Merge series from Peter Ujfalusi <peter.ujfalusi@linux.intel.com>:
The current version of delay reporting code can report incorrect
values when paired with a firmware which enables this feature.
Unfortunately there are several smaller issues that needed to be addressed
to correct the behavior:
Wrong information was used for the host side of counter
For MTL/LNL used incorrect (in a sense that it was verified only on MTL)
link side counter function.
The link side counter needs compensation logic if pause/resume is used.
The offset values were not refreshed from firmware.
Finally, not strictly connected, but the ALSA buffer size needs to be
constrained to avoid constant xrun from media players (like mpv)
The series applies cleanly for 6.9 and 6.8.y stable, but older stable
would need manual backport, but it is questionable if it is needed as
MTL/LNL is missing features.
Clock direction is marked as 'unsigned int' but only two values are
currently used in practice. Note these down in the documentation.
Signed-off-by: John Watts <contact@jookia.org>
Link: https://msgid.link/r/20240326-dai_mclk_hint-v1-1-653cbd2d78d9@jookia.org
Signed-off-by: Mark Brown <broonie@kernel.org>
It has been reported the commit cf3986f8c0 introduced a regression
when the temperature is wavering in the hysteresis region. The
mitigation stops leading to an uncontrolled temperature increase until
reaching the critical trip point.
Here what happens:
* 'throttle' is when the current temperature is greater than the trip
point temperature
* 'target' is the mitigation level
* 'passive' is positive when there is a mitigation, zero otherwise
* these values are computed in the step_wise governor
Configuration:
trip point 1: temp=95°C, hyst=5°C (passive)
trip point 2: temp=115°C, hyst=0°C (critical)
governor: step_wise
1. The temperature crosses the way up the trip point 1 at 95°C
- trend=raising
- throttle=1, target=1
- passive=1
- set_trips: low=90°C, high=115°C
2. The temperature decreases but stays in the hysteresis region at
93°C
- trend=dropping
- throttle=0, target=0
- passive=1
Before cf3986f8c0
- set_trips: low=90°C, high=95°C
After cf3986f8c0
- set_trips: low=90°C, high=115°C
3. The temperature increases a bit but stays in the hysteresis region
at 94°C (so below the trip point 1 temp 95°C)
- trend=raising
- throttle=0, target=0
- passive=1
Before cf3986f8c0
- set_trips: low=90°C, high=95°C
After cf3986f8c0
- set_trips: low=90°C, high=115°C
4. The temperature decreases but stays in the hysteresis region at
93°C
- trend=dropping
- throttle=0, target=THERMAL_NO_TARGET
- passive=0
Before cf3986f8c0
- set_trips: low=90°C, high=95°C
After cf3986f8c0
- set_trips: low=90°C, high=115°C
At this point, the 'passive' value is zero, there is no mitigation,
the temperature is in the hysteresis region, the next trip point is
115°C. As 'passive' is zero, the timer to monitor the thermal zone is
disabled. Consequently if the temperature continues to increase, no
mitigation will happen and it will reach the 115°C trip point and
reboot.
Before the optimization, the high boundary would have been 95°C, thus
triggering the mitigation again and rearming the polling timer.
The optimization make sense but given the current implementation of
the step_wise governor collaborating via this 'passive' flag with the
core framework it can not work.
From a higher perspective it seems like there is a problem between the
governor which sets a variable to be used by the core framework. That
sounds akward and it would make much more sense if the core framework
controls the governor and not the opposite. But as the devil hides in
the details, there are some subtilities to be addressed before.
Elaborating those would be out of the scope this changelog. So let's
stay simple and revert the change first to fixup all broken mobile
platforms.
This reverts commit cf3986f8c0 ("thermal: core: Don't update trip
points inside the hysteresis range") and takes a conflict with commit
0c0c4740c9 ("0c0c4740c9 thermal: trip: Use for_each_trip() in
__thermal_zone_set_trips()") in drivers/thermal/thermal_trip.c into
account.
Fixes: cf3986f8c0 ("thermal: core: Don't update trip points inside the hysteresis range")
Reported-by: Manaf Meethalavalappu Pallikunhi <quic_manafm@quicinc.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>
Cc: 6.7+ <stable@vger.kernel.org> # 6.7+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Improve the error logging in the driver probe failure paths.
Also use dev_err_probe which is probe error check and log helper to
prevent logging in case of probe deferral.
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Link: https://lore.kernel.org/r/20240325204620.1437237-6-cristian.marussi@arm.com
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
ACPICA commit 9061cd9aa131205657c811a52a9f8325a040c6c9
Errors in acpi_evaluate_object() can lead to incorrect state of buffer.
