ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See | | Logs at
ZaEarl has joined #linux-sunxi
servili007_ has joined #linux-sunxi
servili007 has quit [Ping timeout: 256 seconds]
Turl has quit [Excess Flood]
Turl has joined #linux-sunxi
uminded has quit [Quit: Leaving]
anunnakiii has joined #linux-sunxi
anunnakiii has quit [Read error: Connection reset by peer]
hipboi has quit [Ping timeout: 256 seconds]
hipboi has joined #linux-sunxi
hipboi has quit [Quit: Leaving]
hipboi has joined #linux-sunxi
rellla has joined #linux-sunxi
ZaEarl has quit [Ping timeout: 246 seconds]
<libv> uminded is quite right, there's a lot of useless stuff in there and it is rather impossible to find answers to your actual questions in there
<libv> ah yes, my odroid is stuck at customs.
<libv> Fedex: "Package not due for delivery"
servili007_ has quit [Read error: Connection reset by peer]
rz2k has joined #linux-sunxi
hipboi has quit [Ping timeout: 260 seconds]
bfree has joined #linux-sunxi
bfree__ has quit [Ping timeout: 252 seconds]
hipboi has joined #linux-sunxi
rz2k has quit []
anunnaki has joined #linux-sunxi
<libv> is there a reason why the usb gadget ethernet driver is not compiled?
<libv> it would actually provide a useful backdoor for some devices
<libv> well, those devices with an otg or device port
<libv> and i am going to kick the prefix of first steps out to the top level.
<libv> heh, roman mamedov, jan 30th
<libv> mnemoc: what about enabling g_ether module building per default?
<mnemoc> sure, patches to improve the defconfigs are obviusly welcomed
<hramrach> does it work?
<hramrach> I recall somebody complaining about the gadgets being broken
<hramrach> also how would you structure the wiki to make sense?
<hramrach> thre is one way that puts (nearly) everything into one huge page which is then quite ahrd to navigate
<mnemoc> first step is to add categories to the pages that lacks of them
<libv> ok
<libv> i will add the tutorial category to the new page about usb gadget support then
<hramrach> or you make dedicated page for every simple thing but then 'you need to pick through page links to find something simple
<mnemoc> i personally think of wiki pages as of code functions
<mnemoc> as small as it still has meaning, as large as it can't be meaningfully split
<hramrach> the problem then becomes as with code. you don't know what function does the thing you want
<libv> currently, i just see little coherence and too much prose in our wiki
<libv> there's tons of stuff there, but little that actually helps anyone set things up
<mnemoc> bloat :)
<mnemoc> if it does do anything useful, cut it out
<libv> look at our front page from the pov of someone who just got a sunxi device, or who is intending to get a sunxi device
<libv> it will seem like a rather impossible task
<libv> even though we have quite a few things in place
<hramrach> I guess it reflects the state of things
<hramrach> linux for sunxi is under development and there is no definite answer
<hramrach> but the front page links are quite likely obsolete
<hramrach> because you don't get autogenerated list of category hardware/software/tutorial/..
<hramrach> but a snapshot of something picked by somebody in who knows how distant past
<hramrach> does the wiki support autogenerateing sections like that?
<hramrach> yes, the CryptoDev/TrustZone page is totally useless
<hramrach> does not tell how to enable, how to check, nothing
<hramrach> just generic babble
<hramrach> heh, the page says there is CESA enging and that is not possible to enable on sunxi
<hramrach> so I guess that part is completely unusable
mUniKeS has joined #linux-sunxi
<mUniKeS> hi
<mUniKeS> Is it possible to connect DC-IN 9-12V to a mk802+?
<mnemoc> mUniKeS: asking the same in 3 channels qualifies as spam
<mUniKeS> sorry :(
<hramrach> plus you should refer to your mk802+ manual for dc requirements
<xenoxaos> just reply different things on different channels
<libv> 3 time round, just say "yes, give it a go"
<libv> and then the person doing so will not try it a second time round
<mnemoc> :)
<xenoxaos> lol
<Turl> yay, productive morning
<Turl> uart code is now clock-aware and the gating implementation seems to work
<Turl> hipboi: how well do you know the sunxi clocks?
<hipboi> little
<mnemoc> Turl: hi, i've pushed a wip/lichee3-sunxi/a13 branch to my github. it includes pretty complete and recent mach-sun5i code, including clock stuff
<hipboi> Turl: what's your question?
<Turl> hipboi: I was wondering if this "DRAM_AXI_GATING" gate does anything, the system doesn't seem to malfunction when it's off
<Turl> bit 0 at offset 0x5C, the only module gate on AXI clock
<libv> guys, is the statement at the top here correct?
