Always set the RMDevidCheckIgnore registry key for GSP-RM so that it
will continue support newer variants of already supported GPUs.
GSP-RM maintains an internal list of PCI IDs of GPUs that it supports,
and checks if the current GPU is on this list. While the actual GPU
architecture (as specified in the BOOT_0/BOOT_42 registers) determines
how to enable the GPU, the PCI ID is used for the product name, e.g.
"NVIDIA GeForce RTX 5090".
Unfortunately, if there is no match, GSP-RM will refuse to initialize,
even if the device is fully supported. Nouveau will get an error
return code, but by then it's too late. This behavior may be corrected
in a future version of GSP-RM, but that does not help Nouveau today.
Fortunately, GSP-RM supports an undocumented registry key that tells it
to ignore the mismatch. In such cases, the product name returned will
be a blank string, but otherwise GSP-RM will continue.
Unlike Nvidia's proprietary driver, Nouveau cannot update to newer
firmware versions to keep up with every new hardware release. Instead,
we can permanently set this registry key, and GSP-RM will continue
to function the same with known hardware.
RDC PCI to PCIe bridges, present on Vortex86DX3 and Vortex86EX2 SoCs, do
not support MSIs. If enabled, interrupts generated by PCIe devices never
reach the processor.
I have contacted the manufacturer (DM&P) and they confirmed that PCI MSIs
need to be disabled for them.
Use DRM's shadow-plane helper to map and access the GEM object's buffer
within kernel address space. Encapsulates the vmap logic in the GEM-DMA
helpers.
The sharp-memory driver currently reads the vaddr field from the GME
buffer object directly. This only works because GEM code 'automagically'
sets vaddr.
Shadow-plane helpers perform the same steps, but with correct abstraction
behind drm_gem_vmap(). The shadow-plane state provides the buffer address
in kernel address space and the format-conversion state.
While we do need at least 3.6 for kernel-doc to work, and at least
3.7 for it to output functions and structs with parameters at the
right order, let the python binary be compatible with legacy
versions.
The rationale is that the Kernel build nowadays calls kernel-doc
with -none on some places. Better not to bail out when older
versions are found.
With that, potentially this will run with python 2.7 and 3.2+,
according with vermin:
The minimal version I could check was version 3.4
(using anaconda). Anaconda doesn't support 3.2 or 3.3
anymore, and 3.2 doesn't even compile (I tested compiling
Python 3.2 on Fedora 42 and on Fedora 32 - no show).
With 3.4, the script didn't crash and emitted the right warning:
$ conda create -n py34 python=3.4
$ conda activate py34
python --version
Python 3.4.5
$ python ./scripts/kernel-doc --none include/media
Error: Python 3.6 or later is required by kernel-doc
$ conda deactivate
$ python --version
Python 3.13.5
$ python ./scripts/kernel-doc --none include/media
(no warnings and script ran properly)
Supporting 2.7 is out of scope, as it is EOL for 5 years, and
changing shebang to point to "python" instead of "python3"
would have a wider impact.
I did some extra checks about the differences from 3.2 and
3.4, and didn't find anything that would cause troubles:
For non-leaf paging structures we end up selecting a random index
between [0, 3], depending on the first user if the page-table is shared,
since non-leaf structures only have two bits in the HW for encoding the
PAT index, and here we are just passing along the full user provided
index, which can be an index as large as ~31 on xe2+. The user provided
index is meant for the leaf node, which maps the actual BO pages where
we have more PAT bits, and not the non-leaf nodes which are only mapping
other paging structures, and so only needs a minimal PAT index range.
Also the chosen index might need to consider how the driver mapped the
paging structures on the host side, like wc vs wb, which is separate
from the user provided index.
With that move the PDE PAT index selection under driver control. For now
just use a coherent index on platforms with page-tables that are cached
on host side, and incoherent otherwise. Using a coherent index could
potentially be expensive, and would be overkill if we know the page-table
is always uncached on host side.
v2 (Stuart):
- Add some documentation and split into separate helper.
BSpec: 59510 Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Stuart Summers <stuart.summers@intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Stuart Summers <stuart.summers@intel.com> Link: https://lore.kernel.org/r/20250808103455.462424-2-matthew.auld@intel.com Signed-off-by: Sasha Levin <sashal@kernel.org>
[Why]
Customer reported an issue that OS starts and stops device multiple times
during driver installation. Frequently disabling and enabling OTG may
prevent OTG from being safely disabled and cause incorrect configuration
upon the next enablement.
[How]
Add a wait until OTG_CURRENT_MASTER_EN_STATE is cleared as a short term
solution.
