The vsp1_du_setup_lif() function is used to configure and enable a
pipeline, as well as disable it, depending on the cfg argument being a
valid pointer or NULL. This creates a confusing API. Improve it by
splitting the function in two, a vsp1_du_enable() function to configure
a pipeline, and a vsp1_du_disable() function to disaple it.
Keep vsp1_du_setup_lif() as an inline wrapper for existing callers in
the DRM subsystem, to simplify merging. The callers will be updated
separately and the old API will then be removed.
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Link: https://patch.msgid.link/20260511235637.3468558-3-laurent.pinchart+renesas@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The rzv2h-ivc driver fails to handle back-2-back streaming sessions that
do not go through a peripheral reset. As the driver uses an autosuspend
delay of 2 seconds, it is quite possible that two consecutive streaming
sessions won't go through a suspend/resume sequence.
If the peripheral is not reset the second streaming session hangs and no
frames are delivered to the ISP.
This is because the stop_streaming() procedure implemented in the driver
doesn't match what's prescribed by the chip datasheet:
1) The chip manual suggests to poll the RZV2H_IVC_FM_INT_STAT_STPEND bit
of RZV2H_IVC_REG_FM_INT_STA instead of polling on RZV2H_IVC_REG_FM_STOP
and prescribes to clear the bit after polling has completed
2) More importantly: the RZV2H_IVC_REG_FM_STOP_FSTOP bit has to be set
on RZV2H_IVC_REG_FM_STOP -only- if a frame transfer to the ISP is in
progress. Setting the RZV2H_IVC_REG_FM_STOP_FSTOP bit when no frame is
being transferred causes the polling routine to timeout and the next
streaming session fails to start
As a frame transfer of an image in 1920x1080@10bi takes 5 milliseconds
at most, it is quite possible that the frame transfer completion interrupt
races with the stop procedure.
Instead of forcing a frame transfer abort, simply wait for the
in-progress transfer to complete by polling the ivc->vvalid_ifp status
variable in an hand-rolled loop that allows to inspect the variable
while holding the spinlock, to allow the irq handler to complete the
current buffer.
With this change, streaming back-2-back without suspending the
peripheral works successfully.
Cc: stable@vger.kernel.org
Fixes: f0b3984d82 ("media: platform: Add Renesas Input Video Control block driver")
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Using dev_dbg() in irq handlers to debug per-frame events is marginally
useful and possibly not the best idea, as using printk-based helpers
introduce latencies that impact the drivers operations.
If any tracing/debugging has to be performed around frame events
in interrupt handlers, the tracing subsystem offers better alternatives.
Drop dev_dgb() calls from the CRU interrupt handlers.
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The cru driver uses a 'state' variable for debugging purpose in the
interrupt handler. The state is used to detect invalid usage conditions
that are not meant to happen unless the driver has a bug in handling the
stop and start conditions.
Remove the state variable which seems to be a debugging leftover.
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Tested-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The current implementation of rzg2l_cru_fill_hw_slot() results in the
artificial loss of frames. At present whenever a frame-complete IRQ
is received the driver fills the hardware slot that was just written
to with the address of the next buffer in the driver's queue. If the
queue is empty, that hardware slot's address is set to the address of
the scratch buffer to enable the capture loop to keep running. There
is a minimum of a two-frame delay before that slot will be written to
however, and in the intervening period userspace may queue more
buffers which could be used.
To resolve the issue rework rzg2l_cru_fill_hw_slot() so that it
iteratively fills all slots from the queue which currently do not
have a buffer assigned, until the queue is empty. The scratch
buffer is only resorted to in the event that the queue is empty and
the next slot that will be written to does not already have a buffer
assigned.
Signed-off-by: Daniel Scally <dan.scally+renesas@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Tested-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Buffers are programmed into slots in queueing order.
When returning pending buffers we can't simply start from the first slot
but we should actually iterate slots starting from the one is use. The
rzg3e_cru_irq() handler already uses 'active_slot', make rzg2l_cru_irq()
use it as well to know where to start iterating from.
As the pattern of iterating over slots in order will be used for slots
programming in the next patches, provide an helper macro to do that.
While at it, rename return_unused_buffers() to rzg2l_cru_return_buffers().
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Tested-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The CRU cycles over the hardware slots where the destination address for
the next frame has to be programmed.