<libv> "These should work on any Allwinner A10 based device, despite the author targetting some particular device (like the Cubieboard or the MK802). At most, to make the image compatible with your device, you will need to add the u-boot and script.bin from your particular hwpack."
<Turl> mnemoc: can you diff ccmu_regs.h with our sun4i's please?
<libv> that sounds quite misleading
<libv> u-boot and script.bin are _always_ device specific
<Turl> libv: sounds like the truth to me
<Turl> well, they often work
<Turl> maybe not optimally :)
<libv> the mele one points to a full tutorial
mUniKeS has quit [Remote host closed the connection]
<hipboi> Turl: just looked at the clock chapter of the A10 manual
<Turl> libv: it could be worded differently though, "these images will work... as long as you replace uboot ..." maybe
<hipboi> it seems the axi dram gate is for dram clock 6.4.38 at the manual
<libv> i would very much prefer to bury this one somewhere down there, it contains questionable content, and it is better to work with hwpacks from the start
<mnemoc> Turl: arch/arm/mach-sun5i/include/mach/ccmu_regs.h is identical
<Turl> mnemoc: to sun4i's? awesome
<Turl> at a quick glance, I see nothing major, other than missing clocks/gates
<Turl> hipboi: hm, so that dram axi gate controls the other gates then
<hipboi> yes
<Turl> hipboi: but it doesn't have any relation to RAM does it?
<hipboi> it will be used when other module access dram directly
<Turl> like mali or ve?
<hipboi> yes
<Turl> good then, I understand now
<hipboi> BE, FE
<Turl> hipboi: what do all of those stand for? ACE, BE, FE, DE?
<hipboi> be, fe is image processing
<hipboi> front end back end
<libv> grmbl.
<hipboi> front end engin, back end engin
<libv> 2 more things need to get fixed in firststeps...
<Turl> ACE is Audio Codec Engine then?
<hipboi> de: display engin
<hipboi> Turl: it seems, yes
<Turl> hipboi: and AVS?
<hipboi> audio video sync
<Turl> ok, thanks hipboi :)
_whitelogger has joined #linux-sunxi
<libv> who makes such illogical changes to wikis :(
<hramrach> we have an A10 manual?
<mnemoc> hramrach: see
<hramrach> thanks
<hramrach> I like these watermarks ;-)
<hramrach> hmm, the TZ part is very descriptive
<hramrach> looks liek it is just memory protection and no crypto described
<mripard> oh, so we already had the A10 User Manual ?
<mnemoc> mripard: yes, but it's very complete
<hramrach> +in
<mripard> mnemoc: yeah, I know, I got it from Allwinner some time ago, and they told me last week that it was OK to share it last week
<mripard> I was planning to push it tonight :)
<mripard> anyway, I have the one for the A13 as well...
<mripard> but it would be great to mention it on the wiki somewhere :)
<rellla> mripard: ^^ any other docs from allwinner ^^
<rellla> s/ ^^/? ;)/
<rellla> ah, no ibot :)
<mripard> not yet
<mripard> I asked them for the A31 user manual
<mripard> but got nothing so far
<hramrach> well, now it's mentioned ;-)
<mnemoc> mripard: i don't they have any A31 documentation in english atm
<Turl> hi mripard :) I'm preparing v2 and will be sending it in a bit
<mripard> mnemoc: yeah, I think so as well, but it's worth asking :)
<mripard> Turl: great :)
<Turl> mripard: I made 8250_dw clocks-aware too and I'm running with it set now :)
<Turl> without having to (ab)use the ignore clock flag
<libv> *sigh*
<hramrach> why is firststeps another copy of all the tutorials?
<hramrach> also wrong because mmcblk is not available when you are setting up the SD card in an USB card reader
<Turl> mripard: v2 sent
<Turl> is this a valid use of dev_set_drvdata?
<mripard> it's not done through of_serial.c?
<mripard> ah, no, it's not
<mripard> grmbl, this serial dt support is a mess
<mripard> but why don't you store it in 8250_data?
<mripard> and for you clock driver
<mripard> you added a dummy clock
<mripard> what is it for ?
<Turl> to keep the positions on the mux
<Turl> I added comments that explain it
<mripard> yeah, but it wasn't on the v1
<mripard> so, somehow, the muxes worked in the v1 without the dummy clock
<Turl> yes, because the only mux had the missing clock on the end
<Turl> so "0 1 2 <missing>" and "0 1 2" doesn't change anything
<mripard> ah, yes, and apb1 needs it
<mripard> try to be a bit more expressive than just "implement APB1" in your cover letter next time when you have to do such changes
<Turl> ok, will do
<Turl> mripard: 8250_data doesn't seem to be available until later
<hramrach> now first steps has an example of auto-generated section, cool
<Turl> mripard: maybe I can use a managed function to get the clock and forget about the rest
<mripard> Turl: ?