This can lead to access to data in previously ACPI_FREEd buffer and
secondary ACPI_FREE to the same buffer later.
Handle errors in acpi_evaluate_object the same way it is done earlier
with acpi_ns_handle_to_pathname.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Link: 9061cd9a
Fixes: 5fd033288a ("ACPICA: debugger: add command to dump all fields of particular subtype")
Signed-off-by: Nikita Kiryushin <kiryushin@ancud.ru>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
-----BEGIN PGP SIGNATURE-----
iHUEABYIAB0WIQTFp0I1jqZrAX+hPRXbK58LschIgwUCZgHmTAAKCRDbK58LschI
g1gWAP9HjAWE/Sy0B2t9opIiTqRzdMJLYs2B4OFeHRI6+qQg0gD6A4jsKEh/xmtG
Hhjw+AElJRFZ3SUIT4mZlljzUHIYYAA=
=T0lM
-----END PGP SIGNATURE-----
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Daniel Borkmann says:
====================
pull-request: bpf 2024-03-25
The following pull-request contains BPF updates for your *net* tree.
We've added 17 non-merge commits during the last 12 day(s) which contain
a total of 19 files changed, 184 insertions(+), 61 deletions(-).
The main changes are:
1) Fix an arm64 BPF JIT bug in BPF_LDX_MEMSX implementation's offset handling
found via test_bpf module, from Puranjay Mohan.
2) Various fixups to the BPF arena code in particular in the BPF verifier and
around BPF selftests to match latest corresponding LLVM implementation,
from Puranjay Mohan and Alexei Starovoitov.
3) Fix xsk to not assume that metadata is always requested in TX completion,
from Stanislav Fomichev.
4) Fix riscv BPF JIT's kfunc parameter incompatibility between BPF and the riscv
ABI which requires sign-extension on int/uint, from Pu Lehui.
5) Fix s390x BPF JIT's bpf_plt pointer arithmetic which triggered a crash when
testing struct_ops, from Ilya Leoshkevich.
6) Fix libbpf's arena mmap handling which had incorrect u64-to-pointer cast on
32-bit architectures, from Andrii Nakryiko.
7) Fix libbpf to define MFD_CLOEXEC when not available, from Arnaldo Carvalho de Melo.
8) Fix arm64 BPF JIT implementation for 32bit unconditional bswap which
resulted in an incorrect swap as indicated by test_bpf, from Artem Savkov.
9) Fix BPF man page build script to use silent mode, from Hangbin Liu.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
riscv, bpf: Fix kfunc parameters incompatibility between bpf and riscv abi
bpf: verifier: reject addr_space_cast insn without arena
selftests/bpf: verifier_arena: fix mmap address for arm64
bpf: verifier: fix addr_space_cast from as(1) to as(0)
libbpf: Define MFD_CLOEXEC if not available
arm64: bpf: fix 32bit unconditional bswap
bpf, arm64: fix bug in BPF_LDX_MEMSX
libbpf: fix u64-to-pointer cast on 32-bit arches
s390/bpf: Fix bpf_plt pointer arithmetic
xsk: Don't assume metadata is always requested in TX completion
selftests/bpf: Add arena test case for 4Gbyte corner case
selftests/bpf: Remove hard coded PAGE_SIZE macro.
libbpf, selftests/bpf: Adjust libbpf, bpftool, selftests to match LLVM
bpf: Clarify bpf_arena comments.
MAINTAINERS: Update email address for Quentin Monnet
scripts/bpf_doc: Use silent mode when exec make cmd
bpf: Temporarily disable atomic operations in BPF arena
====================
Link: https://lore.kernel.org/r/20240325213520.26688-1-daniel@iogearbox.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Merge series from Bo Liu <liubo03@inspur.com>:
The maple tree register cache is based on a much more modern data structure
than the rbtree cache and makes optimisation choices which are probably
more appropriate for modern systems than those made by the rbtree cache.
