The uboot config for am65 is moving to a merged build for both GP and
HS. Until the yocto configs catch up, we need to keep am65xx-hs-evm
working.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Update the QA checks done on the build artifacts to include [buildpaths] since
the PRU linker includes these paths in its output. This avoids build warnings
or errors of the type:
pru-icss-6.2.0-r1 do_package_qa: QA Issue: File /lib/firmware/pru/PRU_Halt.out in package pru-icss-halt contains reference to TMPDIR [buildpaths]
Signed-off-by: Trevor Woerner <twoerner@gmail.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Sometimes on am335x devices, no more output is printed after
"Starting kernel..." message. Modify UBOOT_EXTLINUX_KERNEL_ARGS
to enable the console before standard serial driver is probed.
Signed-off-by: Julien Panis <jpanis@baylibre.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Previously k3 mcdepends for k3r5 artifacts has been changed from do_image_wic
to support non-wic builds in other layers and do_image_complete was believed
to be sufficient. Switching k3r5 to baremetal builds has exposed another race
when some pieces come from sstate and hence requires using do_image instead.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
The latest kernel has made the change to add vendor subdirectories into
the arch/arm/boot/dts directory. This effectively breaks the
KERNEL_DEVICETREE settings in the machine configs for 32bit platforms,
so switch them over to using the PREFIX variable. Also change the
linux-ti-next to use the PREFIX setting for KERNEL_DEVICETREE.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
With the change to the vendored arch/arm/boot/dts directory, we need to
add support to the KERNEL_DEVICETREE_DTBMERGE to make it a search
pattern instead of a fixed list.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Add in missing *-sec links pointing to signed files especially for
j784s4. Reformat file a little to make the groupings easier to
manage in the future when looking to see if -sec links are missing.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
The following MACHINE_FEATURES are not matched against anywhere, remove:
* kernel26
* ethernet
* mmc
Signed-off-by: Andrew Davis <afd@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
commit 5f9be22839 ("machine: Add AM62Q HS-SE evm configuration")
has added am62xx-lp.inc which is used by am62xx-lp-evm machine.
The features listed in am62xx.inc are common between machines
am62xx-evm and am62xx-lp-evm. (Ex: MACHINE_FEATURES = "gpu")
Hence updating the necessary include file to have
common features to be enabled between these two platform variants.
Signed-off-by: Praneeth Bajjuri <praneeth@ti.com>
Suggested-by: Randolph Sapp <rs@ti.com>
Reviewed-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Bump the SRCREV for the umlibs and driver package to include some fixes
for Jacinto devices. These patches do the following:
- Fix virtual memory addressing
- Resolve an issue with APM and debug requests
- Attempt to resolve most of the cache coherency related issues on
Jacinto devices
- Introduce some extra cache maintenance operations in the kernel
module
It also introduces the following known issues:
- Visual artifacting on the second-to-last EGL/GLES context on BXS
based devices (J721S2/J784S4)
We're still working on that last one, but the stability improvements
make this worth while for now.
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
The address 0x80008000 was used in pre-K3 devices. ARM64 requires the
kernel loaded to a 2MB aligned address when not using KASLR. For FIT
set the address to 0x81000000 which gives the kernel 32MB before running
into the FDT at 0x83000000.
If you get this error someday then your kernel is too large and either it
or FDT needs relocated in memory:
ERROR: image is not a fdt - must RESET the board to recover.
Signed-off-by: Andrew Davis <afd@ti.com>
Tested-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Upgrade to v6.2.0 version.
This release supports Linux kernel 6.1.
The RPMsg library in previous versions of the PSSP will not work with
the PRU RPMsg drivers in Linux kernel 6.1.
Other changes in this version:
* Bug fixes:
- Getting Started Labs, fix "zero" command clearing too many bytes
- Include files, add a missing parentheses to pru_cfg, pru_ctrl, pru_ecap
* Update ReadMe files to point to the latest URLs
Signed-off-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Update ti-rpmsg-char library to fix wrong rpmsg-ctrl
device ids observed when multiple remote procs are present
with multiple channels opened.
Signed-off-by: Hari Nagalla <hnagalla@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
The following patch was dropped from the SRC_URI one of my previous
graphics overhaul patches:
0001-compiler-support-OpenEmbedded-nodistro-internal-aarc.patch
Actually remove it from the source tree now that upstream is carrying
the equivalent to help enable other tools.
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Upstream changed to not be so picky about compilers. Bump SRCREV so the
odd things using our recipes are happy again.
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Edit the file paths so they point at files in the source directory
explicitly. For some reason shell glob were previously only respected by
some machines.
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
* rpmsg_client_sample is a test module and should not be autoloaded
* Both k3_r5 and k3_dsp remoteproc modules depend on virtio_rpmsg_bus
Cc: Denys Dmytriyenko <denys@ti.com>
Cc: Ryan Eatmon <reatmon@ti.com>
Cc: Praneeth Bajjuri <praneeth@ti.com>
Cc: Apurva Nandan <a-nandan@ti.com>
Cc: Hari Nagalla <hnagalla@ti.com>
Signed-off-by: Udit Kumar <u-kumar1@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Use RRECOMMENDS for components. Apparently I made an resolvable
dependency cycle that Yocto silently clipped. My best guess at the
culprit was: mesa-megadriver (RR) -> rogue-km (RR) -> rogue-um (RD) ->
libgles1-rogue (RD) -> mesa-megadriver
That killed the ti-img-rogue-umlibs recipe during do_rootfs. The rogue
GLES libs do depend on mesa-megadriver at runtime but I guess the UM
components should only really recommend the rogue GLES components be
present at runtime to make the cycle resolvable by opkg.
Fixes:
1705f6d3 (ti-img-rogue-umlibs: make all components optional, 2023-06-21)
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Ryan Eatmon
LCPD Automation Script
Trevor Woerner
Julien Panis
Denys Dmytriyenko
Andrew Davis
Kamlesh Gurudasani
Praneeth Bajjuri
Randolph Sapp
Hari Nagalla
Aniket Limaye
Udit Kumar