hno changed the topic of #linux-sunxi to: /Allwinner/sunxi development discussion - Don't ask to ask. Just ask and wait! - See | | Logs at
<Turl> leviathanch: sounds like fun :)
naobsd has quit [Quit: Page closed]
vicenteH has quit [Read error: Connection reset by peer]
vicenteH has joined #linux-sunxi
jdeisenberg has joined #linux-sunxi
<jdeisenberg> Hi. Where would be an appropriate forum to post questions about Ethernet on Mele M5 when running Linux?
Soru has quit [Ping timeout: 245 seconds]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
vinifr has joined #linux-sunxi
popolon has quit [Quit: Quitte]
geecko has quit [Quit: Quitte]
<steev> woo the cb2 is somewhere in the US... it left chicago a few days ago, although it did spend 4 days there
<WarheadsSE> steev: they are floating it down the Mississippi on a raft, all Twain style
<steev> WarheadsSE: probably
<steev> sadly, my utilite was shipped yesterday, and will be here tomorrow
<steev> and it's coming from Israel
<WarheadsSE> Remember how fast the AX3/4's arrived?
<steev> nope
<WarheadsSE> you .. don't have one?
<steev> because i didnt' even know they shipped mine, and i thought FedEx/UPS (whichever it was) had a wrong number when they kept calling telling me they had a package for me from Japan
<steev> so i ignored it for like 4 weeks
<steev> and finally they were like this message is for steev - your package from japan is about to be shipped back if you don't call us back
<steev> so i did, and yeah...
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
jdeisenberg has left #linux-sunxi [#linux-sunxi]
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<bamvor> arokux: hi
vinifr has quit [Remote host closed the connection]
\\Mr_C\\ has quit []
<wens> arokux: i see it now. should've just waited
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
hipboi has joined #linux-sunxi
<steev> steev@oswin /storage/sandbox/sunxi-bsp $ ./configure cubieboard2
<steev> a20: unsupport SoC
<steev> why offer it?
<mnemoc> the bsp uses sunxi-boards, which is external, to "offer" boards
<steev> ahh
<mnemoc> even if the bsp itself doesn't know yet about a20 = sun7i
<steev> tis fine, i can build me a mini-x hwpack
<steev> still need to do something with it
<mnemoc> yes, you have commit access iirc :p
<steev> shhh
<steev> i think i do... but i meant i need to do something with my mini-x
<steev> and the cb2 still hasn't arrived
<steev> fingers crossed that it's sometime this week, but USPS... who knows when it will show up
<steev> i had something ship from indiana -> me overnight via usps... and i've had stuff take 2 months from a city away
<mnemoc> ...
<steev> usps isn't particularly..... reliable
<steev> it won't get stolen... usually... but you also have no idea when it will actually arrive
<steev> hm, is there any way to do a... non-standard video display on these things? like i have a monitor that does 1680x1050, but when i plug in the mini-x, it doesn't show anything as it attempts 720p
<steev> or maybe it's attempting 1080x
<steev> 1080p*
<mnemoc> as long as packages arrive, and without an extra invoice, it's good for me
<steev> haha
<steev> well, yeah
<mnemoc> modern sunxi kernels read edid without troubles
<mnemoc> and you can even override via kernel command line
<steev> what is a modern sunxi kernel, and does that mean not using the sunxi bsp?
<steev> :D
<steev> i've been away from sunxi for a while, but i'd like to get back to it
<mnemoc> good plan
<mnemoc> modern sunxi kernel is either the head of sunxi-3.0 or sunxi-3.4 in github's linux-sunxi
<mnemoc> but 3.0 is deprecated, and iirc the bsp is still using that one :(
<mnemoc> but not sure
<mnemoc> 3.4 has A20 support, 3.0 doesn't
<oliv3r> gooooood morning
<mnemoc> vieeeeeetnaaaammmm
n01 has joined #linux-sunxi
<mnemoc> Turl: thanks!
BluesBoy has joined #linux-sunxi
panda84kde has joined #linux-sunxi
rellla has joined #linux-sunxi
<hno> morning
<mnemoc> good morning hno
Soru has joined #linux-sunxi
<ojn> Copyright (C) 2012-2016 Allwinner Ltd.
<ojn> Hah! :)
<mnemoc> future thinking
<ojn> :-) Just glancing at the a20 smp patchset. I'll let mripard handle the first pass of reviews I think -- it's great to see new contributors!
<steev> mnemoc: yeah looks like the the bsp uses 3.0 still, okay, i'll look into the 3.4 kernel
<mnemoc> steev: just switch branches
<steev> s/kernel/branch/
bamvor has quit [Quit: Leaving]
<mnemoc> make linux-sunxi/.git; cd linux-sunxi; git checkout -b sunxi-3.4
<mnemoc> make linux-sunxi/.git; cd linux-sunxi; git checkout -b sunxi-3.4 origin/sunxi-3.4
notmart has joined #linux-sunxi
<panda84kde> mnemoc: even better "git checkout --track origin/sunxi-3.4"
bamvor has joined #linux-sunxi
tzafrir has quit [Ping timeout: 264 seconds]
<mnemoc> panda84kde: isn't track assumed by default these days?
<panda84kde> mnemoc: It's equivalent to "-b blahblah origin/blahblah" but I find it more concise and less error prone if you want the local branch be named as the remote one
<mnemoc> true, foo origin/foo is annoying
vicenteH has quit [Read error: Operation timed out]
<oliv3r> good morning sunxi :p
<mnemoc> it would be fun to see how different people prononunces "sunxi" :p
<oliv3r> Sun Ksi
<oliv3r> though i don't know how to write pronanciation
<oliv3r> i suppose yanks would say sun-si
<oliv3r> hopefully today will be better to do some hacking
<JohnDoe_71Rus> san hi
<JohnDoe_71Rus> :)
<moofree_> sun-ecks-eye
<oliv3r> oh, i didn't even think of the 'san' bit, but san-si
<rm> Sun Ksi: the ARM of War
panda84kde has quit [Quit: Konversation terminated!]
<oliv3r> rm: hahah yeah
popolon has joined #linux-sunxi
Black_Horseman has quit [Remote host closed the connection]
drachensun has quit [Ping timeout: 260 seconds]
drachensun has joined #linux-sunxi
<eebrah> haha "<rm> [11:22:33] Sun Ksi: the ARM of War"
<arokux> bamvor, hi, nice to see you here on an ultimate sunxi land
<bamvor> arokux: hi
<arokux> morning oliv3r
<bamvor> i am trying to make a bootable sd card by myself.
<bamvor> i could boot boot0 and boot1 from sd card successfull, but boot1 boot into nand or sdcard.
<arokux> bamvor, have you followed:
<arokux> bamvor, ?
<arokux> bamvor, in addition
<bamvor> i want to, but i do not know how to get the u-boot spl work for sun6i.
<bamvor> sorry, i mean for sun6i.
<arokux> bamvor, what do you mean?
<bamvor> it seems that there is not a31 support in that wiki?
<arokux> bamvor, maybe, pppl do not like A31 here because of its GPU being powervr
<arokux> bamvor, but see this: If your board is not listed then see "Adding a new A1x board" below
<arokux> bamvor, A1x should be really changed to AXXx
<mnemoc> i don't think anyone has an a31 with uart yet...
<mripard> at least bamvor, ojn, drachensun and myself have one
<mripard> arokux: the point is that the A31 is pretty much completely different when it comes to u-boot
<mripard> so you can't just follow what's on the wiki
<bamvor> mnemoc: mele M9 include uart. i am using it.
<arokux> mripard, oh..
<mripard> you have to write support for the new RAM controller, the new I2C bus that talks to the new PMIC, etc.
<bamvor> ah...
<arokux> bamvor, can you please post some nice images of the board to our wiki?
