mmc: sdhci-iproc: Fix vmmc regulators (pre-bcm2711)

The Linux support for controlling card power via regulators appears to
be contentious. I would argue that the default behaviour is contrary to
the SDHCI spec - turning off the power writes a reserved value to the
SD Bus Voltage Select field of the Power Control Register, which
seems to kill the Arasan/iProc controller - but fortunately there is a
hook in sdhci_ops to override the behaviour.

Signed-off-by: Juerg Haefliger <juergh@canonical.com>
Signed-off-by: Phil Elwell <phil@raspberrypi.com>

Revert "mmc: sdhci-iproc: Fix vmmc regulators on iProc"

This reverts commit aed19399a01733dbad9be8bf026a4f7dd823b04f.

Commit 6c92ae1e452f ("mmc: sdhci: Introduce sdhci_set_power_and_bus_voltage()")
introduced a generic helper that does the same thing so use that instead in
the following commit.

Signed-off-by: Juerg Haefliger <juergh@canonical.com>

drm/vc4: hdmi: Fix bvb clock enable error checking

Check for errors only if we actually tried to enable the bvb clock.

Fixes: 01a6d727b407 ("vc4/drm: hdmi: Handle case when bvb clock is null")
Signed-off-by: Juerg Haefliger <juergh@canonical.com>

drm/vc4: Fix out of order frames during asynchronous page flips

When doing an asynchronous page flip (PAGE_FLIP ioctl with the
DRM_MODE_PAGE_FLIP_ASYNC flag set), the current code waits for the
possible GPU buffer being rendered through a call to
vc4_queue_seqno_cb().

On the BCM2835-37, the GPU driver is part of the vc4 driver and that
function is defined in vc4_gem.c to wait for the buffer to be rendered,
and once it's done, call a callback.

However, on the BCM2711 used on the RaspberryPi4, the GPU driver is
separate (v3d) and that function won't do anything. This was working
because we were going into a path, due to uninitialized variables, that
was always scheduling the callback.

However, we were never actually waiting for the buffer to be rendered
which was resulting in frames being displayed out of order.

The generic API to signal those kind of completion in the kernel are the
DMA fences, and fortunately the v3d drivers supports them and signal
when its job is done. That API also provides an equivalent function that
allows to have a callback being executed when the fence is signalled as
done.

Let's change our driver a bit to rely on the previous function for the
older SoCs, and on DMA fences for the BCM2711.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

Revert "firmware: Updated mailbox header"

This reverts commit 02038a75fb5ae9e0b8dac3de3f30f37ea613201e.

Revert "raspberrypi-firmware: Export the general transaction function."

This reverts commit 121592a1f507d2b5900db5f7c6510b532563dc09.

bcm2835_smi_dev: Fix handling of word-odd lengths

The read and write functions did not use the correct pointer offset
when dealing with an odd number of bytes after a DMA transfer. Also,
only handle the remaining odd bytes if the DMA transfer completed
successfully.

Submitted-by: @madimario (GitHub)
Signed-off-by: Phil Elwell <phil@raspberrypi.com>

regulator: rpi-panel: Remove get_brightness hook

The driver was implementing a get_brightness function that
tried to read back the PWM setting of the display to report
as the current brightness.
The controller on the display does not support that, therefore
we end up reporting a brightness of 0, and that confuses
systemd's backlight service.

Remove the hook so that the framework returns the current
brightness automatically.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

overlays: Add generic mcp2515 overlay

Can configure mcp2515 on spi0/1/2 without the need for multiple overlays.

overlays: add support for the MLX90640 thermal camera

This allows using the video-i2c camera driver with MLX90640 thermal
infrared sensors, connected to Raspberry Pi. CONFIG_VIDEO_V4L2_I2C
has to be selected to use the camera.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>

staging: bcm2835-codec: Add support for H&V Flips to ISP

The ISP can do H & V flips whilst resizing or converting
the image, so expose that via V4L2_CID_[H|V]FLIP.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Ensure all ctrls are set on streamon

Currently the code was only setting some controls from
bcm2835_codec_set_ctrls, but it's simpler to use
v4l2_ctrl_handler_setup to avoid forgetting to adding new
controls to the list.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835_codec: Correct flushing code for refcounting

Completions don't reference count, so setting the completion
on the first buffer returned and then not reinitialising it
means that the flush function doesn't behave as intended.

Signal the completion when the last buffer is returned.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: mmal-vchiq: Reset buffers_with_vpu on port_enable

Should we go through the timeout failure case with port_disable
not returning all buffers for whatever reason, the
buffers_with_vpu counter gets left at a non-zero value, which
will cause reference counting issues should the instance be
reused.

Reset the count when the port is enabled again, but before
any buffers have been sent to the VPU.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Queue flushed buffers instead of completing

When a buffer is returned on a port that is disabled, return it
to the videobuf2 QUEUED state instead of DONE which returns it
to the client.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/mmal-vchiq: Add module parameter to enable logging.

Adds a module parameter "debug" to enable various logging levels.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Signal the firmware to stop on all changes

The firmware defaults to not stopping video decode if only the
pixel aspect ratio or colourspace change. V4L2 requires us
to stop decoding on any change, therefore tell the firmware
of the desire for this alternate behaviour.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Format changed should trigger drain

When a format changed event occurs, the spec says that it
triggers an implicit drain, and that needs to be signalled
via -EPIPE.