The RZ/G2L version of the IP has a register that tells which is the
last used slot by the hardware but, unfortunately, such register is not
available on RZ/G3E and RZ/V2H(P).
The driver currently compares the value of the AMnMADRSL/H register
which report "the memory address which the current video data was
written to" and compares it with the address programmed in the slots.
This heuristic requires a bit of book keeping and proper locking. As the
driver handles the FrameEnd interrupt, it's way easier to keep track
of the slot that has been used by ourselves with a driver variable.
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Tested-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Reviewed-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Split the locking between a spinlock dedicated to protect the hardware
slots programming (hw_lock) and one lock (qlock) to protect the queue of
buffers submitted by userspace.
Do not rework the locking strategy yet but start simply by splitting the
locking in two.
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The CRU driver uses a single spinlock to protect the buffers queue and
the hardware operations.
This single spinlock is held for the whole duration of the interrupt
handler, causing all other driver's operations to freeze.
Under heavy system stress conditions with userspace not providing
buffers fast enough, this causes loss of frames.
Prepare to re-work the driver locking by introducing (but not using yet)
a new spinlock to protect the hardware registers programming.
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The start/stop streaming routines do not need to lock the whole function
body against possible concurrent accesses to the CRU buffers or hardware
registers.
The stop function starts by disabling interrupts, and only this portion
needs to be protected not to race against a possible IRQ.
Once interrupts are disabled, nothing in the video device driver can race
and once the peripheral has been disabled we can release all pending
buffers.
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Tested-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Reviewed-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
On RZ/G3E the CRU driver relies on the frame end interrupt to detect the
completion of an active frame transfer when stopping DMA.
Update the driver to enable only frame end interrupts (CRUnIE2_FExE),
dropping the usage of the frame start interrupts, which is not required
for this operations flow.
Fix the interrupt status handling in the DMA stopping state by checking
the correct frame end status bits (FExS) instead of the frame start one
(FSxS). Add a dedicated CRUnINTS2_FExS() macro to reflect the actual
register bit layout and drop the now unused CRUnIE2_FSxE() and
CRUnINTS2_FSxS() macros.
This ensures that DMA stopping is triggered by the intended frame end
events and avoids incorrect interrupt handling.
Signed-off-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
When the CRU is configured to use ICnSVC for virtual channel mapping,
as on the RZ/{G3E, V2H/P} SoC, the ICnMC register must not be
programmed.
Return early after setting up ICnSVC to avoid overriding the ICnMC
register, which is not applicable in this mode.
This prevents unintended register programming when ICnSVC is enabled.
Cc: stable@vger.kernel.org
Fixes: 3c5ca0a48b ("media: rzg2l-cru: Drop function pointer to configure CSI")
Signed-off-by: Tommaso Merciai <tommaso.merciai.xr@bp.renesas.com>
[Rework to not break image format programming]
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Reviewed-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The histogram .set_fmt() handler suffers from two problems:
- When operating on the source pad, it returns correct information to
userspace, but does not store the format in the subdev state.
Subsequent calls to .get_fmt(), handled by the
vsp1_subdev_get_pad_format() helper, will not return the correct
information.
- When operating on the sink pad, it uses the
vsp1_subdev_set_pad_format(), which propagates the sink format to the
source, incorrectly overwriting the fixed source format.
The first issue could be fixed by implementing the set format operation
with vsp1_subdev_get_pad_format() on the source pad, if it wasn't that
.set_fmt() is also used to initialize the subdev state in
vsp1_entity_init_state(). The histogram would need a custom
.init_state() handler.
As the second issue would anyway overwrite the format, and therefore
requires a custom .set_fmt() implementation, fix both issues without
using the helpers.
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # HiHope RZ/G2M
Link: https://patch.msgid.link/20260318235907.831556-12-laurent.pinchart+renesas@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The RPF doesn't enforces the alignment constraint on the sink pad
format, which could have an odd size, possibly down to 1x1. In that
case, the upper bounds for the left and top coordinates clamping would
become negative, cast to a very large positive value. Incorrect crop
rectangle coordinates would then be incorrectly accepted.
A second issue can occur when the requested left and top coordinates are
negative. They are cast to a large unsigned value, clamped to the
maximum. While the calculation will produce valid values for the
hardware, this is not compliant with the V4L2 specification that
requires values to be adjusted to the closest valid value.
Fix both issues by switching to signed clamping, with an explicit
minimum to adjust negative values, and adjusting the clamp bounds to
avoid negative upper bounds.