<bamvor> yes. i will.
<arokux> mripard, didn't know you are into u-boot stuff too :)
<mripard> arokux: I'm not, that's why we don't have any ongoing effort to support it yet :)
<mripard> but I don't have enough time to work on it
<mripard> it's a small item on my todo list
<mripard> oliv3r started working on it quite some time ago
<arokux> mripard, as you are here I wanted to ask you to cc linus-sunxi on everything related to sunxi
<mripard> but since he doesn't have an A31 device, and doesn't want to have one, it's been pretty dead ever since
<arokux> mripard, following linux-arm takes lots of time
<mripard> hmmm, I usually do, except when I forget
<mripard> what patches have I forgotten to send there?
<arokux> and this
<arokux> mripard, ^
<mripard> ah, indeed
<mripard> and yet, when it comes to mainline kernel, patches are expected to show up on LAKML, not anywhere else, so you can't expect to have all the patches sent to linux-sunxi.
<arokux> mripard, yep. one can add as many CCs as needed or?
massi_ has joined #linux-sunxi
bobby_ has joined #linux-sunxi
hipboi has quit [Quit: Leaving]
<mripard> arokux: I don't understand your question
<arokux> mripard, ok. I mean I totally understand the primary recipient should be LAKML, but one can add as many CCs as needed. it is just nice if everything related to sunxi shows up on linux-sunxi.
<arokux> mripard, so that I (and maybe others here) can keep their list of mailing lists at minimum :)
<mripard> arokux: yep
<mripard> my point was, you can't expect anyone to know and be aware of
<arokux> mripard, yes. you are right. I'll ask all those ppl to CC linux-sunxi in the future. (as I asked cini) but I'm surprised not everybody knows about linux-sunxi :)
<mripard> why should they?
tzafrir has joined #linux-sunxi
<arokux> mripard, I do not know, they want to build the kernel or smth sunxi related then they google and appears to be the fist link :)
<Turl> mripard: ping
panda84kde has joined #linux-sunxi
<oliv3r> set term_charset UTF-8
<Turl> oliv3r: error: UTF-8 is too awesome for this terminal
<oliv3r> actually, that is an irssi setting
<oliv3r> :)
<oliv3r> i'm getting compiler errors that are hard to read :(
<Turl> clang all the things
<oliv3r> invalid type argument of â->â
<oliv3r> its just really annoying :(
<Turl> oliv3r: what lang?
<arokux> compiling sunxi-3.4 code?
<oliv3r> Turl: en_US.UTF-8
<oliv3r> arokux: mainline :p
<Turl> oliv3r: I mean code
<Turl> so C
<Turl> weird :p
<Turl> is it your own code?
<oliv3r> havent compiled 3.4 in months :p
<oliv3r> ynezz: yes
<oliv3r> Turl: yes
<Turl> maybe you mistakenly typed some hidden utf8 on that line
<Turl> has happened to me
<Turl> remove it entirely and rewrite by hand
<oliv3r> Turl: i'm guessing as my terminals/screen aren't doing the utf-8 dance
<oliv3r> and that line i typed at home using gvim
<oliv3r> but very good tip; hidden charactesr *sigh*
<oliv3r> arokux: as mripard said, a31 is a completly new beast, and i will put effort into it, if it turns out that a40 will use the same i2c/memcontroller
<oliv3r> arokux: here's my very intial stuff
<arokux> oliv3r, I see, good to know
<oliv3r> but yes, i dont' want a powerVR based device
<oliv3r> it'll just frustrate and annoy me
<oliv3r> and i really don't want to have to use the blob
<oliv3r> especially on how far we are with a20 stuff
<oliv3r> i should ask benn this
<arokux> oliv3r, but for headless set ups it is still a quadcore..
<arokux> oliv3r, we shouln'd necessarily support GPU, should we?
rz2k has joined #linux-sunxi
<oliv3r> ssvb: have you seen peter chu's reply on the ML :)
<oliv3r> anybody google translated that? :)
<oliv3r> arokux: yeah but we'd put efforts into half a platform, personally, i'd say just skip the a31 and focus on whatever comes next :)
<oliv3r> we don't do anything for sun3i either :p
<oliv3r> IF however a40 or whatever, is almost like a31, but bigger/better/stronger, then yeah a31 will get 'free' support
<arokux> oliv3r, I see.
<Turl> oliv3r: chinese is so compact :)
<mripard> Turl: pong
<rm> oliv3r, Gōngnéng méi chābié; duōxiè;
<rm> HTH
<oliv3r> i see only hari-kiri here
<oliv3r> all ascii with ~ and (c) stuff
<oliv3r> defo no utf
<oliv3r> oh whitequark
<oliv3r> i see different combinations, but still the same crap :)
<bobby_> hello
<bobby_> can someone "translate" me the following error:
<Turl> mripard: I rewrote my mem driver as per comments
<bobby_> I'm trying to build my own image based on ALARM for Cubieboard 2
<oliv3r> bobby_: try updating your u-boot :)
<Turl> mripard: what do you think of "simple-memory-controller" as a compatible for it?
<oliv3r> bobby_: looks like your booting 'falcon' mode
<bobby_> ok, thanks I'll compile a new u-boot then
<arokux> bobby_, hey
<arokux> bobby_, this is problem with u-boot installation
<arokux> bobby_, make sure you follow:
<oliv3r> Turl: doesn't seem to be a hidden character issue
<oliv3r> oh wait i think is see now
<mripard> Turl: it looks fine for me
\\Mr_C\\ has joined #linux-sunxi
massi_ has quit [Remote host closed the connection]
<arokux> can't AW ppl actually speak english???
<arokux> how do they read data sheets?
<Turl> mripard: I still need to test it, but this is what I'll send if it works
<Turl> arokux: he probably didn't notice the cc'd lists were in english :p
<mnemoc> arokux: reading english is VERY different than feeling confident enough to speak it
<mnemoc> arokux: and it's specially scary for them because they don't have grammar, and fear to write something shameful
<arokux> Turl, why don't you use devm_*?
<arokux> my office colleague is Chinese
<Turl> arokux: because? :) dunno
<Turl> I don't request any resource other than the clock, and I clean it up early
Black_Horseman has joined #linux-sunxi
<mripard> Turl: looks fine :)
<arokux> Turl, why your code doesn't have an .exit func?
<mripard> why would you want to remove a memory controller driver from the system?
<oliv3r> arokux: they use chinese datasheets :)
<oliv3r> hotpluggable memory
<arokux> mripard, Turl ah, ok, I've thought it is something for mmc stuff..
<arokux> but then what is the point of it being a module?
<mnemoc> multiplatform support
<Turl> arokux: it's a boolean module
<mripard> Turl: you don't need the of_find_matchning _node though
<Turl> mripard: how do I get the node then?
<mripard> pdev->dev.of_node
<mnemoc> the world seriously needs an ldd4
<arokux> mnemoc, yes!
<Turl> mripard: great then :)
<arokux> Turl, what is meant by "boolean module"?
<mnemoc> ...
<mnemoc> but there is hope
<Turl> arokux: kconfig can only be y/n
<Turl> mripard: IRCv2
<mripard> Turl: you can then drop the of_node_put :)
<Turl> arokux: so you can never build it as module
<ssvb> oliv3r: yes, though I pretty much already knew that :)
<Turl> mripard: good catch :p forgot about that one
<mripard> Turl: and the of_match_table pointer should be in a of_match_ptr()
<mripard> other than that, it's fine, I guess you can send it
<ssvb> oliv3r: but it's surely good to have an official confirmation
<Turl> mripard: any reason why of_match_ptr ?
<Turl> mripard: looks like a macro to unuse stuff on the !OF case
<Turl> but the module depends on OF
deasy has joined #linux-sunxi
<mripard> Turl: just to be safe.