Reviewed-by: Dillon Varone <dillon.varone@amd.com> Signed-off-by: TungYu Lu <tungyu.lu@amd.com> Signed-off-by: Tom Chung <chiahsuan.chung@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
User queues are disabled before GEM objects are released
(protecting against user app crashes).
No races with PCI hot-unplug (because drm_dev_enter prevents cleanup
if iewdevice is being removed).
Cc: Christian König <christian.koenig@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[WHY&HOW]
The user closed the lid while the system was powering on and opened it
again before the “apply_seamless_boot_optimization” was set to false,
resulting in the eDP remaining blank.
Reset the “apply_seamless_boot_optimization” to false when dpms off.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Danny Wang <Danny.Wang@amd.com> Signed-off-by: Tom Chung <chiahsuan.chung@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The SND_JACK_AVOUT is a more specific jack type for HDMI and DisplayPort.
Updatae the MediaTek drivers to use such jack type, allowing system to
determine the device type based on jack event.
Since version 0x105, the keyword 'init' was introduced into the profile,
which is used for chip initialization, particularly to store common
settings for other non-initialization profiles.
snprintf() returns the number of characters that *would* have been
written, which can overestimate how much you actually wrote to the
buffer in case of truncation. That leads to 'data += this' advancing
the pointer past the end of the buffer and size going negative.
Switching to scnprintf() prevents potential buffer overflows and ensures
consistent behavior when building the output string.
Previously, the LMTT directory pointer was only programmed for primary GT
within a tile. However, to ensure correct Local Memory access by VFs,
the LMTT configuration must be programmed on all GTs within the tile.
Lets program the LMTT directory pointer on every GT of the tile
to guarantee proper LMEM access across all GTs on VFs.
HSD: 18042797646
Bspec: 67468 Signed-off-by: Piotr Piórkowski <piotr.piorkowski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Michał Winiarski <michal.winiarski@intel.com> Reviewed-by: Stuart Summers <stuart.summers@intel.com> Link: https://lore.kernel.org/r/20250805091850.1508240-1-piotr.piorkowski@intel.com Signed-off-by: Michał Winiarski <michal.winiarski@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Acquire jpeg_pg_lock before changes to jpeg power state
and release it after power off from idle work handler.
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com> Reviewed-by: Leo Liu <leo.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The suspend and resume callbacks for pm and runtime pm should be same.
During suspending, it needs to stop all hardware contexts first. And
the hardware contexts will be restarted after the device is resumed.
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org> Reviewed-by: Maciej Falkowski <maciej.falkowski@linux.intel.com> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com> Link: https://lore.kernel.org/r/20250803191450.1568851-1-lizhi.hou@amd.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Cached metrics data validity is 1ms on arcturus. It's not reasonable for
any client to query gpu_metrics at a faster rate and constantly
interrupt PMFW.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Cached metrics data validity is 1ms on aldebaran. It's not reasonable
for any client to query gpu_metrics at a faster rate and constantly
interrupt PMFW.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[Why]
The reason some high-resolution monitors fail to display properly
is that this platform does not support sufficiently high DPP and
DISP clock frequencies
[How]
Update DISP and DPP clocks from the smu clock table then DML can
filter these mode if not support.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Paul Hsieh <Paul.Hsieh@amd.com> Signed-off-by: Roman Li <roman.li@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Accessing DC from amdgpu_dm is usually preceded by acquisition of
dc_lock mutex. Most of the DC API that DM calls are under a DC lock.
However, there are a few that are not. Some DC API called from interrupt
context end up sending DMUB commands via a DC API, while other threads were
using DMUB. This was apparent from a race between calls for setting idle
optimization enable/disable and the DC API to set vmin/vmax.
Offload the call to dc_stream_adjust_vmin_vmax() to a thread instead
of directly calling them from the interrupt handler such that it waits
for dc_lock.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Signed-off-by: Roman Li <roman.li@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Move amdgpu_device_health_check into amdgpu_device_gpu_recover to
ensure that if the device is present can be checked before reset
The reason is:
1.During the dpc event, the device where the dpc event occurs is not
present on the bus
2.When both dpc event and ATHUB event occur simultaneously,the dpc thread
holds the reset domain lock when detecting error,and the gpu recover thread
acquires the hive lock.The device is simultaneously in the states of
amdgpu_ras_in_recovery and occurs_dpc,so gpu recover thread will not go to
amdgpu_device_health_check.It waits for the reset domain lock held by the
dpc thread, but dpc thread has not released the reset domain lock.In the dpc
callback slot_reset,to obtain the hive lock, the hive lock is held by the
gpu recover thread at this time.So a deadlock occurred
Signed-off-by: Ce Sun <cesun102@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Try to ensure poison creation handle is completed in time
to set device rma value.
