11:14 <amitk> Hi all, I'm trying to get wifi working on the Exynos 5800-based Chromebook2 (http://linux-exynos.org/wiki/Samsung_Chromebook_2_XE503C32) with a mainline kernel. Compiling in the mwifiex sdio driver and copying the sd8897 firmware in /lib/firmware doesn't do the trick.

11:15 <amitk> I suspect some DT changes might be required too? (similar to https://github.com/linux-exynos/linux/commit/9d0094739f89ff8cb0a6f801c15ddafd4eb687ab)

11:15 <amitk> can anybody confirm wifi works with mainline kernel on this chromebook?

11:15 <Wizzup> I am more interesting in the other parts

11:16 <Wizzup> you got mainline to work (regardless of wifi?)

11:16 <Wizzup> I still gives me a black screen on boot

11:16 <Wizzup> There are several different firmwares

11:16 <Wizzup> Swabbles: we really need to remove or update or github repo

11:16 <Wizzup> s/or/our/2

11:17 <amitk> Wizzup: yes

11:17 <amitk> well, almost, with some patches

11:18 <amitk> we're maintaining and using this branch: https://git.kernel.org/cgit/linux/kernel/git/khilman/linux.git/log/?h=wip/exynos/v3.17-rc3-integ

11:18 <amitk> we'll rebase onto 3.18 soon

11:19 <Wizzup> nice!

11:19 <Wizzup> Too bad that the initial idea of the channel -- to centralise some efforts hasn't worked out too much, but I guess that's also because of a huge lack of time :)

11:19 <Wizzup> I'll try that branch in a few days :)

11:20 <amitk> Here are the instructions to prepare the sd card, but I think your wikis are a bit better: https://wiki.linaro.org/WorkingGroups/PowerManagement/Doc/Chromebook2_boot

11:20 <javier__> most of those patches already made to mainline though

11:20 <javier__> Wizzup: for exynos 5420/5800 you should only need https://git.kernel.org/cgit/linux/kernel/git/khilman/linux.git/commit/?h=wip/exynos/v3.17-rc3-integ&id=039e36380b51c4a76c6f1a5fac577d092bbf2e1e afaict

11:20 <javier__> Wizzup: I meant for 5422/5800

11:21 <javier__> and disable CONFIG_FB_SIMPLE if you have a u-boot that injects the simplefb device node

11:21 <Wizzup> I think I did that last time but it didn't work yet

11:21 <Wizzup> I think I applied that

11:23 <javier__> amitk: I haven't tested wifi myself but yes I believe the DTS snippet you shared are needed too

11:23 <javier__> amitk: also you need to grab the max77802 en32khz_cp clock or boot with clk_ignore_unused

11:24 <javier__> amitk: since in the patch you posted it mentions "vin-supply = <&en32khz_cp>" but that is because the chromeos kernel didn't have a proper clock driver for the max77802 clocks

11:28 <javier__> so that DTS needs some mangling, I've testing wifi in my (very long) TODO list

11:31 <amitk> javier__: hmm, let me check if there is already DT bindings for the max77802 clocks

11:33 <javier__> amitk: yes, there is in Documentation/devicetree/bindings/clock/maxim,max77802.txt

11:34 <javier__> amitk: MAX77802_CLK_32K_AP is used as the source clock of the s3c RTC in arch/arm/boot/dts/exynos5800-peach-pi.dts

11:35 <javier__> MAX77802_CLK_32K_CP is used by the wifi chip but since there is no DTS snippet for that is of course unused

11:39 <javier__> amitk: and you also need to enable CONFIG_COMMON_CLK_MAX77802=y which is not enabled in 3.18

11:40 <javier__> amitk: sorry if I'm telling you obvious things but just saying what I believe may save you some time :)

11:42 <amitk> javier__: no, this is very useful. My main focus is to work on the big.LITTLE chip on the device, so getting the device enablement out of the way quickly is a lot of help.

11:45 <javier__> amitk: great, let me know if you find any issues since I've been hacking on the Exynos5420 version of the chromebook2 (the 11" one) so I may help

11:46 <amitk> javier__: did we talk before on #linaro? about not being able to get all 8 cores booting on the 5420?