<Turl> it won't hurt, I can add it if it's just for the sake of consistency :)
FDCX_ has joined #linux-sunxi
<mripard> not only that, but say if someone needs your driver for a board-style architecture, make the patch and send it to you
<mripard> I'm pretty sure you'll have forgotten about it by then :)
<arokux> what starring means on github? "Like" ?
<mripard> (I know I will)
<Turl> mripard: good point
<Turl> mripard: yeah, I tend to forget about a lot of stuff too :)
<bobby_> I compiled the latest sunxi u-boot and it is still the same:
<bobby_> what I'm missing?
<Turl> mripard: who maintains drivers/memory?
<arokux> bobby_, how have you burned it to the sd card?
<arokux> bobby_, ah, I see
<arokux> bobby_, so you haven't followed doc on wiki
<arokux> bobby_, this one is wrong: sudo dd if=u-boot.bin of=/dev/sdb bs=1024 seek=32
<oliv3r> bobby_: i think you are trying to boot 'falcon mode' and didn't supply a direct kernel image
<oliv3r> bobby_: try flashing the u-boot-spl-inonebinary-orsomelongnameiforgot.bin
<bobby_> so it should be: dd if=u-boot.img of=/dev/sdX bs=1024 seek=40 ?
<oliv3r> bobby_: yes, for u-boot
<oliv3r> SPL is at 8; u-boot.img is at 40
<oliv3r> but if you take the 'combined' file, you can just put it at 8 and ti works
<bobby_> ok, I'll try that
<oliv3r> try u-boot-with-spl.img i think the name was
<oliv3r> success guaranteed
<arokux> oliv3r, u-boot-sunxi-with-spl.bin
<arokux> not sure if *.bin == *.img
<oliv3r> that sounds likely
<oliv3r> i figured by making obvious i'm guessing bobby_ would figure it out ;)
<oliv3r> yay, finalyl compile error free
rz2k has quit []
rz2k has joined #linux-sunxi
<oliv3r> who is bored?
<mripard> Turl: no idea
<mripard> what MAINTAINERS say?
<Turl> mripard: nothing
<oliv3r> i didn't know until yesterday that wingrime wanted to do this; so I allready appologized to him
<oliv3r> only compiled tested it as i belive there's still required bits missing
<oliv3r> well turl's stuff is missing i think
<oliv3r> oh i forgot to commit a file :S
<arokux> oliv3r, you can use __set_bit/__clear_bit
<oliv3r> ah was looking for those macro's
<oliv3r> wait, i haven't pushed my file yet :p
<arokux> there are also set_bit/clear_bit, but those do a bit more
<mripard> Turl: then maybe andrew morton
<mripard> I don't really know
<oliv3r> there, that's better
<oliv3r> arokux: i rather not use __ functions unless I really have to
<arokux> oliv3r, why?
<arokux> ppl do it.
<oliv3r> __ are concisderd 'internal unmanaged' functions
<arokux> oliv3r, are they subject to change any time without a notice?
<oliv3r> well look at clear_bit for example
<oliv3r> clear_bit does an _atomic_spin_lock_irqsave before and after it
<oliv3r> i bet __clear_bit doesn't
<oliv3r> so clear_bit is generally safer
Black_Horseman has quit [Remote host closed the connection]
<arokux> and do you need atomic? you haven't used it.
<arokux> oliv3r, onboard AHCI SATA --> on-chip AHCI SATA
<oliv3r> heh, i just copy pasted that from imx :p
<oliv3r> it may be reordered, and things may fail
<arokux> oliv3r, my point is you used only readl/writel and __ func does just that.
<oliv3r> arokux: as for the other changes, those modifications i'll keep 'as the rest of the file is' as for ahci_sunxi.c, i put in some functions, might use set_bit for it instead; not a horrible idea :p
<oliv3r> arokux: i'm missing your point
<arokux> oliv3r, you need atomic if there are two or more simultaneous attempts to set bit
vinifr has joined #linux-sunxi
<arokux> oliv3r, but in your case nobody else will ever want to touch that bit, or?
<oliv3r> arokux: multiple drivers :p
<oliv3r> i dunno yet
<arokux> oliv3r, and so far you have used readl/writel which also do not care about atomicity. so i you want to replace your code literally you need to use __set_bit.
<arokux> oliv3r, anyway, I do not know if it is a big deal to use set_bit just in case. I just wanted to do save some CPU cycles...
<bobby_> thanks guys, it booted with the u-boot-sunxi-with-spl.bin :)
<oliv3r> arokux: saving cpu-cycles is always good :)
<oliv3r> arokux: i probably should use ioread32 anyway
<oliv3r> arokux: as those are the higher lvl fucntions for reading/writing :)
<oliv3r> arokux: also, doesn't set_bit write only a single bit??
<arokux> oliv3r, why not switch to (__)set_bit?
<oliv3r> e.g. isn't set_bit() the same as BIT() sort of?
<arokux> oliv3r, yes, it does
<arokux> BIT doesn't write anything
<oliv3r> e.g. set_bit(5, addr) is the same as BIT(5) which is (0x1 << 5)?
<arokux> bobby_, cool. what hardware do you own?
<oliv3r> yeah i know, but that's besides the point
<oliv3r> arokux: what i'm saying is, set_bit modifies a single bit
<oliv3r> so how do I use setbit for 0x1f << 3?
<arokux> oliv3r, correct, only one bit.
<oliv3r> set_bit(3, addr), but what happens to 0x1f
<arokux> oliv3r, well, if you figure it out, tell me :)
<bobby_> arokux: Olimex A20-OLinuXino-MICRO
<oliv3r> arokux: is the later commit, look at that :)
<oliv3r> bobby_: i use that one too
<oliv3r> arokux: and then link the line you'd like to see 'set_bit()' for
<arokux> oliv3r, I see your point.
vicenteH has joined #linux-sunxi
<arokux> oliv3r, and I've also asked myself why there is no set_mask or smth ((
<Turl> oliv3r: you could make those sunxi_* inline
<Turl> oliv3r: but you basically reimplemented the regmap api I think :p
<Turl> oliv3r: the ifdef on drivers/ata/libahci.c needs to go too, won't work with multiplatform
<oliv3r> Turl: compiler opitimzies those, don't inline! :)
<Turl> oliv3r: btw, talk with wingrime, I believe he had some of SATA implemented and was waiting on me for clocks
<Turl> oliv3r: if you mark them 'static inline', sure it will
<Turl> if you don't.. dunno :) better mark them as such imo
<Turl> oliv3r: marking as inline doesn't guarantee you that they will as far as know, it's just a suggestion to the compiler
<Turl> oliv3r: like 'register int a;'
<oliv3r> Turl: yeah but i don't know the proper way to fix the #ifdef yet, first it needs to be confirmed working etc, then we can start moving bits around, i tried ot keep it as close to the AW code as possible
<oliv3r> Turl: wingrime said he was going to start on sata once you did the clocks; talked to him yesterday when i found out he wanted to do sata
<oliv3r> Turl: if i'm not mistaken, you should never mark something inline, the compiler will do it automatically and will do it smarter then you; mripard even agreed to that yesterday :)
<oliv3r> Turl: same goes for regsiter, one is not supposed to define it as such,t he compile knows better (most of the time)
<Turl> oliv3r: yes, agreed on register
<Turl> oliv3r: but SubmittingPatches says 3) 'static inline' is better than a macro
<oliv3r> Turl: but i dont' do either ;)
<oliv3r> Turl: that LWN article is good read
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier.]