Tested-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # HiHope RZ/G2M
Link: https://patch.msgid.link/20260318235907.831556-5-laurent.pinchart+renesas@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The media bus code passed to the .enum_frame_size() operation for the
sink pad is required to be supported by the device, but not to match the
current format. All entities that use the vsp1_subdev_enum_frame_size()
helper, as well as the SRU and UDS entities that implement the operation
manually, perform the check incorrectly.
Fix the issue by implementing the correct code check in the
vsp1_subdev_enum_frame_size(). For the SRU and UDS, to avoid duplicating
code, use the vsp1_subdev_enum_frame_size() as a base and override the
enumerated size on the source pad with entity-specific constraints.
While at it, include the missing <linux/mutex.h> as the code locks
mutexes.
Tested-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # HiHope RZ/G2M
Link: https://patch.msgid.link/20260318235907.831556-4-laurent.pinchart+renesas@ideasonboard.com
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Scheduling of work items with an async workqueue opens the door to
potential races between multiple instances of a work item.
While the frame transfer function is now protected against races, using
a workqueue doesn't provide much benefit considering the limited cost of
creating a job transfer.
Replace the usage of the work queue with direct function calls.
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The scheduling of a new buffer transfer in the IVC driver is triggered
by two occurrences of the "frame completed" interrupt.
The first interrupt occurrence identifies when all image data have been
transferred to the ISP, the second occurrence identifies when the
post-transfer VBLANK has completed and a new buffer can be transferred.
Under heavy system load conditions the actual execution of the workqueue
item might be delayed and two items might happen to run concurrently,
leading to a new frame transfer being triggered while the previous one
has not yet finished.
This error condition is only visible because the driver maintains a
status variable that counts the number of interrupts since the last
transfer, and warns in case an IRQ happens before the counter has been
reset.
To ensure sequential execution of the worqueue items and avoid a double
buffer transfer to run concurrently, protect the whole function body
with the spinlock that so far was solely used to reset the counter and
inspect the interrupt counter variable at the beginning of the buffer
transfer function.
As soon as the ongoing transfer completes, the workqueue item will be
re-scheduled and will consume the pending buffer.
Cc: stable@vger.kernel.org
Fixes: f0b3984d82 ("media: platform: Add Renesas Input Video Control block driver")
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The list of buffers (`rzv2h_ivc::buffers.queue`) is protected by a
spinlock (`rzv2h_ivc::buffers.lock`). However, in
`rzv2h_ivc_transfer_buffer()`, which runs in a separate workqueue, the
`list_del()` call is executed without holding the spinlock, which makes
it possible for the list to be concurrently modified
Fix that by removing a buffer from the list in the lock protected section.
Cc: stable@vger.kernel.org
Fixes: f0b3984d82 ("media: platform: Add Renesas Input Video Control block driver")
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Signed-off-by: Barnabás Pőcze <barnabas.pocze+renesas@ideasonboard.com>
[assign ivc->buffers.curr in critical section as reported by Barnabas]
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
The vertical blanking settings for the IVC block are dependent on
settings in the ISP. This was originally set to calculate as the
worst-case possible value, but it seems that this can cause the IVC
block to hang. Instead calculate the vblank to match the default
settings (which are currently all the driver sets anyway).
Cc: stable@vger.kernel.org
Fixes: f0b3984d82 ("media: platform: Add Renesas Input Video Control block driver")
Signed-off-by: Daniel Scally <dan.scally+renesas@ideasonboard.com>
Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Commit e7376745ad ("media: rcar-vin: Fix stride setting for RAW8
formats") removed dividing the stride by two for RAW8 formats. It is
unclear how this was tested, but in any of the recent tests this does
not seem to work and produces quite distorted images.
However, reverting the patch fixes the issues only partially. VNIS_REG
requires alignment to 16 bytes, and when dividing the stride by 2, in
some cases we end up with a non-aligned stride, producing a tilted
image. This issue has to be fixed in rvin_format_bytesperline() where we
do the alignment for bytesperline.
Adding back the stride division and increasing the alignment for RAW8
formats to 0x20 fixes the problems related to RAW8.
Fixes: e7376745ad ("media: rcar-vin: Fix stride setting for RAW8 formats")
Cc: stable@vger.kernel.org
Signed-off-by: Tomi Valkeinen <tomi.valkeinen+renesas@ideasonboard.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