11:46 <jnettlet> amitk, if you are looking at a commit in github go up to the url location and add .patch to the end. That will then download the commit as a standard patch that will work with git am

11:47 <amitk> jnettlet: nifty, I ended up cloning the tree though :)

11:47 <jnettlet> amitk, okay. good to know for future endeavors

11:48 <amitk> jnettlet: indeed

11:48 <javier__> amitk: yes we did. I disabled the b.L switcher and I got the 8 cores working on my 5420 chromebook, I know is not working on arndale octa board

11:48 <javier__> amitk: olof said that it may work but if you start stressing you may find issues. I don't know what is the problem though

11:49 <javier__> but afaiu whatever the stability problem is, it was fixed in the Exynos5422

11:49 <amitk> javier__: it is a HW bug

11:49 <jnettlet> you are bringing the cores on and off individually or just have all of them on all the time?

11:50 <amitk> and yes, it was fixed in 5422/5800

11:50 <amitk> jnettlet: all 8 active at the same time

11:50 <jnettlet> what was doing the scheduling?

11:50 <javier__> amitk: yeah, I know is a HW bug but I meant that I don't know what the bug really is and why is working on the chromebook but not in arndale octa

11:51 <javier__> seems that some bootloader magic is needed but I also don't know what that is :)

11:51 <amitk> jnettlet: the scheduler is what we're trying to fix upstream

11:54 <jnettlet> does the 13" chromebook with the 5800 in it have all 8 cores enabled? Or any big.LITTLE switching for that matter?

11:55 <amitk> jnettlet: the chromeos kernel has cluster switching enabled

11:55 <jnettlet> amitk, for the 11" ?

11:55 <amitk> jnettlet: I'm using mainline with all 8 cores enabled and changes to the scheduler

11:56 <amitk> jnettlet: the 11" has the 5420, right?

11:56 <jnettlet> yes

11:56 <sjoerd> 13" on mainline boots all cores properly

11:56 <amitk> jnettlet: that one uses cluster switching as well, and getting all cores working is a problem on that one

11:56 <sjoerd> which is a 5422.. But there are some folks having issues owth the odroid XU3 to get all cores going

11:56 <amitk> its not stable

11:57 <amitk> sjoerd: yeah, though there seems to be some solutions for the 6/8 cores bug now

11:57 <sjoerd> Right, some writes to magical registers :/

11:57 <jnettlet> I will need to look at their kernel code then. Nothing in linux reports anything other the big cores being used.

11:58 <javier__> jnettlet: do you have the 11" (Exynos 5420 Peach Pit)?

11:58 <jnettlet> javier__, yep

11:58 <javier__> jnettlet: peach pit boots with 8 cores but there are reports that has stability issues

11:58 <javier__> I didn't have any issues but didn't stress all the cores either

11:59 <jnettlet> I need to get some time to test that out.

11:59 <jnettlet> been too busy with other hardware.

11:59 <javier__> jnettlet: -next with CONFIG_BL_SWITCHER disabled should be enough

11:59 <sjoerd> javier__: Is that due to the little cores not getting a high enough voltage like what was going on the 13"

12:00 <jnettlet> javier__, did you see a major difference in battery usage?

12:01 <javier__> sjoerd: I don't really know, seems to be a hw bug in 5420 that got fixed in 5422/5800 but don't have more details

12:01 <javier__> sjoerd: maybe amitk knows more

12:01 <sjoerd> ah different issue then

12:02 <javier__> jnettlet: I didn't measure consumption, sorry

12:02 <sjoerd> jnettlet: last time i played with these things mainline didn't do voltage scaling yet for the cpus, so power usage was high anyway

12:03 <javier__> sjoerd: oh, I thought you could unbrick that one :(

12:03 <amitk> javier__: sjoerd: SoC bug related to CCI

12:03 <sjoerd> javier__: i might try again, but i doubt it

12:04 <sjoerd> javier__: I don't even see power on the debug pads where it should have reference volgae & the power led just turns of right away

12:05 <javier__> sjoerd: I see, so seems to be worse than we thought...

12:06 <sjoerd> well i can't really work out what went wrong which is a tad annoying

12:06 <sjoerd> my best theory is that i accidentally screwed up the ec ro firmware

12:08 <amitk> jnettlet: that was supposed to be the case, yet

12:08 <amitk> *yes