The demo extension is overridable so the new install behavior needs to
use the variable.
Fixes#2056.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
Add missing include file. Otherwise with latest master we see:
| ../git/vpu_wrapper_hantro_encoder.c:953:3: error: implicit declaration of function 'gettimeofday' [-Wimplicit-function-declaration]
| 953 | gettimeofday (&pObj->tvEnd, NULL);
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Bump revision to 1.2.5 and remove old patches. This also switches to new
upstream location at github. The old location uses git-protocol which can
cause problems in corporate network environments.
Signed-off-by: Markus Niebel <Markus.Niebel@ew.tq-group.com>
Make tuningext a packageconfig as it no longer builds with latest
master (walnascar):
isp-imx contains the 3rd party cpp-netlib package. Compiling that
package against boost 1.87 results in build errors.
boost 1.87 is provided with latest master (walnascar).
Compare also with meta-oe commit a46de2745b30 ("cpp-netlib:
Skip recipe, not buildable with boost >= 1.87")
Drop patch. Fix now present in package sources.
Add jsoncpp to DEPENDS as the package no longer compiles its own
version thereoff and dynamically links against the OE provided one.
Sync CMake options with upstream recipe.
The binary blobs provided now link against libtinyxml2.so.10.
Drop the fix.
The binary blobs provided still link against libjsoncpp.so.25, but OE
provides libjsoncpp.so.26. Fixup the binaries.
The package now provides a Makefile with a install task. Use that.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
This pulls in all other dev packages. It looks like the unversioned
shared object is not (no longer?) needed, i.e. the driver points to
the versioned object.
| readelf --dynamic opt/imx8-isp/bin/daA3840_30mc.drv
| ...
| 0x000000000000000e (SONAME) Library soname: [libdaA3840_30mc.so.1]
(/usr/lib/libdaA3840_30mc.so is the only file in the -dev package.)
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
mx8mq uses 0x30860000 as the base address for boot uart base.
Signed-off-by: Ricardo Salveti <ricardo@foundries.io>
(cherry picked from commit b6a666659c)
Some hardware vendors provide a customized Optional Executable Image
(OEI) with changes specific to their boards (e.g. Toradex SMARC i.MX95).
Add a virtual provider to avoid conflicts between the OEI recipes from
other layers. This follows the pattern used for the related
imx-system-manager.
Signed-off-by: Hiago De Franco <hiago.franco@toradex.com>
(cherry picked from commit 7cf349d287)
The current COMPATIBLE_MACHINE is set to build every i.MX SoC with both
BSP variants. However the firmware-nxp-wifi deploys firmware related to
the NXP Wi-Fi Modules (e.g. Maya IW416), which can be used with other
SoCs as well, as example the Toradex Verdin-AM62 wich uses the TI SoC
and the NXP Wi-Fi Module.
With the current COMPATIBLE_MACHINE it is not possible to build this
recipe and deploy the NXP firmware with different SoCs, therefore remove
it.
Signed-off-by: Hiago De Franco <hiago.franco@toradex.com>
(cherry picked from commit 8e7ebc096b)
The upstream kernel Wi-Fi driver expects the sdiouart8997_combo_v4 and
sdiouartiw416_combo_v0 binaries to be located under firmware/mrvl [1],
different from the out-of-tree driver, which places them under
firmware/nxp. Create symbolic links to deploy these binaries in the
correct location for the upstream driver.
Due to this change, bump the recipe version to 1.1.
[1] https://github.com/torvalds/linux/blob/master/drivers/net/wireless/marvell/mwifiex/sdio.h#L31
Signed-off-by: Hiago De Franco <hiago.franco@toradex.com>
(cherry picked from commit deebc27a3a)
Using --size in the imx-imx-boot-bootpart.wks.in to set size of /boot
partition does not result in a 256MiB partition as expected but in a
~332MiB one. Indeed, --size ensures --extra-space (default 10MiB) is
added to actual data size and furthermore applies --overhead-factor
(default 1.3) to the wanted size. This is exactly the case here as 256
* 1.3 = 332.8.
Use --fixed-size instead to force the /boot partition to be exactly
256MiB just as it is indicated in the comments of this file.