For BCM2835, the format changed event happens at the point
the format change occurs, so no further buffers exist from
before the resolution changed point. We therefore signal the
last buffer immediately.
We don't have a V4L2 available to us at this point, so set
the videobuf2 queue last_buffer_dequeued flag directly.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/mmal-vchiq: Rationalise included headers

The list of includes was slightly over generic, and wasn't
in alphabetical order. Clean it up.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835-codec: Do not send buffers to the VPU unless streaming

With video decode we now enable both input and output ports on
the component. This means that buffers will get passed to the VPU
earlier than desired if they are queued befoer STREAMON.

Check that the queue is streaming before sending buffers to the VPU.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Allow decode res changed before STREAMON(CAPTURE)

The V4L2 stateful video decoder API requires that you can STREAMON
on only the OUTPUT queue, feed in buffers, and wait for the
SOURCE_CHANGE event.
This requires that we enable the MMAL output port at the same time
as the input port, because the output port is the one that creates
the SOURCE_CHANGED event.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

media: rpivid: Ensure IRQs have completed before uniniting context

Before uniniting the decode context sync with the IRQ queues to ensure
that decode no longer has any buffers in use.  This fixes a problem that
manifested as ffmpeg leaking CMA buffers when it did a stream off on
OUTPUT before CAPTURE, though in reality it was probably much more
dangerous than that.

Signed-off-by: John Cox <jc@kynesim.co.uk>

media: rpivid: Remove unused ctx state variable and defines

Remove unused ctx state tracking variable and associated defines.
Their presence implies they might be used, but they aren't.

Signed-off-by: John Cox <jc@kynesim.co.uk>

net: phy: lan87xx: Allow more time for link detect

With EDPWRDOWN set in idle, it must be cleared before checking for
ENERGYON going high, indicating that a link is being established.
The existing code allows 640ms for ENERGYON to go high, but on
Raspberry Pis that appears not to be enough, causing link detection
to fail.

Increase the polling timeout to 1500ms - with a polling interval of
10ms it shouldn't cause unnecessary delays.

See: https://github.com/raspberrypi/linux/issues/4393

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

media: rpivid: Avoid returning EINVAL to a G_FMT ioctl

V4L2 spec says that G/S/TRY_FMT IOCTLs should never return errors for
anything other than wrong buffer types. Improve the capture format
function such that this is so and unsupported values get converted
to supported ones properly.

Signed-off-by: John Cox <jc@kynesim.co.uk>

ARM: proc-v7: Retry uncached stmia if necessary

A failure of some CPU cores to come online has been traced to the
failure of a stm instruction while the cache is disabled. The symptom
is that the saved values read back as zeroes, a catastrophic error since
one of the values is a return address.

This patch forces a readback and retry until the correct value is
returned,

Notes:

At this stage in the boot process the core is running with its cache
disabled. Before enabling the cache its contents must be explicitly
invalidated, a process that requires quite a few registers that the
caller must preserve. Evidence suggests that something is writing a
block of zeroes over that space at a time when all other cores should
be idle, possibly some kind of write-combiner, and retrying is an
attempt to avoid the problem.

The previous attempted fix (forcing the accesses to only be 4-byte
aligned) appears to have only worked for a while and likely for less
obvious reasons such as a change in code alignment.

See: https://github.com/Hexxeh/rpi-firmware/issues/232

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

Revert "ARM: proc-v7: Force misalignment of early stmia"

This reverts commit fe4cc0ed5923d3b91d4cb25f8f6cbbf5a6f8c06f.

The speculative patch that this commit reverts is proving to not be
effective any more, so revert it and try a new approach.

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

configs: Add CAN_PEAK_USB=m

Add the driver for the PEAK Systems USB CAN interface.

See: https://github.com/raspberrypi/linux/issues/4583

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

drm/amdgpu: Add missing conversion of is_hdr_metadata_different for amdgpu

staging/bcm2835-v4l2_codec: Fix for encode selection API

Matches correct behaviour from DECODE and DEINTERLACE

Signed-off-by: Dom Cobley <popcornmix@gmail.com>

staging/bcm2835_codec: Add support for image_fx to deinterlace

Adds another /dev/video node wrapping image_fx doing deinterlace.

Co-developed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Dom Cobley <popcornmix@gmail.com>

staging/vchiq-mmal: Add the deinterlace image effects enums

As we're wanting to wrap the image_fx component for deinterlacing,
add the deinterlace algorithm values to enum mmal_parameter_imagefx

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging: bcm2835-codec: Allow custom specified strides/bytesperline.

If the client provides a bytesperline value in try_fmt/s_fmt then
validate it and correct if necessary.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835_codec: Log MMAL flags in hex

The flags is a bitmask, so it's far easier to interpret as hex
data instead of decimal.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835-codec: Return buffers to QUEUED not ERROR state

Should start_streaming fail, or buffers be queued during
stop_streaming, they should be returned to the core as QUEUED
and not (as currently) as ERROR.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835-codec: Correct ENUM_FRAMESIZES stepsize to 2

Being YUV420 formats, the step size is always 2 to avoid part
chroma subsampling.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835-codec: Add support for decoding interlaced streams

The video decoder can support decoding interlaced streams, so add
the required plumbing to signal this correctly.

The encoder and ISP do NOT support interlaced data, so trying to
configure an interlaced format on those nodes will be rejected.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/vchiq-mmal: Add parameters for interlaced video support

Adds enum mmal_interlace_type and struct
mmal_parameter_video_interlace_type to allow for querying the
interlacing mode on decoders.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/vchiq-mmal: Add buffer flags for interlaced video

Add the buffer flags that the firmware uses to identify fields
on interlaced video

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

staging/bcm2835-codec: Change the default codec res to 32x32

In order to effectively guarantee that a V4L2_EVENT_SOURCE_CHANGE
event occurs, adopt a default resolution of 32x32 so that it
is incredibly unlikely to be decoding a stream of that resolution
and therefore failing to note a "change" requiring the event.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/vc4: hdmi: Simplify exit path of vc4_hdmi_encoder_pre_crtc_configure

The exit path of vc4_hdmi_encoder_pre_crtc_configure() is fairly hard to
maintain given its numerous error conditions.