Bo Liu (13):
regulator: da9121: convert to use maple tree register cache
regulator: da9211: convert to use maple tree register cache
regulator: isl9305: convert to use maple tree register cache
regulator: max8973: convert to use maple tree register cache
regulator: mt6311: convert to use maple tree register cache
regulator: pca9450: convert to use maple tree register cache
regulator: pf8x00: convert to use maple tree register cache
regulator: pfuze100: convert to use maple tree register cache
regulator: rtmv20: convert to use maple tree register cache
regulator: rtq6752: convert to use maple tree register cache
regulator: tps51632: convert to use maple tree register cache
regulator: tps62360: convert to use maple tree register cache
regulator: rpi-panel-attiny: convert to use maple tree register cache
drivers/regulator/da9121-regulator.c | 4 ++--
drivers/regulator/da9211-regulator.c | 2 +-
drivers/regulator/isl9305.c | 2 +-
drivers/regulator/max8973-regulator.c | 2 +-
drivers/regulator/mt6311-regulator.c | 2 +-
drivers/regulator/pca9450-regulator.c | 2 +-
drivers/regulator/pf8x00-regulator.c | 2 +-
drivers/regulator/pfuze100-regulator.c | 2 +-
drivers/regulator/rpi-panel-attiny-regulator.c | 2 +-
drivers/regulator/rtmv20-regulator.c | 2 +-
drivers/regulator/rtq6752-regulator.c | 2 +-
drivers/regulator/tps51632-regulator.c | 2 +-
drivers/regulator/tps62360-regulator.c | 2 +-
13 files changed, 14 insertions(+), 14 deletions(-)
--
2.18.2
When consolidating GPIO lookups in ACPI code, the debug messaging
had been reworked that the user may see
[ 13.401147] (NULL device *): using ACPI '\_SB.LEDS.led-0' for '(null)' GPIO lookup
[ 13.401378] gpio gpiochip0: Persistence not supported for GPIO 40
[ 13.401402] gpio-40 (?): no flags found for (null)
instead of
[ 14.182962] gpio gpiochip0: Persistence not supported for GPIO 40
[ 14.182994] gpio-40 (?): no flags found for gpios
The '(null)' parts are less informative and likely scare the users.
Replace them by '(default)' which can point out to the default connection
IDs, such as 'gpios'.
While at it, amend other places where con_id is used in the messages.
Reported-by: Ferry Toth <ftoth@exalondelft.nl>
Fixes: 8eb1f71e7a ("gpiolib: consolidate GPIO lookups")
Suggested-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Tested-by: Ferry Toth <ftoth@exalondelft.nl>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Merge series from Andy Shevchenko <andriy.shevchenko@linux.intel.com>:
Fix kernel documentation and inclusion block, and dropping the size
of the num_chipselect.
Merge series from Andy Shevchenko <andriy.shevchenko@linux.intel.com>:
A couple of cleanups against linux/spi/pxa2xx_spi.h.
I'm sending this as v3 to land in the SPI subsystem. Meanwhile I'm
preparing an update to make linux/spi/pxa2xx_spi.h private to the
subsystem (PXA2xx driver). But the second part will be presented later
on (likely after v6.9-rc1). That said, this can be routed either via
SoC tree or SPI, up to respective maintainers.
The gpio_device_find_by_() functions do not have stubs which means that if
they are referenced from code with an optiona dependency on gpiolib then
the code will fail to link. Add stubs for lookups via fwnode and label. I
have not added a stub for plain gpio_device_find() since it seems harder to
see a use case for that which does not depend on gpiolib.
With the addition of the GPIO reset controller (which lacks a gpiolib
dependency) to the arm64 defconfig this is causing build breaks for arm64
virtconfig in -next:
aarch64-linux-gnu-ld: drivers/reset/core.o: in function `__reset_add_reset_gpio_lookup':
/build/stage/linux/drivers/reset/core.c:861:(.text+0xccc): undefined reference to `gpio_device_find_by_fwnode'
Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
When an interrupt is requested, a procfs directory is created under
"/proc/irq/<irqnum>/<label>" where <label> is the string passed to one of
the request_irq() variants.
What follows is that the string must not contain the "/" character or
the procfs mkdir operation will fail. We don't have such constraints for
GPIO consumer labels which are used verbatim as interrupt labels for
GPIO irqs. We must therefore sanitize the consumer string before
requesting the interrupt.
Let's replace all "/" with ":".
Cc: stable@vger.kernel.org
Reported-by: Stefan Wahren <wahrenst@gmx.net>
Closes: https://lore.kernel.org/linux-gpio/39fe95cb-aa83-4b8b-8cab-63947a726754@gmx.net/
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Reviewed-by: Kent Gibson <warthog618@gmail.com>
Calling i915_gem_object_get_dma_address() from the vblank
evade critical section triggers might_sleep().
While we know that we've already pinned the framebuffer
and thus i915_gem_object_get_dma_address() will in fact
not sleep in this case, it seems reasonable to keep the
unconditional might_sleep() for maximum coverage.
