,-Wunused-result]
207 | mpam_mon_sel_lock(_T), mpam_mon_sel_unlock(_T));
mpam_mon_sel_lock() is __must_check and can fail: for an MSC accessed
through the MPAM-Fb firmware interface it returns false when called
from a context that cannot sleep. The mon_sel guard was defined with
DEFINE_GUARD(), which calls it unconditionally and discards the
result, so a guard(mon_sel) user would carry on without the lock and
release it on scope exit. The unconditional guard only exists as the
base for the conditional mon_sel_lock variant, which is the one
actually used.
Define mon_sel_lock directly as a conditional guard class instead, as
posix-timers does for lock_timer: the constructor returns NULL when
the lock cannot be taken, and the destructor only releases a lock
that was taken. ACQUIRE(mon_sel_lock, ...) works as before, including
ACQUIRE_ERR() returning -EBUSY on failure, and there is no
unconditional variant left to misuse.
Reported-by: kernelci.org bot <bot@kernelci.org>
Closes: https://d.kernelci.org/i/maestro:66f869ecbbb5b8592562ffd89f049c5ddf682547
Closes: https://d.kernelci.org/i/maestro:647269216758644837afcaac8e67d8dbe713a0e2
Fixes: 0db1715e4c9c ("arm_mpam: propagate MSC access errors for hw_probe functions")
Assisted-by: LLM
Signed-off-by: Sasha Levin <sashal@kernel.org>
Reviewed-by: Ben Horgan <ben.horgan@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
|
|
Audio can be triggered by the HPD handler and start up before video has
been restored, calling vc4_hdmi_write_infoframe() for the audio infoframe.
vc4_hdmi_write_infoframe() needs Packet RAM to be on to store the
infoframe, however, if the video hasn't been enabled,
vc4_hdmi_encoder_post_crtc_enable() hasn't been called and therefore, the
Packet RAM can be disabled by the time vc4_hdmi_write_infoframe() executes,
causing the following splat:
[ 201.570266] ------------[ cut here ]------------
[ 201.570293] Packet RAM has to be on to store the packet.
[ 201.570367] WARNING: drivers/gpu/drm/vc4/vc4_hdmi.c:683 at vc4_hdmi_write_infoframe+0x404/0x494 [vc4], CPU#0: wireplumber/1606
[...]
[ 201.658650] Hardware name: Raspberry Pi 4 Model B Rev 1.5 (DT)
[...]
[ 201.758499] Call trace:
[ 201.760967] vc4_hdmi_write_infoframe+0x404/0x494 [vc4] (P)
[ 201.766607] vc4_hdmi_write_audio_infoframe+0x1c/0x28 [vc4]
[ 201.772247] write_infoframe+0xbc/0x140 [drm_display_helper]
[ 201.777975] drm_atomic_helper_connector_hdmi_update_audio_infoframe+0x80/0xdc [drm_display_helper]
[ 201.787140] vc4_hdmi_audio_prepare+0x4ac/0x570 [vc4]
[ 201.792252] drm_connector_hdmi_audio_prepare+0x24/0x30 [drm_display_helper]
[ 201.799390] hdmi_codec_prepare+0x100/0x160 [snd_soc_hdmi_codec]
[ 201.805471] snd_soc_pcm_dai_prepare+0xd4/0x100 [snd_soc_core]
[ 201.811375] __soc_pcm_prepare+0x60/0x12c [snd_soc_core]
[ 201.816752] soc_pcm_prepare+0x34/0x58 [snd_soc_core]
[ 201.821863] snd_pcm_do_prepare+0x70/0x8c [snd_pcm]
[ 201.826798] snd_pcm_action_nonatomic+0xe4/0x174 [snd_pcm]
[ 201.832350] snd_pcm_prepare+0xc8/0xd8 [snd_pcm]
[ 201.837021] snd_pcm_common_ioctl+0x458/0x614 [snd_pcm]
[ 201.842308] snd_pcm_ioctl+0x30/0x48 [snd_pcm]
[ 201.846802] __arm64_sys_ioctl+0x88/0xd8
[ 201.850769] invoke_syscall+0x48/0xf4
[ 201.854470] el0_svc_common+0x80/0xd8
[ 201.858171] do_el0_svc+0x1c/0x28
[ 201.861520] el0_svc+0x3c/0xb0
[ 201.864605] el0t_64_sync_handler+0x68/0xdc
[ 201.868834] el0t_64_sync+0x198/0x19c
[ 201.872536] ---[ end trace 0000000000000000 ]---
Considering that Packet RAM enable is essentially a power management
configuration, move its enabling/disabling to the Runtime PM hooks.
