Declare power-domain names "top", "ife0" and "ife1" eponymously for the
power-domains TITAN_TOP_GDSC, IFE_0_GDSC and IFE_1_GDSC respectively.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Some platforms such as SC7280 have 3 CSIDs and 2 CSID-lites but current
code has hardcoded 2 as the maximum number of CSIDs. Remove the hardcoded
maximum number of VFEs to handle all possible combinations of CSIDs and
CSID-lites.
Signed-off-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Right now we use fixed indexes to assign power-domains, with a
requirement for the TOP GDSC to come last in the list.
Adding support for named power-domains means the declaration in the dtsi
can come in any order.
After this change we continue to support the old indexing - if a SoC
resource declaration or the in-use dtb doesn't declare power-domain names
we fall back to the default legacy indexing.
From this point on though new SoC additions should contain named
power-domains, eventually we will drop support for legacy indexing.
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Moving the location of the hooks to VFE power domains has several
advantages.
1. Separation of concerns and functional decomposition.
vfe.c should be responsible for and know best how manage
power-domains for a VFE, excising from camss.c follows this
principle.
2. Embedding a pointer to genpd in struct camss_vfe{} meas that we can
dispense with a bunch of kmalloc array inside of camss.c.
3. Splitting up titan top gdsc from vfe/ife gdsc provides a base for
breaking up magic indexes in dtsi.
Suggested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
For the various versions of VFE we have a boiler-plate
pm_domain_on/pm_domain_off callback pair of the general form.
- Error check.
Not always done but applicable to all.
- device_link_add (DL_FLAG_STATELESS | DL_FLAG_PM_RUNTIME |
DL_FLAG_RPM_ACTIVE);
- Error check returning -EINVAL on error.
- Return 0
Reduce the pattern down to a common callback. VFE 4.1 is a special case
which to me also indicates that it is worthwhile maintaining an indirection
for the vfe_pm_domain_{on|off} for now.
Otherwise lets chuck out a bunch of needlessly replicated code.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Suggested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Right now we use the top-level camss structure to provide pointers via
VFE id index back to genpd linkages.
In effect this hard-codes VFE indexes to power-domain indexes in the
dtsi and mandates a very particular ordering of power domains in the
dtsi, which bears no relationship to a real hardware dependency.
As a first step to rationalising the VFE power-domain code and breaking
the magic indexing in dtsi use per-VFE pointers to genpd linkages.
The top-level index in msm_vfe_subdev_init is still used to attain the
initial so no functional or logical change arises from this change.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
At the moment we have some complex code for determining if a VFE requires a
power-domain attachment. Particularly discordant in this scheme is the
subtle reliance on VFE and VFE Lite declaration ordering in our resources.
VFE id is used to determine if a VFE is lite or not and consequently if a
VFE requires power-domain attachment. VFE Lite though is not a correct
delineation between power-domain and non power-domain state since early
SoCs have neither VFE Lite nor power-domains attached to VFEs.
Introduce has_pd to the VFE resource structure to allow the CAMSS code to
understand if it needs to try to attach a power-domain for a given VFE.
As a side-effect from this we no longer need to care about VFE Lite or
non-Lite or the id number associated with either and which order the
VFE/VFE Lite was declared in.
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Tested-by: Matti Lehtimäki <matti.lehtimaki@gmail.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
V4L2 subdev operations have moved from operating on a
v4l2_subdev_pad_config to a v4l2_subdev_state a long time ago. Fix
remaining incorrect references to pad config in comments.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Digging into the documentation we find that the DT_ID bitfield is used to
map the six bit DT to a two bit ID code. This value is concatenated to the
VC bitfield to create a CID value. DT_ID is the two least significant bits
of CID and VC the most significant bits.
Originally we set dt_id = vc * 4 in and then subsequently set dt_id = vc.
commit 3c4ed72a16 ("media: camss: sm8250: Virtual channels for CSID")
silently fixed the multiplication by four which would give a better
value for the generated CID without mentioning what was being done or why.
Next up I haplessly changed the value back to "dt_id = vc * 4" since there
didn't appear to be any logic behind it.
Hans asked what the change was for and I honestly couldn't remember the
provenance of it, so I dug in.
Link: https://lore.kernel.org/linux-arm-msm/edd4bf9b-0e1b-883c-1a4d-50f4102c3924@xs4all.nl/
Add a comment so the next hapless programmer doesn't make this same
mistake.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
A previous patch I had removed the returns from vfe_disable() since we
didn't trap any meaningful errors. Konrad pointed out vfe_reset() could
return an error, which is true.