So let's instead pre-populate the dma address during
fb pinning, which all happens before we enter the
vblank evade critical section.
We can use u32 for the dma address as this class of
hardware doesn't support >32bit addresses.
Cc: stable@vger.kernel.org
Fixes: 0225a90981 ("drm/i915: Make cursor plane registers unlocked")
Reported-by: Borislav Petkov <bp@alien8.de>
Closes: https://lore.kernel.org/intel-gfx/20240227100342.GAZd2zfmYcPS_SndtO@fat_crate.local/
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240325175738.3440-1-ville.syrjala@linux.intel.com
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Systems with a large number of CPUs may generate a large number of
machine check records when things go seriously wrong. But Linux has
a fixed-size buffer that can only capture a few dozen errors.
Allocate space based on the number of CPUs (with a minimum value based
on the historical fixed buffer that could store 80 records).
[ bp: Rename local var from tmpp to something more telling: gpool. ]
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
Reviewed-by: Avadhut Naik <avadhut.naik@amd.com>
Link: https://lore.kernel.org/r/20240307192704.37213-1-tony.luck@intel.com
The current code is pulling the wrong pointer causing it to disable the
wrong IRQ. Correct the code to pull the correct cs42l43 core data
pointer.
Fixes: 64353af49f ("ASoC: cs42l43: Add system suspend ops to disable IRQ")
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://msgid.link/r/20240326105434.852907-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Unregistering SCMI notifications using the managed devres interface can be
done providing as a reference simply the previously successfully registered
notification block since it could have been registered only on one kernel
notification_chain: drop any reference to SCMI protocol, events and
sources.
Devres internal helpers can search for the provided notification block
reference and, once found, the associated devres object will already
provide the above SCMI references for the event.
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Link: https://lore.kernel.org/r/20240325204620.1437237-5-cristian.marussi@arm.com
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
It is useful to have message dump traces for any invalid/bad/unexpected
replies. Let us add traces for the same as well as late-timed-out,
out-of-order and unexpected/spurious messages.
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Link: https://lore.kernel.org/r/20240325204620.1437237-4-cristian.marussi@arm.com
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Locally generated IP multicast packets (such as the ones used in the
test) do not perform routing and simply egress the bound device.
However, as explained in commit 8bcfb4ae4d ("selftests: forwarding:
Fix failing tests with old libnet"), old versions of libnet (used by
mausezahn) do not use the "SO_BINDTODEVICE" socket option. Specifically,
the library started using the option for IPv6 sockets in version 1.1.6
and for IPv4 sockets in version 1.2. This explains why on Ubuntu - which
uses version 1.1.6 - the IPv4 overlay tests are failing whereas the IPv6
ones are passing.
Fix by specifying the source and destination MAC of the packets which
will cause mausezahn to use a packet socket instead of an IP socket.
Fixes: 62199e3f16 ("selftests: net: Add VXLAN MDB test")
Reported-by: Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr>
Closes: https://lore.kernel.org/netdev/5bb50349-196d-4892-8ed2-f37543aa863f@alu.unizg.hr/
Tested-by: Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20240325075030.2379513-1-idosch@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
We have a couple of duplicate logs and missing information, add
__func__ consistently and make sure useful error codes are logged.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Link: https://msgid.link/r/20240325221925.206507-1-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The gpio field is not used anymore, remove the leftover.
This also fixes the compilation error after the ...
Fixes: 9855f05e55 ("ASoC: fsl: imx-es8328: Switch to using gpiod API")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://msgid.link/r/20240325191341.3977321-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Upon reception of malformed and unexpected timed-out SCMI messages, it is
not possible to trace those bad messages in their entirety, because usually
we cannot even retrieve the payload, or it is just not reliable.
Add a helper to trace at least the content of the header of the received
message while associating a meaningful tag and error code.
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Link: https://lore.kernel.org/r/20240325204620.1437237-3-cristian.marussi@arm.com
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The dreamcastcard->timer could schedule the spu_dma_work and the
spu_dma_work could also arm the dreamcastcard->timer.
When the snd_pcm_substream is closing, the aica_channel will be
deallocated. But it could still be dereferenced in the worker
thread. The reason is that del_timer() will return directly
regardless of whether the timer handler is running or not and
the worker could be rescheduled in the timer handler. As a result,
the UAF bug will happen. The racy situation is shown below:
(Thread 1) | (Thread 2)
snd_aicapcm_pcm_close() |
... | run_spu_dma() //worker
| mod_timer()
flush_work() |
del_timer() | aica_period_elapsed() //timer
kfree(dreamcastcard->channel) | schedule_work()
| run_spu_dma() //worker
... | dreamcastcard->channel-> //USE
In order to mitigate this bug and other possible corner cases,
call mod_timer() conditionally in run_spu_dma(), then implement
PCM sync_stop op to cancel both the timer and worker. The sync_stop
op will be called from PCM core appropriately when needed.
