Tenkawa has joined #linux-exynos
<Tenkawa> Hey all
<Tenkawa> anything new?
<Wizzup> at this time of the night - no :) going to bed
<Wizzup> you can check the logs though
<Tenkawa> heheh
prahal has quit [Quit: prahal]
Tenkawa has quit [Quit: leaving]
dlezcano has quit [Ping timeout: 250 seconds]
amitk has joined #linux-exynos
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 252 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 252 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 240 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 246 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 244 seconds]
dlezcano has joined #linux-exynos
ssvb has quit [Quit: Leaving]
dlezcano has quit [Ping timeout: 245 seconds]
dlezcano has joined #linux-exynos
aballier has joined #linux-exynos
dlezcano has quit [Ping timeout: 245 seconds]
dlezcano has joined #linux-exynos
<Wizzup> nice!
dlezcano has quit [Ping timeout: 252 seconds]
dlezcano has joined #linux-exynos
gladiac has joined #linux-exynos
dlezcano has quit [Ping timeout: 265 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 250 seconds]
Zotan has joined #linux-exynos
<daniels> libv: it's unfair to slate the kernel status based on an ancient 3.4 fork you've used, without ever having tried actual current mainline
<libv> daniels: i believe that i have clearly explained the situation
<libv> this is an issue of integration.
<libv> the old kernel has working mali, but horrible display driver
<libv> while mainline might have a working display driver, it probably does not have a working mali
<daniels> libv: as i told you, it does have working mali
<libv> daniels: perhaps, perhaps not
<libv> daniels: i would have to waste time to go find out.
<daniels> libv: download r5p0 source from malideveloper.arm.com -> compile for kernel -> it works
<Wizzup> and the instructions were somewhat lacking, and I think only very recently linux-next has working hdmi
<libv> daniels: 3.4 display driver was shit, samsung idiot display people & arm idiotic binary only policy waste my time, i believe i have every right to complain
<Wizzup> the first part of course falls partially to blame to us, not writing enough wiki pages yet... :)
<libv> plus, i clearly explained the situation
<libv> Wizzup: blame is not the right word
<libv> Wizzup: samsung is to blame
<Wizzup> I guess that's true
<daniels> libv: personally i think spending time with a three-year old kernel that has a known broken display driver is more of a waste than compiling a current mainline kernel, changing one line in defconfig and then compiling the same external module you'd be compiling anyway, but i guess we have different priorities
* daniels goes back to looking at mali's rendering working just fine on hdmi on current mainline
<libv> daniels: again. i would have had to spend extra time to make sure of that
<libv> with the added benefit of the possibility of having had nothing to demo at all.
<libv> i believe that i am perfectly clear and logical here
<libv> as i was in the blog post
<libv> and i believe that i also explained this reasoning extensively in the last two weeks of january when i was hitting these issues and when this was mentioned here
<libv> yes, it supposedly works on mainline. great. thank you for that information.
<daniels> to me, your blog post reads 'i'm stuck on 3.4 and can't move off and everything is shit and there are some people but also everything is shit'
<libv> but 3.4 was horribly broken, and i did not want to waste time on fixing it and risk not having a demo
<libv> i had working mali, i had working ptrace, i had a broken display driver
<libv> daniels: did you also read the bit where i talk about linux-exynos and the pain of having to deal with many SoC families?
<libv> perhaps you should focus on that
<libv> This sentence in particular: "The information for doing a proper linux installation on an android or chrome device is usually dispersed, not up to date, and not too good, and i get to do a lot of the legwork for myself every time, knowing full well that a lot of others have done so already but couldn't be bothered to document thing"
<libv> so please stop trying to berate me, and evaluate your own actions
<daniels> libv: good luck with 3.4
<libv> daniels: who says that i will stick with it forever?
<libv> could it be that i have updating the installation for my chromebook and finally putting my odroid xu3 to good use, are high up my todo list?
<libv> but not as high as "produce a demo for fosdem" was 2-3 weeks ago?
<libv> daniels: do you need me to explain my logic again?
<daniels> libv: i'd rather you didn't
<libv> i'm not sure whether you will now finally relent and admit that my logic in this specific situation actually makes sense, instead of just trying to drag this out endlessly.