Trap the vfe_reset() error code and throw it up the callstack.
Suggested-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
We can move vfe_disable() into a common routine in the core VFE file
provided we make wm_stop() a VFE specific callback.
The callback is required to capture the case where VFE 17x currently isn't
VC enabled where as VFE 480 is.
Suggested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Some VFEs have four RDIs apiece. Right now the ISR code has a hard-coded
value which tops-out at RDI2 meaning only three RDIs can be utilised in
practice.
Extend out the various routines in camss-vfe-17x.c to support the higher
RDI count.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Several of our upstream and soon-to-be upstream SoC CAMSS dtsi declare
csiphyX as opposed to the older clock name csiX_phy.
Right now the CAMSS code will fail to set the csiphyX clock even if we have
declared it in our list of clocks. For sdm845 and sm8250 we appear to "get
away" with this error, however on sc8280xp we don't.
The right approach here is to set the clock when it is declared. If a SoC
doesn't require or a SoC driver implementer doesn't think we need, then the
clock ought to simply be omitted from the clock list.
Include csiphyX in the set of permissible strings which will subsequently
lead to the csiphyX clock being set during csiphy_set_clock_rates() phase.
sdm845 and sm8250 will work with the code as-is so I've omitted this from a
suggested Fixes list.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The csiphyX_timer and csiX_phy values need not be hard-coded. We can
functionally decompose the string matching inside of a loop.
Static string values are brittle, difficult to extend and not required
anyway since the camss->res->csiphy_num value informs us of the number
of CSIPHYs and hence the set of potential clocks for a given CSIPHY.
In simple terms if we have five CSIPHYs we can have no more and no less
than five csiphy_timer clocks. Similarly csi_phy core clocks have a 1:1
relationship with the PHY they clock.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The number of Video Front End - VFE or Image Front End - IFE supported
with new SoCs can vary both for the full and lite cases.
For example sdm845 has one vfe_lite and two vfe interfaces with the vfe
clock called simply "vfe_lite" with no integer postfix. sc8280xp has four
vfe and four vfe lite blocks.
At the moment we declare vfe_lite0 and vfe_lite1 for sm8250 but never set
those clocks because we don't match the strings.
We need to support the following clock name formats
- vfeN
- vfe_liteN
- vfe_lite
with N being any reasonably sized integer.
There are two sites in this code which need to do the same thing,
constructing and matching strings with the pattern above, so encapsulate
the logic in one function.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
We have a very convoluted if/else legacy here which needs to be
rationalised to make it more sustainable.
Adding in another Soc or two will make some of these if statements into
increasingly large multi-clause behemoths.
Introduce switches in the obvious places to despaghetiify.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
From sdm845 onwards we need to ensure the VFE is powered on prior to
switching on the CSID.
Currently the code tests for sdm845, sm8250 and then does get/set. This is
not extensible and it turns out is not necessary either since vfe_get and
vfe_set reference count.
Remove the over-conservative SoC version check.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Suggested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Tested-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org> # rb3 # db410c
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Each Video Front End - VFE - has a variable number of Raw Data Interfaces -
RDIs associated with it.
The CAMSS code started from a naive implementation where a fixed define was
used as a control in a for(){} loop iterating through RDIs.
That model scales badly. An attempt was made with VFE_LINE_NUM_GEN2 and
VFE_LINE_NUM_GEN1 to differentiate between SoCs but, the problem with that
is "gen1" and "gen2" have no meaning in the silicon. There is no fixed
constraint in the silicon between VFE and RDI, it is entirely up to the SoC
designers how many VFEs are populated and how many RDIs to associate with
each VFE.
As an example sdm845 has VFE version 175 and sm8250 VFE version 480.
sdm845 has 2 VFEs with 4 RDIs and 1 VFE Lite with 4 RDIs.
sm8250 has 2 VFEs with 3 RDIs and 2 VFE Lite with 4 RDIs.
Clearly then we need a more granular model to capture the necessary data.
The defines have gone away to be replaced with per-SoC data but, we haven't
populated the parameter data with the real values.