Fixes: 198de43d75 ("[ALSA] Add ALSA support for the SEGA Dreamcast PCM device")
Suggested-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Message-ID: <20240326094238.95442-1-duoming@zju.edu.cn>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Sebastian Andrzej Siewior says:
====================
net: Provide SMP threads for backlog NAPI
The RPS code and "deferred skb free" both send IPI/ function call
to a remote CPU in which a softirq is raised. This leads to a warning on
PREEMPT_RT because raising softiqrs from function call led to undesired
behaviour in the past. I had duct tape in RT for the "deferred skb free"
and Wander Lairson Costa reported the RPS case.
This series only provides support for SMP threads for backlog NAPI, I
did not attach a patch to make it default and remove the IPI related
code to avoid confusion. I can post it for reference it asked.
The RedHat performance team was so kind to provide some testing here.
The series (with the IPI code removed) has been tested and no regression
vs without the series has been found. For testing iperf3 was used on 25G
interface, provided by mlx5, ix40e or ice driver and RPS was enabled. I
can provide the individual test results if needed.
====================
Link: https://lore.kernel.org/r/20240325074943.289909-1-bigeasy@linutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The rps_lock.*() functions use the inner lock of a sk_buff_head for
locking. This lock is used if RPS is enabled, otherwise the list is
accessed lockless and disabling interrupts is enough for the
synchronisation because it is only accessed CPU local. Not only the list
is protected but also the NAPI state protected.
With the addition of backlog threads, the lock is also needed because of
the cross CPU access even without RPS. The clean up of the defer_list
list is also done via backlog threads (if enabled).
It has been suggested to rename the locking function since it is no
longer just RPS.
Rename the rps_lock*() functions to backlog_lock*().
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The defer_list is a per-CPU list which is used to free skbs outside of
the socket lock and on the CPU on which they have been allocated.
The list is processed during NAPI callbacks so ideally the list is
cleaned up.
Should the amount of skbs on the list exceed a certain water mark then
the softirq is triggered remotely on the target CPU by invoking a remote
function call. The raise of the softirqs via a remote function call
leads to waking the ksoftirqd on PREEMPT_RT which is undesired.
The backlog-NAPI threads already provide the infrastructure which can be
utilized to perform the cleanup of the defer_list.
The NAPI state is updated with the input_pkt_queue.lock acquired. It
order not to break the state, it is needed to also wake the backlog-NAPI
thread with the lock held. This requires to acquire the use the lock in
rps_lock_irq*() if the backlog-NAPI threads are used even with RPS
disabled.
Move the logic of remotely starting softirqs to clean up the defer_list
into kick_defer_list_purge(). Make sure a lock is held in
rps_lock_irq*() if backlog-NAPI threads are used. Schedule backlog-NAPI
for defer_list cleanup if backlog-NAPI is available.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Backlog NAPI is a per-CPU NAPI struct only (with no device behind it)
used by drivers which don't do NAPI them self, RPS and parts of the
stack which need to avoid recursive deadlocks while processing a packet.
The non-NAPI driver use the CPU local backlog NAPI. If RPS is enabled
then a flow for the skb is computed and based on the flow the skb can be
enqueued on a remote CPU. Scheduling/ raising the softirq (for backlog's
NAPI) on the remote CPU isn't trivial because the softirq is only
scheduled on the local CPU and performed after the hardirq is done.
In order to schedule a softirq on the remote CPU, an IPI is sent to the
remote CPU which schedules the backlog-NAPI on the then local CPU.
On PREEMPT_RT interrupts are force-threaded. The soft interrupts are
raised within the interrupt thread and processed after the interrupt
handler completed still within the context of the interrupt thread. The
softirq is handled in the context where it originated.
With force-threaded interrupts enabled, ksoftirqd is woken up if a
softirq is raised from hardirq context. This is the case if it is raised
from an IPI. Additionally there is a warning on PREEMPT_RT if the
softirq is raised from the idle thread.
This was done for two reasons:
- With threaded interrupts the processing should happen in thread
context (where it originated) and ksoftirqd is the only thread for
this context if raised from hardirq. Using the currently running task
instead would "punish" a random task.
- Once ksoftirqd is active it consumes all further softirqs until it
stops running. This changed recently and is no longer the case.