Switch to a goto based approach to simplify it.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

drm/vc4: hdmi: Remove unnecessary pm_runtime_put

Unlike pm_runtime_get_sync(), pm_runtime_resume_and_get() doesn't take a
reference on failure, so we don't need to call pm_runtime_put() on
failure.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

drm/vc4: hdmi: Remove redundant HSM clk enable/disable in detect

Our detect function calls pm_runtime_resume_and_get() and
pm_runtime_put() to make sure the device is properly powered before
trying to access the controller.

However, it also makes sure the HSM clock is properly enabled (and
disabled), which is redundant with what runtime_pm is doing already.
Let's just remove it.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

dtoverlays: Update 7inch DSI display overlay to use newer drivers

The older panel-raspberrypi-touchscreen driver had issues in
that it also controlled the power for the touchscreen without
having an appropriate hook for the touchscreen driver to control
that.

Mainline has now added a Toshiba TC358762 bridge driver, and
a regulator/backlight driver for the ATTiny microcontroller on
the board. That allows clean integration with the touchscreen
driver.

Switch the overlays over to using newer drivers.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

defconfig: Add DRM_TOSHIBA_TC358762 and REGULATOR_RASPBERRYPI_TOUCHSCREEN_ATTINY

Add the two newer drivers for the Pi 7" touchscreen to the defconfigs

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Add GPIO control for panel and touch resets

We need independent control of the resets for the panel&bridge,
vs the touch controller.

Expose the reset lines that are on the Atmel's port C via the GPIO
API so that they can be controlled appropriately.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Convert to drive lines directly

The Atmel was doing a load of automatic sequencing of
control lines, however it was combining the touch controller's
reset with the bridge/panel control.

Change to control the control signals directly rather than
through the automatic POWERON control.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Ensure the backlight is off during probe.

The initial state of the Atmel is not defined, so ensure the
backlight PWM is set to 0 by default.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Serialise operations.

The driver was using the regmap lock to serialise the
individual accesses, but we really need to protect the
timings of enabling the regulators, including any communication
with the Atmel.

Use a mutex within the driver to control overall accesses to
the Atmel, instead of the regmap lock.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Handle I2C errors/timing to the Atmel

The Atmel is doing some things in the I2C ISR, during which
period it will not respond to further commands. This is
particularly true of the POWERON command.

Increase delays appropriately, and retry should I2C errors be
reported.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

regulator: rpi-panel: Register with a unique backlight name

There's no reason why 2 Raspberry Pi DSI displays can't be
attached to a Pi Compute Module, so the backlight names need to
be unique.

Use the parent dev_name. It's not as readable, but is unique.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

Input: edt-ft54x6: Clean up timer and workqueue on remove

If no interrupt is defined then a timer and workqueue are used
to poll the controller.
On remove these were not being cleaned up correctly.

Fixes: ca61fdaba79f "Input: edt-ft5x06: Poll the device if no interrupt is
configured."

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/panel-simple: Add a timing for the Raspberry Pi 7" panel

The Raspberry Pi 7" 800x480 panel uses a Toshiba TC358762 DSI
to DPI bridge chip, so there is a requirement for the timings
to be specified for the end panel. Add such a definition.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/panel/raspberrypi-touchscreen: Handle I2C errors.

rpi_touchscreen_i2c_read returns any errors from i2c_transfer,
or the 8 bit received value.
Check for error values before trying to process the data as
valid.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/panel/raspberrypi-touchscreen: Initialise the bridge in prepare

The panel has a prepare call which is before video starts, and an
enable call which is after.
The Toshiba bridge should be configured before video, so move
the relevant power and initialisation calls to prepare.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/panel/raspberrypi-touchscreen: Avoid NULL deref if not initialised

If a call to rpi_touchscreen_i2c_write from rpi_touchscreen_probe
fails before mipi_dsi_device_register_full is called, then
in trying to log the error message if uses ts->dsi->dev when
it is still NULL.

Use ts->i2c->dev instead, which is initialised earlier in probe.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/vc4: Correct DSI divider calculations

The divider calculations tried to find the divider
just faster than the clock requested. However if
it required a divider of 7 then the for loop
aborted without handling the "error" case, and could
end up with a clock lower than requested.

Correct the loop so that we always have a clock greater
than requested.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/vc4: Release workaround buffer and DMA in error paths and unbind

On Pi0-3 the driver allocates a buffer and requests a DMA channel
because the ARM can't write to DSI1's registers directly.
However unbind and the error paths in bind don't release the buffer or
the DMA channel.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

drm/vc4: Reset HDMI MISC_CONTROL register.

The HDMI block can repeat pixels for double clocked modes,
and the firmware is now configuring the block to do this as
the PV is doing it incorrectly when at 2pixels/clock.
If the kernel doesn't reset it then we end up with strange
modes.

Reset MISC_CONTROL.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>

gpio-fsm: Clamp the delay time to zero

The sysfs delay_ms value is calculated live, and it is possible for
the time left to appear to be negative briefly if the timer handling
hasn't completed. Ensure the displayed value never goes below zero,
for the sake of appearances.

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

media: i2c: imx477: Allow control of on-sensor DPC

A module parameter "dpc_enable" is added to allow the control of the
sensor's on-board DPC (Defective Pixel Correction) function.

This is a global setting to be configured before using the sensor;
there is no intention that this would ever be changed on-the-fly.