Signed-off-by: Ce Sun <cesun102@amd.com> Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The variable `pm_suspend_target_state` is conditionally defined only when
`CONFIG_SUSPEND` is enabled (see `include/linux/suspend.h`). Directly
referencing it without guarding by `#ifdef CONFIG_SUSPEND` causes build
failures when suspend functionality is disabled (e.g., `CONFIG_SUSPEND=n`).
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Perry Yuan <perry.yuan@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The PF's restart worker shouldn't attempt to resume the device on
its own, since its goal is to finish PF and VFs reprovisioning on
the recently reset GuC. Take extra RPM reference while scheduling
a work and release it from the worker or when we cancel a work.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com> Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com> Link: https://lore.kernel.org/r/20250801142822.180530-4-michal.wajdeczko@intel.com Signed-off-by: Sasha Levin <sashal@kernel.org>
I saw an oops in xe_gem_fault when running the xe-fast-feedback
testlist against the realtime kernel without debug options enabled.
The panic happens after core_hotunplug unbind-rebind finishes.
Presumably what happens is that a process mmaps, unlocks because
of the FAULT_FLAG_RETRY_NOWAIT logic, has no process memory left,
causing ttm_bo_vm_dummy_page() to return VM_FAULT_NOPAGE, since
there was nothing left to populate, and then oopses in
"mem_type_is_vram(tbo->resource->mem_type)" because tbo->resource
is NULL.
It's convoluted, but fits the data and explains the oops after
the test exits.
The GuC load process will abort if certain status codes (which are
indicative of a fatal error) are reported. Otherwise, it keeps waiting
until the 'success' code is returned. New error codes have been added
in recent GuC releases, so add support for aborting on those as well.
v2: Shuffle HWCONFIG_START to the front of the switch to keep the
ordering as per the enum define for clarity (review feedback by
Jonathan). Also add a description for the basic 'invalid init data'
code which was missing.
[WHY]
Last LT automation update can cause crash by referencing current_state and
calling into dc_update_planes_and_stream which may clobber current_state.
[HOW]
Cache relevant stream pointers and iterate through them instead of relying
on the current_state.
Reviewed-by: Wenjing Liu <wenjing.liu@amd.com> Signed-off-by: Michael Strauss <michael.strauss@amd.com> Signed-off-by: Ivan Lipski <ivan.lipski@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[why & how]
Header misalignment in struct dmub_cmd_replay_copy_settings_data and
struct dmub_alpm_auxless_data causes incorrect data read between driver
and dmub.
Fix the misalignment and ensure that everything is aligned to 4-byte
boundaries.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Ovidiu Bunea <ovidiu.bunea@amd.com> Signed-off-by: Ivan Lipski <ivan.lipski@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[WHY]
In the worst case, AUX intra-hop done can take hundreds of milliseconds as
each retimer in a link might have to wait a full AUX_RD_INTERVAL to send
LT abort downstream.
[HOW]
Wait 300ms for each retimer in a link to allow time to propagate a LT abort
without infinitely waiting on intra-hop done.
For no-retimer case, keep the max duration at 10ms.
Reviewed-by: Wenjing Liu <wenjing.liu@amd.com> Signed-off-by: Michael Strauss <michael.strauss@amd.com> Signed-off-by: Ivan Lipski <ivan.lipski@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[WHY]
If symclk RCO is enabled, stream encoder may not be receiving an ungated
clock by the time we attempt to set stream attributes when setting dpms
on. Since the clock is gated, register writes to the stream encoder fail.
[HOW]
Move set_stream_attribute call into enable_stream, just after the point
where symclk32_se is ungated.
Logically there is no need to set stream attributes as early as is
currently done in link_set_dpms_on, so this should have no impact beyond
the RCO fix.
Reviewed-by: Ovidiu (Ovi) Bunea <ovidiu.bunea@amd.com> Signed-off-by: Michael Strauss <michael.strauss@amd.com> Signed-off-by: Ivan Lipski <ivan.lipski@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Log an error when vcn sram load fails in indirect mode
and return the same error value.
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com> Reviewed-by: Leo Liu <leo.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[Why]
When transitioning between topologies such as multi-display to single
display ODM 2:1, pipes might not be freed before use.
[How]
In dc_commit_streams, commit an additional, minimal transition if
original transition is not seamless to ensure pipes are freed.
Reviewed-by: Alvin Lee <alvin.lee2@amd.com> Signed-off-by: Clay King <clayking@amd.com> Signed-off-by: Wayne Lin <wayne.lin@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
timing_adjust_pending is used to defer certain programming sequences
when OTG timing is about to be changed, like with VRR. Insufficient
checking for timing change in this case caused a regression which
reduces PSR Replay residency.