Consequently, the `do_image_wic` task will also fail if data copied in
this partition are too large at creation.
Signed-off-by: Jonathan GUILLOT <jonathan@joggee.fr>
(cherry picked from commit 0ee7c736c7)
U-Boot fork has been updated to include tagged version v2025.01 from
upstream.
Following commits were ported from 2024.07+fslc to this version:
----
aa4bc52d08 ("imx: iMX6: Define disconnect_from_pc function")
a6f371cad5 ("MLK-21854-2 imx: uuu: Update autoboot to run uuu or fastboot")
bd90432d6c ("pico-imx6ul, pico-imx7d: Explain how to use eMMC user partition")
49f4fa81f9 ("mx6sabre_common: Dynamically change the video output")
9f3a98873d ("mx6sabresd: Enable video interfaces in bootargs")
----
Signed-off-by: Andrey Zhizhikin <andrey.z@gmail.com>
(cherry picked from commit 26bccf046e)
* vpu: Add more options, flags, and frameskipping support to encoder API
* vpu: Deprecate min_intra_refresh_mb_count and replace it with flag
This is more practical, since the macroblock count if anyway usually set
to resemble the GOP length, so this might as well be automatically
calculated.
* vpu: encoder: Add GOP size checks
* vpu: encoder: Add imx_vpu_api_enc_get_encoded_frame_ext()
This is a preparation for reworked intra refresh support. In intra
refresh encoding, only the first frame is an I/IDR frame; all followup
frames are P frames. The start of an intra refresh interval qualifies
as a sync point. But since only P frames are present, callers callers
can't rely on the frame type to detect sync points. To fix this, the
imx_vpu_api_enc_get_encoded_frame_ext() is added to the API to be able
to communicate to callers that a sync point was produced.
* imx8m-hantro: Rewrite encoder to directly use the Hantro H1 API
This allows for more fine grained control of the H1 encoder.
Most notably, intra refresh works properly, and skipped frames
are correctly handled.
* imx8m-hantro: encoder: Add closed GOP interval support
* imx8m-hantro: encoder: Rework intra refresh
Use codec specific features for intra refresh. For h.264, use the Hantro
GDR (Gradual Decoder Refresh) feature, since it also takes care of
filling SEI message and SPS/PPS NALUs appropriately. For VP8, implement
a manual method. (Intra refresh is not officially supported in VP8
though. See the note about VP8 intra refresh in the comment blocks in
this commit.)
* imx6-coda: return IMX_VPU_API_ENC_RETURN_CODE_INVALID_CALL from
imx_vpu_api_enc_get_skipped_frame_info()
This keeps the imx6 backend consistent with the API. Previously, that
backend's implementation of that function was not returning anything at
all.
* imx8m-hantro: implement imx_vpu_api_enc_get_skipped_frame_info() stubs
for all encoders
Implement this function for all encoders to fix a linker error in
gstreamer-imx.
These encoders never return IMX_VPU_API_ENC_OUTPUT_CODE_FRAME_SKIPPED,
so we always return IMX_VPU_API_ENC_RETURN_CODE_INVALID_CALL from
imx_vpu_api_enc_get_skipped_frame_info().
* imx8m-hantro: dummy_encoder: add imx_vpu_api_enc_get_encoded_frame_ext()
Linking gstreamer-imx will fail if imx_vpu_api_enc_get_encoded_frame_ext)
is not available.
Signed-off-by: Carlos Rafael Giani <crg7475@mailbox.org>
(cherry picked from commit 71bef353dc)
The recipe is missing PROVIDES virtual/libgles3 for i.MX 6 and 7 GPU.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com>
(cherry picked from commit ed203f9d54)
Port some patches from 6.6-2.1.x to 6.6-2.2.x and update the recipe
documentation to include them.
Signed-off-by: Ernest Van Hoecke <ernest.vanhoecke@toradex.com>
(cherry picked from commit b50af8befa)
It allows a correct kernel load address when using FIT image with
i.MX93. The value 0x80400000 has been retrieved from
imx93_11x11_evk_defconfig in uboot-imx.
Signed-off-by: Jonathan GUILLOT <jonathan@joggee.fr>
(cherry picked from commit 1285232175)
