restrack: ------------[ cut here ]------------
infiniband hfi1_0: BUG: RESTRACK detected leak of resources
restrack: Kernel PD object allocated by ib_isert is not freed
restrack: Kernel CQ object allocated by ib_core is not freed
restrack: Kernel QP object allocated by rdma_cm is not freed
restrack: ------------[ cut here ]------------
Fixes: 699826f4e30a ("IB/isert: Fix incorrect release of isert connection") Reported-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Closes: https://lore.kernel.org/all/921cd1d9-2879-f455-1f50-0053fe6a6655@cornelisnetworks.com Link: https://lore.kernel.org/r/a27982d3235005c58f6d321f3fad5eb6e1beaf9e.1692604607.git.leonro@nvidia.com Tested-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 5de1540b7bc4 ("drivers/amba: create devices from device tree")
increases the refcount of of_node, but not releases it in
amba_device_release, so there is refcount leak. By using of_node_put
to avoid refcount leak.
Fixes: 5de1540b7bc4 ("drivers/amba: create devices from device tree") Signed-off-by: Peng Fan <peng.fan@nxp.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20230821023928.3324283-1-peng.fan@oss.nxp.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
BCMs with an enable_mask expect to only have that specific value written
to them. The current implementation only works by miracle for BCMs with
enable mask == BIT(0), as the minimal vote we've been using so far just
so happens to be equal to that.
Use the correct value with keepalive voting.
Fixes: d8630f050d3f ("interconnect: qcom: Add support for mask-based BCMs") Reported-by: Bjorn Andersson <quic_bjorande@quicinc.com> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org> Link: https://lore.kernel.org/r/20230811-topic-icc_fix_1he-v2-2-0620af8ac133@linaro.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
We don't need all the complex arithmetic for BCMs utilizing enable_mask,
as all we need to do is to determine whether there's any user (or
keepalive) asking for it to be on.
Separate the logic for such BCMs for a small speed boost.
Suggested-by: Bjorn Andersson <quic_bjorande@quicinc.com> Reviewed-by: Bjorn Andersson <quic_bjorande@quicinc.com> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org> Link: https://lore.kernel.org/r/20230811-topic-icc_fix_1he-v2-1-0620af8ac133@linaro.org Signed-off-by: Georgi Djakov <djakov@kernel.org>
Stable-dep-of: 1a70ca71547b ("interconnect: qcom: bcm-voter: Use enable_maks for keepalive voting") Signed-off-by: Sasha Levin <sashal@kernel.org>
There is a long call chain that &fip->ctlr_lock is acquired by isr
fnic_isr_msix_wq_copy() under hard IRQ context. Thus other process context
code acquiring the lock should disable IRQ, otherwise deadlock could happen
if the IRQ preempts the execution while the lock is held in process context
on the same CPU.
Change scsi_host_lookup() hostnum argument type from unsigned short to
unsigned int to match the type used everywhere else.
Fixes: 6d49f63b415c ("[SCSI] Make host_no an unsigned int") Signed-off-by: Tony Battersby <tonyb@cybernetics.com> Link: https://lore.kernel.org/r/a02497e7-c12b-ef15-47fc-3f0a0b00ffce@cybernetics.com Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently irdma allows zero-length STAGs to be programmed in HW during
the kernel mode fast register flow. Zero-length MR or STAG registration
disable HW memory length checks.
Improve gaps in bounds checking in irdma by preventing zero-length STAG or
MR registrations except if the IB_PD_UNSAFE_GLOBAL_RKEY is set.
smp_call_function_single() will allocate an IPI interrupt vector to
the target processor and send a function call request to the interrupt
vector. After the target processor receives the IPI interrupt, it will
execute arm_trbe_remove_coresight_cpu() call request in the interrupt
handler.
According to the device_unregister() stack information, if other process
is useing the device, the down_write() may sleep, and trigger deadlocks
or unexpected errors.
arm_trbe_remove_coresight_cpu
coresight_unregister
device_unregister
device_del
kobject_del
__kobject_del
sysfs_remove_dir
kernfs_remove
down_write ---------> it may sleep
Add a helper arm_trbe_disable_cpu() to disable TRBE precpu irq and reset
per TRBE.
Simply call arm_trbe_remove_coresight_cpu() directly without useing the
smp_call_function_single(), which is the same as registering the TRBE
coresight device.
For cp error case, there will be dirty meta/node pages remained after
f2fs_write_checkpoint() in f2fs_put_super(), drop them explicitly, and
do sanity check on reference count of dirty pages and inflight IOs.
Total GC calls is larger than BGGC calls, the reason is:
- f2fs_stat_info.call_count accounts total migrated section count
by f2fs_gc()
- f2fs_stat_info.bg_gc accounts total call times of f2fs_gc() from
background gc_thread
Another issue is gc_foreground_calls sysfs entry shows total GC call
count rather than FGGC call count.
This patch changes as below for fix:
- account GC calls and migrated segment count separately
- support to account migrated section count if it enables large section
mode
- fix to show correct value in gc_foreground_calls sysfs entry
Fixes: fc7100ea2a52 ("f2fs: Add f2fs stats to sysfs") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch provides a large number of variants of F2FS_RW_ATTR
and F2FS_RO_ATTR macros, reducing the number of parameters required
to initialize the f2fs_attr structure.