Reviewed-by: Tom Chung <chiahsuan.chung@amd.com> Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com> Signed-off-by: Robin Chen <robin.chen@amd.com> Signed-off-by: Wayne Lin <wayne.lin@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[Why & How]
Tiled command rect dimensions is 1 based, do rect_x/y - 1 internally
Reviewed-by: Alvin Lee <alvin.lee2@amd.com> Signed-off-by: Ostrowski Rafal <rostrows@amd.com> Signed-off-by: Wayne Lin <wayne.lin@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
While we expect config directory names to match PCI device name,
currently we are only scanning provided names for domain, bus,
device and function numbers, without checking their format.
This would pass slightly broken entries like:
To avoid such mistakes, check if the name provided exactly matches
the canonical PCI device address format, which we recreated from
the parsed BDF data. Also simplify scanf format as it can't really
catch all formatting errors.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Lucas De Marchi <lucas.demarchi@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://lore.kernel.org/r/20250722141059.30707-3-michal.wajdeczko@intel.com Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
As pm_runtime_force_suspend() will force the device state to suspend,
the driver needs to ensure no IRQ handlers are currently running. If not
those handlers may find they are now running on suspended hardware
despite holding a PM runtime reference. disable_irq() will sync any
currently running handlers, so move the IRQ disabling to cover the whole
of the forced suspend state to avoid such race conditions.
When an MFD device is added, a platform_device is allocated. If this
device is linked to a DT description, the corresponding OF node is linked
to the new platform device but the OF node's refcount isn't incremented.
As of_node_put() is called during the platform device release, it leads
to a refcount underflow.
Call of_node_get() to increment the OF node's refcount when the node is
linked to the newly created platform device.
After discussion with the devicetree maintainers we agreed to not extend
lists with the generic compatible "apple,smc" anymore [1]. Use
"apple,t8103-smc" as base compatible as it is the SoC the driver and
bindings were written for.
We observed the initial probe of the da9063 failing in
da9063_get_device_type in about 30% of boots on a Xilinx ZynqMP based
board. The problem originates in da9063_i2c_blockreg_read, which uses
a single bus transaction to turn the register page and then read a
register. On the bus, this should translate to a write to register 0,
followed by a read to the target register, separated by a repeated
start. However, we found that after the write to register 0, the
controller sometimes continues directly with the register address of
the read request, without sending the chip address or a repeated start
in between, which makes the read request invalid.
To fix this, separate turning the page and reading the register into
two separate transactions. This brings the initialization code in line
with the rest of the driver, which uses register maps (which to my
knowledge do not use repeated starts after turning the page). This has
been included in our kernel for several months and was recently
included in a shipped product. For us, it reliably fixes the issue,
and we have not observed any new issues.
While the underlying problem is probably with the i2c controller or
its driver, I still propose a change here in the interest of
robustness: First, I'm not sure this issue can be fixed on the
controller side, since there are other issues related to repeated
start which can't (AR# 60695, AR# 61664). Second, similar problems
might exist with other controllers.
clang-21 warns about one uninitialized variable getting dereferenced
in madera_dev_init:
drivers/mfd/madera-core.c:739:10: error: variable 'mfd_devs' is uninitialized when used here [-Werror,-Wuninitialized]
739 | mfd_devs, n_devs,
| ^~~~~~~~
drivers/mfd/madera-core.c:459:33: note: initialize the variable 'mfd_devs' to silence this warning
459 | const struct mfd_cell *mfd_devs;
| ^
| = NULL
The code is actually correct here because n_devs is only nonzero
when mfd_devs is a valid pointer, but this is impossible for the
compiler to see reliably.
Change the logic to check for the pointer as well, to make this easier
for the compiler to follow.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Richard Fitzgerald <rf@opensource.cirrus.com> Link: https://lore.kernel.org/r/20250807071932.4085458-1-arnd@kernel.org Signed-off-by: Lee Jones <lee@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The QIXIS FPGA found on Layerscape boards such as LX2160AQDS, LS1028AQDS
etc deals with power-on-reset timing, muxing etc. Use the simple-mfd-i2c
as its core driver by adding its compatible string (already found in
some dt files). By using the simple-mfd-i2c driver, any child device
will have access to the i2c regmap created by it.
When we are successful in using cpufreq min/max limits,
skip setting the raw MSR limits entirely.
This is necessary to avoid undoing any modification that
the cpufreq driver makes to our sysfs request.
eg. intel_pstate may take our request for a limit
that is valid according to HWP.CAP.MIN/MAX and clip
it to be within the range available in PLATFORM_INFO.
Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The fopen_or_die() function was previously hardcoded
to open files in read-only mode ("r"), ignoring the
mode parameter passed to it. This patch corrects
fopen_or_die() to use the provided mode argument,
allowing for flexible file access as intended.