<mripard> it's the first function called in the probe that allocates it
<hramrach> wtf is linaro?
<hramrach> that's ubuntu version for embedded devices, right?
<mripard> a consortium of ARM, most ARM silicon vendors, and some software companies
<Turl> mripard: look around line 318, dw8250_probe_of is first caled, then "data" is kzalloc'd later
<Turl> called*
<mripard> that put together engineers to enhance linux support for ARM in general
<mripard> so they work on the kernel, gcc, etc.
<mripard> the ubuntu thing is just a side effect.
<hramrach> oh, so not really Ubuntu, they just develop Ubuntu variant as their Linux distro
<mripard> yes, they have an Android distro as well
<mripard> Turl: I don't have 318 lines in the code I'm looking at :)
<hramrach> ok, thanks
<mripard> Turl: what version are you working on?
<Turl> mripard: are you looking at drivers/tty/serial/8250/8250_dw.c?
<Turl> ah, looks like they added a bunch of stuff on master already
<Turl> ACPI and DMA support
<mripard> well, the comment still applies
<mripard> just move around the code to have the kzalloc before you call probe_of
<Turl> mripard: what about using devm_clk_get?
<mripard> what about it?
<Turl> nevermind
<mripard> didn't get your question sorry :)
<Turl> I thought about using it to avoid having to clean up, but it doesn't disable the clock I think
<mripard> ah, I don't know
<mripard> use the source, luke :)
<Turl> mripard: :)
<Turl> mripard: btw, I pushed the sunxi-clk-submit-2 tag with working gates if you want to test it on sun5i
<mripard> I'll use it in a short while I guess
<mripard> almost finished the I2C driver
<Turl> nice
<Turl> mripard: did you get any response from Arnd regarding the led dt bits?
<mripard> nope
<mnemoc> mripard: is the diff between the start point of i2c in sunxi-3.0/3.4 has and allwinner's latest code
<mripard> but I talked with Olof at the ELC
<mripard> I'll resend a mail, and we'll see...
rz2k has joined #linux-sunxi
<mripard> mnemoc: thanks, I'll take a look at it
grevaillot has joined #linux-sunxi
merbzt has joined #linux-sunxi
rellla has quit [Remote host closed the connection]
rm has joined #linux-sunxi
<mnemoc> rm: welcome!
<hramrach> you should get another surname startirng with S ;-)
<mnemoc> o_O
* mnemoc hadn't noticed rm has fsf member
paulk-desktop has joined #linux-sunxi
user135 has quit [Ping timeout: 260 seconds]
<rm> hramrach, people ask me if I'm RMS even without that :)
<rm> (when I talk on #fsf)
<rm> mnemoc, thanks :) the member cloak is for quite some time, several years I think
<mnemoc> rm: I suppose it was before you stay on the channels long enough to never see you joining/leaving :)
<mnemoc> s/before/because/
<rm> right, I connect via ZNC and it can be online for weeks or months
<Turl> mripard: I've been thinking about the pll-axi-ahb-apb clock relations and cpufreq, but I cannot find any way on the framework to implement it nicely :(
<Turl> I found this email on the list, seems they have a similar issue as ours
<Turl> "[PATCH/RFC 2/4] ARM: shmobile: sh73a0: add support for adjusting CPU frequency"
user has joined #linux-sunxi
<user> how do I disable early suspend in the kernel? I want to use the framebuffer.
<techn_> user: it should work with latest stock kernel.. you just need to modify u-boot params
<techn_> what you are trying to enable?
<user> the framebuffer console
<hramrach> user: you can try to search your .config for something like EARLY and disable it
<hramrach> but if it is selected by other option it will be reenabled silently
<user> the only switch with early is CONFIG_ANDROID_RAM_CONSOLE_EARLY_INIT=y
<user> :/
<hramrach> odd
<hramrach> you can always try to build 3.4 which does not have that option at all ;-)
<user> has 3.4 got serial over usb?
<user> could I get a serial console over usb, even?
<hramrach> not likely
<hramrach> last time I heard there were half-broken USB gadgets on 3.0 and totally broken on 3.4
<mnemoc> techn_: btw, tsvetan confirmed allwinner considers the linux code in the A10s SDK (with the CEC driver) GPL and just don't care about adding proper headers.