Instead of keeping the backlog NAPI in ksoftirqd (in force-threaded/
PREEMPT_RT setups) I am proposing NAPI-threads for backlog.
The "proper" setup with threaded-NAPI is not doable because the threads
are not pinned to an individual CPU and can be modified by the user.
Additionally a dummy network device would have to be assigned. Also
CPU-hotplug has to be considered if additional CPUs show up.
All this can be probably done/ solved but the smpboot-threads already
provide this infrastructure.
Sending UDP packets over loopback expects that the packet is processed
within the call. Delaying it by handing it over to the thread hurts
performance. It is not beneficial to the outcome if the context switch
happens immediately after enqueue or after a while to process a few
packets in a batch.
There is no need to always use the thread if the backlog NAPI is
requested on the local CPU. This restores the loopback throuput. The
performance drops mostly to the same value after enabling RPS on the
loopback comparing the IPI and the tread result.
Create NAPI-threads for backlog if request during boot. The thread runs
the inner loop from napi_threaded_poll(), the wait part is different. It
checks for NAPI_STATE_SCHED (the backlog NAPI can not be disabled).
The NAPI threads for backlog are optional, it has to be enabled via the boot
argument "thread_backlog_napi". It is mandatory for PREEMPT_RT to avoid the
wakeup of ksoftirqd from the IPI.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
A NAPI thread is scheduled by first setting NAPI_STATE_SCHED bit. If
successful (the bit was not yet set) then the NAPI_STATE_SCHED_THREADED
is set but only if thread's state is not TASK_INTERRUPTIBLE (is
TASK_RUNNING) followed by task wakeup.
If the task is idle (TASK_INTERRUPTIBLE) then the
NAPI_STATE_SCHED_THREADED bit is not set. The thread is no relying on
the bit but always leaving the wait-loop after returning from schedule()
because there must have been a wakeup.
The smpboot-threads implementation for per-CPU threads requires an
explicit condition and does not support "if we get out of schedule()
then there must be something to do".
Removing this optimisation simplifies the following integration.
Set NAPI_STATE_SCHED_THREADED unconditionally on wakeup and rely on it
in the wait path by removing the `woken' condition.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Use the `DEFINE_FLEX()` helper for an on-stack definition of a
flexible structure where the size of the flexible-array member
is known at compile-time, and refactor the rest of the code,
accordingly.
So, with these changes, fix the following warning:
sound/firewire/amdtp-stream.c:1184:46: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
Link: https://github.com/KSPP/linux/issues/202
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Message-ID: <ZgIsBqoMb7p3fMDr@neat>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The remove callback of a device is called whenever it is unbound,
which may happen during runtime e.g. through sysfs, so this is not
allowed to be dropped from the binary:
WARNING: modpost: vmlinux: section mismatch in reference: einj_driver+0x8 (section: .data) -> einj_remove (section: .exit.text)
ERROR: modpost: Section mismatches detected.
Remove that annotation.
Fixes: 12fb28ea6b ("EINJ: Add CXL error type support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Currently, the performance domain names are not logged whenever any
error occurs when processing the OPPs by adding to local data structures
or to the OPP library.
It would be easier to locate the problem if domain name is printed out.
So let us add the performance domain names to the other information
logged already in the error paths.
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Link: https://lore.kernel.org/r/20240322080531.3365016-1-peng.fan@oss.nxp.com
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
STE attributes(NSCFG, PRIVCFG, INSTCFG) use value 0 for "Use Icomming",
for some reason SHCFG doesn't follow that, and it is defined as "0b01".
Currently the driver sets SHCFG to Use Incoming for stage-2 and bypass
domains.
However according to the User Manual (ARM IHI 0070 F.b):
When SMMU_IDR1.ATTR_TYPES_OVR == 0, this field is RES0 and the
incoming Shareability attribute is used.
This patch adds a condition for writing SHCFG to Use incoming to be
compliant with the architecture, and defines ATTR_TYPE_OVR as a new
feature discovered from IDR1.
This also required to propagate the SMMU through some functions args.
There is no need to add similar condition for the newly introduced function
arm_smmu_get_ste_used() as the values of the STE are the same before and
after any transition, so this will not trigger any change. (we already
do the same for the VMID).
Although this is a misconfiguration from the driver, this has been there
for a long time, so probably no HW running Linux is affected by it.