Additionally, the call to fopen_or_die() in
err_on_hypervisor() incorrectly used the mode
"ro", which is not a valid fopen mode. This is
fixed to use the correct "r" mode.
Signed-off-by: Kaushlendra Kumar <kaushlendra.kumar@intel.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The pmt_telemdir_sort() comparison function was returning a boolean
value (0 or 1) instead of the required negative, zero, or positive
value for proper sorting. This caused unpredictable and incorrect
ordering of telemetry directories named telem0, telem1, ..., telemN.
Update the comparison logic to return -1, 0, or 1 based on the
numerical value extracted from the directory name, ensuring correct
numerical ordering when using scandir.
This change improves stability and correctness when iterating PMT
telemetry directories.
Signed-off-by: Kaushlendra Kumar <kaushlendra.kumar@intel.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
bpf_cookie can fail on perf_event_open(), when it runs after the task_work
selftest. The task_work test causes perf to lower
sysctl_perf_event_sample_rate, and bpf_cookie uses sample_freq,
which is validated against that sysctl. As a result,
perf_event_open() rejects the attr if the (now tighter) limit is
exceeded.
>From perf_event_open():
if (attr.freq) {
if (attr.sample_freq > sysctl_perf_event_sample_rate)
return -EINVAL;
} else {
if (attr.sample_period & (1ULL << 63))
return -EINVAL;
}
Switch bpf_cookie to use sample_period, which is not checked against
sysctl_perf_event_sample_rate.
Fix incorrect size parameter passed to cpuidle_state_write_file() in
cpuidle_state_disable().
The function was incorrectly using sizeof(disable) which returns the
size of the unsigned int variable (4 bytes) instead of the actual
length of the string stored in the 'value' buffer.
Since 'value' is populated with snprintf() to contain the string
representation of the disable value, we should use the length
returned by snprintf() to get the correct string length for
writing to the sysfs file.
This ensures the correct number of bytes is written to the cpuidle
state disable file in sysfs.
The Dell OptiPlex 7040 supports the legacy SMM interface for reading
sensors and performing fan control. Whitelist this machine so that
this driver loads automatically.
It turns out the second fan on the Dell Precision 490 does not
really support I8K_FAN_TURBO. Setting the fan state to 3 enables
automatic fan control, just like on the other two fans.
The reason why this was misinterpreted as turbo mode was that
the second fan normally spins faster in automatic mode than
in the previous fan states. Yet when in state 3, the fan speed
reacts to heat exposure, exposing the automatic mode setting.
Some motherboards require more time to acquire the ACPI mutex,
causing "Failed to acquire mutex" messages to appear in the kernel log.
Increase the timeout from 500ms to 800ms to accommodate these cases.
If uprobe handler changes instruction pointer we still execute single
step) or emulate the original instruction and increment the (new) ip
with its length.
This makes the new instruction pointer bogus and application will
likely crash on illegal instruction execution.
If user decided to take execution elsewhere, it makes little sense
to execute the original instruction, so let's skip it.
"""
A host requests an explicit persistent connection ... by specifying a
non-zero Keep Alive Timer value in the Connect command.
"""
As such if we are starting a persistent connection to a discovery
controller and the KATO is currently 0 we need to update KATO to a non
zero value to avoid continuous timeouts on the target.
Signed-off-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is possible for drivers to generate xdp packets with data residing
entirely in fragments. To keep parsing headers using direct packet
access, call bpf_xdp_pull_data() to pull headers into the linear data
area.
It is possible for bpf_xdp_adjust_tail() to free all fragments. The
kfunc currently clears the XDP_FLAGS_HAS_FRAGS bit, but not
XDP_FLAGS_FRAGS_PF_MEMALLOC. So far, this has not caused a issue when
building sk_buff from xdp_buff since all readers of xdp_buff->flags
use the flag only when there are fragments. Clear the
XDP_FLAGS_FRAGS_PF_MEMALLOC bit as well to make the flags correct.
Signed-off-by: Amery Hung <ameryhung@gmail.com> Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Link: https://patch.msgid.link/20250922233356.3356453-2-ameryhung@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
In the __arch_prepare_bpf_trampoline() function, retval_off is only
meaningful when save_ret is true, so the current logic is correct.
However, in the original logic, retval_off is only initialized under
certain conditions; for example, in the fmod_ret logic, the compiler is
not aware that the flags of the fmod_ret program (prog) have set
BPF_TRAMP_F_CALL_ORIG, which results in an uninitialized symbol
compilation warning.
So initialize retval_off unconditionally to fix it.
Root cause is that queue_usage_counter is grabbed with rq_qos_mutex
held in blkg_conf_prep(), while queue should be freezed before
rq_qos_mutex from other context.