Signed-off-by: David Plowman <david.plowman@raspberrypi.com>

overlays: README: Refer to WLAN where possible

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

Merge remote-tracking branch 'stable/linux-5.10.y' into rpi-5.10.y

drm/probe: Fix drm_connector_helper_hpd_irq_event deadlock

drm_connector_helper_hpd_irq_event() calls
drm_kms_helper_hotplug_event() with the mode-setting lock taken while
it's supposed to be called without that lock taken.

This results in a lockdep warning, and a deadlock if we were to wake up
a TV through CEC (and possibly other cases).

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

Linux 5.10.63

Link: https://lore.kernel.org/r/20210906125449.756437409@linuxfoundation.org
Tested-by: Pavel Machek (CIP) <pavel@denx.de>
Tested-by: Fox Chen <foxhlchen@gmail.com>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Hulk Robot <hulkrobot@huawei.com>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>

media: stkwebcam: fix memory leak in stk_camera_probe

commit 514e97674400462cc09c459a1ddfb9bf39017223 upstream.

My local syzbot instance hit memory leak in usb_set_configuration().
The problem was in unputted usb interface. In case of errors after
usb_get_intf() the reference should be putted to correclty free memory
allocated for this interface.

Fixes: ec16dae5453e ("V4L/DVB (7019): V4L: add support for Syntek DC1125 webcams")
Cc: stable@vger.kernel.org
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fuse: fix illegal access to inode with reused nodeid

commit 15db16837a35d8007cb8563358787412213db25e upstream.

Server responds to LOOKUP and other ops (READDIRPLUS/CREATE/MKNOD/...)
with ourarg containing nodeid and generation.

If a fuse inode is found in inode cache with the same nodeid but different
generation, the existing fuse inode should be unhashed and marked "bad" and
a new inode with the new generation should be hashed instead.

This can happen, for example, with passhrough fuse filesystem that returns
the real filesystem ino/generation on lookup and where real inode numbers
can get recycled due to real files being unlinked not via the fuse
passthrough filesystem.

With current code, this situation will not be detected and an old fuse
dentry that used to point to an older generation real inode, can be used to
access a completely new inode, which should be accessed only via the new
dentry.

Note that because the FORGET message carries the nodeid w/o generation, the
server should wait to get FORGET counts for the nlookup counts of the old
and reused inodes combined, before it can free the resources associated to
that nodeid.

Stable backport notes:
* This is not a regression. The bug has been in fuse forever, but only
  a certain class of low level fuse filesystems can trigger this bug
* Because there is no way to check if this fix is applied in runtime,
  libfuse test_examples.py tests this fix with hardcoded check for
  kernel version >= 5.14
* After backport to stable kernel(s), the libfuse test can be updated
  to also check minimal stable kernel version(s)
* Depends on "fuse: fix bad inode" which is already applied to stable
  kernels v5.4.y and v5.10.y
* Required backporting helper inode_wrong_type()

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/linux-fsdevel/CAOQ4uxi8DymG=JO_sAU+wS8akFdzh+PuXwW3Ebgahd2Nwnh7zA@mail.gmail.com/
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

new helper: inode_wrong_type()

commit 6e3e2c4362e41a2f18e3f7a5ad81bd2f49a47b85 upstream.

inode_wrong_type(inode, mode) returns true if setting inode->i_mode
to given value would've changed the inode type.  We have enough of
those checks open-coded to make a helper worthwhile.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

spi: Switch to signed types for *_native_cs SPI controller fields

commit 35f3f8504c3b60a1ae5576e178b27fc0ddd6157d upstream.

While fixing undefined behaviour the commit f60d7270c8a3 ("spi: Avoid
undefined behaviour when counting unused native CSs") missed the case
when all CSs are GPIOs and thus unused_native_cs will be evaluated to
-1 in unsigned representation. This will falsely trigger a condition
in the spi_get_gpio_descs().

Switch to signed types for *_native_cs SPI controller fields to fix above.

Fixes: f60d7270c8a3 ("spi: Avoid undefined behaviour when counting unused native CSs")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20210510131242.49455-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Nobuhiro Iwamatsu (CIP) <nobuhiro1.iwamatsu@toshiba.co.jp>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

serial: 8250: 8250_omap: Fix possible array out of bounds access

commit d4548b14dd7e5c698f81ce23ce7b69a896373b45 upstream.

k3_soc_devices array is missing a sentinel entry which may result in out
of bounds access as reported by kernel KASAN.

Fix this by adding a sentinel entry.

Fixes: 439c7183e5b9 ("serial: 8250: 8250_omap: Disable RX interrupt after DMA enable")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Link: https://lore.kernel.org/r/20201111112653.2710-1-vigneshr@ti.com
Signed-off-by: Nobuhiro Iwamatsu (CIP) <nobuhiro1.iwamatsu@toshiba.co.jp>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: pcm: fix divide error in snd_pcm_lib_ioctl

commit f3eef46f0518a2b32ca1244015820c35a22cfe4a upstream.

Syzkaller reported a divide error in snd_pcm_lib_ioctl. fifo_size
is of type snd_pcm_uframes_t(unsigned long). If frame_size
is 0x100000000, the error occurs.

Fixes: a9960e6a293e ("ALSA: pcm: fix fifo_size frame calculation")
Signed-off-by: Zubin Mithra <zsm@chromium.org>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210827153735.789452-1-zsm@chromium.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek: Workaround for conflicting SSID on ASUS ROG Strix G17

commit 13d9c6b998aaa76fd098133277a28a21f2cc2264 upstream.