Reported-by: kernel test robot <lkp@intel.com> Link: https://lore.kernel.org/oe-kbuild-all/202304152234.wjaY3IYm-lkp@intel.com/ Signed-off-by: Yangtao Li <frank.li@vivo.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Stable-dep-of: 9bf1dcbdfdc8 ("f2fs: fix to account gc stats correctly") Signed-off-by: Sasha Levin <sashal@kernel.org>
f2fs: fix to do sanity check on extent cache correctly
The root cause is we applied both v1 and v2 of the patch, v2 is the right
fix, so it needs to revert v1 in order to fix reported issue.
v1:
commit d48a7b3a72f1 ("f2fs: fix to do sanity check on extent cache correctly")
https://lore.kernel.org/lkml/20230109034920.492914-1-chao@kernel.org/
v2:
commit 269d11948100 ("f2fs: fix to do sanity check on extent cache correctly")
https://lore.kernel.org/lkml/20230207134808.1827869-1-chao@kernel.org/
Reported-by: syzbot+601018296973a481f302@syzkaller.appspotmail.com Closes: https://lore.kernel.org/linux-f2fs-devel/000000000000fcf0690600e4d04d@google.com/ Fixes: d48a7b3a72f1 ("f2fs: fix to do sanity check on extent cache correctly") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Now f2fs support four block allocation modes: lfs, adaptive,
fragment:segment, fragment:block. Only lfs mode is allowed with zoned block
device feature.
Fixes: 6691d940b0e0 ("f2fs: introduce fragment allocation mode mount option") Signed-off-by: Chunhai Guo <guochunhai@vivo.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the ov2680_power_on() "sensor soft reset failed" path is hit during
probe() the WARN() about putting an enabled regulator at
drivers/regulator/core.c:2398 triggers 3 times (once for each regulator),
filling dmesg with backtraces.
Fix this by properly disabling the regulators on ov2680_power_on() errors.
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
ov2680_set_fmt() which == V4L2_SUBDEV_FORMAT_TRY was getting
the try_fmt v4l2_mbus_framefmt struct from the passed in sd_state
and then storing the contents of that into the return by reference
format->format struct.
While the right thing to do would be filling format->format based on
the just looked up mode and then store the results of that in
sd_state->pads[0].try_fmt .
Before the previous change introducing ov2680_fill_format() this
resulted in ov2680_set_fmt() which == V4L2_SUBDEV_FORMAT_TRY always
returning the zero-ed out sd_state->pads[0].try_fmt in format->format
breaking callers using this.
After the introduction of ov2680_fill_format() which at least
initializes sd_state->pads[0].try_fmt properly, format->format
is now always being filled with the default 800x600 mode set by
ov2680_init_cfg() independent of the actual requested mode.
Move the filling of format->format with ov2680_fill_format() to
before the if (which == V4L2_SUBDEV_FORMAT_TRY) and then store
the filled in format->format in sd_state->pads[0].try_fmt to
fix this.
Note this removes the fmt local variable because IMHO having a local
variable which points to a sub-struct of one of the function arguments
just leads to confusion when reading the code.
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Add a ov2680_fill_format() helper function and use this everywhere were
a v4l2_mbus_framefmt struct needs to be filled in so that the driver always
fills it consistently.
This is a preparation patch for fixing ov2680_set_fmt()
which == V4L2_SUBDEV_FORMAT_TRY calls not properly filling in
the passed in v4l2_mbus_framefmt struct.
Note that for ov2680_init_cfg() this now simply always fills
the try_fmt struct of the passed in sd_state. This is correct because
ov2680_init_cfg() is never called with a NULL sd_state so the old
sd_state check is not necessary.
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
VIDEO_V4L2_SUBDEV_API is now automatically selected in Kconfig
for all sensor drivers. Remove the ifdef CONFIG_VIDEO_V4L2_SUBDEV_API
checks.
This is a preparation patch for fixing ov2680_set_fmt()
which == V4L2_SUBDEV_FORMAT_TRY calls not properly filling in
the passed in v4l2_mbus_framefmt struct.
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
ov2680_vflip_disable() / ov2680_hflip_disable() pass BIT(0) instead of
0 as value to ov2680_mod_reg().
While fixing this also:
1. Stop having separate enable/disable functions for hflip / vflip
2. Move the is_streaming check, which is unique to hflip / vflip
into the ov2680_set_?flip() functions.
for a nice code cleanup.
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The index into ov2680_hv_flip_bayer_order[] should be 0-3, but
ov2680_bayer_order() was using 0 + BIT(2) + (BIT(2) << 1) as
max index, while the intention was to use: 0 + 1 + 2 as max index.
Fix the index calculation in ov2680_bayer_order(), while at it
also just use the ctrl values rather then reading them back using
a slow i2c-read transaction.
This also allows making the function void, since there now are
no more i2c-reads to error check.
Note the check for the ctrls being NULL is there to allow
adding an ov2680_fill_format() helper later, which will call
ov2680_set_bayer_order() during probe() before the ctrls are created.
[Sakari Ailus: Change all users of ov2680_set_bayer_order() here]
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the OV2680, the exposure time and gain are set manually from an external
controller. The OV2680 supports manual gain and exposure control only for
normal applications, no auto mode."