Note that post_crtc_disable() still stops all infoframes, but keeps Packet
RAM enabled, leaving it to be disabled by runtime suspend.
Link: https://github.com/raspberrypi/linux/issues/7069
Fixes: b47a72f30768 ("drm/vc4: hdmi: Switch to HDMI connector")
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Acked-by: Maxime Ripard <mripard@kernel.org>
Link: https://patch.msgid.link/20261002-vc4-hdmi-packet-ram-fix-v1-1-3d73148096c9@igalia.com
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Signed-off-by: Maíra Canal <mcanal@igalia.com>
|
|
Commit 8581b19eb2c5 ("dm-snapshot: fix 'scheduling while atomic' on
real-time kernels") replaced hlist_bl_for_each_entry with
hlist_for_each_entry, but the two iteration macros differ after the loop
ends: hlist_bl_for_each_entry leaves "e" pointing at the last element,
while hlist_for_each_entry() sets it to NULL.
The result is that the list is no longer sorted by chunk number as it was
intended to be. This bug doesn't cause data corruption because
dm_lookup_exception walks the full list.
Restore the invariant by tracking the list tail in the "prev" variable
and appending after it with hlist_add_behind.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Fixes: 8581b19eb2c5 ("dm-snapshot: fix 'scheduling while atomic' on real-time kernels")
Assisted-by: Claude:claude-opus-5
|
|
If dm_io_client_create() fails, it stores error into the ps->io_client
variable. The code would try to free it later with
dm_io_client_destroy(ps->io_client), causing an invalid pointer
dereference. Fix this by setting ps->io_client = NULL.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Assisted-by: Claude:claude-opus-5
Cc: stable@vger.kernel.org
|
|
The device probe passes zeroes for all seven argument registers to
arm_smccc_1_1_invoke(). On ARM32 Thumb-2, this binds an input to r7 and
GCC 8.5 reports "r7 cannot be used in asm here".
The current smccc_devices entries need at most one argument after the
function ID: TRNG_VERSION has none and RSI_ABI_VERSION takes a requested
version. Pass zero for that argument and omit the unused arguments,
avoiding the r7 constraint. Check argument requirements when adding new
probe functions to smccc_devices.
Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Sudeep Holla <sudeep.holla@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Steven Price <steven.price@arm.com>
Cc: Suzuki K Poulose <Suzuki.Poulose@arm.com>
Cc: Andre Przywara <andre.przywara@arm.com>
Cc: Russell King <linux@armlinux.org.uk>
Fixes: d8263a131bb0 ("firmware: hwrng: arm_smccc_trng: Register as an SMCCC device")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202610070641.B8hxBJcf-lkp@intel.com/
Signed-off-by: Aneesh Kumar K.V (Arm) <aneesh.kumar@kernel.org>
Reviewed-by: Sudeep Holla <sudeep.holla@kernel.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
|
|
init_pins_table() adds to npins in the driver's static descriptor, so
after a probe deferral the count doubles and the next probe fails with
"pin 0 already registered". Reset it first.
Signed-off-by: Pablo Mazzini <pmazzini@gmail.com>
Fixes: f5e2cd34b12f ("pinctrl: sunxi: allow reading mux values from DT")
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Linus Walleij <linusw@kernel.org>
|
|
GSI completions queue notify_work to invoke the CQ completion handler.
CQ destruction drains hardware events and core polling, but does not
cancel this software notifier.