Let's call those values out now
msm8916:
1 x VFE
3 x RDI per VFE (not 4)
msm8996:
2 x VFE
3 x RDI per VFE (not 4)
sdm660:
2 x VFE
3 x RDI per VFE (not 4)
sdm845:
2 x VFE
4 x RDI per VFE (not 3)
1 x VFE Lite
4 x RDI per VFE Lite (not 3)
sm8250:
2 x VFE
3 x RDI per VFE (not 4)
2 x VFE Lite
4 x RDI per VFE Lite
This more complex and correct mapping was not possible prior to passing
values via driver data. Now that we have that change in place we can
correctly map VFEs to RDIs for each VFE.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
It is possible to pass all of the CAMSS subdevice internal operations
pointers from the controlling resources structure with an additional
pointer added to the resources structure.
This allows for the removal of most of the probe-time control structures.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
line_num indicates the number of RDI - raw data interface channels which
are associated with a given IFE/VFE - image/video front end.
On several SoCs the RDI number is not static for each VFE - for example
on sm8250 VFE Lite has four RDIs where regular VFE has three.
Assigning line_num statically in the subdev_init() phase initialises
each VFE to the lower number, meaning in practical terms that we are
lobbing off one RDI on some VFEs.
Interrupt handling uses static for (i = RDI0; i < RDI2; i++) {} in some
of our VFE blocks but this can't work for situations where we have a
mixture of VFE @ 3 RDI and VFE-lite @ 4 RDI blocks.
First step to remediate is to pass line_num from a compat string
controlled data-structure and do so on a per-VFE basis.
Later patches will assign the correct number of RDI blocks per VFE.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
The following variables are being assigned statically based on
compatible strings in the probe path.
* enum camss_version version;
* unsigned int csiphy_num;
* unsigned int csid_num;
* unsigned int vfe_num;
* unsigned int vfe_lite_num;
* unsigned int vfe_total_num;
Migrate those variables to resource parameters passed in on platform
probe arguments. The one caveat is for VFE it has been necessary to
intoduce a new variable vfe_total_num to capture the aggregate value of
vfe_num + vfe_lite_num.
All the rest of the changes are rote camss->variable to
camss->res->variable with the parameter tables now populating the listed
variables.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
There is a lot of unnecessary if/elsing in this code that arguably
should never have made it upstream when adding a second let alone
subsequent SoC.
I'm guilty of not fixing the mess myself when adding in the sm8250.
Before adding in any new SoCs or resources lets take the time to cleanup
the resource passing.
First step is to pass the generic struct camss_resources as a parameter
per the compatible list.
Subsequent patches will address the other somewhat disparate strutures
which we are also doing if/else on and assigning statically.
Squashed down a commit to drop useless NULL assignment for ispif resources.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Rename non-specific struct resources {} to struct camss_subdev_resources {}
Each logical block in CAMSS has a number of regulators, clocks and resets
associated with it. We represent these blocks as v4l subdevices.
The name "struct camss_subdev_resources" is a more descriptive and accurate
name.
Suggested-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
There is no good reason to differentiate the two resource structures
here. As part of a general tidyup of the declaration and passing of
resources within in the CAMSS driver it will be advantageous to have
one unified resource structure.
The two structures are very similar anyway thus leading more credence
still to the argument there should be only one.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Platforms with qcom-camss use CSI-2, which means 1X16 is more correct
than 2X8. The fact that qcom-camss supported only 2X8 meant it was
incompatible with camera sensors it should have been compatible with.
For example, the ov5645 driver (correctly) reports that its format is
UYVY8_1X16 (after ba449bb562).
Since qcom-camss only supports 2X8, the qcom-camss system is
incompatible with the ov5645 driver, even though they should be
compatible.
This patch replaces all uses of UYVY8_2X8/VYUY8_2X8/YUYV8_2X8/YVYU8_2X8
with the equivalent 1X16 formats.
Signed-off-by: Martin Dørum <dorum@noisolation.com>
Acked-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
In the current driver csid Test Pattern Generator (TPG) doesn't work.
This change:
- fixes writing frame width and height values into CSID_TPG_DT_n_CFG_0
- fixes the shift by one between test_pattern control value and the
actual pattern.
- drops fixed VC of 0x0a which testing showed prohibited some test
patterns in the CSID to produce output.
So that TPG starts working, but with the below limitations:
- only test_pattern=9 works as it should
- test_pattern=8 and test_pattern=7 produce black frame (all zeroes)
- the rest of test_pattern's don't work (yavta doesn't get the data)
- regardless of the CFA pattern set by 'media-ctl -V' the actual pixel
order is always the same (RGGB for any RAW8 or RAW10P format in
4608x2592 resolution).