And indeed testing with the atomisp_ov2680 fork of ov2680.c has shown that
auto-exposure and auto-gain do not work.
Note that the code setting the auto-exposure flag was broken, callers
of ov2680_exposure_set() were directly passing !!ctrls->auto_exp->val as
"bool auto_exp" value, but ctrls->auto_exp is a menu control with:
So instead of passing !!ctrls->auto_exp->val they should have been passing
ctrls->auto_exp->val == V4L2_EXPOSURE_AUTO, iow the passed value was
inverted of what it should have been.
Also remove ov2680_g_volatile_ctrl() since without auto support the gain
and exposure controls are not volatile.
This also fixes the control values not being properly applied in
ov2680_mode_set(). The 800x600 mode register-list also sets gain,
exposure and vflip overriding the last set ctrl values.
ov2680_mode_set() does call ov2680_gain_set() and ov2680_exposure_set()
but did this before writing the mode register-list, so these values
would still be overridden by the mode register-list.
Add a v4l2_ctrl_handler_setup() call after writing the mode register-list
to restore all ctrl values. Also remove the ctrls->gain->is_new check from
ov2680_gain_set() so that the gain always gets restored properly.
Last since ov2680_mode_set() now calls v4l2_ctrl_handler_setup(), remove
the v4l2_ctrl_handler_setup() call after ov2680_mode_restore() since
ov2680_mode_restore() calls ov2680_mode_set().
Fixes: 3ee47cad3e69 ("media: ov2680: Add Omnivision OV2680 sensor driver") Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Acked-by: Rui Miguel Silva <rmfrfs@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The initial state of RESETB input signal of OV5640 should be
asserted, i.e. the sensor should be in reset. This is not the
case, make it so.
Since the subsequent assertion of RESETB signal is no longer
necessary and the timing of the power sequencing could be
slightly adjusted, add annotations to the delays which match
OV5640 datasheet rev. 2.03, both:
figure 2-3 power up timing with internal DVDD
figure 2-4 power up timing with external DVDD source
The 5..10ms delay between PWDN assertion and RESETB assertion
is not even documented in the power sequencing diagram, and
with this reset fix, it is no longer even necessary.
Set OV5640_REG_IO_MIPI_CTRL00 bit 2 to 1 instead of 0, since 1 means
MIPI CSI2 interface, while 0 means CPI parallel interface.
In the ov5640_set_power_mipi() the interface should obviously be set
to MIPI CSI2 since this functions is used to power up the sensor when
operated in MIPI CSI2 mode. The sensor should not be in CPI mode in
that case.
This fixes a corner case where capturing the first frame on i.MX8MN
with CSI/ISI resulted in corrupted frame.
Fixes: aa4bb8b8838f ("media: ov5640: Re-work MIPI startup sequence") Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com> Tested-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com> # [Test on imx6q] Signed-off-by: Marek Vasut <marex@denx.de> Tested-by: Jai Luthra <j-luthra@ti.com> # [Test on bplay, sk-am62] Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some digitizers (notably XP-Pen Artist 24) do not report the Invert
usage when erasing. This causes the device to be permanently stuck with
the BTN_TOOL_RUBBER tool after sending Eraser, as Invert is the only
usage that can release the tool. In this state, Touch and Inrange are
no longer reported to userspace, rendering the pen unusable.
Prior to commit 87562fcd1342 ("HID: input: remove the need for
HID_QUIRK_INVERT"), BTN_TOOL_RUBBER was never set and Eraser events were
simply translated into BTN_TOUCH without causing an inconsistent state.
Introduce HID_QUIRK_NOINVERT for such digitizers and detect them during
hidinput_configure_usage(). This quirk causes the tool to be released
as soon as Eraser is reported as not set. Set BTN_TOOL_RUBBER in
input->keybit when mapping Eraser.
Fixes: 87562fcd1342 ("HID: input: remove the need for HID_QUIRK_INVERT") Co-developed-by: Nils Fuhler <nils@nilsfuhler.de> Signed-off-by: Nils Fuhler <nils@nilsfuhler.de> Signed-off-by: Illia Ostapyshyn <ostapyshyn@sra.uni-hannover.de> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the current code, devres_release_all() only gets called if the device
has a bus and has been probed.
This leads to issues when using bus-less or driver-less devices where
the device might never get freed if a managed resource holds a reference
to the device. This is happening in the DRM framework for example.
We should thus call devres_release_all() in the device_del() function to
make sure that the device-managed actions are properly executed when the
device is unregistered, even if it has neither a bus nor a driver.
This is effectively the same change than commit 2f8d16a996da ("devres:
release resources on device_del()") that got reverted by commit a525a3ddeaca ("driver core: free devres in device_release") over
memory leaks concerns.
This patch effectively combines the two commits mentioned above to
release the resources both on device_del() and device_release() and get
the best of both worlds.
Correct spelling problems as identified by codespell.
Correct one grammar error.
Fixes: 9a93de620e0a ("docs: ABI: testing: Document the SBEFIFO timeout interface") Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: Eddie James <eajames@linux.ibm.com> Cc: Joel Stanley <joel@jms.id.au> Link: https://lore.kernel.org/r/20230710052305.29611-1-rdunlap@infradead.org Signed-off-by: Joel Stanley <joel@jms.id.au> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 567f97bd381f ("media: ipu3-cio2: support multiple sensors and VCMs
with same HID") introduced an on stack vcm_name and then uses this for
the name field of the software_node struct used for the vcm.