<Turl> oliv3r: well yout bit functions could easily be in a macro :p
<oliv3r> yeah
<mripard> the LWN article is not really relevant to the discussion
<oliv3r> i wanted to do first do a simple macro, then i rememberd to better use static inline; but then i remembered it's better to not inline at all ;)
<oliv3r> The problem with this is that to be effective, inline functions must be defined in header files
<oliv3r> strange, i read why it's done by the compiler and not the linker, but that's compiler stuff i don't want to understand ;)
<mripard> no.
<mripard> functions in the headers are inlined because otherwise it would be plain broken.
<oliv3r> that's a copy/paste from
<mripard> it's not the other way around.
vicenteH has quit [Ping timeout: 264 seconds]
<oliv3r> mripard: well i have no clue whether it should or should not be inlined, i can mark it inlined for now and let the maintainers decide later what they want?
<oliv3r> to me its either or :p
<oliv3r> mripard: that's the bit I read! thanks for re-finding it :p
panda84kde has quit [Ping timeout: 240 seconds]
<oliv3r> mripard: so the only reason to mark it inline is because it has 3 or less lines of code
<oliv3r> ignoring the variable declaration
<oliv3r> Turl: about the ifdef again, once we know the code works etc, this could either be moved to some sunxi init bit, or it might some other solution, I honestly don't even know what AW did there and why, since the ahci bit is supposed to be pretty standard, so is this a bugfix for the AW IP? is this some thought process error, I don't know
<Turl> oliv3r: let's not worry for now
<Turl> lets make it work first :p
<oliv3r> Turl: that was my thought ;)
<oliv3r> Turl: is my useage of 'clocks' correct here though?
<oliv3r> Turl: never used clocks :p
<oliv3r> oh, i reliaze i forgot to add the 'dt'
<Turl> oliv3r: looks sane
<oliv3r> Turl: are those cocks implemented yet?
<Turl> oliv3r: I don't think linux implements cocks yet :)
<Turl> oliv3r: but yeah, I have pll6 implemented
<Turl> not yet merged though
<Turl> the other one should be in there already in torvalds' tree
<n01> cocks?
<oliv3r> n01: cLocks!
<oliv3r> Turl: and the prepare/enable/disabling bit looks sane and at the sane positions?
<n01> ;)
<Turl> oliv3r: yeah, lgtm
<mnemoc> sunxi-v3.4.43-r2 and sunxi-v3.4.61-r0 tags up, reference-3.4, sunxi-3.4 and stage/sunxi-3.4 branches updated
<Turl> mnemoc: :D
<vinifr> mripard, from the perspective of IIO framework, what can be used as trigger?
<Turl> bbl
FDCX_ has quit [Ping timeout: 245 seconds]
<rellla> mnemoc: good!
<mripard> vinifr: anything that can trigger a conversion on your ADC
<rellla> btw is anyone busy pushing mali r3p2 into kernel? i saw, libv + benn updated mali-proprietary
<libv> rellla: ssvb had r3p2 working yesterday
<libv> rellla: about to send an email out about the updated repo
<rellla> ok :)
<mripard> vinifr: but i don't think there's any for the LRADC
<vinifr> mripard, hm, then depends controller
<mripard> yes
<arokux> mnemoc, I want to understand something. suppose a new user comes with normal hardware i.e. (A10/A20), what kernel should he use?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<rellla> i just pushed the new cedar blob into cedarx-libs. without reorganizing directory structure and testing. only dos2unix'ing :p
nove has joined #linux-sunxi
<rellla> just in case someone wants to try it
<vinifr> at91 has software and hardware triggers from what I saw, it is very complete :)
<nove> can someone confirm at A13 that usleep(10), takes 10ms?
<ssvb> libv: the addition of r3p2-01rel2 kernel module is a very hackish patch at the moment, first I need to clean it up a bit before submitting
<mnemoc> arokux: 3.4
<arokux> mnemoc, but which tag? or what is the purpose of tagging?
<libv> ssvb: did you get any noise on shutdown of X11 applications?
<mnemoc> arokux: -r0 means "just after the version jump, might have regressions"
<libv> like we had before i hacked the r3p0 binary?
<mripard> vinifr: yep, and it's even a touchscreen controller
<ssvb> libv: did not have time to test it yet
<mnemoc> 3.4.43-r2 is known to work, and the last 3.4.42 before the version jump
<mripard> but that's expected, the at91 are industrial chips
<arokux> mnemoc, so tags are more for us then for end users?
<mripard> having good ADCs is always nice for these applications
<libv> ssvb: that's ok :)
<mnemoc> arokux: if user has troubles with sunxi-v3.4.61-r0 first suggestion will be, please try with 3.4.43-r2
<arokux> mnemoc, I see, but first the user needs to git clone default branch, which is sunxi-3.4?
<mnemoc> yes
<mnemoc> and currently matching sunxi-v3.4.61-r0
<libv> ssvb: just reply that the kernel side is still a bit hackish, and that r3p0 just works at 2/3rds the speed for the time being
<arokux> mnemoc, ok, thanks
<mnemoc> arokux: sunxi-v3.4.61-r1 will be the first known to work fine sunxi-v3.4.61..... because -r0 only has built time tests
<mnemoc> + known to work features, fixes, etc .... coming from stage/sunxi-3.4
<arokux> mnemoc, I just thought that end users should continue to use 3.4.43-r2, and that -r0 is for internal testing only
<mnemoc> sunxi-v3.4.43-r2 is known to work. sunxi-v3.4.61-r0 is only known to compile
<mnemoc> arokux: we aren't that big... we need people testing HEAD :p
<mnemoc> not hidden behind the latest "known to work fine" tag
<arokux> mnemoc, yes, I also want the end users to help us = give something back
<arokux> mnemoc, thanks for answers
<mnemoc> stage/sunxi-3.4 otoh is intended only for devs
<mnemoc> and patches approved on the ML go there, for real life testing before merging them to the sunxi-3.4 branch (for users)
atiti has quit [Read error: Connection reset by peer]
vinifr_ has joined #linux-sunxi
atiti has joined #linux-sunxi
<arokux> mnemoc, so we here really need to run stage/sunxi-3.4 :)
<mnemoc> yes
vinifr has quit [Ping timeout: 245 seconds]
vini__ has joined #linux-sunxi
vini__ has quit [Read error: Connection reset by peer]
vinifr_ has quit [Ping timeout: 245 seconds]
<libv> it seems that the sunxi-mali repo is quite popular for other projects as well
<libv> none of them are bringing any of their changes back to sunxi of course...
<mnemoc> :(
<libv> things like .spec files and a few .pc files
<mnemoc> libv: btw, please don't forget to send your address to benn for the CT
<libv> mnemoc: i already did :)
<mnemoc> gut :)
<libv> mnemoc: but thanks for reminding me :)
<libv> mnemoc: this does seem to be a trend for sunxi
<libv> there are a load of projects which seem to depend on us
<libv> but only few actually contribute back
<libv> in my view, fewer than with the individual software components that i know of
<mnemoc> with mali and the new open source video decoder we are the most open cortex-a platform
<mnemoc> lima
<libv> well, if lima was worth something already :p
<mnemoc> more than any other RE GPU project can tell
<libv> freedreno is further, but it lacks an SoC with the level of support of sunxi :)
<libv> so in a way lima depends on sunxi as well :)
<specing> qucikly relicense lima under "contribute back or we will sue your ass off for $10000000"
<libv> specing: i think i am mostly whining about how bad things are with the fish that just crawled out of the android swamp i guess
<mnemoc> :)
<nove> wingrime, jemk, vdpau is working fine at A13, the slowness was caused by usleep that takes about 10ms
jemk has joined #linux-sunxi
rzk has joined #linux-sunxi
<jemk> nove: that's strange, the usleep is supposed to sleep 10us, not 10ms
tzafrir has quit [Ping timeout: 260 seconds]
rz2k has quit [Ping timeout: 260 seconds]
tzafrir has joined #linux-sunxi
jelly-home has quit [Ping timeout: 264 seconds]
<nove> jemk, here A13 kernel 3.0, nanosleep also takes minimum 10ms
<nove> jemk, i removed the sleeps, and until now is decoding ok
rzk has quit [Client Quit]
<jemk> nove: removing should work, but i don't find out which bit to check if it is finished, so i added some sleep to be sure
<jemk> nove: in traces there is a status read, but that doesn't tell me the right bits
hurtigbuffer has joined #linux-sunxi
<nove> jemk, lets check, i will lie to the blob the value of the status register and see what happens
hurtigbuffer is now known as jelly-home
notmart has quit [Quit: notmart terminated!]