ASUS ROG Strix G17 has the very same PCI and codec SSID (1043:103f) as
ASUS TX300, and unfortunately, the existing quirk for TX300 is broken
on ASUS ROG.  Actually the device works without the quirk, so we'll
need to clear the quirk before applying for this device.
Since ASUS ROG has a different codec (ALC294 - while TX300 has
ALC282), this patch adds a workaround for the device, just clearing
the codec->fixup_id by checking the codec vendor_id.

It's a bit ugly to add such a workaround there, but it seems to be the
simplest way.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=214101
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210820143214.3654-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: hda/realtek: Quirk for HP Spectre x360 14 amp setup

commit 93ab3eafb0b3551c54175cb38afed3b82356a047 upstream.

This patch extends support for the HP Spectre x360 14
amp enable quirk to support a model of the device with
an additional subdevice ID.

Signed-off-by: Johnathon Clark <john.clark@cantab.net>
Link: https://lore.kernel.org/r/20210823162110.8870-1-john.clark@cantab.net
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cryptoloop: add a deprecation warning

[ Upstream commit 222013f9ac30b9cec44301daa8dbd0aae38abffb ]

Support for cryptoloop has been officially marked broken and deprecated
in favor of dm-crypt (which supports the same broken algorithms if
needed) in Linux 2.6.4 (released in March 2004), and support for it has
been entirely removed from losetup in util-linux 2.23 (released in April
2013).  Add a warning and a deprecation schedule.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20210827163250.255325-1-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/x86/amd/power: Assign pmu.module

[ Upstream commit ccf26483416a339c114409f6e7cd02abdeaf8052 ]

Assign pmu.module so the driver can't be unloaded whilst in use.

Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20210817221048.88063-4-kim.phillips@amd.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/x86/amd/ibs: Work around erratum #1197

[ Upstream commit 26db2e0c51fe83e1dd852c1321407835b481806e ]

Erratum #1197 "IBS (Instruction Based Sampling) Register State May be
Incorrect After Restore From CC6" is published in a document:

  "Revision Guide for AMD Family 19h Models 00h-0Fh Processors" 56683 Rev. 1.04 July 2021

  https://bugzilla.kernel.org/show_bug.cgi?id=206537

Implement the erratum's suggested workaround and ignore IBS samples if
MSRC001_1031 == 0.

Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20210817221048.88063-3-kim.phillips@amd.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

ceph: fix possible null-pointer dereference in ceph_mdsmap_decode()

[ Upstream commit a9e6ffbc5b7324b6639ee89028908b1e91ceed51 ]

kcalloc() is called to allocate memory for m->m_info, and if it fails,
ceph_mdsmap_destroy() behind the label out_err will be called:
  ceph_mdsmap_destroy(m);

In ceph_mdsmap_destroy(), m->m_info is dereferenced through:
  kfree(m->m_info[i].export_targets);

To fix this possible null-pointer dereference, check m->m_info before the
for loop to free m->m_info[i].export_targets.

[ jlayton: fix up whitespace damage
   only kfree(m->m_info) if it's non-NULL ]

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Tuo Li <islituo@gmail.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/x86/intel/pt: Fix mask of num_address_ranges

[ Upstream commit c53c6b7409f4cd9e542991b53d597fbe2751d7db ]

Per SDM, bit 2:0 of CPUID(0x14,1).EAX[2:0] reports the number of
configurable address ranges for filtering, not bit 1:0.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://lkml.kernel.org/r/20210824040622.4081502-1-xiaoyao.li@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

qede: Fix memset corruption

[ Upstream commit e543468869e2532f5d7926e8f417782b48eca3dc ]

Thanks to Kees Cook who detected the problem of memset that starting
from not the first member, but sized for the whole struct.
The better change will be to remove the redundant memset and to clear
only the msix_cnt member.

Signed-off-by: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Reported-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: macb: Add a NULL check on desc_ptp

[ Upstream commit 85520079afce885b80647fbd0d13d8f03d057167 ]

macb_ptp_desc will not return NULL under most circumstances with correct
Kconfig and IP design config register. But for the sake of the extreme
corner case, check for NULL when using the helper. In case of rx_tstamp,
no action is necessary except to return (similar to timestamp disabled)
and warn. In case of TX, return -EINVAL to let the skb be free. Perform
this check before marking skb in progress.
Fixes coverity warning:
(4) Event dereference:
Dereferencing a null pointer "desc_ptp"

Signed-off-by: Harini Katakam <harini.katakam@xilinx.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

qed: Fix the VF msix vectors flow

[ Upstream commit b0cd08537db8d2fbb227cdb2e5835209db295a24 ]

For VFs we should return with an error in case we didn't get the exact
number of msix vectors as we requested.
Not doing that will lead to a crash when starting queues for this VF.

Signed-off-by: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

reset: reset-zynqmp: Fixed the argument data type

[ Upstream commit ed104ca4bd9c405b41e968ad4ece51f6462e90b6 ]

This patch changes the data type of the variable 'val' from
int to u32.

Addresses-Coverity: argument of type "int *" is incompatible with parameter of type "u32 *"
Signed-off-by: Sai Krishna Potthuri <lakshmi.sai.krishna.potthuri@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Link: https://lore.kernel.org/r/925cebbe4eb73c7d0a536da204748d33c7100d8c.1624448778.git.michal.simek@xilinx.com
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpu: ipu-v3: Fix i.MX IPU-v3 offset calculations for (semi)planar U/V formats

[ Upstream commit 7cca7c8096e2c8a4149405438329b5035d0744f0 ]

Video captured in 1400x1050 resolution (bytesperline aka stride = 1408
bytes) is invalid. Fix it.