But the software_node struct is much longer lived then the current
stack-frame, so this is no good.
Instead extend the ipu_node_names struct with an extra field to store
the vcm software_node name and use that.
Note this also changes the length of the allocated buffer from
ACPI_ID_LEN + 4 to 16. the name is filled with "<ipu_vcm_types[x]>-%u"
where ipu_vcm_types[x] is not an ACPI_ID. The maximum length of
the strings in the ipu_vcm_types[] array is 11 + 5 bytes for "-255\0"
means 16 bytes are needed in the worst case scenario.
Fixes: 567f97bd381f ("media: ipu3-cio2: support multiple sensors and VCMs with same HID") Cc: Bingbu Cao <bingbu.cao@intel.com> Reviewed-by: Andy Shevchenko <andy@kernel.org> Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
cio2 bridge was involved along with IPU3. However, in fact all Intel IPUs
besides IPU3 CIO2 need this bridge driver. This patch move bridge driver
out of ipu3 directory and rename as ipu-bridge. Then it can be worked with
IPU3 and other Intel IPUs.
Signed-off-by: Bingbu Cao <bingbu.cao@intel.com> Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Stable-dep-of: 11e0a7c8e04e ("media: ipu-bridge: Do not use on stack memory for software_node.name field") Signed-off-by: Sasha Levin <sashal@kernel.org>
CONFIG_VIDEO_IMX_MEDIA isn't needed on arm64 platforms since commit 9f257f502c2e ("media: imx: Unstage the imx7-media-csi driver") which
moved the last arm64 driver depending on that Kconfig symbol out of
staging. Drop it from the arm64 defconfig.
Fixes: 9f257f502c2e ("media: imx: Unstage the imx7-media-csi driver") Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
As per information from Qualcomm [1], this property is not really
supported beyond msm8916 (HFI V1) and some newer HFI versions really
dislike receiving it, going as far as crashing the device.
Only consider toggling it (via the module option) on HFIV1.
While at it, get rid of the global static variable (which defaulted
to zero) which was never explicitly assigned to for V1.
Note: [1] is a reply to the actual message in question, as lore did not
properly receive some of the emails..
[1] https://lore.kernel.org/lkml/955cd520-3881-0c22-d818-13fe9a47e124@linaro.org/ Fixes: 7ed9e0b3393c ("media: venus: hfi, vdec: v6 Add IS_V6() to existing IS_V4() if locations") Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org> Signed-off-by: Stanimir Varbanov <stanimir.k.varbanov@gmail.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
The if statement that compares msgs[i].len != 3 is always false because
it is in a code block where msg[i].len is equal to 3. The check is
redundant and can be removed.
As detected by cppcheck static analysis:
drivers/media/usb/go7007/go7007-i2c.c:168:20: warning: Opposite inner
'if' condition leads to a dead code block. [oppositeInnerCondition]
The adap_configured() callback was called with the adap->lock mutex
held if the 'configured' argument was false, and without the adap->lock
mutex held if that argument was true.
That was very confusing, and so split this up in a adap_unconfigured()
callback and a high-level configured() callback.
This also makes it easier to understand when the mutex is held: all
low-level adap_* callbacks are called with the mutex held. All other
callbacks are called without that mutex held.
A potential deadlock was found by Zheng Zhang with a local syzkaller
instance.
The problem is that when a non-blocking CEC transmit is canceled by calling
cec_data_cancel, that in turn can call the high-level received() driver
callback, which can call cec_transmit_msg() to transmit a new message.
The cec_data_cancel() function is called with the adap->lock mutex held,
and cec_transmit_msg() tries to take that same lock.
The root cause is that the received() callback can either be used to pass
on a received message (and then adap->lock is not held), or to report a
canceled transmit (and then adap->lock is held).
This is confusing, so create a new low-level adap_nb_transmit_canceled
callback that reports back that a non-blocking transmit was canceled.
And the received() callback is only called when a message is received,
as was the case before commit f9d0ecbf56f4 ("media: cec: correctly pass
on reply results") complicated matters.
If a duplicate attribute is found using kset_find_obj(),
a reference to that attribute is returned. This means
that we need to dispose it accordingly. Use kobject_put()
to dispose the duplicate attribute in such a case.
Compile-tested only.
Fixes: e8a60aa7404b ("platform/x86: Introduce support for Systems Management Driver over WMI for Dell Systems") Signed-off-by: Armin Wolf <W_Armin@gmx.de> Link: https://lore.kernel.org/r/20230805053610.7106-1-W_Armin@gmx.de Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Even the PCI devices don't support pasid capability, PASID table is
mandatory for a PCI device in scalable mode. However flushing cache
of pasid directory table for these devices are not taken after pasid
table is allocated as the "size" of table is zero. Fix it by
calculating the size by page order.
Found this when reading the code, no real problem encountered for now.
TX is handled by primary sequencer. After cancelling primary command, poll
primary sequencer's irq status instead of that of secondary.
While at it, also remove a couple of redundant lines that read from IRQ_EN
register and write back same.