<Wizzup> The compromise would be to note in the blog post that the situation has since gotten better in recent kernels, I guess
<javier__> Wizzup: or at least test a newer kernel before taking conclusions...
<libv> what conclusion?
<Wizzup> javier__: he didn't draw a conclusion in the blog post; he just mentions 3.4.0 didn't work well
<libv> 3.4 has a horribly broken display driver. period.
<libv> do i state there that it is now oh so much better?
<daniels> 'The ARM chromebook and its broken kms driver is much of the same.'
<libv> daniels: are you deliberately taking this out of context just so you can drag this out?
<daniels> i mean, that bit is entirely missing a qualifier about how you picked a random three-year old fork rather than using the chromebook vendor tree which actually shipped on the device (really shockingly stable), or mainline
<libv> daniels: do you really have nothing better to do with your time?
<Wizzup> 3.4.0 was the chromebook vendor tree
<daniels> libv: you're right - this is actually detracting from working on kms in a current mainline kernel. so i'll go back to that, and you can go back to ... whatever it is.
<javier__> libv: for me your conclusion seems to be "Samsung Chromebooks is badly support with any other kernel than the vendor one but now things could get better now since there is a linux-exynos.org wiki"
<Wizzup> and shipped on snow at the time
<daniels> Wizzup: oh right, for snow, sorry - i was thinking pi(t)
<libv> are you now going to finish off stating that your life is too short for all of this?
<javier__> libv: but I don't really care tbh, I prefer spending my time trying to fix the situation than complaining about it
<libv> javier__: are you thereby implying that i whine instead of code?
* Wizzup often codes and whines about his own code
<libv> for some context which might explain why daniels feels he needs to pursue this as he does, and why i respond like i do (and this is some really limited context when the last 11 years are taken into account): https://bugs.freedesktop.org/show_bug.cgi?id=45232
<javier__> libv: no, I was talking about the Exynos SoC in general and Samsung Chromebooks in particular
<afaerber> ftr, the HP Chromebook vendor tree is 3.8 based, so "Samsung" is worth pointing out
<libv> for all those complaining: http://linux-exynos.org/wiki/Special:RecentChanges is severely empty.
<afaerber> javier__, I've tried adding the pwrseq node for Spring and my dmesg gets spammed with "mmc1: error -110 whilst initialising SDIO card"
<afaerber> any idea why?
<Wizzup> libv: yeah, the changes that are listed there (the wishlist one) are ones that I plan to pick up soon, but right now I have to rush getting work done before departing for a week to bulgaria and then spain
<Wizzup> at this point I think mainline (or linux-next) is actually getting in pretty good shape
<Wizzup> I'd definitely like some advice a few weeks from now on how to make all of this easily accessible (providing a kernel for people, or just u-boot binaries)
<Wizzup> we already have some bits that are quite OK, but as a whole it just doesn't come together yet
<libv> Wizzup: that RecentChanges link was not aimed at you
<Wizzup> I realised
<Wizzup> but it just reminded me
<javier__> afaerber: yes, I get the same error on Peach Pit/Pi (Exynos5420/5422/5800) but didn't get it for Snow (Exynos5250)
<libv> just pointing out that people should not spend their time taking a few small statements in that blog post out of context, to artificially get a discussion going or to artificially sustain it, while they could instead spend time documenting the fact that mainline has working hdmi and mali on exynos
<libv> which actually also is a part of that blog post.
<Wizzup> I added a support table some time ago: http://linux-exynos.org/wiki/Samsung_Chromebook_XE303C12
<Wizzup> now already outdated
<Wizzup> since linux-next doesn't need those patches anymore
<Wizzup> updated
<javier__> afaerber: but I thought the problem only happens with the dw_mmc 250A IP version while Exynos5250 has 241A
<afaerber> javier__, do you have a branch with that patch(es)?