<hramrach> plus using the OTG as serial console might not be the most reliable thing ever
<hramrach> do you have ethernet ?
<user> no, I have wireless though
<hramrach> hm
<user> I'm just trying to get console output. I'm blind atm.
<hramrach> not sure you can get console over wireless
<user> well, might have to use the new kernel and see if I can get framebuffer.
<hramrach> the serial port does not work for you or do you have no pins soldered for it?
<user> I have no pins soldered for it.
<hramrach> the framebuffer should work but is initialized rather late so if anything goes wrong early you still don't see anything
<hramrach> plus u-boot supportes serial only
<user> I think I've managed to uboot, because when I change my init.rc I get different behaviour from the tablet
<user> but that's working blind.
<hramrach> getting OTG working properly would be nice project if you had another reliable console but without one ..
<hramrach> with a tablet this is more difficult because you don't have EDID and have to figure out the proper tablet mode
<hramrach> also does the backlight light up?
<user> yes
<user> it does
<hramrach> so you would see the fb console if it worked ;-)
<hramrach> try to set different modes for the display
<hramrach> there is some info on doing that in one of the commit messages
<mnemoc> which should be in the wiki... :<
<hramrach> there is an example here in the ekrnel arguments:
<hramrach> but I don't have an exhaustive explanation of what you can set up
<user> ok brb
* mnemoc cries
<mnemoc> why does BuildingOnDebian have not-debian-specific info? :<
* mnemoc cries
<hramrach> it's simply a list of steps I did
<mnemoc> fine, but BuildingOnFoo should only covert how to get the toolchain and needed packages installed
<hramrach> I picked that example from the commit messages. I don't know if there is a better source
<hramrach> also I am not sure how that would work for LVDS
<hramrach> is that configured the same way or with script parameters?
<hramrach> you can configure screen0 to be LVDS in script presumably but I have no idea where the mode would come from then. Possibly from the kernel command line aas well but possibly from script
<hramrach> well, there is USB gadget page on the wiki which shows how to use USB ethernet
<hramrach> presumably it's *the* working option on 3.0
<hramrach> user: thre are staging/sunxi-3.0 and sunxi-3.0 branches. At least one should work. Try rm's configs which do build and then add other options that you want.
gzamboni has joined #linux-sunxi
<mripard> Turl: I've never written anything related to clocks or cpufreq, so I'm probably not the best person to ask
<Turl> mripard: I'll email Gregory and see if he has any ideas
<user> hramrach, I will build with the 3.4 kernel, and see if that works
<user> I have an a13 device
<user> *a13_mid
<hramrach> I have no experience with building images for devices with LVDS
<hramrach> presumably the a13_mid fex file selects the LVDS as screen0 and you then configure the mode using the kernel commandline
<libv> mnemoc: i got really really worked up this afternoon about our wiki
<libv> mnemoc: even those things which once made sense just got muddled up
<libv> what's the point of trying to get it in shape
<libv> 1 month from now it will be a useless mess again
<hramrach> well, people add useful info they could not find there in random places
<hramrach> presumably when most useful info is there it will get more stable ;-)
<techn_> mnemoc: nice to hear :)
<mnemoc> libv: people wants to contribute, and when they can't find content, they contribute it where they find more correct
<mnemoc> libv: if people finds the content, they won't duplicate it. if people understands the structure, they'll contribute on the right place
<rz2k> structures is hard for people
<rz2k> *are
<rz2k> btw mali page needs complete wipeout since we dont need the complete info about each libraries set and etc.
<user> This is what I'm getting whilst compiling with the a13_mid_nuclear_defconfig
<user> with the 3.4 kernel
<hramrach> I like the inof about eac libraries set
<hramrach> *info
<mnemoc> user: "internal compiler error:", nice one
<hramrach> user: what compiler?
<techn_> user: I have that a13b tablet too.. a13_mid hwpack and linaro rootfs should work without problems..
<hramrach> also are you sure your build machine hardware is OK?
<hramrach> gcc likes to throw ICEs on bad memory
<Turl> your toolchain and/or pc look broken :)
<user> gcc (Debian 4.4.5-8) 4.4.5
<user> This is free software; see the source for copying conditions. There is NO
<user> Copyright (C) 2010 Free Software Foundation, Inc.
<hramrach> yes, the linux kernel sources should not cause ICEs
<Turl> try again user
<user> ok
<Turl> if it doesn't crash this time, it's probably bad ram
<hramrach> or creshes in different place
<user> it crashed
<user> hmm
<hramrach> the same file?