Tested with:
RAW10P format, VC0:
media-ctl -V '"msm_csid0":0[fmt:SRGGB10/4608x2592 field:none]'
media-ctl -V '"msm_vfe0_rdi0":0[fmt:SRGGB10/4608x2592 field:none]'
media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
v4l2-ctl -d /dev/v4l-subdev6 -c test_pattern=9
yavta -B capture-mplane --capture=3 -n 3 -f SRGGB10P -s 4608x2592 /dev/video0
RAW10P format, VC1:
media-ctl -V '"msm_csid0":2[fmt:SRGGB10/4608x2592 field:none]'
media-ctl -V '"msm_vfe0_rdi1":0[fmt:SRGGB10/4608x2592 field:none]'
media-ctl -l '"msm_csid0":2->"msm_vfe0_rdi1":0[1]'
v4l2-ctl -d /dev/v4l-subdev6 -c test_pattern=9
yavta -B capture-mplane --capture=3 -n 3 -f SRGGB10P -s 4608x2592 /dev/video1
RAW8 format, VC0:
media-ctl --reset
media-ctl -V '"msm_csid0":0[fmt:SRGGB8/4608x2592 field:none]'
media-ctl -V '"msm_vfe0_rdi0":0[fmt:SRGGB8/4608x2592 field:none]'
media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
yavta -B capture-mplane --capture=3 -n 3 -f SRGGB8 -s 4608x2592 /dev/video0
Fixes: eebe6d00e9 ("media: camss: Add support for CSID hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
vfe-480 is copied from vfe-17x and has the same racy idle timeout bug as in
17x.
Fix the vfe_disable_output() logic to no longer be racy and to conform
to the 17x way of quiescing and then resetting the VFE.
Fixes: 4edc8eae71 ("media: camss: Add initial support for VFE hardware version Titan 480")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
There are two problems with the current vfe_disable_output() routine.
Firstly we rightly use a spinlock to protect output->gen2.active_num
everywhere except for in the IDLE timeout path of vfe_disable_output().
Even if that is not racy "in practice" somehow it is by happenstance not
by design.
Secondly we do not get consistent behaviour from this routine. On
sc8280xp 50% of the time I get "VFE idle timeout - resetting". In this
case the subsequent capture will succeed. The other 50% of the time, we
don't hit the idle timeout, never do the VFE reset and subsequent
captures stall indefinitely.
Rewrite the vfe_disable_output() routine to
- Quiesce write masters with vfe_wm_stop()
- Set active_num = 0
remembering to hold the spinlock when we do so followed by
- Reset the VFE
Testing on sc8280xp and sdm845 shows this to be a valid fix.
Fixes: 7319cdf189 ("media: camss: Add support for VFE hardware version Titan 170")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Right now we never release the power-domains properly on the error path.
Add a routine to be reused for this purpose and appropriate jumps in
probe() to run that routine where necessary.
Fixes: 2f6f8af672 ("media: camss: Refactor VFE power domain toggling")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Previously the jump label err_cleanup was used higher in the probe()
function to release the async notifier however the async notifier
registration was moved later in the code rendering the previous four jumps
redundant.
Rename the label from err_cleanup to err_v4l2_device_unregister to capture
what the jump does.
Fixes: 51397a4ec75d ("media: qcom: Initialise V4L2 async notifier later")
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil: fix old name in commit log: err_v4l2_device_register -> err_v4l2_device_unregister]
We need to make sure camss_configure_pd() happens before
camss_register_entities() as the vfe_get() path relies on the pointer
provided by camss_configure_pd().
Fix the ordering sequence in probe to ensure the pointers vfe_get() demands
are present by the time camss_register_entities() runs.
In order to facilitate backporting to stable kernels I've moved the
configure_pd() call pretty early on the probe() function so that
irrespective of the existence of the old error handling jump labels this
patch should still apply to -next circa Aug 2023 to v5.13 inclusive.
Fixes: 2f6f8af672 ("media: camss: Refactor VFE power domain toggling")
Cc: stable@vger.kernel.org
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Set the v4l2_device already in async notifier init, so struct device
related to it will be available before the notifier is registered. This
requires separating notifier initialisation into two functions, one that
takes v4l2_device as its argument, v4l2_async_nf_init and
v4l2_async_subdev_nf_init, for sub-device notifiers. Registering the
notifier will use a single function, v4l2_async_nf_register.
This is done in order to make struct device available earlier, during
construction of the async connections, for sensible debug prints.
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Initialise V4L2 async notifier and parse DT for async sub-devices later,
just before registering the notifier. This way the device can be made
available to the V4L2 async framework from the notifier init time onwards.
