* Kernel source migration from git.ti.com to omapzoom.org. The development
tree is shared between the Linux and Android teams.This enables a common
kernel development platform for both the teams for better alignment.
* Add support for VIP and VPE
* Add wilink/wlan support
* Enable rpmsg/remoteproc driver for multimedia
Signed-off-by: Mrinmayee Hingolikar <mrinmayee@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Now the DRA7xx support is fully merged into upstream u-boot
then we can remove setting the PREFERRED_PROVIDER for
u-boot and virtual/bootloader to u-boot-dra7xx and instead
pick up the default setting from the omap-a15.inc file.
Signed-off-by: Chase Maupin <Chase.Maupin@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Update the U-Boot LICENSE md5sum for the new release.
* Update the PV for the new U-Boot release
* Add the SRCREV that points to the 2013.10 release
* Use the ti-u-boot-2013.10 branch
Signed-off-by: Chase Maupin <Chase.Maupin@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Eliminate need for BBMASK - remove images requiring systemd/angstrom dependency
* Move ti-test packagegroup to proper location and update from latest Arago
* Move BoneScript and Bone GSG recipes to recipes-ti/beagleboard
* README in recipes-ti warns about unsupported nature of those packages
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* For am335x and am437x, the kernel and u-boot expects the ubifs volume name
to be set to "rootfs".
* Set UBI_VOLNAME variable to "rootfs" so that the ubifs files generated by OE
will work.
Signed-off-by: Franklin S. Cooper Jr <fcooper@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Use SOC family value for am437x instead of machine name for DTB files.
* This mimics the current approach taken for am335x.
Signed-off-by: Franklin S. Cooper Jr <fcooper@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
To be used with latest u-boot that prefers zImage over uImage
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
* The PM for TI parts was merged in
* Platform and connectivity got updated
* Since we now have separate DTB for BeagleBone Black, install it too
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Set up the run-time dependency on the package with devicetree files to be
installed in /boot directory of the rootfs. The new U-boot prefers to load
them from /boot directory of the rootfs instead of the separate boot partition.
Also, there are some specific requirements for naming DTB files in /boot
directory (like no extra prefixes), but that should be addressed separately.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Bump the SRCREV for 2013.10-rc4 pre-release
* The Licenses/README file has been changed with new licenses added, but the
general license hasn't changed from GPLv2, so re-hash the file
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Cc: Tom Rini <trini@ti.com>
Essentially, machine_kernel_pr.bbclass does this one line assignment, but
introduces a new unnecessary dependency on meta-openembedded/meta-oe layer.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Many machines are supported by newer 3.x kernel recipes
* Few old platforms are still supported by 2.6.37 recipes
* Some platforms are not supported in meta-ti, like Crane, Touchbook, Overo
* For any platform still requiring this outdated kernel, the recipe can be picked
up from Danny or Denzil
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Hawkboard platform is not supported in meta-ti.
* Since there no new revs of the kernel, the old recipe can be picked up from
Danny or Denzil.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Franklin Cooper Jr. <fcooper@ti.com>
Beagleboard is supported by newer 3.2 and 3.3.7 kernel recipes.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Franklin Cooper Jr. <fcooper@ti.com>
This adds a new mandatory dependency on meta-openembedded layer. While we are
still trying to resolve an existing optional dependency, not bring a new one.
The MACHINE_KERNEL_PR requirement will need to be handled differently.
This reverts commit 1450f19380.
* MACHINE_KERNEL_PR was originally defined in the kernel class located in
meta-oe which overrode the version of the class in oe-core.
* Some time ago kernel.bbclass was removed from meta-oe and a separate class
machine_kernel_pr was created to add support for MACHINE_KERNEL_PR.
* Since this variable is used significantly in both meta-ti and meta-arago make
sure all the kernel recipes directly or indirectly inherit this new class.
Signed-off-by: Franklin S. Cooper Jr <fcooper@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
We rely on kernel-base having run-time dependency on kernel-image, which
normally would pull uImage/zImage into /boot directory of rootfs.
Thanks to Chase for pointing me in the right direction by noticing that our
addition of cm3 firmware breaks that dependency for ti33x.
The reason for that is RDEPENDS is being conditionally assigned with ?=
in kernel.bbclass and using += with machine-override clears that value due
to the order of processing. Hence _append should be used instead of += to
process it after the machine-override takes place.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
Acked-by: Franklin Cooper Jr. <fcooper@ti.com>
* To be used with latest u-boot, that prefers zImage over uImage
* ti33x will follow as soon as u-boot changes its default for that platform
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Cc: Tom Rini <trini@ti.com>
* DTB for OMAP5 sEVM was removed upstream, drop it from the recipe
* Don't mask the default preference to be picked up normally by supported machines
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Franklin Cooper Jr. <fcooper@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
Some kernel versions (and corresponding u-boot) may still need to set uImage,
as preferred image type, since the machine config changes preference to zImage
for newer kernels - keep that for ti33x and omap-a15 stuck on 3.2 and 3.8.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Franklin Cooper Jr. <fcooper@ti.com>
Checksums are handled as var attributes and require immediate expansion,
so use := operation as in the original patches, that got lost in the update.
Otherwise do_fetch will fail to validate the checksums.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
This is based on the work by Franklin Cooper, with following changes:
* Cleanup and simplify the logic
* Reduce the number of Python calls
* Expand support from Raw to X11 variant
The original description:
* The Graphics SDK includes some binaries that are prebuilt.
* These prebuilt binaries come in two versions of the Graphics SDK. By default
the prebuilt binaries are built using a softfp toolchain. A separate hardfp
Graphics SDK exist that contains binaries that are prebuilt with a hardfp
toolchain.
* Add support for both by checking to see if the toolchain used in the OE build
is using a hardfp or softfp toolchain.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Cc: Franklin S. Cooper Jr <fcooper@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
This is based on the work by Franklin Cooper, with following changes:
* Cleanup and simplify the logic
* Reduce the number of Python calls
* Expand support from Raw to X11 variant
The original description:
* The Graphics SDK includes some binaries that are prebuilt.
* These prebuilt binaries come in two versions of the Graphics SDK. By default
the prebuilt binaries are built using a softfp toolchain. A separate hardfp
Graphics SDK exist that contains binaries that are prebuilt with a hardfp
toolchain.
* Add support for both by checking to see if the toolchain used in the OE build
is using a hardfp or softfp toolchain.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Cc: Franklin S. Cooper Jr <fcooper@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
The new method is to automatically package modules into kernel-module-<modname>
and make the custom name depend on the standard unified one. Messing with
FILES_${PN} usually causes problems, so avoid it and let module.bbclass do
it's magic.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
The new method is to automatically package modules into kernel-module-<modname>
and make the custom name depend on the standard unified one. Messing with
FILES_${PN} usually causes problems, so avoid it and let module.bbclass do
it's magic.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
* Do not hardcode kernel source location and name.
* Fix some discrepancies between GFX SDK versions as related to DRM.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Acked-by: Chase Maupin <chase.maupin@ti.com>
Use properly FILES_${PN}-dev to avoid warning about replacing value with default.
Signed-off-by: Valentin Cobelea <valentin.cobelea@enea.com>
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Mrinmayee Hingolikar
Denys Dmytriyenko
Chase Maupin
Franklin S. Cooper Jr
Valentin Cobelea