Fixes: 2aaa43c70778 ("tty: serial: qcom-geni-serial: add support for serial engine DMA") Signed-off-by: Vijaya Krishna Nivarthi <quic_vnivarth@quicinc.com> Link: https://lore.kernel.org/r/1691578393-9891-1-git-send-email-quic_vnivarth@quicinc.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Writing the new TTBRs, TCRs and MAIRs on a previously enabled
context bank may trigger a context fault, resulting in firmware
driven AP resets: change the domain initialization programming
sequence to disable the context bank(s) and to also clear the
related fault address (CB_FAR) and fault status (CB_FSR)
registers before writing new values to TTBR0/1, TCR/TCR2, MAIR0/1.
Fixes: 0ae349a0f33f ("iommu/qcom: Add qcom_iommu") Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Link: https://lore.kernel.org/r/20230622092742.74819-4-angelogioacchino.delregno@collabora.com Signed-off-by: Will Deacon <will@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
It has been observed that sometimes the FSI master will return all 0xffs
after a CFAM has been taken out of reset, without presenting any error.
Resetting the FSI master errors resolves the issue.
Currently, direct wqe is not supported for wr-list. RoCE driver excludes
direct wqe for wr-list by judging whether the number of wr is 1.
For a wr-list where the second wr is a length-error atomic wr, the
post-send driver handles the first wr and adds 1 to the wr number counter
firstly. While handling the second wr, the driver finds out a length error
and terminates the wr handle process, remaining the counter at 1. This
causes the driver mistakenly judges there is only 1 wr and thus enters
the direct wqe process, carrying the current length-error atomic wqe.
This patch fixes the error by adding a judgement whether the current wr
is a bad wr. If so, use the normal doorbell process but not direct wqe
despite the wr number is 1.
drivers/iommu/iommu.c:3261: warning: Function parameter or member 'group'
not described in 'iommu_group_release_dma_owner'
drivers/iommu/iommu.c:3261: warning: Excess function parameter 'dev'
description in 'iommu_group_release_dma_owner'
drivers/iommu/iommu.c:3275: warning: Function parameter or member 'dev'
not described in 'iommu_device_release_dma_owner'
drivers/iommu/iommu.c:3275: warning: Excess function parameter 'group'
description in 'iommu_device_release_dma_owner'
force_aperture was intended to false only by GART drivers that have an
identity translation outside the aperture. This does not describe sprd, so
add the missing 'force_aperture = true'.
Prepare for mt8188 to fix a two IOMMU HWs share pagetable issue.
We have two MM IOMMU HWs in mt8188, one is VPP-IOMMU, the other is
VDO-IOMMU. The 2 MM IOMMU HWs share pagetable don't work in this case:
a) VPP-IOMMU probe firstly.
b) VDO-IOMMU probe.
c) The master for VDO-IOMMU probe (means frstdata is vpp-iommu).
d) The master in another domain probe. No matter it is vdo or vpp.
Then it still create a new pagetable in step d). The problem is
"frstdata->bank[0]->m4u_dom" was not initialized. Then when d) enter, it
still create a new one.
In this patch, we create a new variable "share_dom" for this share
pgtable case, it should be helpful for readable. and put all the share
pgtable logic in the mtk_iommu_domain_finalise.
In mt8195, the master of VPP-IOMMU probes before than VDO-IOMMU
from its dtsi node sequence, we don't see this issue in it. Prepare for
mt8188.
The driver fails to link when CONFIG_POWER_SUPPLY is disabled:
x86_64-linux-ld: vmlinux.o: in function `cht_wc_extcon_psy_get_prop':
extcon-intel-cht-wc.c:(.text+0x15ccda7): undefined reference to `power_supply_get_drvdata'
x86_64-linux-ld: vmlinux.o: in function `cht_wc_extcon_pwrsrc_event':
extcon-intel-cht-wc.c:(.text+0x15cd3e9): undefined reference to `power_supply_changed'
x86_64-linux-ld: vmlinux.o: in function `cht_wc_extcon_probe':
extcon-intel-cht-wc.c:(.text+0x15cd596): undefined reference to `devm_power_supply_register'
It should be possible to change the driver to not require this at
compile time and still provide other functions, but adding a hard
Kconfig dependency does not seem to have any practical downsides
and is simpler since the option is normally enabled anyway.
Fixes: 66e31186cd2aa ("extcon: intel-cht-wc: Add support for registering a power_supply class-device") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In some randconfig builds, kernfs ends up being disabled, so there is no prototype
for kernfs_generic_poll()
In file included from kernel/sched/build_utility.c:97:
kernel/sched/psi.c:1479:3: error: implicit declaration of function 'kernfs_generic_poll' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
kernfs_generic_poll(t->of, wait);
^
Add a stub helper for it, as we have it for other kernfs functions.
When test_remove is enabled really_probe() does not properly pair
dma_configure() with dma_remove(), it will end up calling dma_configure()
twice. This corrupts the owner_cnt and renders the group unusable with
VFIO/etc.
Add the missing cleanup before going back to re_probe.
Enable the generic .sync_state callback to ensure there are no
outstanding votes that would waste power.
Generally one would need a bunch of interface clocks to access the QoS
registers when trying to go over all possible nodes during sync_state,
but QCM2290 surprisingly does not seem to require any such handling.