BluesBoy has quit [Quit: HydraIRC -> <- The alternative IRC client]
bobby_ has left #linux-sunxi [#linux-sunxi]
<ssvb> jemk: can you possibly have a look at the last read after which the status register polling stopped?
<ssvb> jemk: it would be natural to assume that the needed bit would change its status
<jemk> ssvb: it gets only polled once ;)
<jemk> ssvb: cedar is too fast
<ssvb> jemk: downclock it :)
<jemk> ssvb: im trying, but divider isn't enough and i don't know what will happen when changing pll4
<ssvb> jemk: also is there no irq involved there?
<jemk> ssvb: no, it isn't, also for reading a single byte interrupt would be overkill i think
wingrime has joined #linux-sunxi
<ssvb> jemk: sleeping for reading a single byte is definitely an overkill (if I understand the current situation right)
<jemk> ssvb: yes, of course, but this is left over from the time everything didn't work and i wanted to be sure the errors aren't because of that
<jemk> wow, cedar even docodes pretty fast with only 24MHz
naobsd has joined #linux-sunxi
rz2k has joined #linux-sunxi
<wingrime> jemk: no
<wingrime> jemk: driver aproximate freq
<jemk> wingrime: yes, sure, i used devmem
<wingrime> jemk: are you changed pll?
<wingrime> jemk: pll can get wrong freq if you change it when enabled
<jemk> wingrime: yes, after i noticed that it isn't used anywhere else i thougt it would be safe
<jemk> wingrime: but enabling bypass ("If the bypass is enabled, the PLL4 output is 24MHz.") should work
<wingrime> jemk: so, it fast even on such speed?
<jemk> wingrime: 5-10fps i think, with 854x480 video
<jemk> wingrime: so not *really* fast...
vicenteH has joined #linux-sunxi
rellla has quit [Quit: Nettalk6 -]
FDCX_ has joined #linux-sunxi
woprr has quit [Remote host closed the connection]
n01 has quit [Quit: Lost terminal]
notmart has joined #linux-sunxi
hipboi has joined #linux-sunxi
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
hipboi has quit [Client Quit]
vinifr has joined #linux-sunxi
woprr has joined #linux-sunxi
Paullo612 has joined #linux-sunxi
<nove> jemk, the blob gets in infinite loop pulling the status register, when lied that is 0xffffffff
<Paullo612> Hello. I just pushed VP8 decoding PoC to
<jemk> nove: surprise ;)
<Paullo612> Now it does not use any proprietary code
tzafrir has quit [Ping timeout: 264 seconds]
<Paullo612> Just rewritten everything from scratch based on jemk's MPEG2 decoding PoC code.
<jemk> nove: i'm trying to find the right bit by downclocking, but either there is no such bit or it is still too fast
<jemk> jemk: well, there has to be one, otherwise you wouldn't have got infinite loop
<arokux> Paullo612, can you please create a quick wiki page with requirements how to test it?
massi has joined #linux-sunxi
<nove> Paullo612, what you know about MACC_H264_STATUS?
wingrime has quit [Ping timeout: 260 seconds]
<Paullo612> nove, there is 8'th bit that indicates H264 engine busyness when it reads basic bits.
<nove> Paullo612, jemk and i are trying to see with bit means function finished
popolon has quit [Quit: Quitte]
<nove> Paullo612, i saw that in the vp8 poc, so i asked
deasy has joined #linux-sunxi
<Paullo612> nove, the 8'th bit is set on while the function is in process. But i saw it in this state just few times. I think it is because of slow (?) mmap.
<deasy> plop
<Paullo612> nove, another problem is that there is cached access to ve memory in kernel module. I don't know how to sync mmap'ed memory from userspace before triggering any engine functions that works with probabilities table.
wingrime has joined #linux-sunxi
<wingrime> jemk: clear-room concept
<wingrime> jemk: we should not broke it
<wingrime> mnemoc: ping
<Paullo612> nove, tryed msync with no luck. I actually don't know how libvecore works without syncing.
<wingrime> Paullo612: can you help with h264 encoder?
<nove> Paullo612, i don't know any about memory caching, and kernel way of working
ssvb has quit [Ping timeout: 248 seconds]
<Paullo612> wingrime: I can try. Actually i can do the same that i did with VP8. I can reverse all asm code to C and write open realization based on it.
<wingrime> Paullo612: you have hex rays arm plugin?
<wingrime> Paullo612: you can't write open realization, as CLEAR ROOM here
notmart has quit [Quit: notmart terminated!]
<wingrime> Paullo612: also VP8 have no vdpau support :(
<Paullo612> wingrime: no, i does not have hex rays for arm.
<wingrime> Paullo612: what decompiller are you used?
<Paullo612> wingrime: By brain and hands as decompiller. It's monotonic, but helps to learn\remember asm.
<wingrime> Paullo612: yeh, but I can't do it in this amount
<wingrime> Paullo612: mostly not required
<nove> i knew it, hand made
<wingrime> nove: time waste
<wingrime> libv: ping
<Paullo612> wingrime: there is VP8 support in staging of libva (VAAPI)
<wingrime> Paullo612: ...
<wingrime> Paullo612: bad idea , mostly
<wingrime> Paullo612: we have one api
<wingrime> Paullo612: but more one
<wingrime> nove: I can do work simualr, but this mostly useless work
<wingrime> nove: trace give more information usualy
<nove> wingrime, i know, the important is to know how the hardware works, not some binary blob
<Paullo612> wingrime: but there is no support of VP8 in it. By the way, I fixed at least three bugs of libvecore in VP8 PoC. libvecore can't play VP8 coded videos with segmentation enabled.
ssvb has joined #linux-sunxi
<wingrime> Paullo612: I interested only in mpeg4 and h264
<wingrime> Paullo612: mpeg4 -- easy , I thik jemk do it quick
<wingrime> Paullo612: but audio codecs and encoding not handled by that blob
<wingrime> Paullo612: strange android binaries, I can't easy trace
<wingrime> Paullo612: I can send binaries
<wingrime> Paullo612: and worst thing
<rz2k> did those binaries have source in the big SDKs?
<rz2k> s/did/didnt/
<rz2k> because I remember reading some video handling stuff in a20-sdk
<wingrime> rz2k: yeah with andoroid files
<wingrime> rz2k: I can't trace it
<Paullo612> nove: libvecore's VP8 part does a lots of useless work. For example, it builds quantizer tables that never used anywhere.
<wingrime> Paullo612: worst thing that we have parts that blob not covers
<wingrime> Paullo612: for example AVC module can do zlib's deflate
<wingrime> Paullo612: and can speed up PNG decoding
<wingrime> Paullo612: also, we have new china codec -- AVS jizun, but blob have only studs
<Paullo612> wingrime: Yes. And there is no docs on all of this stuff. And how we can implement it without any docs or code that uses this stuff?
<wingrime> Paullo612: thats problem, but we can ask ex allwinner enginers about it
<wingrime> Paullo612: also, beyond cedar
<wingrime> Paullo612: we have blob with GPS support
<Paullo612> wingrime: It would be cool.