A subsequent patch will add struct v4l2_device as an argument to
v4l2_async_nf_init().
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Rename v4l2_async_subdev as v4l2_async_connection, in order to
differentiate between the sub-devices and their connections: one
sub-device can have many connections but the V4L2 async framework has so
far allowed just a single one. Connections in this context will later
translate into either MC ancillary or data links.
This patch prepares changing that relation by changing existing users of
v4l2_async_subdev to switch to v4l2_async_connection. Async sub-devices
themselves will not be needed anymore
Additionally, __v4l2_async_nf_add_subdev() has been renamed
__v4l2_async_nf_add_connection().
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp
Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x
Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743
Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Linux 6.4-rc5
* tag 'v6.4-rc5': (919 commits)
Linux 6.4-rc5
leds: qcom-lpg: Fix PWM period limits
selftests/ftrace: Choose target function for filter test from samples
KVM: selftests: Add test for race in kvm_recalculate_apic_map()
KVM: x86: Bail from kvm_recalculate_phys_map() if x2APIC ID is out-of-bounds
KVM: x86: Account fastpath-only VM-Exits in vCPU stats
KVM: SVM: vNMI pending bit is V_NMI_PENDING_MASK not V_NMI_BLOCKING_MASK
KVM: x86/mmu: Grab memslot for correct address space in NX recovery worker
tpm, tpm_tis: correct tpm_tis_flags enumeration values
Revert "ext4: remove ac->ac_found > sbi->s_mb_min_to_scan dead check in ext4_mb_check_limits"
media: uvcvideo: Don't expose unsupported formats to userspace
media: v4l2-subdev: Fix missing kerneldoc for client_caps
media: staging: media: imx: initialize hs_settle to avoid warning
media: v4l2-mc: Drop subdev check in v4l2_create_fwnode_links_to_pad()
riscv: Implement missing huge_ptep_get
riscv: Fix huge_ptep_set_wrprotect when PTE is a NAPOT
module/decompress: Fix error checking on zstd decompression
fork, vhost: Use CLONE_THREAD to fix freezer/ps regression
dt-bindings: serial: 8250_omap: add rs485-rts-active-high
selinux: don't use make's grouped targets feature yet
...
In an earlier commit, setting the which field of the subdev format struct
in video_get_subdev_format was moved to a designated initializer that also
zeroes all other fields. However, the memset call that was zeroing the
fields earlier was left in place, causing the which field to be cleared
after being set in the initializer.
Remove the memset call from video_get_subdev_format to avoid clearing the
initialized which field.
Fixes: ecefa105cc ("media: Zero-initialize all structures passed to subdev pad operations")
Signed-off-by: Yassine Oudjana <y.oudjana@protonmail.com>
Acked-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Tested-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
From the experiments with camera sensors using SGRBG10_1X10/3280x2464 and
SRGGB10_1X10/3280x2464 formats, it becomes clear that on sdm845 and sm8250
VFE outputs the lines padded to a length multiple of 16 bytes. As in the
current driver the value of the bpl_alignment is set to 8 bytes, the frames
captured in formats with the bytes-per-line value being not a multiple of
16 get corrupted.
Set the bpl_alignment of the camss video output device to 16 for sdm845 and
sm8250 to fix that.
Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Tested-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Several drivers call subdev pad operations, passing structures that are
not fully zeroed. While the drivers initialize the fields they care
about explicitly, this results in reserved fields having uninitialized
values. Future kernel API changes that make use of those fields thus
risk breaking proper driver operation in ways that could be hard to
detect.
To avoid this, make the code more robust by zero-initializing all the
structures passed to subdev pad operation. Maintain a consistent coding
style by preferring designated initializers (which zero-initialize all
the fields that are not specified) over memset() where possible, and
make variable declarations local to inner scopes where applicable. One
notable exception to this rule is in the ipu3 driver, where a memset()
is needed as the structure is not a local variable but a function
parameter provided by the caller.
Not all fields of those structures can be initialized when declaring the
variables, as the values for those fields are computed later in the
code. Initialize the 'which' field in all cases, and other fields when
the variable declaration is so close to the v4l2_subdev_call() call that
it keeps all the context easily visible when reading the code, to avoid
hindering readability.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Shuah Khan <skhan@linuxfoundation.org> # For vimc
Reviewed-by: Lad Prabhakar <prabhakar.csengg@gmail.com> # For am437x
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de> # For drivers/staging/media/imx/
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Use the multistream series function video_device_pipeline_alloc_start
to allows multiple clients of the same pipeline.