Perf cs_etm session executed unexpectedly when AUX buffer > 1G.
perf record -C 0 -m ,2G -e cs_etm// -- <workload>
[ perf record: Captured and wrote 2.615 MB perf.data ]
Perf only collect about 2M perf data rather than 2G. This is becasuse
the operation, "nr_pages << PAGE_SHIFT", in coresight tmc driver, will
overflow when nr_pages >= 0x80000(correspond to 1G AUX buffer). The
overflow cause buffer allocation to fail, and TMC driver will alloc
minimal buffer size(1M). You can just get about 2M perf data(1M AUX
buffer + perf data header) at least.
Explicit convert nr_pages to 64 bit to avoid overflow.
Fixes: 22f429f19c41 ("coresight: etm-perf: Add support for ETR backend") Fixes: 99443ea19e8b ("coresight: Add generic TMC sg table framework") Fixes: 2e499bbc1a92 ("coresight: tmc: implementing TMC-ETF AUX space API") Signed-off-by: Ruidong Tian <tianruidong@linux.alibaba.com> Reviewed-by: James Clark <james.clark@arm.com> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com> Link: https://lore.kernel.org/r/20230804081514.120171-2-tianruidong@linux.alibaba.com Signed-off-by: Sasha Levin <sashal@kernel.org>
One-element and zero-length arrays are deprecated. So, replace
one-element array in struct irdma_qvlist_info with flexible-array
member.
A patch for this was sent a while ago[1]. However, it seems that, at
the time, the changes were partially folded[2][3], and the actual
flexible-array transformation was omitted. This patch fixes that.
The only binary difference seen before/after changes is shown below:
If a send packet is dropped by the IP layer in rxe_requester()
the call to rxe_xmit_packet() can fail with err == -EAGAIN.
To recover, the state of the wqe is restored to the state before
the packet was sent so it can be resent. However, the routines
that save and restore the state miss a significnt part of the
variable state in the wqe, the dma struct which is used to process
through the sge table. And, the state is not saved before the packet
is built which modifies the dma struct.
Under heavy stress testing with many QPs on a fast node sending
large messages to a slow node dropped packets are observed and
the resent packets are corrupted because the dma struct was not
restored. This patch fixes this behavior and allows the test cases
to succeed.
Fixes: 3050b9985024 ("IB/rxe: Fix race condition between requester and completer") Link: https://lore.kernel.org/r/20230721200748.4604-1-rpearsonhpe@gmail.com Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch corrects an error in rxe_modify_srq where if the
caller changes the srq size the actual new value is not returned
to the caller since it may be larger than what is requested.
Additionally it open codes the subroutine rcv_wqe_size() which
adds very little value, and makes some whitespace changes.
If create_qp does not fully succeed it is possible for qp cleanup
code to attempt to drain the send or recv work queues before the
queues have been created causing a seg fault. This patch checks
to see if the queues exist before attempting to drain them.
Link: https://lore.kernel.org/r/20230620135519.9365-3-rpearsonhpe@gmail.com Reported-by: syzbot+2da1965168e7dbcba136@syzkaller.appspotmail.com Closes: https://lore.kernel.org/linux-rdma/00000000000012d89205fe7cfe00@google.com/raw Fixes: 49dc9c1f0c7e ("RDMA/rxe: Cleanup reset state handling in rxe_resp.c") Fixes: fbdeb828a21f ("RDMA/rxe: Cleanup error state handling in rxe_comp.c") Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch:
- Moves code to initialize a qp send work queue to a
subroutine named rxe_init_sq.
- Moves code to initialize a qp recv work queue to a
subroutine named rxe_init_rq.
- Moves initialization of qp request and response packet
queues ahead of work queue initialization so that cleanup
of a qp if it is not fully completed can successfully
attempt to drain the packet queues without a seg fault.
- Makes minor whitespace cleanups.
Fixes: 8700e3e7c485 ("Soft RoCE driver") Link: https://lore.kernel.org/r/20230620135519.9365-2-rpearsonhpe@gmail.com Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com> Acked-by: Zhu Yanjun <zyjzyj2000@gmail.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The global pointer 'sprd_port' may not zero when sprd_probe returns
failure, that is a risk for sprd_port to be accessed afterward, and
may lead to unexpected errors.
For example:
There are two UART ports, UART1 is used for console and configured in
kernel command line, i.e. "console=";
The UART1 probe failed and the memory allocated to sprd_port[1] was
released, but sprd_port[1] was not set to NULL;
In UART2 probe, the same virtual address was allocated to sprd_port[2],
and UART2 probe process finally will go into sprd_console_setup() to
register UART1 as console since it is configured as preferred console
(filled to console_cmdline[]), but the console parameters (sprd_port[1])
belong to UART2.
So move the sprd_port[] assignment to where the port already initialized
can avoid the above issue.
There are three places that qla4xxx parses nlattrs:
- qla4xxx_set_chap_entry()
- qla4xxx_iface_set_param()
- qla4xxx_sysfs_ddb_set_param()
and each of them directly converts the nlattr to specific pointer of
structure without length checking. This could be dangerous as those
attributes are not validated and a malformed nlattr (e.g., length 0) could
result in an OOB read that leaks heap dirty data.