<wingrime> Paullo612: don't expect miracles
<wingrime> Paullo612: we have no driver/no code for PATA
<Paullo612> wingrime: Hmmm. I don't think that there is GPS in my A10.
<wingrime> Paullo612: there interface for radio
<wingrime> Paullo612: also, Trust Zone
<wingrime> Paullo612: no driver no code for CAN
<wingrime> Paullo612: chip have strange things that not documented
<rz2k> TZ is arm ip core based
<rz2k> i bet you can launch it, but you need the blob from ARM
<rz2k> atleast on odroid there is a blob loaded by spl
<wingrime> rz2k, it have firmware?
<rz2k> yes, also on odroid once you enabled TZ, you cant launch various things without using the TZ, like you cant launch lx20 cache without invoking TZ call
<rz2k> good night
rz2k has quit []
Paullo612 has quit [Ping timeout: 250 seconds]
massi has quit [Remote host closed the connection]
sanka has joined #linux-sunxi
<techn_> anyone got r3p2-01rel1 android binaries?
n01 has joined #linux-sunxi
jemk has quit [Ping timeout: 260 seconds]
Soru_ has joined #linux-sunxi
Soru has quit [Ping timeout: 264 seconds]
<wingrime> oliv3r: kill me please
<wingrime> oliv3r: 01c0e0f0: 16230055 00000000 00000000 00000000 U.#.............
<mnemoc> wingrime: pong
<wingrime> mnemoc: I just builded kernel for a20
<wingrime> mnemoc: yes, emac works
<wingrime> mnemoc: but not selected by default
<wingrime> mnemoc: fix it?
<wingrime> mnemoc: also mistery , a20 cedar have 16230055 in version register
<mnemoc> wingrime: if emac works on a20, let's enable it until we get gmac
<mnemoc> same IP version.... funny
<mnemoc> i assumed they had upgraded cedarx too
<wingrime> mnemoc: most intersted that a13 have 16250055
<mnemoc> good we did't use the cedarx version for soc-detect :p
<wingrime> hno: ping
jemk has joined #linux-sunxi
<wingrime> jemk: 01c0e0f0: 16230055 00000000 00000000 00000000 U.#.............
<wingrime> jemk: in a20
<jemk> wingrime: good (maybe), that means nothing different i hope
<wingrime> jemk: 16230055
<wingrime> jemk: a10 equals a20 here
vrga has joined #linux-sunxi
focus_it has joined #linux-sunxi
<vrga> okay, if my tablet is skipping the uboot from the sd card, the symptom of it would be it simply going on to android, right?
<mnemoc> g2d too?
<wingrime> mnemoc: waybe (need investegation)
<jemk> wingrime: a13 works now too? can i clean up and add the patch from yesterday?
<wingrime> jemk: a13 works but very slow
<jemk> wingrime: nove says thats because of the usleeps in get_*()
<wingrime> jemk: and you tested patch on a10
<wingrime> jemk: need it ve check?
<wingrime> mnemoc: also no cedar on a20
<jemk> wingrime: patch didn't do any harm on a10, but i think i add check, we will need it sometime
<wingrime> jemk: I think you can add fix now, becose it works
<wingrime> jemk: in a20 case, driver not builded for a20
<wingrime> jemk: I will make cedar driver unification after I find what wrong with a13
<wingrime> jemk: and a20 will be too
<wingrime> jemk: also nice defaults 01c0e080: 40302010 80706050 c0b0a090 00f0e0d0 . 0@P`p.........
<jemk> wingrime: yes, sad we don't know much about this first block 0x0-0xff, seems like it is used more than we know
<wingrime> jemk: any idea what ISP does?
<wingrime> ISP Engine
<ssvb> mnemoc: g2d is identical in a10 and a20
<jemk> wingrime: i'd guess thats jpeg encoder and such things, at least on other SoCs this means "Image Signal Processor"
<wingrime> jemk: but we have encoder engine already
<nove> wingrime, in A13, usleep(10) take minimum 10ms, that is why is slow
<wingrime> nove: how it possible?
sanka has quit [Remote host closed the connection]
<nove> wingrime, i timed with clock_gettime, around 9ms
<wingrime> jemk: any idea how avoid it?
* vrga crosses fingers that his toolchain compiles...
<vrga> nope -.-
<wingrime> jemk: you can try use AVS2 API
<jemk> wingrime: see git, thats the right way to do it.
arokux2 has joined #linux-sunxi
<wingrime> jemk: nice, but add 13 fix too
<jemk> wingrime: yes, but i want to do it right, needs some time
<wingrime> mnemoc: seems I find way how ddr3 priority can be changed
<wingrime> mnemoc: yes, we looks like can change memory access priority
vrga has left #linux-sunxi ["Leaving"]
<wingrime> oliv3r: ping
<hno> wingrime, yes? Just ask whatever you want to ask instead of pinging.
<wingrime> hno: uboot looks hung if any unknown IRQ comes
<hno> wingrime, that's likely. u-boot don't use any IRQ and I don't think the IRQ vectors are even set up proper.
<hno> why are you getting interrupts in u-boot?
<wingrime> hno: I trying discover unknown things in cedar
<wingrime> hno: write 1 to some regs get IRQ,
<wingrime> hno: I mostly like to know witch register cause it
<hno> ok.
<wingrime> hno: but I have do clock setup procedure again and again
<hno> or make sure that the IRQ is masked.
akaizen has quit [Remote host closed the connection]
<jemk> wingrime: a13 is 0x1625 right?
<hno> you really do not want unknown IRQs delivered to the CPU. It needs to deal with theom somehow and clear the IRQ source before it can continue.
<wingrime> jemk: yes
<wingrime> jemk: also, make debug print for version
<wingrime> lkcl_: ping
<wingrime> ssvb: ping
ZetaNeta has quit [Ping timeout: 260 seconds]
<ssvb> wingrime: ?
<wingrime> ssvb: take a look
<wingrime> ssvb: looks we can change dram priority over devices
<ssvb> wingrime: yes, if it helps to resolve the screen shaking problem, that would be great
<wingrime> ssvb: also, not see shakes on a20 , (but without video)
<wingrime> ssvb: I need check with video
<techn_> wingrime: nice finding
arokux2 has quit [Quit: Leaving]
arokux2 has joined #linux-sunxi
<wingrime> techn_: oliver have documented it
<wingrime> techn_: but not noticed
<wingrime> 0x250 HPCR[32] host port configure register
<ssvb> wingrime: trying to increasing the priority for BE and FE might be a good idea
<jemk> wingrime: would you like to test latest vdpau on a13, should work now if i didn't mess it up
<ssvb> wingrime: and I wonder if right now the priority for the dispc FE is higher than for BE, which might explain why scaler mode behaves better for heavy resolutions such as 1080p32
<wingrime> ssvb: that I get from a20 (uboot)
<ssvb> wingrime: 5 in the lowest digit means that bit 2 is set, which probably means high priority
<wingrime> ssvb: correct
<wingrime> ssvb: I need someone do it on a10
<ssvb> wingrime: and it is set for BE, FE, CSI, USB, IEP
<ssvb> now we need to check the same on a10
<ssvb> :)
<wingrime> ssvb: sun7i# md 0x01c01250 50
<wingrime> in uboot
<oliv3r> wingrime: why should I kill you? 1623 is a13?
<oliv3r> (backreading, might take an hour :p)
<wingrime> oliv3r: find why in backlog
<wingrime> hno: have you uart and a10 nearly?
<ssvb> wingrime: heh, my a10 is currently unplugged (I'm working with cubieboard2 now)
<ssvb> wingrime: probably need to find more space on the desk and more wires to have them both operable simultaneously
<wingrime> ssvb: can priority be usefull somesomethig else?