The blk_queue_enter() from blkg_conf_prep() is used to protect against
policy deactivation, which is already protected with blkcg_mutex, hence
convert blk_queue_enter() to blkcg_mutex to fix this problem. Meanwhile,
consider that blkcg_mutex is held after queue is freezed from policy
deactivation, also convert blkg_alloc() to use GFP_NOIO.
Signed-off-by: Yu Kuai <yukuai3@huawei.com> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
During normal operation the timers are reprogrammed including an
interrupt acknowledgement. This has no effect as the whole timer
is setup from scratch afterwards. Especially in an interrupt this
has already been done by rttm_timer_interrupt().
Change the behaviour as follows:
- Use rttm_disable_timer() during reprogramming
- Keep rttm_stop_timer() for all other use cases.
Downstream has already tested and confirmed a patch. See
https://github.com/openwrt/openwrt/pull/19468
https://forum.openwrt.org/t/support-for-rtl838x-based-managed-switches/57875/3788
Signed-off-by: Markus Stockhausen <markus.stockhausen@gmx.de> Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Tested-by: Stephen Howell <howels@allthatwemight.be> Tested-by: Bjørn Mork <bjorn@mork.no> Link: https://lore.kernel.org/r/20250804080328.2609287-4-markus.stockhausen@gmx.de Signed-off-by: Sasha Levin <sashal@kernel.org>
The OpenWrt distribution has switched from kernel longterm 6.6 to
6.12. Reports show that devices with the Realtek Otto switch platform
die during operation and are rebooted by the watchdog. Sorting out
other possible reasons the Otto timer is to blame. The platform
currently consists of 4 targets with different hardware revisions.
It is not 100% clear which devices and revisions are affected.
Analysis shows:
A more aggressive sched/deadline handling leads to more timer starts
with small intervals. This increases the bug chances. See
https://marc.info/?l=linux-kernel&m=175276556023276&w=2
Focusing on the real issue a hardware limitation on some devices was
found. There is a minimal chance that a timer ends without firing an
interrupt if it is reprogrammed within the 5us before its expiration
time. Work around this issue by introducing a bounce() function. It
restarts the timer directly before the normal restart functions as
follows:
- Stop timer
- Restart timer with a slow frequency.
- Target time will be >5us
- The subsequent normal restart is outside the critical window
Downstream has already tested and confirmed a patch. See
https://github.com/openwrt/openwrt/pull/19468
https://forum.openwrt.org/t/support-for-rtl838x-based-managed-switches/57875/3788
Signed-off-by: Markus Stockhausen <markus.stockhausen@gmx.de> Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Tested-by: Stephen Howell <howels@allthatwemight.be> Tested-by: Bjørn Mork <bjorn@mork.no> Link: https://lore.kernel.org/r/20250804080328.2609287-2-markus.stockhausen@gmx.de Signed-off-by: Sasha Levin <sashal@kernel.org>
The Microsoft Serial Port Console Redirection (SPCR) specification
revision 1.09 comprises additional field: Precise Baud Rate [1].
It is used to describe non-traditional baud rates (such as those
used by high-speed UARTs).
It contains a specific non-zero baud rate which overrides the value
of the Configured Baud Rate field. If this field is zero or not
present, Configured Baud Rate is used.
Some ublk selftests have strange behavior when fio is not installed.
While most tests behave correctly (run if they don't need fio, or skip
if they need fio), the following tests have different behavior:
- test_null_01, test_null_02, test_generic_01, test_generic_02, and
test_generic_12 try to run fio without checking if it exists first,
and fail on any failure of the fio command (including "fio command
not found"). So these tests fail when they should skip.
- test_stress_05 runs fio without checking if it exists first, but
doesn't fail on fio command failure. This test passes, but that pass
is misleading as the test doesn't do anything useful without fio
installed. So this test passes when it should skip.
Fix these issues by adding _have_program fio checks to the top of all of
these tests.
Signed-off-by: Uday Shankar <ushankar@purestorage.com> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
The main reason is that 'r8' in insn '70' is not an arena pointer.
Further debugging at llvm side shows that llvm commit ([1]) caused
the failure. For the original code:
page[i] = NULL;
page[i + 1] = NULL;
the llvm transformed it to something like below at source level:
__builtin_memset(&page[i], 0, 16)
Such transformation prevents llvm BPFCheckAndAdjustIR pass from
generating proper addr_space_cast insns ([2]).
Adding support in llvm BPFCheckAndAdjustIR pass should work, but
not sure that such a pattern exists or not in real applications.
At the same time, simply adding a memory barrier between two 'page'
assignment can fix the issue.
sys_get_robust_list() and compat_get_robust_list() use ptrace_may_access()
to check if the calling task is allowed to access another task's
robust_list pointer. This check is racy against a concurrent exec() in the
target process.