If the VFE entity is used by another instance of the pipeline,
the pipeline won't be stopped. This allows for stopping and starting
streams at any point without disrupting the other running streams.
To prepare and start multiple virtual channels each CSID source pad
corresponding to a virtual channel must be linked to the corresponding
IFE entity. CSID pad 1 (1st source pad) corresponds to virtual
channel 0, CSID pad 2 corresponds to virtual channel 1 and so on.
Each of these must be linked to corresponding IFE RDI port.
E.g. to enable vc 0 on CSID0:
media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
To enable vc1 on CSID0:
media-ctl -l '"msm_csid0":2->"msm_vfe0_rdi1":0[1]'
And so on. Note that on SM8250 each CSID is connected, at the
hardware level, to only one IFE. Thus, you must link CSID0
with IFE0, you can't link it with IFE1.
Example: the following media controller setup expects multiplexed
sensor data on CSIPHY2. Data will be passed on to CSID0, which will
demux it to 2 streams - for RDI0 and RD1 ports of IFE0:
media-ctl -v -d /dev/media0 -V '"imx577 '22-001a'":0[fmt:SRGGB10/3840x2160 field:none]'
media-ctl -V '"msm_csiphy2":0[fmt:SRGGB10/3840x2160]'
media-ctl -V '"msm_csid0":0[fmt:SRGGB10/3840x2160]'
media-ctl -V '"msm_csid0":1[fmt:SRGGB10/3840x2160]'
media-ctl -V '"msm_csid0":2[fmt:SRGGB10/3840x2160]'
media-ctl -V '"msm_vfe0_rdi0":0[fmt:SRGGB10/3840x2160]'
media-ctl -V '"msm_vfe0_rdi1":0[fmt:SRGGB10/3840x2160]'
media-ctl -l '"msm_csiphy2":1->"msm_csid0":0[1]'
media-ctl -l '"msm_csid0":1->"msm_vfe0_rdi0":0[1]'
media-ctl -l '"msm_csid0":2->"msm_vfe0_rdi1":0[1]'
Note: CSID's entity pad 0 is a sink pad, pads 1..4 are source pads
To start streaming a v4l2 client must open the corresponding
/dev/videoN node. For example, with yavta:
yavta -B capture-mplane -c -I -n 5 -f SRGGB10P -s 3840x2160 -F /dev/video0
yavta -B capture-mplane -c -I -n 5 -f SRGGB10P -s 3840x2160 -F /dev/video1
Note that IFEs (vfe0, vfe1) on SM8250 have 3 RDI ports and a single
PIX port and IFELites (vfe2, vfe3) have 4 RDI ports and no PIX port.
Signed-off-by: Milen Mitkov <quic_mmitkov@quicinc.com>
Reviewed-by: Robert Foss <robert.foss@linaro.org>
Tested-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Robert Foss <robert.foss@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
For multiple virtual channels support, each VFE line can be in either
ON, RESERVED or OFF states. This allows the starting and stopping
of a VFE line independently of other active VFE lines (e.g. already-
running lines stay in ON state, and newly-added lines are RESERVED)
Also, link the CSID entity's source ports to corresponding VFE lines.
Signed-off-by: Milen Mitkov <quic_mmitkov@quicinc.com>
Reviewed-by: Robert Foss <robert.foss@linaro.org>
Tested-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Robert Foss <robert.foss@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
CSID hardware on SM8250 can demux up to 4 simultaneous streams
based on virtual channel (vc) or datatype (dt).
The CSID subdevice entity now has 4 source ports that can be
enabled/disabled and thus can control which virtual channels
are enabled. Datatype demuxing not tested.
In order to keep a valid internal state of the subdevice,
implicit format propagation from the sink to the source pads
has been preserved. However, the format on each source pad
can be different and in that case it must be configured explicitly.
CSID's s_stream is called when any stream is started or stopped.
It will call configure_streams() that will rewrite IRQ settings to HW.
When multiple streams are running simultaneously there is an issue
when writing IRQ settings for one stream while another is still
running, thus avoid re-writing settings if they were not changed
in link setup, or by fully powering off the CSID hardware.
Signed-off-by: Milen Mitkov <quic_mmitkov@quicinc.com>
Reviewed-by: Robert Foss <robert.foss@linaro.org>
Tested-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Acked-by: Robert Foss <robert.foss@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