Signed-off-by: Krzysztof Halasa <khalasa@piap.pl>
Link: https://lore.kernel.org/r/m3y2bmq7a4.fsf@t19.piap.pl
[p.zabel@pengutronix.de: added "gpu: ipu-v3:" prefix to commit description]
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: OMAP1: ams-delta: remove unused function ams_delta_camera_power

commit bae989c4bc53f861cc1b706aab0194703e9907a8 upstream.

The ams_delta_camera_power() function is unused as reports
Clang compilation with omap1_defconfig on linux-next:

arch/arm/mach-omap1/board-ams-delta.c:462:12: warning: unused function 'ams_delta_camera_power' [-Wunused-function]
static int ams_delta_camera_power(struct device *dev, int power)
           ^
1 warning generated.

The soc_camera support was dropped without removing
ams_delta_camera_power() function, making it unused.

Fixes: ce548396a433 ("media: mach-omap1: board-ams-delta.c: remove soc_camera dependencies")
Signed-off-by: Maciej Falkowski <maciej.falkowski9@gmail.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/1326
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xtensa: fix kconfig unmet dependency warning for HAVE_FUTEX_CMPXCHG

commit ed5aacc81cd41efc4d561e14af408d1003f7b855 upstream.

XTENSA should only select HAVE_FUTEX_CMPXCHG when FUTEX is
set/enabled. This prevents a kconfig warning.

WARNING: unmet direct dependencies detected for HAVE_FUTEX_CMPXCHG
  Depends on [n]: FUTEX [=n]
  Selected by [y]:
  - XTENSA [=y] && !MMU [=n]

Fixes: d951ba21b959 ("xtensa: nommu: select HAVE_FUTEX_CMPXCHG")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: linux-xtensa@linux-xtensa.org
Message-Id: <20210526070337.28130-1-rdunlap@infradead.org>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

static_call: Fix unused variable warn w/o MODULE

commit 7d95f22798ecea513f37b792b39fec4bcf20fec3 upstream.

Here is the warning converted as error and reported by GCC:

  kernel/static_call.c: In function ‘__static_call_update’:
  kernel/static_call.c:153:18: error: unused variable ‘mod’ [-Werror=unused-variable]
    153 |   struct module *mod = site_mod->mod;
        |                  ^~~
  cc1: all warnings being treated as errors
  make[1]: *** [scripts/Makefile.build:271: kernel/static_call.o] Error 1

This is simply because since recently, we no longer use 'mod' variable
elsewhere if MODULE is unset.

When using 'make tinyconfig' to generate the default kconfig, MODULE is
unset.

There are different ways to fix this warning. Here I tried to minimised
the number of modified lines and not add more #ifdef. We could also move
the declaration of the 'mod' variable inside the if-statement or
directly use site_mod->mod.

Fixes: 698bacefe993 ("static_call: Align static_call_is_init() patching condition")
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210326105023.2058860-1-matthieu.baerts@tessares.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "Add a reference to ucounts for each cred"

This reverts commit b2c4d9a33cc2dec7466f97eba2c4dd571ad798a5 which is
commit 905ae01c4ae2ae3df05bb141801b1db4b7d83c61 upstream.

This commit should not have been applied to the 5.10.y stable tree, so
revert it.

Reported-by: "Eric W. Biederman" <ebiederm@xmission.com>
Link: https://lore.kernel.org/r/87v93k4bl6.fsf@disp2133
Cc: Alexey Gladkov <legion@kernel.org>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "cred: add missing return error code when set_cred_ucounts() failed"

This reverts commit 0855952ed4f1a6861fbb0e5d684efd447d7347c9 which is
commit 5e6b8a50a7cec5686ee2c4bda1d49899c79a7eae upstream.

The "original" commit 905ae01c4ae2 ("Add a reference to ucounts for each
cred"), should not have been applied to the 5.10.y tree, so revert it,
and the follow-on fixup patches as well.

Reported-by: "Eric W. Biederman" <ebiederm@xmission.com>
Link: https://lore.kernel.org/r/87v93k4bl6.fsf@disp2133
Cc: Yang Yingliang <yangyingliang@huawei.com>
Cc: Alexey Gladkov <legion@kernel.org>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "ucounts: Increase ucounts reference counter before the security hook"

This reverts commit b493af3a66e067f93e5e03465507866ddeabff9e which is
commit bbb6d0f3e1feb43d663af089c7dedb23be6a04fb upstream.

The "original" commit 905ae01c4ae2 ("Add a reference to ucounts for each
cred"), should not have been applied to the 5.10.y tree, so revert it,
and the follow-on fixup patches as well.

Reported-by: "Eric W. Biederman" <ebiederm@xmission.com>
Link: https://lore.kernel.org/r/87v93k4bl6.fsf@disp2133
Cc: Alexey Gladkov <legion@kernel.org>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ubifs: report correct st_size for encrypted symlinks

commit 064c734986011390b4d111f1a99372b7f26c3850 upstream.

The stat() family of syscalls report the wrong size for encrypted
symlinks, which has caused breakage in several userspace programs.

Fix this by calling fscrypt_symlink_getattr() after ubifs_getattr() for
encrypted symlinks.  This function computes the correct size by reading
and decrypting the symlink target (if it's not already cached).

For more details, see the commit which added fscrypt_symlink_getattr().

Fixes: ca7f85be8d6c ("ubifs: Add support for encrypted symlinks")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210702065350.209646-5-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

f2fs: report correct st_size for encrypted symlinks

commit 461b43a8f92e68e96c4424b31e15f2b35f1bbfa9 upstream.

The stat() family of syscalls report the wrong size for encrypted
symlinks, which has caused breakage in several userspace programs.