Reported-by: Will Deacon <will@kernel.org>
Closes: https://lore.kernel.org/all/20240215134952.GA690@willie-the-truck/
Signed-off-by: Mostafa Saleh <smostafa@google.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20240323134658.464743-1-smostafa@google.com
Signed-off-by: Will Deacon <will@kernel.org>
Since the Renesas Ethernet Switch driver was added by Yoshihiro Shimoda,
I started receiving the patches to review for it -- which I was unable to
do, as I don't know this hardware and don't even have the manuals for it.
Fortunately, Shimoda-san has volunteered to be a reviewer for this new
driver, thus let's now split the single entry into 3 per-driver entries,
each with its own reviewer...
Signed-off-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/de0ccc1d-6fc0-583f-4f80-f70e6461d62d@omp.ru
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Jason Xing says:
====================
trace: use TP_STORE_ADDRS macro
Using the macro for other tracepoints use to be more concise.
No functional change.
====================
Link: https://lore.kernel.org/r/20240325034347.19522-1-kerneljasonxing@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As the title said, use the macro directly like the patch[1] did
to avoid those duplications. No functional change.
[1]
commit 6a6b0b9914 ("tcp: Avoid preprocessor directives in tracepoint macro args")
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As the title said, use the macro directly like the patch[1] did
to avoid those duplications. No functional change.
[1]
commit 6a6b0b9914 ("tcp: Avoid preprocessor directives in tracepoint macro args")
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Put the macro into another standalone file for better extension.
Some tracepoints can use this common part in the future.
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Among the attribute file callback routines in
drivers/usb/core/sysfs.c, the interface_authorized_store() function is
the only one which acquires a device lock on an ancestor device: It
calls usb_deauthorize_interface(), which locks the interface's parent
USB device.
The will lead to deadlock if another process already owns that lock
and tries to remove the interface, whether through a configuration
change or because the device has been disconnected. As part of the
removal procedure, device_del() waits for all ongoing sysfs attribute
callbacks to complete. But usb_deauthorize_interface() can't complete
until the device lock has been released, and the lock won't be
released until the removal has finished.
The mechanism provided by sysfs to prevent this kind of deadlock is
to use the sysfs_break_active_protection() function, which tells sysfs
not to wait for the attribute callback.
Reported-and-tested by: Yue Sun <samsun1006219@gmail.com>
Reported by: xingwei lee <xrivendell7@gmail.com>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://lore.kernel.org/linux-usb/CAEkJfYO6jRVC8Tfrd_R=cjO0hguhrV31fDPrLrNOOHocDkPoAA@mail.gmail.com/#r
Fixes: 310d2b4124 ("usb: interface authorization: SysFS part of USB interface authorization")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/1c37eea1-9f56-4534-b9d8-b443438dc869@rowland.harvard.edu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The PD of Type-C port needs to be updated in pd_set. Unlink the Type-C
port device to the old PD before linking it to a new one.
Fixes: cd099cde4e ("usb: typec: tcpm: Support multiple capabilities")
Cc: stable@vger.kernel.org
Signed-off-by: Kyle Tso <kyletso@google.com>
Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20240311172306.3911309-1-kyletso@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When unregister pd capabilitie in tcpm, KASAN will capture below double
-free issue. The root cause is the same capabilitiy will be kfreed twice,
the first time is kfreed by pd_capabilities_release() and the second time
is explicitly kfreed by tcpm_port_unregister_pd().
[ 3.988059] BUG: KASAN: double-free in tcpm_port_unregister_pd+0x1a4/0x3dc
[ 3.995001] Free of addr ffff0008164d3000 by task kworker/u16:0/10
[ 4.001206]
[ 4.002712] CPU: 2 PID: 10 Comm: kworker/u16:0 Not tainted 6.8.0-rc5-next-20240220-05616-g52728c567a55 #53
[ 4.012402] Hardware name: Freescale i.MX8QXP MEK (DT)
[ 4.017569] Workqueue: events_unbound deferred_probe_work_func
[ 4.023456] Call trace:
[ 4.025920] dump_backtrace+0x94/0xec
[ 4.029629] show_stack+0x18/0x24
[ 4.032974] dump_stack_lvl+0x78/0x90
[ 4.036675] print_report+0xfc/0x5c0
[ 4.040289] kasan_report_invalid_free+0xa0/0xc0
[ 4.044937] __kasan_slab_free+0x124/0x154
[ 4.049072] kfree+0xb4/0x1e8
[ 4.052069] tcpm_port_unregister_pd+0x1a4/0x3dc
[ 4.056725] tcpm_register_port+0x1dd0/0x2558
[ 4.061121] tcpci_register_port+0x420/0x71c
[ 4.065430] tcpci_probe+0x118/0x2e0
To fix the issue, this will remove kree() from tcpm_port_unregister_pd().