<javier__> afaerber: I just pushed a wip http://cgit.collabora.com/git/user/javier/linux.git/log/?h=wip/exynos/wifi-peach
<javier__> afaerber: please keep in mind that is a wip, for example addy's modified patch always retries in mci_send_cmd() instead of giving up after a couple of times and also had to revert a patch that broke fw loading instead of finding the cause
<afaerber> javier__, btw does it matter which of the 3 s5m8767 32khz clocks I pass to the pwrseq? I just took the first one (0)
<javier__> afaerber: yes it does, it has to be the one that is routed to the wifi chip
<javier__> which may explain the timeout if the clock is not the correct one
<afaerber> javier__, can you look that up for me then? :)
<javier__> afaerber: I don't have a schematic for spring sorry
<javier__> afaerber: you have to ask to one of the ChromiumOS folks
<afaerber> javier__, hm, is there some way to tell from downstream DT?
<afaerber> given there's only 3 I could just try them all ;)
<javier__> afaerber: indeed, you can do that :)
<afaerber> javier__, as far as WIP, your sd1_* nodes are still extended the old way
<afaerber> for Spring I was asked to do it via &sd1_clk {...}; etc.
dlezcano has joined #linux-exynos
<javier__> afaerber: thanks for the feedback, I'll change before posting once I've it working
<Wizzup> also updated http://linux-exynos.org/wiki/Samsung_Chromebook_2_XE503C32 ; but only briefly
<Wizzup> now it needs actual links to commits and soforth, some binaries perhaps
<javier__> afaerber: about your question of looking for the clocks at the downstream kernel, it models the s5m8767 as regulators that are always-on
<afaerber> javier__, also the binding docs request "ext_clock" as name, not "ref_clock"
<afaerber> javier__, so that means to clock necessary due to always-on?
<afaerber> s/to/no/
<javier__> afaerber: I know, I wrote that binding :)
<javier__> I just created a branch to share you and got an old patch
<javier__> afaerber: the clock has to be enabled and the GPIO power sequencing toggled to reset the chip
<afaerber> sure, just pointing out so that you can clean it up :)
<javier__> afaerber: yes, thanks a lot!
<javier__> afaerber: so what the downstream kernel does is to toggle the GPIO using a set of fake fixed-regulators
<javier__> as if the GPIO is to power on a regulator but really is to reset the wifi chip
<javier__> but they also needed a clock so the hack they found is to model the clock as regulators
<javier__> instead of having a proper clock driver since there wasn't a way to grab that clock and enable it
<javier__> so the clock is a "regulator" that is always on and the reset and power down GPIO are "regulators" too
<javier__> afaerber: but now the simple pwrseq driver grabs the correct GPIOs for power sequencing and clock to enable it
* afaerber reverted to next-20150213 btw where things still boot
<javier__> afaerber: tl; dr try with all the 3 clocks as you said and probably that will fix the timeout :)
dme has quit [Ping timeout: 252 seconds]
<afaerber> javier__, hah, it's the third clock (2) :) thanks
dme has joined #linux-exynos
<afaerber> but it seems deferred firmware loading does not work
<Wizzup> for wifi?
<Wizzup> I think you want to do the same as this commit, if required http://cgit.collabora.com/git/user/javier/linux.git/commit/?h=wip/exynos/wifi-peach&id=83b3713104d4e0f32658c27285f1220e7f78256d
<afaerber> [ +0,003772] mmc1: new high speed SDIO card at address 0001
<afaerber> [ +0,000442] mwifiex: rx work enabled, cpus 2
<afaerber> [ +0,000168] mwifiex_sdio mmc1:0001:1: Direct firmware load for mrvl/sd8797_uapsta.bin failed with error -2
<afaerber> [ +0,000019] mwifiex_sdio mmc1:0001:1: Failed to get firmware mrvl/sd8797_uapsta.bin
<Wizzup> yes
<Wizzup> I also get that on snow when the patch that javier reverted in above link is not reverted
<Wizzup> so fw loading for wifi is/was broken on linux-next, at least a few days ago
<javier__> afaerber: awesome and yeah I had to revert that commit
<javier__> it would be good to root cause but I didn't have time and I'm busy with other stuff now
<afaerber> reverting did not help
<Wizzup> did you enable the config to enable userspace to load the fw?
<afaerber> which config?
<afaerber> CONFIG_FW_LOADER=y
<afaerber> # CONFIG_FW_LOADER_USER_HELPER_FALLBACK is not set
<afaerber> Wizzup, you mean the latter?