Add the nla_len check before accessing the nlattr data and return EINVAL if
the length check fails.
Fixes: 26ffd7b45fe9 ("[SCSI] qla4xxx: Add support to set CHAP entries") Fixes: 1e9e2be3ee03 ("[SCSI] qla4xxx: Add flash node mgmt support") Fixes: 00c31889f751 ("[SCSI] qla4xxx: fix data alignment and use nl helpers") Signed-off-by: Lin Ma <linma@zju.edu.cn> Link: https://lore.kernel.org/r/20230723080053.3714534-1-linma@zju.edu.cn Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
beiscsi_iface_set_param() parses nlattr with nla_for_each_attr and assumes
every attributes can be viewed as struct iscsi_iface_param_info.
This is not true because there is no any nla_policy to validate the
attributes passed from the upper function iscsi_set_iface_params().
Add the nla_len check before accessing the nlattr data and return EINVAL if
the length check fails.
Fixes: 0e43895ec1f4 ("[SCSI] be2iscsi: adding functionality to change network settings using iscsiadm") Signed-off-by: Lin Ma <linma@zju.edu.cn> Link: https://lore.kernel.org/r/20230723075938.3713864-1-linma@zju.edu.cn Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The functions iscsi_if_set_param() and iscsi_if_set_host_param() convert an
nlattr payload to type char* and then call C string handling functions like
sscanf and kstrdup:
However, since the nlattr is provided by the user-space program and the
nlmsg skb is allocated with GFP_KERNEL instead of GFP_ZERO flag (see
netlink_alloc_large_skb() in netlink_sendmsg()), dirty data on the heap can
lead to an OOB access for those string handling functions.
By investigating how the bug is introduced, we find it is really
interesting as the old version parsing code starting from commit fd7255f51a13 ("[SCSI] iscsi: add sysfs attrs for uspace sync up") treated
the nlattr as integer bytes instead of string and had length check in
iscsi_copy_param():
if (ev->u.set_param.len != sizeof(uint32_t))
BUG();
But, since the commit a54a52caad4b ("[SCSI] iscsi: fixup set/get param
functions"), the code treated the nlattr as C string while forgetting to
add any strlen checks(), opening the possibility of an OOB access.
Fix the potential OOB by adding the strlen() check before accessing the
buf. If the data passes this check, all low-level set_param handlers can
safely treat this buf as legal C string.
Fixes: fd7255f51a13 ("[SCSI] iscsi: add sysfs attrs for uspace sync up") Fixes: 1d9bf13a9cf9 ("[SCSI] iscsi class: add iscsi host set param event") Signed-off-by: Lin Ma <linma@zju.edu.cn> Link: https://lore.kernel.org/r/20230723075820.3713119-1-linma@zju.edu.cn Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The current NETLINK_ISCSI netlink parsing loop checks every nlmsg to make
sure the length is bigger than sizeof(struct iscsi_uevent) and then calls
iscsi_if_recv_msg().
Hence, in iscsi_if_recv_msg() the nlmsg_data can be safely converted to
iscsi_uevent as the length is already checked.
However, in other cases the length of nlattr payload is not checked before
the payload is converted to other data structures. One example is
iscsi_set_path() which converts the payload to type iscsi_path without any
checks:
To sum up, some code paths called in iscsi_if_recv_msg() do not check the
length of the data (see below picture) and directly convert the data to
another data structure. This could result in an out-of-bound reads and heap
dirty data leakage.
Fix the issue by adding the length check before accessing it. To clean up
the code, an additional parameter named rlen is added. The rlen is
calculated at the beginning of iscsi_if_recv_msg() which avoids duplicated
calculation.
Fixes: ac20c7bf070d ("[SCSI] iscsi_transport: Added Ping support") Fixes: 43514774ff40 ("[SCSI] iscsi class: Add new NETLINK_ISCSI messages for cnic/bnx2i driver.") Fixes: 1d9bf13a9cf9 ("[SCSI] iscsi class: add iscsi host set param event") Fixes: 01cb225dad8d ("[SCSI] iscsi: add target discvery event to transport class") Fixes: 264faaaa1254 ("[SCSI] iscsi: add transport end point callbacks") Fixes: fd7255f51a13 ("[SCSI] iscsi: add sysfs attrs for uspace sync up") Signed-off-by: Lin Ma <linma@zju.edu.cn> Link: https://lore.kernel.org/r/20230725024529.428311-1-linma@zju.edu.cn Reviewed-by: Chris Leech <cleech@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Although the code for residual handling in the SRP initiator follows the
SCSI documentation, that documentation has never been correct. Because
scsi_finish_command() starts from the data buffer length and subtracts the
residual, scsi_set_resid() must not be called if a residual overflow
occurs. Hence remove the scsi_set_resid() calls from the SRP initiator if a
residual overflow occurrs.
Cc: Leon Romanovsky <leon@kernel.org> Cc: Jason Gunthorpe <jgg@nvidia.com> Fixes: 9237f04e12cc ("scsi: core: Fix scsi_get/set_resid() interface") Fixes: e714531a349f ("IB/srp: Fix residual handling") Signed-off-by: Bart Van Assche <bvanassche@acm.org> Link: https://lore.kernel.org/r/20230724200843.3376570-3-bvanassche@acm.org Acked-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The function mxs_phy_is_otg_host() will return true if OTG_ID_VALUE is
0 at USBPHY_CTRL register. However, OTG_ID_VALUE will not reflect the real
state if the ID pin is float, such as Host-only or Type-C cases. The value
of OTG_ID_VALUE is always 1 which means device mode.
