bgamari- has quit [Ping timeout: 244 seconds]
dlan has quit [Ping timeout: 265 seconds]
dlan has joined #linux-exynos
dlan has joined #linux-exynos
dlan has quit [Changing host]
<steev> javier__: at some point could you go through the exynos_defconfig and fix it so.... it actually just works, like people claim it does? it seems that it just works but only if you have some kernel config that apparently i don't have
<steev> javier__: or better, a list of "this must be on for everything to work" - we're kinda working on a script to make sure all the various options are on
amitk has joined #linux-exynos
D1337d has quit [Quit: leaving]
<Wizzup> steev: we could do that too
<steev> Wizzup: well i don't see anything *exynos* from your config that isn't in the defconfig, so i dunno
<steev> maybe it's because i'm not using an nvuboot, i'm just doing a kpart
<Wizzup> it is possible I disabled some things
<Wizzup> I use nv-uboot, yeah
<Wizzup> bbiab.
<steev> i don't because i'd like these to be fairly generic, instead of having to do an image for peach pi, and image for peach pit, an image for snow, an image for spring
<Wizzup> I have exactly that
<Wizzup> I just mkimage for zImage + dts for various machines
<Wizzup> And that "just works"
<Wizzup> (at least last time I did that, around 4.0-rcsomething)
<steev> uh, how? considering u-boot for spring and snow requires separate display issues
<steev> i'm cloning linux-stable's linux-4.1.y branch and will compare the output of exynos_defconfig's .config to yours
<steev> you turn off fbsimple
<steev> and enable an errata
<Wizzup> well, yes, like I said, separate images, but it's not a separate config
<Wizzup> Ah, I missed an errata?
<Wizzup> are you sure you're not comparing it the other way around?
<Wizzup> I think specifically enabled that errata :)
<Wizzup> (and disabled simplefb)
<steev> Wizzup: no, i said you enable an errata
<steev> and as i said, i want a generic *image* - not a generic kernel image, the full image
<steev> generic kernel image is pointless if i have to have 4 separate images
<Wizzup> right.
<Wizzup> afk
<steev> ah bah, i forgot to remove your initramfs from the config
bgamari- has joined #linux-exynos
bgamari- has quit [Ping timeout: 244 seconds]
* javier__ reads the backlog
<javier__> steev: well I try to keep exynos_defconfig so things just works
<javier__> steev: but you can't have a FIT image for all since the FIT .its has the definition of the DTB anyways
<javier__> or you can bundle several config/dtb on a single FIT image? I'm not expert on FIT
<steev> javier__: except it doesn't just work? i'm checking now, but i'm thinking the fbsimple is breaking things
<steev> you can bundle several dtbs into one fit image
<steev> kernels too for that matter, but i'm not sure what the specifics of that are
<javier__> steev: no, fbsimple will only break things if you have an nv-uboot that mangles the FDT to add the simplefb node
<javier__> steev: right but kernel is not needed in this case since the same image works on all boards
<steev> javier__: hm, then something else is causing me to have backlight only (still)
<steev> does mainline *require* using nvuboot? that would be sad times
<javier__> steev: it certainly didn't the last time I tried
<steev> i cloned linux-stable's tree, linux-4.1.y branch, make exynos_defconfig, create fit image from the results, and no bueno
<javier__> steev: which board?
<steev> backlight powers on, but without any type of serial, on peach pi
<steev> i can't tell what's going on
<javier__> steev: ok, let me give a try today and will report here
<javier__> Wizzup: what's the thing with this errata? did you experience something or just enabled because it says is for A15?
<javier__> althoug it says up to r0p4, I onder what part version the A15 on Exynos5250 and Exynos5800 have
<steev> fwiw, chromeos seems to enable errata 766421, 773022, 773769, 774769, 798181
<steev> not even sure those are in mainline
<steev> oh, that's the 773022 that Wizzup turns on
<steev> looks like, of those that chromeos turns on, only 798181 and 773022 are in mainline
<javier__> steev: I see, I haven't experienced anything so I didn't dig which erratas were needed
<javier__> steev: btw, this is how I build FIT just to make sure we are on th same page http://paste.debian.net/plain/261502
<steev> Strongly-Ordered/Device load or NC LDREX could return incorrect data, incorrect instructions may be executed from loop buffer, Large data RAM latencies can lead to rare data corruption
<steev> javier__: pretty much yes, but i use root=PARTUUID=%U/PARTNROFF=1
<steev> http://pastebin.com/zNDUurqr that's my kernel-peach.its
* sjoerd never got how errata's for the defconfigs are selected
<sjoerd> multi_v7 doesn't have any turned on
<steev> sjoerd: they can't turn on all of them for multi_v7, because some affect others negatively
<steev> i.e. something that fixes a15, may aversely affect a8
<javier__> sjoerd: for OMAP we enabled some of them when people started trying with hard float binaries
<javier__> I don't remmeber the details but there were issues with hard float, maybe eballetbo remembers
<steev> basically, you want to enable it if you know for sure your board(s) are affected
<javier__> steev: did you looked at the commit log why chromiumos folks enabled those?