During exec(), a task may transition from a non-privileged binary to a
privileged one (e.g., setuid binary) and its credentials/memory mappings
may change. If get_robust_list() performs ptrace_may_access() before
this transition, it may erroneously allow access to sensitive information
after the target becomes privileged.
A racy access allows an attacker to exploit a window during which
ptrace_may_access() passes before a target process transitions to a
privileged state via exec().
For example, consider a non-privileged task T that is about to execute a
setuid-root binary. An attacker task A calls get_robust_list(T) while T
is still unprivileged. Since ptrace_may_access() checks permissions
based on current credentials, it succeeds. However, if T begins exec
immediately afterwards, it becomes privileged and may change its memory
mappings. Because get_robust_list() proceeds to access T->robust_list
without synchronizing with exec() it may read user-space pointers from a
now-privileged process.
This violates the intended post-exec access restrictions and could
expose sensitive memory addresses or be used as a primitive in a larger
exploit chain. Consequently, the race can lead to unauthorized
disclosure of information across privilege boundaries and poses a
potential security risk.
Take a read lock on signal->exec_update_lock prior to invoking
ptrace_may_access() and accessing the robust_list/compat_robust_list.
This ensures that the target task's exec state remains stable during the
check, allowing for consistent and synchronized validation of
credentials.
Refuse to register a cpuidle device if the given CPU has a cpuidle
device already and print a message regarding it.
Without this, an attempt to register a new cpuidle device without
unregistering the existing one leads to the removal of the existing
cpuidle device without removing its sysfs interface.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
btf_dumper.c:71:31: error: variable 'finfo' is uninitialized when passed as a const pointer argument here [-Werror,-Wuninitialized-const-pointer]
71 | info.func_info = ptr_to_u64(&finfo);
| ^~~~~
prog.c:2294:31: error: variable 'func_info' is uninitialized when passed as a const pointer argument here [-Werror,-Wuninitialized-const-pointer]
2294 | info.func_info = ptr_to_u64(&func_info);
|
The X1E80100 battery management firmware sends a notification with
code 0x83 when the battery charging state changes, such as switching
between fast charge, taper charge, end of charge, or any other error
charging states.
The same notification code is used with bit[8] set when charging stops
because the charge control end threshold is reached. Additionally,
a 2-bit value is included in bit[10:9] with the same code to indicate
the charging source capability, which is determined by the calculated
power from voltage and current readings from PDOs: 2 means a strong
charger over 60W, 1 indicates a weak charger, and 0 means there is no
charging source.
These 3-MSB [10:8] in the notification code is not much useful for now,
hence just ignore them and trigger a power supply change event whenever
0x83 notification code is received. This helps to eliminate the unknown
notification error messages.
Reported-by: Sebastian Reichel <sebastian.reichel@collabora.com> Closes: https://lore.kernel.org/all/r65idyc4of5obo6untebw4iqfj2zteiggnnzabrqtlcinvtddx@xc4aig5abesu/ Signed-off-by: Fenglin Wu <fenglin.wu@oss.qualcomm.com> Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
After discussion with the devicetree maintainers we agreed to not extend
lists with the generic compatible "apple,pmgr-pwrstate" anymore [1]. Use
"apple,t8103-pmgr-pwrstate" as base compatible as it is the SoC the
driver and bindings were written for.
The cpupower_write_sysfs() function currently returns -1 on
write failure, but the function signature indicates it should
return an unsigned int. Returning -1 from an unsigned function
results in a large positive value rather than indicating
an error condition.
Fix this by returning 0 on failure, which is more appropriate
for an unsigned return type and maintains consistency with typical
success/failure semantics where 0 indicates failure and non-zero
indicates success (bytes written).
The `kunit_test` proc macro only checks for the `test` attribute
immediately preceding a `fn`. If the function is disabled via a `cfg`,
the generated code would result in a compile error referencing a
non-existent function [1].
This collects attributes and specifically cherry-picks `cfg` attributes
to be duplicated inside KUnit wrapper functions such that a test function
disabled via `cfg` compiles and is marked as skipped in KUnit correctly.
The bpf_cgroup_from_id kfunc relies on cgroup_get_from_id to obtain the
cgroup corresponding to a given cgroup ID. This helper can be called in
a lot of contexts where the current thread can be random. A recent
example was its use in sched_ext's ops.tick(), to obtain the root cgroup
pointer. Since the current task can be whatever random user space task
preempted by the timer tick, this makes the behavior of the helper
unreliable.
Refactor out __cgroup_get_from_id as the non-namespace aware version of
cgroup_get_from_id, and change bpf_cgroup_from_id to make use of it.