<jemk> wingrime, ssvb: a10 is
ykchavan has joined #linux-sunxi
<wingrime> jemk: thanks
<ssvb> at least looks like both DE and BE have this bit set also on A10
<techn_> ssvb: there is 2bits reserved for that
<techn_> or may be
<wingrime> ssvb: yes, but it different
<ssvb> now I wonder why we have "00000735 18 BE" but "00001035 19 FE", is BE discriminated somehow?
<wingrime> command numer
<wingrime> command number - 10 or 7
<wingrime> command number - 0x10 or 0x7
<ssvb> if 10 is better than 7, we should go for it :)
mturquette has quit [Quit: leaving]
<wingrime> ssvb: arch/arm/plat-sunxi/include/plat/dram.h: DRAM_HOST_FE = 19,
<wingrime> ssvb: we can try change it boot
<wingrime> *at boot
<leviathanch> mripard: do I need to do anything specific about pinctl?
mturquette has joined #linux-sunxi
<arokux2> hi leviathanch
<arokux2> leviathanch, got mmc working?
<leviathanch> hi arokux2
<leviathanch> unfortunately not yet
<leviathanch> the sd card still gives me a timeout
<leviathanch> even when I remove and reinsert it from the slot
<leviathanch> so maybe, now that I successfully have enabled the clock
<techn_> sun7i has mctl_configure_hostport which configures those
<leviathanch> another question is: does the clock make it to the output pin?
<techn_> in dram_init.c
<leviathanch> techn_: ah, ok
<leviathanch> hmm
<leviathanch> ah
<leviathanch> techn_: was that for me?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<techn_> leviathanch: sorry.. I was participating dram discussion :)
<leviathanch> ah, ok
<leviathanch> sorry
<Turl> howdy
<leviathanch> Turl: hi
<leviathanch> Turl: do I need to do something about pinctl?
<leviathanch> because the driver now appears to be setting all clocks correctly
<leviathanch> and the host controller even generated interrupts now
<leviathanch> but the sdcard times out
<leviathanch> even when I remove and reenter the card
<leviathanch> 400kHz SPI mode should work immediately after insertion
<leviathanch> I think the pinmuxing isn't right
<hno> wingrime, what you want me to test?
n01 has quit [Read error: Connection reset by peer]
n01 has joined #linux-sunxi
<wingrime> hno: too late
<Turl> leviathanch: did you define the correct pinmuxing on the dt?
<wingrime> hno: I discoved that we can change dramc prioroty over devices
<hno> in the hpcr I presume.
<wingrime> hno: are you do something with it in uboot
<hno> I am only loading the known default tables as blobs. Not actively changing anything.
<wingrime> jemk: still slow
Paullo612 has joined #linux-sunxi
<hno> the structure of the hpcr registers became known long after we had those tables.
<hno> and still unknown what many of the ports are used for.
<wingrime> jemk: but maybe better
<jemk> wingrime: now i don't have more ideas.
<wingrime> nove: ping
<jemk> nove: you only removed usleep and it was faster?
<Turl> hno: do you recall if mmc pins take any special drive or pull parameter?
<hno> Turl, no I don't.
<Turl> leviathanch: your sdc node seems to lack pinctrl-names = "default";
<wingrime> hno: you have different blob for a10/a20 ?
<nove> jemk, yes like that, also with vdpau:fps=100 it speed up playback
<hno> wingrime, yes. See u-boot/arch/arm/cpu/armv7/sunxi/dram.c
<techn_> Turl: do we want to preserve old lcd backlight interface.. or is new /sys/class/backlight/pwm-backlight/brightness enough?
<arokux2> hno, some hcpr bits are documented in usb related code
<techn_> Turl: on android point of view ^
FDCX_ has quit [Ping timeout: 245 seconds]
<techn_> gn
<leviathanch> Turl: sdc0: sdc@01C0f000 { .. pinctrl-names = "default"; .. } ?
<Turl> techn_: as long as it works I suppose either is ok
<Turl> leviathanch: yes
<wingrime> ssvb: uboot magic
<arokux2> 0x0301 is familiar to me, don't touch it guys :)
<Paullo612> And what about all this CedarX mmap stuff? Anybody knows how to sync mmap'ed memory fom userspace?
<hno> arokux2, that looks identical to what I have.
<arokux2> hno, yes, sorry, I've compared only to DRAM_HOST_*
<hno> Those too are identical. USB1 = 4, USB2 = 5.
<hno> two lines up in the sw_hci_sunxi.h file.
<wingrime> Paullo612: there ioctl
<arokux2> hno, yes
<arokux2> but the thing I reference is also on the page i've given
<arokux2> you've*
<arokux2> hno, unsigned int AcsEn:1;
<arokux2> hno, etc.
<wingrime> hno: can you remake using known bits fields
<nove> jemk, wingrime, just tested with the new changes, works ok
<arokux2> hno, grepping for HPCR in sunxi-3.4 will also reveal some spots in sdram.c (haven't looked further)
<jemk> nove: thanks, but then we still don't know why it doesn't work for wingrime
<nove> jemk, wingrime 1080p, mplayer 2%, xorg 23%
<arokux2> hno, dram.c, sorry.
<hno> arokux2, yes there is HPCR references in the standby code. But not much details on what it does..
<arokux2> yes, that is it.
<nove> jemk, put a print only once, for that message
<hno> wingrime, patches welcome.
<hno> Hm... looking at the table I am not sure I have the last HPCR correct on A20.
<wingrime> hno: I thinking about
<wingrime> hno: also, remove prints form spl, should make it realy light
<wingrime> nove: without any additional options ?
<wingrime> nove: works correctly?
<mripard> leviathanch: if you're trying to use the MMC you booted on, the muxing probably isn't the issue
<arokux2> [RFC] how should we call allwinner drivers? "Allwinner EHCI Controller" vs. "SUNXI EHCI Controller"?
<wingrime> nove: and fast
<arokux2> mripard, ^
<mripard> since you actually have the muxing already set up by u-boot
<arokux2> leviathanch, can you show your code?
<mripard> arokux2: sunxi is probably more accurate
<nove> wingrime, yes, is right now playing with -loop 0, perfectly fine
<arokux2> mripard, thought the same. everything in capital?
<Paullo612> wingrime: Is it on /dev/cedar_dev, or anywhere else? Just can't find it.
<mripard> arokux2: the best being Allwinner sunXi EHCI controller :)
<arokux2> mripard, yes, this is really the best, thanks
<wingrime> Paullo612: cedar_dev
<wingrime> Paullo612: CACHE something
<hno> wingrime, there is plenty of space in SPL now, no need to shrink it further.
n01 has quit [Ping timeout: 260 seconds]
<hno> Hm... don't have consistent data on the last four HPCRs on A20. have three different sets of values for those.
<wingrime> hno, HPCR have nice ability to protect userspace dram from strange devices -- secuity nice
Black_Horseman has joined #linux-sunxi
Black_Horseman has joined #linux-sunxi
<hno> 1. What lkcl added to u-boot. 2. Cubieboard2 when booted from NAND. 3. boot0 sources. All three differ in how the last four is configured (28-31).
<arokux2> :)
<leviathanch> arokux2: yes
<leviathanch> wait, I commit
<mnemoc> hno: try poking benn
<hno> Anyone here having the latest A20 SDK nearby? Can you please paste hpcr_value from lichee/boot/boot0/drv/init_dram/dram_init.c
<arokux2> hno, a second...
<arokux2> hno,
<wingrime> hno: ^
<Turl> I have an old sdk if you want to have a look
<hno> arokux2 & wingrime, which sdk release do you have?
<arokux2> hno, me: 2.0
<wingrime> hno: too
<wingrime> hno: but I have more one sdk
<leviathanch> arokux2:;a=shortlog;h=refs/heads/sunxi-mmc-for-next-plsdsif
<Paullo612> wingrime: ok, found it, thanks.