Fixes: cd099cde4e ("usb: typec: tcpm: Support multiple capabilities")
cc: stable@vger.kernel.org
Suggested-by: Aisheng Dong <aisheng.dong@nxp.com>
Signed-off-by: Xu Yang <xu.yang_2@nxp.com>
Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20240311065219.777037-1-xu.yang_2@nxp.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When auxiliary_device_add() returns error and then calls
auxiliary_device_uninit(), callback function ljca_auxdev_release
calls kfree(auxdev->dev.platform_data) to free the parameter data
of the function ljca_new_client_device. The callers of
ljca_new_client_device shouldn't call kfree() again
in the error handling path to free the platform data.
Fix this by cleaning up the redundant kfree() in all callers and
adding kfree() the passed in platform_data on errors which happen
before auxiliary_device_init() succeeds .
Fixes: acd6199f19 ("usb: Add support for Intel LJCA device")
Cc: stable <stable@kernel.org>
Signed-off-by: Yongzhi Liu <hyperlyzcs@gmail.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20240311125748.28198-1-hyperlyzcs@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When orientation switch is enabled in ucsi glink, there is a xhci
probe failure seen when booting up in host mode in reverse
orientation.
During bootup the following things happen in multiple drivers:
a) DWC3 controller driver initializes the core in device mode when the
dr_mode is set to DRD. It relies on role_switch call to change role to
host.
b) QMP driver initializes the lanes to TYPEC_ORIENTATION_NORMAL as a
normal routine. It relies on the typec_switch_set call to get notified
of orientation changes.
c) UCSI core reads the UCSI_GET_CONNECTOR_STATUS via the glink and
provides initial role switch to dwc3 controller.
When booting up in host mode with orientation TYPEC_ORIENTATION_REVERSE,
then we see the following things happening in order:
a) UCSI gives initial role as host to dwc3 controller ucsi_register_port.
Upon receiving this notification, the dwc3 core needs to program GCTL from
PRTCAP_DEVICE to PRTCAP_HOST and as part of this change, it asserts GCTL
Core soft reset and waits for it to be completed before shifting it to
host. Only after the reset is done will the dwc3_host_init be invoked and
xhci is probed. DWC3 controller expects that the usb phy's are stable
during this process i.e., the phy init is already done.
b) During the 100ms wait for GCTL core soft reset, the actual notification
from PPM is received by ucsi_glink via pmic glink for changing role to
host. The pmic_glink_ucsi_notify routine first sends the orientation
change to QMP and then sends role to dwc3 via ucsi framework. This is
happening exactly at the time GCTL core soft reset is being processed.
c) When QMP driver receives typec switch to TYPEC_ORIENTATION_REVERSE, it
then re-programs the phy at the instant GCTL core soft reset has been
asserted by dwc3 controller due to which the QMP PLL lock fails in
qmp_combo_usb_power_on.
d) After the 100ms of GCTL core soft reset is completed, the dwc3 core
goes for initializing the host mode and invokes xhci probe. But at this
point the QMP is non-responsive and as a result, the xhci plat probe fails
during xhci_reset.
Fix this by passing orientation switch to available ucsi instances if
their gpio configuration is available before ucsi_register is invoked so
that by the time, the pmic_glink_ucsi_notify provides typec_switch to QMP,
the lane is already configured and the call would be a NOP thus not racing
with role switch.
Cc: stable@vger.kernel.org
Fixes: c6165ed2f4 ("usb: ucsi: glink: use the connector orientation GPIO to provide switch events")
Suggested-by: Wesley Cheng <quic_wcheng@quicinc.com>
Signed-off-by: Krishna Kurapati <quic_kriskura@quicinc.com>
Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20240301040914.458492-1-quic_kriskura@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In tcpm_pd_set, the array of port source capabilities is port->src_pdo,
not port->snk_pdo.
Fixes: cd099cde4e ("usb: typec: tcpm: Support multiple capabilities")
Cc: stable@vger.kernel.org
Signed-off-by: Kyle Tso <kyletso@google.com>
Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://lore.kernel.org/r/20240311144500.3694849-1-kyletso@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use the `DEFINE_FLEX()` helper for an on-stack definition of a
flexible structure where the size of the flexible-array member
is known at compile-time, and refactor the rest of the code,
accordingly.
So, with these changes, fix the following warning:
fs/notify/fdinfo.c:45:36: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
Link: https://github.com/KSPP/linux/issues/202
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <ZgImguNzJBiis9Mj@neat>