There is no compatibility breakage here, since changing the namespace
against which the lookup is being done to the root cgroup namespace only
permits a wider set of lookups to succeed now. The cgroup IDs across
namespaces are globally unique, and thus don't need to be retranslated.
For systems having CONFIG_NR_CPUS set to > 1024 in kernel config
the selftest fails as arena_spin_lock_irqsave() returns EOPNOTSUPP.
(eg - incase of powerpc default value for CONFIG_NR_CPUS is 8192)
The selftest is skipped incase bpf program returns EOPNOTSUPP,
with a descriptive message logged.
Various KUnit tests require PCI infrastructure to work. All normal
platforms enable PCI by default, but UML does not. Enabling PCI from
.kunitconfig files is problematic as it would not be portable. So in
commit 6fc3a8636a7b ("kunit: tool: Enable virtio/PCI by default on UML")
PCI was enabled by way of CONFIG_UML_PCI_OVER_VIRTIO=y. However
CONFIG_UML_PCI_OVER_VIRTIO requires additional configuration of
CONFIG_UML_PCI_OVER_VIRTIO_DEVICE_ID or will otherwise trigger a WARN() in
virtio_pcidev_init(). However there is no one correct value for
UML_PCI_OVER_VIRTIO_DEVICE_ID which could be used by default.
This warning is confusing when debugging test failures.
On the other hand, the functionality of CONFIG_UML_PCI_OVER_VIRTIO is not
used at all, given that it is completely non-functional as indicated by
the WARN() in question. Instead it is only used as a way to enable
CONFIG_UML_PCI which itself is not directly configurable.
Instead of going through CONFIG_UML_PCI_OVER_VIRTIO, introduce a custom
configuration option which enables CONFIG_UML_PCI without triggering
warnings or building dead code.
Link: https://lore.kernel.org/r/20250908-kunit-uml-pci-v2-1-d8eba5f73c9d@linutronix.de Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de> Reviewed-by: Johannes Berg <johannes@sipsolutions.net> Reviewed-by: David Gow <davidgow@google.com> Signed-off-by: Shuah Khan <skhan@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
nvme_fc_unregister_remote removes the remote port on a lport object at
any point in time when there is no active association. This races with
with the reconnect logic, because nvme_fc_create_association is not
taking a lock to check the port_state and atomically increase the
active count on the rport.
Reported-by: Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com> Closes: https://lore.kernel.org/all/u4ttvhnn7lark5w3sgrbuy2rxupcvosp4qmvj46nwzgeo5ausc@uyrkdls2muwx Signed-off-by: Daniel Wagner <wagi@kernel.org> Reviewed-by: Hannes Reinecke <hare@suse.de> Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When forcefully shutting down a port via the configfs interface,
nvmet_port_subsys_drop_link() first calls nvmet_port_del_ctrls() and
then nvmet_disable_port(). Both functions will eventually schedule all
remaining associations for deletion.
The current implementation checks whether an association is about to be
removed, but only after the work item has already been scheduled. As a
result, it is possible for the first scheduled work item to free all
resources, and then for the same work item to be scheduled again for
deletion.
Because the association list is an RCU list, it is not possible to take
a lock and remove the list entry directly, so it cannot be looked up
again. Instead, a flag (terminating) must be used to determine whether
the association is already in the process of being deleted.
Reported-by: Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com> Closes: https://lore.kernel.org/all/rsdinhafrtlguauhesmrrzkybpnvwantwmyfq2ih5aregghax5@mhr7v3eryci3/ Reviewed-by: Hannes Reinecke <hare@suse.de> Signed-off-by: Daniel Wagner <wagi@kernel.org> Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
A TEE driver doesn't always need to provide a pool if it doesn't
support memory sharing ioctls and can allocate memory for TEE
messages in another way. Although this is mentioned in the
documentation for tee_device_alloc(), it is not handled correctly.
When deleting the previous walkstate operand stack
acpi_ds_call_control_method() was deleting obj_desc->Method.param_count
operands. But Method.param_count does not necessarily match
this_walk_state->num_operands, it may be either less or more.
After correcting the for loop to check `i < this_walk_state->num_operands`
the code is identical to acpi_ds_clear_operands(), so just outright
replace the code with acpi_ds_clear_operands() to fix this.
Link: https://github.com/acpica/acpica/commit/53fc0220 Signed-off-by: Hans de Goede <hansg@kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
For Qualcomm SoCs which needs level shifter for SD card, extra delay is
seen on receiver data path.
To compensate this delay enable tuning for SDR50 mode for targets which
has level shifter. SDHCI_SDR50_NEEDS_TUNING caps will be set for targets
with level shifter on Qualcomm SOC's.
projects / thirdparty / kernel / stable.git / log