Fixes: 25361e02c448 ("IB/mlx5: Generate completions in software")
Signed-off-by: Prathamesh Deshpande <prathameshdeshpande7@gmail.com>
Link: https://patch.msgid.link/20261008141209.6923-1-prathameshdeshpande7@gmail.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
|
|
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Page-fault EQ work can remain queued after the notifier is disabled.
Destroying the EQ before cancelling that work can let the worker access
freed EQ memory.
Cancel the deferred work before destroying the EQ.
Fixes: d5d284b829a6 ("{net,IB}/mlx5: Move Page fault EQ and ODP logic to RDMA")
Signed-off-by: Prathamesh Deshpande <prathameshdeshpande7@gmail.com>
Link: https://patch.msgid.link/20261007234216.2765-1-prathameshdeshpande7@gmail.com
Reviewed-by: Chenguang Zhao <zhaochenguang@kylinos.cn>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into arm/fixes
soc/tegra: Fixes for v7.3
This contains two minor fixes for a missing newline in an error message
and an uninitialized clock rate used during suspend.
* tag 'tegra-for-7.3-soc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
soc/tegra: pmc: Fix uninitialised clock rate
soc/tegra: fuse: Add missing newline to APBMISC error message
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
Make sure to drop the reference taken by usb_anchor_urb() also on
successful submission of delayed write URBs during runtime resume to
prevent them from leaking on completion.
Fixes: e6929a9020ac ("USB: support for autosuspend in sierra while online")
Reported-by: Wentao Liang <vulab@iscas.ac.cn>
Link: https://lore.kernel.org/20260916165352.2085480-1-vulab@iscas.ac.cn
Cc: stable@vger.kernel.org # 2.6.32
Cc: Oliver Neukum <oliver@neukum.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Johan Hovold <johan@kernel.org>
|
|
In preparation for making the devm_kmalloc family of allocators type
aware, we need to make sure that the returned type from the allocation
matches the type of the variable being assigned. (Before, the allocator
would always return "void *", which can be implicitly cast to any
pointer type.)
The assigned type is "struct type_c_cfg *", but the converted allocation
type would be "const struct type_c_cfg *", as the size was taken from
the local "type_c_cfg", which points to const as it comes from
of_device_get_match_data(). As there is no general way to remove const
qualifiers, take the size from the assignment target instead. No change
in allocation size results.
Build tested ARCH=x86_64 allmodconfig with GCC 16.2.0:
drivers/extcon/extcon-rtk-type-c.o
Assisted-by: LLM coccinelle
Link: https://patch.msgid.link/20260917211246.i.196-kees@kernel.org
Signed-off-by: Kees Cook <kees+treewide@kernel.org>
|
|
In preparation for making the devm_kmalloc family of allocators type
aware, we need to make sure that the returned type from the allocation
matches the type of the variable being assigned. (Before, the allocator
would always return "void *", which can be implicitly cast to any
pointer type.)
The assigned type is "const struct attribute_group **", but the
converted allocation type would be "struct attribute_group **", which is
the same type without the const qualifier. As there is no general way to
safely add const qualifiers, take the size from the assignment target
instead. No change in allocation size results.
Build tested ARCH=arm64 allmodconfig with GCC aarch64-linux-gnu 16.1.0:
drivers/hwtracing/coresight/coresight-cti-sysfs.o
Assisted-by: LLM coccinelle
Reviewed-by: Mike Leach <mike.leach@arm.com>
Reviewed-by: Leo Yan <leo.yan@arm.com>
Link: https://patch.msgid.link/20260917211428.i.947-kees@kernel.org
Signed-off-by: Kees Cook <kees+treewide@kernel.org>
|
|
To avoid DoS from printing too much, make printing in Binder rate
limited.
A big portion of these print statements have been updated to use
binder_debug!, but some still remain to be converted. For now, just
update them to use pr_*_ratelimited! until we get around to converting
them to use binder_debug! too. While we're at it, fix the missing
newlines at the end of some of those println statements.
Four transaction failure pr_warn! calls in thread.rs and transaction.rs
are not converted because they're being deleted by another patch
series.
Acked-by: Carlos Llamas <cmllamas@google.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://patch.msgid.link/20261007-pr-ratelimited-v6-4-f14accd4ea3b@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|