This patch will fix the issue by judging the current mode based on
last_event. The controller will update last_event in time.
If we encounter any error in the vdec_msg_queue_init() then we need
to set "msg_queue->wdma_addr.size = 0;". Normally, this is done
inside the vdec_msg_queue_deinit() function. However, if the
first call to allocate &msg_queue->wdma_addr fails, then the
vdec_msg_queue_deinit() function is a no-op. For that situation, just
set the size to zero explicitly and return.
There were two other error paths which did not clean up before returning.
Change those error paths to goto mem_alloc_err.
Fixes: b199fe46f35c ("media: mtk-vcodec: Add msg queue feature for lat and core architecture") Fixes: 2f5d0aef37c6 ("media: mediatek: vcodec: support stateless AV1 decoder") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
The "lat_buf->private_data" needs to be set to NULL to prevent a
double free. How this would happen is if vdec_msg_queue_init() failed
twice in a row and on the second time it failed earlier than on the
first time.
The vdec_msg_queue_init() function has a loop which does:
for (i = 0; i < NUM_BUFFER_COUNT; i++) {
Each iteration initializes one element in the msg_queue->lat_buf[] array
and then the clean up function vdec_msg_queue_deinit() frees each
element of the msg_queue->lat_buf[] array. This clean up code relies
on the assumption that every element is either initialized or zeroed.
Leaving a freed pointer which is non-zero breaks the assumption.
Fixes: b199fe46f35c ("media: mtk-vcodec: Add msg queue feature for lat and core architecture") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
"fb_use_list" is used to store used or referenced frame buffers for
vp9 stateful decoder. "NULL" should be returned when getting target
frame buffer failed from "fb_use_list", not a random unexpected one.
Fixes: f77e89854b3e ("[media] vcodec: mediatek: Add Mediatek VP9 Video Decoder Driver") Signed-off-by: Irui Wang <irui.wang@mediatek.com> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
the supported_instance_count determine the instance index range,
it shouldn't exceed the bits number of instance_mask,
otherwise the bitops of instance_mask may cross boundaries
Fixes: 9f599f351e86 ("media: amphion: add vpu core driver") Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by: Ming Qian <ming.qian@nxp.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
the firmware only support low latency mode for h264,
but firmware will notify an event to driver
when one frame is decoded,
if V4L2_CID_MPEG_VIDEO_DEC_DISPLAY_DELAY_ENABLE is enabled,
and V4L2_CID_MPEG_VIDEO_DEC_DISPLAY_DELAY is set to 0,
driver can display the decoded frame immediately.
In mtk_jpeg_probe, &jpeg->job_timeout_work is bound with
mtk_jpeg_job_timeout_work. Then mtk_jpeg_dec_device_run
and mtk_jpeg_enc_device_run may be called to start the
work.
If we remove the module which will call mtk_jpeg_remove
to make cleanup, there may be a unfinished work. The
possible sequence is as follows, which will cause a
typical UAF bug.
Fix it by canceling the work before cleanup in the mtk_jpeg_remove
Commit f100ce3bbd6a ("media: verisilicon: Fix crash when probing
encoder") removed vpu_fmt from hantro_try_fmt(), since it was
initialized from vpu_dst_fmt, which may not be initialized, when TRY_FMT
is called. It was replaced by fmt, which is found using the pixelformat.
For the encoder, this changed the fmt to contain the raw format instead
of the coded format. The format constraints as of fmt->frmsize are only
valid for the coded format and are 0 for the raw formats. Therefore, the
size of a encoder OUTPUT device is constrained to 0 and the
v4l2-compliance tests for G_FMT, TRY_FMT, and SET_FMT fail.
Bring back vpu_fmt to use the coded format on an encoder OUTPUT device,
but initialize it using the currently set pixelformat on dst_fmt, which
is the coded format on an encoder.
Fixes: f100ce3bbd6a ("media: verisilicon: Fix crash when probing encoder") Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
according to v4l2 stateful decoder document 4.5.1.3. State Machine,
the state should change from seek to initialization
if call VIDIOC_REQBUFS(OUTPUT, 0).
so reinit the vpu decoder if reqbufs output 0
Fixes: 6de8d628df6e ("media: amphion: add v4l2 m2m vpu decoder stateful driver") Signed-off-by: Ming Qian <ming.qian@nxp.com> Tested-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
The PIO read command has no response frame and the struct iu[1024] won't be
filled. I/Os which are normally completed will be treated as failed in
sas_ata_task_done() when iu contains abnormal dirty data.
Consequently ending_fis should not be filled by iu when the response frame
hasn't been written to memory.
Fixes: d380f55503ed ("scsi: hisi_sas: Don't bother clearing status buffer IU in task prep") Signed-off-by: Xingui Yang <yangxingui@huawei.com> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Link: https://lore.kernel.org/r/1689045300-44318-2-git-send-email-chenxiang66@hisilicon.com Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / thirdparty / kernel / stable.git / log