<javier__> I can look too but since you are already on that repo
<steev> javier__: no, but i can, lemme kick off this dd
<steev> although
<steev> not sure if it will actually show that, i think i may have depth=1
<javier__> steev: no worries, I'll take a look
<steev> yeah, i did depth=1
<steev> sorry
* javier__ prefers to enable what is actually needed instead of just blindly follow the chromeos defconfig
<steev> BUG=chromium:226280 798181
<sjoerd> javier__: which means fun if folks start using the multiplatform kernel on omap
<javier__> sjoerd: yeah
<steev> looks like most were enabled for snow, since that was uh, 2 years ago
<steev> https://chromium-review.googlesource.com/#/c/33608/ that's for the errata that Wizzup enabled, sadly, looks like the bug is private
<steev> so, i don't think any affect pi/pit, but i think 798181 and 773022 might affect snow/spring
<javier__> steev: well, Doug and Olof were not sure if this actually was needed
<steev> could always ask - ojn - care to chime in re https://code.google.com/p/chromium/issues/detail?id=226280 ? since we don't enable that in exynos_defconfig in mainline? :)
<javier__> Doug just says in the bug you shared that "Perhaps this could explain some of the completely random crashes that we have always seen"
<steev> right
<steev> i read through the bug, i'm just passing along the information, ojn would know way more than i would... i've only ever committed stuff to chromiumos as an outsider
<steev> though it was ~2 years ago that it was enabled, so he may not remember
<javier__> yeah, my point is. If we have a real bug report and a confirmation that an errata fix it, then I want to enable
<javier__> otherwise I prefer to leave them disabled
<steev> oh i'm not saying to enable them, i'm just saying that's what chromeos did, and is possibly why Wizzup enabled it
<javier__> steev: yes, I was referring to Wizzup indeed and not you :)
<steev> oh haha
bgamari- has joined #linux-exynos
<javier__> steev: I've some urgent things to finish for work but after that I'll build a v4.1 fit image and test it on my peach pi and pit
<steev> cool
<steev> no rush on it, it's almost bedtime here
<javier__> steev: ok, I'll let you know how it goes. I'll be quite surprised if FIT is not working since I tried some time ago (v4.0 iirc)
<steev> yeah, i'm just waiting on my latest build to write to SDcard and then i'll test it
<steev> it's sorta based off wizzup's config
<steev> that said, flipping through his changes, i really can't see anything that sticks out as "oh yeah, that's why it's not working"
<javier__> steev: yeah, me neither
<steev> i just know that i get backlight on, and no tty
<steev> and since it's a fresh image... no network
<steev> javier__: fwiw, this is the config i'm using http://paste.ubuntu.com/11777455/ if anything glaringly sticks out
<steev> and yeah, still nothing here. giving up for now
<javier__> steev: ok, no worries I'll test today and let you know if works for me or not
<javier__> if it doesn't then I've a serial console for peach pit :)
<steev> javier__: well that config + the .its i gave you should work on pi or pit (snow and spring too but i haven't tested those yet)
<steev> and i don't have a pit, so there's that
zombah has joined #linux-exynos
<javier__> steev: mmm I tried a FIT image and indeed is not booting
<javier__> steev: building a v4.0 kernel now to test, I was pretty sure that it was working and I don't see how the kernel could rely on u-boot
<javier__> steev: right, v4.0 boots correctly but v4.1 doesn't...
<javier__> steev: I'll try a couple of things but I don't have time to do a bisect
<javier__> steev: if you can do a git bisect and point me to the first offending commit, that would be awesome and I can dig more
<javier__> but unfortunately this is a hell of a week for me so I don't have the time :-/
<Wizzup> javier__: the errata is needed to compile gcc
<javier__> Wizzup: I see, then it should be added to exynos_defconfig if does not cause a side effect
<javier__> Wizzup: care to post a patch to linux-samsung-soc@vger.kernel.org cc'ing Krzysztof and Kukjin? :)
<javier__> you can get their email addresses from https://www.kernel.org/doc/linux/MAINTAINERS
<Wizzup> javier__: not today or tomorrow, can do afterwards
<javier__> Wizzup: sure, no rush. Whenever is a good time for you
<javier__> I could do it myself but I lack the context about what's the issue when building GCC and you know ;)
<Wizzup> afk :)
<javier__> Wizzup: interesting, thanks for the info!
afaerber has joined #linux-exynos
dlan has quit [Ping timeout: 264 seconds]
dlan has joined #linux-exynos
amitk has quit [Quit: leaving]
zombah has quit [Quit: Leaving]
afaerber has quit [Quit: Verlassend]
<steev> javier__: huh, okay