Fix this by calling fscrypt_symlink_getattr() after f2fs_getattr() for
encrypted symlinks.  This function computes the correct size by reading
and decrypting the symlink target (if it's not already cached).

For more details, see the commit which added fscrypt_symlink_getattr().

Fixes: cbaf042a3cc6 ("f2fs crypto: add symlink encryption")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210702065350.209646-4-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: report correct st_size for encrypted symlinks

commit 8c4bca10ceafc43b1ca0a9fab5fa27e13cbce99e upstream.

The stat() family of syscalls report the wrong size for encrypted
symlinks, which has caused breakage in several userspace programs.

Fix this by calling fscrypt_symlink_getattr() after ext4_getattr() for
encrypted symlinks.  This function computes the correct size by reading
and decrypting the symlink target (if it's not already cached).

For more details, see the commit which added fscrypt_symlink_getattr().

Fixes: f348c252320b ("ext4 crypto: add symlink encryption")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210702065350.209646-3-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fscrypt: add fscrypt_symlink_getattr() for computing st_size

commit d18760560593e5af921f51a8c9b64b6109d634c2 upstream.

Add a helper function fscrypt_symlink_getattr() which will be called
from the various filesystems' ->getattr() methods to read and decrypt
the target of encrypted symlinks in order to report the correct st_size.

Detailed explanation:

As required by POSIX and as documented in various man pages, st_size for
a symlink is supposed to be the length of the symlink target.
Unfortunately, st_size has always been wrong for encrypted symlinks
because st_size is populated from i_size from disk, which intentionally
contains the length of the encrypted symlink target.  That's slightly
greater than the length of the decrypted symlink target (which is the
symlink target that userspace usually sees), and usually won't match the
length of the no-key encoded symlink target either.

This hadn't been fixed yet because reporting the correct st_size would
require reading the symlink target from disk and decrypting or encoding
it, which historically has been considered too heavyweight to do in
->getattr().  Also historically, the wrong st_size had only broken a
test (LTP lstat03) and there were no known complaints from real users.
(This is probably because the st_size of symlinks isn't used too often,
and when it is, typically it's for a hint for what buffer size to pass
to readlink() -- which a slightly-too-large size still works for.)

However, a couple things have changed now.  First, there have recently
been complaints about the current behavior from real users:

- Breakage in rpmbuild:
  https://github.com/rpm-software-management/rpm/issues/1682
  https://github.com/google/fscrypt/issues/305

- Breakage in toybox cpio:
  https://www.mail-archive.com/toybox@lists.landley.net/msg07193.html

- Breakage in libgit2: https://issuetracker.google.com/issues/189629152
  (on Android public issue tracker, requires login)

Second, we now cache decrypted symlink targets in ->i_link.  Therefore,
taking the performance hit of reading and decrypting the symlink target
in ->getattr() wouldn't be as big a deal as it used to be, since usually
it will just save having to do the same thing later.

Also note that eCryptfs ended up having to read and decrypt symlink
targets in ->getattr() as well, to fix this same issue; see
commit 3a60a1686f0d ("eCryptfs: Decrypt symlink target for stat size").

So, let's just bite the bullet, and read and decrypt the symlink target
in ->getattr() in order to report the correct st_size.  Add a function
fscrypt_symlink_getattr() which the filesystems will call to do this.

(Alternatively, we could store the decrypted size of symlinks on-disk.
But there isn't a great place to do so, and encryption is meant to hide
the original size to some extent; that property would be lost.)

Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210702065350.209646-2-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: fix race writing to an inline_data file while its xattrs are changing

commit a54c4613dac1500b40e4ab55199f7c51f028e848 upstream.

The location of the system.data extended attribute can change whenever
xattr_sem is not taken.  So we need to recalculate the i_inline_off
field since it mgiht have changed between ext4_write_begin() and
ext4_write_end().

This means that caching i_inline_off is probably not helpful, so in
the long run we should probably get rid of it and shrink the in-memory
ext4 inode slightly, but let's fix the race the simple way for now.

Cc: stable@kernel.org
Fixes: f19d5870cbf72 ("ext4: add normal write support for inline data")
Reported-by: syzbot+13146364637c7363a7de@syzkaller.appspotmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "USB: serial: ch341: fix character loss at high transfer rates"

commit df7b16d1c00ecb3da3a30c999cdb39f273c99a2f upstream.

This reverts commit 3c18e9baee0ef97510dcda78c82285f52626764b.

These devices do not appear to send a zero-length packet when the
transfer size is a multiple of the bulk-endpoint max-packet size. This
means that incoming data may not be processed by the driver until a
short packet is received or the receive buffer is full.

Revert back to using endpoint-sized receive buffers to avoid stalled
reads.

Reported-by: Paul Größel <pb.g@gmx.de>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=214131
Fixes: 3c18e9baee0e ("USB: serial: ch341: fix character loss at high transfer rates")
Cc: stable@vger.kernel.org
Cc: Willy Tarreau <w@1wt.eu>
Link: https://lore.kernel.org/r/20210824121926.19311-1-johan@kernel.org
Signed-off-by: Johan Hovold <johan@kernel.org>

configs: Enable the Toneport UX1 Line 6 audio i/f

See: https://github.com/raspberrypi/linux/issues/4562

Signed-off-by: Phil Elwell <phil@raspberrypi.com>

overlays: Add overlay for QCA7000 via UART0

This adds an overlay to connect the QCA7000 in UART mode via UART0.
The qcauart driver uses the serial device bus instead of deprecated
line disciplines.

