On Fri, Jan 20, 2017 at 5:06 PM, Christopher Covington <cov@xxxxxxxxxxxxxx> wrote: > Hi Robert, > > On 01/18/2017 09:02 PM, Robert Moskowitz wrote: >> There is a local (Detroit) IoT meeting showcasing the Dragonboard 410c. >> >> What is the status of support on it now? >> >> Want to know to what degree I try to go to the meeting... > > It looks like the Dragonboard 410c has an APQ8016 chip, which is basically > the MSM8916 with no modem, so broadly equivalent to applications processor > software. > > There has been a lot of work to get those and other Qualcomm Technologies > chips supported in the upstream Linux kernel, but some of it has only > been finished recently, for example with display support for 8016 landing > in 4.9: > > http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=bd3af15a4e4d528bb0a0eea1daca0b818baa9fd8 > > A quick look at Fedora kernel configuration options makes me worried that > the Fedora 25 kernel would not work on the DB410c out of the box. > > # CONFIG_MSM_GCC_8916 is not set > http://pkgs.fedoraproject.org/cgit/rpms/kernel.git/tree/baseconfig/arm/arm64/CONFIG_MSM_GCC_8916 > > GCC is the Global Clock Controller in this context. I have no idea how > it works. It sounds important, but perhaps it's only required for > run-time changes to the clocks, which may or may not be needed for basic > functionality. Basic functionality may be available by coasting along on > the clock setup from firmware. If somebody like Stephen or Andy who knows > this stuff wants to tell me what to change in the configuration I'd be > happy to prepare the patch, although I only have 8074 and 8064 boards > and not 8016 myself for testing. > > One of the barriers to distribution support of these Q mobile devices has > been the firmware/bootloader. The devices come with Little Kernel (LK) as > the primary bootloader. I've heard that it requires special lines in the > device tree file, but since these aren't useful to Linux, they aren't > allowed to be kept in the device tree source on kernel.org. So if you want > to use upstream device tree files, you need to run a tool to insert the > special bootloader-specific lines before or while creating the device tree > blob. > > https://source.codeaurora.org/quic/la/device/qcom/common/tree/dtbtool/dtbtool.txt?h=l_master > https://source.codeaurora.org/quic/kernel/skales/tree/dtbTool > > Once that's complete, the kernel, device-tree-with-bootloader-extras, and > initramfs must be packaged into the Android fastboot/bootimg binary image > format: > > https://android.googlesource.com/platform/system/core/+/master/mkbootimg/bootimg.h#66 > https://source.codeaurora.org/quic/kernel/skales/tree/mkbootimg > > Fedora doesn't support this format at the moment, and I think some or most > developers favor putting a familiar bootloader in the bootimg as a > compatibility shim. As far as I know, that's yet another unfinished project. > > If anyone knows better, please correct me if I got any of this wrong, but > hopefully this is a useful starting point. That's pretty much correct. The kernel config was a best guess based on other people's configs from me. I've not tested it because I've not had the time to smash my head against the bootloader garbage so patches for the kernel stuff are welcome. There's also the little problem of it requiring non redistributable binary blobs. This was suppose to have been fixed, or at least improved, but the last time I went to get them there was still a click through legal agreement so I left and moved on to something else. Peter _______________________________________________ arm mailing list -- arm@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to arm-leave@xxxxxxxxxxxxxxxxxxxxxxx