<hramrach> where did you get the gcc?
<user> yes
<user> apt-get
<user> debian repositories
<hramrach> ther is no cross gcc in debian repositories
<user> wait a second then
<mnemoc> hramrach: that's what BuildingOnDebian was intended for. not to duplicate instructions host-agnostic
<Turl> emdebian?
<user> I'm compiling with replicant
<user> I do have a cross compiler installed though
<mnemoc> how to setup the display driver should go on the page of the display driver. how to compile u-boot or linux, should go the the u-boot or linux pages. how to make a uSD bootable, should be to the Bootable SD page, etc. etc.
<mnemoc> user: then tell your kernel to use the right cross compiler
<user> yes I have an embeddian cross compiler installed
<mnemoc> use it :)
<user> mnemoc, so how do I tell my kernel to use a different compiler?
<mnemoc> user: pass CROSS_COMPILE=foo- assuming you have a foo-gcc
<mnemoc> in your $PATH
<user> okay, thanks
anunnaki has quit [Read error: Connection reset by peer]
anunnaki has joined #linux-sunxi
<mnemoc> uh
<hramrach> mnemoc: afaik very little is duplicated on that page aside form that *other* Debian tutorial which I did not find on the wiki and the yet another Debian tutorial which somebody wrote after
<hramrach> I wrote a KernelArguments page which lists the arguments I used
<hramrach> hopefully somebody can add any extra arguments that are relevant
<hramrach> and any extra values I don't know about, too
<mnemoc> hramrach: more than a kernel arguments page, which is not what people will search for, i would say a page about setting up the display
<libv> rz2k: i also do not want to throw away that table, i am not sure whether there is any useful information in that
<libv> rz2k: everything else, yeah, at first glance, it has become useless/deprecated or was fluff anyway
<libv> mnemoc: about this firststeps page, i would say that half the changes were useful, a quarter was useful, if it had been properly structured, and the other quarter was just improper
<libv> i find that a rather depressing ratio
<mnemoc> narf
<hramrach> mnemoc: well, I searched for kernel arguments and had to pick them from random logs and whatnot
<hramrach> the display should jsut work when you have sane device
<hramrach> when not that's another problem
paulk-desktop has quit [Quit: Ex-Chat]
<user> I tried compiling without debugging symbols, and it got further, but now I get: drivers/built-in.o: In function `Hdmi_init':
<user> lowmemorykiller.c:(.text+0x306f0): undefined reference to `audio_set_hdmi_func'
<mnemoc> user: what defconfig are you using?
<mnemoc> that happens if you build hdmi without hdmi_audio
<user> I'll try.
<mnemoc> hramrach: so you don't see a more useful?
<user> Building modules, stage 2.
<user> MODPOST 57 modules
<user> ERROR: "ve_start" [drivers/media/video/videobuf-dma-contig.ko] undefined!
<user> ERROR: "ve_size" [drivers/media/video/videobuf-dma-contig.ko] undefined!
<user> hmm
<mnemoc> if you disable cedarx you need to disable the whole video thing
<mnemoc> (patches to fix that one are welcomed, it's just a missing #ifdef)
<hramrach> mnemoc: I have very little info for that page. If you have the details to fill in that would be great
<hramrach> I use only the HDMI which by now should work atomagically on all devices that have it
ZaEarl has joined #linux-sunxi
<user> mnemoc, how do I enable cedarx, I cannot find anything in the defconfig about it?
<user> nm, I found it in the sun5i config.
<mnemoc> user: see CONFIG_VIDEO_*
<mnemoc> hramrach: i don't think it's magic yet. but I haven't plugged a display to a sunxi device in a long time
<user> thanks mnemoc, it wasn't in the a13_nuclear_defconfig
<mnemoc> hramrach: even if it's a couple of lines in the right page it's better than having those lines in random places
<mnemoc> user: a13_nuclear_defconfig is an android config
<user> indeed, I'm trying to port replicant
<mnemoc> no one maintains that defconfig, patches are welcomed
<mnemoc> compare with crane's for reference
<user> I've managed to compile now
<user> (but it'd be nice if I didn't have to use cedarx) :)
<user> hmm still no framebuffer output
merbzt has quit [Ping timeout: 245 seconds]
FergusL has joined #linux-sunxi
rz2k has quit [Read error: Connection reset by peer]
rz2k has joined #linux-sunxi
arete74 has quit [Ping timeout: 255 seconds]
anunnaki has quit [Read error: Connection reset by peer]
anunnaki has joined #linux-sunxi
servili007 has joined #linux-sunxi
arete74 has joined #linux-sunxi