Signed-off-by: Stefan Wahren <stefan.wahren@in-tech.com>

Linux 5.10.62

Link: https://lore.kernel.org/r/20210901122300.503008474@linuxfoundation.org
Tested-by: Fox Chen <foxhlchen@gmail.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Pavel Machek (CIP) <pavel@denx.de>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Hulk Robot <hulkrobot@huawei.com>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bpf: Fix potentially incorrect results with bpf_get_local_storage()

commit a2baf4e8bb0f306fbed7b5e6197c02896a638ab5 upstream.

Commit b910eaaaa4b8 ("bpf: Fix NULL pointer dereference in bpf_get_local_storage()
helper") fixed a bug for bpf_get_local_storage() helper so different tasks
won't mess up with each other's percpu local storage.

The percpu data contains 8 slots so it can hold up to 8 contexts (same or
different tasks), for 8 different program runs, at the same time. This in
general is sufficient. But our internal testing showed the following warning
multiple times:

  [...]
  warning: WARNING: CPU: 13 PID: 41661 at include/linux/bpf-cgroup.h:193
     __cgroup_bpf_run_filter_sock_ops+0x13e/0x180
  RIP: 0010:__cgroup_bpf_run_filter_sock_ops+0x13e/0x180
  <IRQ>
   tcp_call_bpf.constprop.99+0x93/0xc0
   tcp_conn_request+0x41e/0xa50
   ? tcp_rcv_state_process+0x203/0xe00
   tcp_rcv_state_process+0x203/0xe00
   ? sk_filter_trim_cap+0xbc/0x210
   ? tcp_v6_inbound_md5_hash.constprop.41+0x44/0x160
   tcp_v6_do_rcv+0x181/0x3e0
   tcp_v6_rcv+0xc65/0xcb0
   ip6_protocol_deliver_rcu+0xbd/0x450
   ip6_input_finish+0x11/0x20
   ip6_input+0xb5/0xc0
   ip6_sublist_rcv_finish+0x37/0x50
   ip6_sublist_rcv+0x1dc/0x270
   ipv6_list_rcv+0x113/0x140
   __netif_receive_skb_list_core+0x1a0/0x210
   netif_receive_skb_list_internal+0x186/0x2a0
   gro_normal_list.part.170+0x19/0x40
   napi_complete_done+0x65/0x150
   mlx5e_napi_poll+0x1ae/0x680
   __napi_poll+0x25/0x120
   net_rx_action+0x11e/0x280
   __do_softirq+0xbb/0x271
   irq_exit_rcu+0x97/0xa0
   common_interrupt+0x7f/0xa0
   </IRQ>
   asm_common_interrupt+0x1e/0x40
  RIP: 0010:bpf_prog_1835a9241238291a_tw_egress+0x5/0xbac
   ? __cgroup_bpf_run_filter_skb+0x378/0x4e0
   ? do_softirq+0x34/0x70
   ? ip6_finish_output2+0x266/0x590
   ? ip6_finish_output+0x66/0xa0
   ? ip6_output+0x6c/0x130
   ? ip6_xmit+0x279/0x550
   ? ip6_dst_check+0x61/0xd0
  [...]

Using drgn [0] to dump the percpu buffer contents showed that on this CPU
slot 0 is still available, but slots 1-7 are occupied and those tasks in
slots 1-7 mostly don't exist any more. So we might have issues in
bpf_cgroup_storage_unset().

Further debugging confirmed that there is a bug in bpf_cgroup_storage_unset().
Currently, it tries to unset "current" slot with searching from the start.
So the following sequence is possible:

  1. A task is running and claims slot 0
  2. Running BPF program is done, and it checked slot 0 has the "task"
     and ready to reset it to NULL (not yet).
  3. An interrupt happens, another BPF program runs and it claims slot 1
     with the *same* task.
  4. The unset() in interrupt context releases slot 0 since it matches "task".
  5. Interrupt is done, the task in process context reset slot 0.

At the end, slot 1 is not reset and the same process can continue to occupy
slots 2-7 and finally, when the above step 1-5 is repeated again, step 3 BPF
program won't be able to claim an empty slot and a warning will be issued.

To fix the issue, for unset() function, we should traverse from the last slot
to the first. This way, the above issue can be avoided.

The same reverse traversal should also be done in bpf_get_local_storage() helper
itself. Otherwise, incorrect local storage may be returned to BPF program.

  [0] https://github.com/osandov/drgn

Fixes: b910eaaaa4b8 ("bpf: Fix NULL pointer dereference in bpf_get_local_storage() helper")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210810010413.1976277-1-yhs@fb.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: move put_tree() to avoid trim_trees refcount underflow and UAF

commit 67d69e9d1a6c889d98951c1d74b19332ce0565af upstream.

AUDIT_TRIM is expected to be idempotent, but multiple executions resulted
in a refcount underflow and use-after-free.

git bisect fingered commit fb041bb7c0a9 ("locking/refcount: Consolidate
implementations of refcount_t") but this patch with its more thorough
checking that wasn't in the x86 assembly code merely exposed a previously
existing tree refcount imbalance in the case of tree trimming code that
was refactored with prune_one() to remove a tree introduced in
commit 8432c7006297 ("audit: Simplify locking around untag_chunk()")

Move the put_tree() to cover only the prune_one() case.

Passes audit-testsuite and 3 passes of "auditctl -t" with at least one
directory watch.

Cc: Jan Kara <jack@suse.cz>
Cc: Will Deacon <will@kernel.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Seiji Nishikawa <snishika@redhat.com>
Cc: stable@vger.kernel.org
Fixes: 8432c7006297 ("audit: Simplify locking around untag_chunk()")
Signed-off-by: Richard Guy Briggs <rgb@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
[PM: reformatted/cleaned-up the commit description]
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>