<javier__> afaerber: with CONFIG_FW_LOADER_USER_HELPER_FALLBACK
<Wizzup> the latter, yes
<javier__> in fact, I found the in-kernel fw loading not working and needed the user-space helper
<javier__> afaerber: so is CONFIG_FW_LOADER_USER_HELPER_FALLBACK + reverting the mentioned commit
<afaerber> at times it's really depressing how little works in linux-next...
<Wizzup> I was actually surprised that a lot of stuff did work
<Wizzup> snow is almost usable now
<Wizzup> basically audio and then I will consider it for daily use again
<Wizzup> now I'm still on peach-pi with chromeos 3.8 kernel
<javier__> afaerber: yeah, the problem is that the regression testing in mainline is basically to boot and get a serial console
<javier__> it will be nice to have a more automated unit test to catch these kind of regressions earlier
<afaerber> [ +0,000348] mwifiex_sdio mmc1:0001:1: Direct firmware load for mrvl/sd8797_uap
<afaerber> sta.bin failed with error -2
<afaerber> [ +0,000017] mwifiex_sdio mmc1:0001:1: Falling back to user helper
<afaerber> followed by late
<afaerber> [ +0,157714] mwifiex_sdio mmc1:0001:1: Failed to get firmware mrvl/sd8797_uapsta.bin
sjoerd has joined #linux-exynos
ssvb has joined #linux-exynos
_whitelogger has joined #linux-exynos
prahal has joined #linux-exynos
liquidAcid has joined #linux-exynos
afaerber_ has joined #linux-exynos
<obbrobbrio> guys, how do i install mali T604 drivers on snow? is it fine taking just graphics-related steps from this guide? http://community.arm.com/docs/DOC-9494
afaerber has quit [Ping timeout: 250 seconds]
afaerber_ is now known as afaerber
bmbeach has joined #linux-exynos
<bmbeach> hello anybody home?
ssvb has quit [Ping timeout: 240 seconds]
ssvb has joined #linux-exynos
<bmbeach> Wizzup are you there?
<bmbeach> I tried booting using the cat ... procedure and it worked fine
<bmbeach> and also noticed the following kernel option with the following text:
<bmbeach> With this option, the boot code will look for a device tree
<bmbeach> binary (DTB) appended to zImage (e.g. cat zImage
<bmbeach> <filename>.dtb > zImage_w_dtb).
<bmbeach> This is meant as a backward compatibility convenience for
<bmbeach> those systems with a bootloader that can't be upgraded to
<bmbeach> accommodate the documented boot protocol using a device
<bmbeach> tree.
<bmbeach> Beware that there is very little in terms of protection
<bmbeach> against this option being confused by leftover garbage in
<bmbeach> memory that migh look like a DTB header after a reboot if no
<bmbeach> actual DTB is appended to zImage. Do not leave this option
<bmbeach> active in a production kernel if you don't intend to always
<bmbeach> append a DTB. Proper passing of the location into r2 of a
<bmbeach> bootloader provided DTB is always preferable to this option.
<bmbeach> thought this might solve the problem but when I disabled it , rebuilt the kernel and rebooted the kernel still crashed
<bmbeach> but anyway I'm am now booting 3.19.0-next-20150216
<bmbeach> notice that I still booted off nv_uboot-snow-simplefb and
<bmbeach> things like clk_ignore_unused and it still worked. theses I will
<bmbeach> eventually fix
<bmbeach> On x-windows everything works fine but it is using fbdev
dme has quit [Quit: Coyote finally caught me]
<bmbeach> shutdown -r works fine and shutdown -h shuts down but the power led stays on.
<bmbeach> also I noticed in /sys/class/power.. I am finding the following values
<bmbeach> voltage_max_design 7500000
<bmbeach> voltage_min_design 7500000
<bmbeach> voltage_now 8559000
<bmbeach> I think voltage_max_design should be around 8559000
<bmbeach> also I get on "dmesg" the following value
<bmbeach> Total of 2 processors activated (96.00 BogoMIPS)
<bmbeach> that value should be around 3000
bmbeach has left #linux-exynos [#linux-exynos]
aballier has quit [Read error: Connection reset by peer]
aballier has joined #linux-exynos
dlezcano has quit [Remote host closed the connection]
liquidAcid has quit [Quit: Leaving]
<afaerber> javier__, disabling exynos thermal resolved my boot issues on next-20150218