Black_Horseman has quit [Quit: Zwi se logou mou!!!]
Black_Horseman has joined #linux-sunxi
<hno> odd.. that's the same as in 1.2 SDK, but Cubieboard2 defaults differ.
Soru_ has quit [Read error: Connection reset by peer]
<arokux2> leviathanch, mmc has buit-in dma?
<leviathanch> arokux2: yes
<arokux2> leviathanch, I needed this for usb to work:
<arokux2> leviathanch, sorry, it was double pasted.
<arokux2> leviathanch, first two ifs.
<leviathanch> ok
<leviathanch> but I don't think that's the issue
<leviathanch> it gives me response timeout
<leviathanch> through an interrupt
soul has joined #linux-sunxi
<leviathanch> after reading the registers
soul is now known as Guest48558
<leviathanch> the response registers are all zero
<leviathanch> most of the time
<arokux2> leviathanch, I see. I also had this error with usb without those dma bits:
<arokux2> leviathanch, just make sure chinese code doesn't have something like this.
<leviathanch> hrhr
<leviathanch> I didn't see any of it
<arokux2> leviathanch, ok, sorry then.
<leviathanch> the internal dma controller of the mmc host just writes onto a target address I pass on to it through a register
<leviathanch> but instead of reading anything
<leviathanch> it already times out at the reset command
<drachensun> anyone run debian on the cubieboard? I'm trying to bring debian over for the A31
<drachensun> but its like /dev isn't mounted
deasy has joined #linux-sunxi
<drachensun> only happens on the debian rootfs I have tried, I can't figure it out
<arokux2> drachensun, which u-boot/kernel do you use for A31 (this is not related to your issue)?
<drachensun> the lichee-dev one from the Allwinner sdk
<arokux2> drachensun, and U-Boot?
ykchavan has quit [Quit: Leaving]
<drachensun> yup
jemk has quit [Quit: night]
<drachensun> specifically I am seeing this
<drachensun> [info] Using makefile-style concurrent boot in runlevel S.
<drachensun> mount: /dev not mounted or bad option
<hno> wingrime & arokux2 boot0 code looks a bit odd... but explains the difference seen between boot0 table and cubieboard2 boot. Last 4 is not set according to the table.
<hno> but looks like boot0 does it wrong to me.
<hno> 28, 30 is skipped. 29,31 is set to table value 28.
leviathanch has quit [Remote host closed the connection]
<Turl> mripard: doesn't pinctrl remux everything to gpio?
<hno> scary.
<Turl> drachensun: what are you trying to mount on /dev?
leviathanch has joined #linux-sunxi
<Turl> if devtmpfs, make sure it's enabled on kernel config
<arokux2> hno, just tell me how can I help :0
<arokux2> :)*
<hno> arokux2, try to figure out what hpcr #31 is..
<wingrime> nove: can you try 1920 video
<hno> and what is 1031, 0301 and 0731 decoded.
<arokux2> hno, 301 -- that one I know.
<drachensun> turl: Well everything y'know :)
<drachensun> turl: I'm exactly sure what goes wrong, that one I pasted is the only error message
<drachensun> turl: but I get no screen and no serial console
<arokux2> hno, this produces 301 for me
<hno> arokux2, that just enables/disables the port.
<drachensun> turl: yeah devtmpfs is in the kernel
<drachensun> one difference
<nove> wingrime, i did, was here 40min playing with no problem
<hno> 301 is already there.
<drachensun> my new setup as an initrd
<wingrime> nove: this is a13 realy?
<arokux2> hno, agree, but that is what you have asked for :P
<wingrime> nove: I still have low speed issue
<drachensun> and all the other systems were using the typical sd card ->uboot->Linux setup
<nove> wingrime, kernel 3.0, libvdpau and mplayer2 from wheezy
<hno> arokux2, no, I asked to have it decoded, i.e spelling out the bitfields as values.
<nove> wingrime, [VDPAU SUNXI] VE version 0x1625 opened
<nove> wingrime, sintel_trailer-1080p.mp4
<hno> 301 is cmd 3, enable.
<wingrime> nove: I have 3.4
<wingrime> nove: can you try
<nove> wingrime, no i can't, next week maybe i can upgrade
<wingrime> nove: than can you try 1920p bunny
<arokux2> hno, "spelling out the bitfields" I'm sorry, what do you mean by that?
<arokux2> hno, I've thought those are register values and you want to know what they enable
<nove> wingrime, link?
<wingrime> nove: so than I have try 3.0, it maybe some problem with timers
<hno> arokux2, see bit field definitions in wiki or usb header file. But is easy to read. 0xCCW1 CC is CmdNum, W is WaitState, and 1 is enabled.
<nove> wingrime, still downloading, but what plays looks ok
<nove> wingrime, did you try vdpau:fps=?
<wingrime> nove my mplayer not know this
<nove> wingrime, -vo vdpau:fps=?
<nove> wingrime, is that really slow, or only the video is not smooth?
focus_it has quit [Quit: Leaving]
<hno> arokux2, can you test if u-boot still works from sdcard on A20? Time for bed here. Have changed A20 HPCR to match what boot0 does, but the boot0 code looks very odd.
<nove> wingrime, is not perfect, the video sometimes stop for 1s, and mplayer counts frames dropped
<arokux2> hno, A10 only here, sorry.
<hno> ok. Will test tomorrow. If I am not mistaken #28 is the SD/MMC controller, and it differs noticeably there.
<hno> #31 also differs, but have no idea what that is used for.
<hno> Ah, nevermind. It's only #31 that differs. Got confused with the boot0 table that isn't actualy used..
<arokux2> hno, you are decoding , right?
<hno> yes
<hno> last two lines is the odd ones.
<arokux2> ok
<hno> #28,#30 is never assigned by that code. And #29,#31 is set to the value that the table have for #28.
<hno> HPCR numbers.
<arokux2> wingrime said some of the SoC has GPS
<hno> Yes, that is #30.
<hno> Kind of suspect that the A20 also have GPS but disabled, but that is another topic..
<nove> wingrime, 278seconds of video, with -framedrop, 94 frames dropped
<nove> wingrime, is also with no audio, -vo null
<nove> wingrime, -ao null
<wingrime> nove: no idea
<wingrime> nove: I have to try 3.0
<Turl> hno: didn't they put a extra i2c bus where gps would be?
vinifr has quit [Quit: Leaving]
<wingrime> nove: but I nearly sure there some regression after a20 merge
akaizen has joined #linux-sunxi
wingrime has quit [Ping timeout: 245 seconds]
<oliv3r> i did some hcrp research a few months ago, see my github
<oliv3r> u-boot hcpr branch
<oliv3r> cant paste url from tablet
<arokux2> hno, ^
<oliv3r> i kind of gave up because it stoppped mmaking sense,but with hno's findings i see it may have been wrong
<oliv3r> anyway i nrver looked at the sdks only uboot code
<oliv3r> arokux2: can you pasye the github link pls :)
<oliv3r> especially the macros are helpful for deoding stuff quickly
<nove> wingrime, this kernel 3.0 is from around 1 of may
<arokux2> oliv3r, no problem, here it is:
<oliv3r> once we find the wrong/writ i cansend a patch
<arokux2> hno, you are probably sleeping, but.. -------------------------^
<oliv3r> thanks arokux2 :) gonna backread more, then sleep
<arokux2> oliv3r, allright. good night!
<oliv3r> bbut thats what you where talking about right? it should help deemistyfying stuff
<arokux2> oliv3r, yes, I think it was.
<arokux2> oliv3r, but only 31 was unknown and you have UNKNOWN_31 :)
<arokux2> oliv3r, the others could be used for checking thins once again.
nove has quit [Quit: nove]
Black_Horseman has quit [Read error: Operation timed out]
naobsd has quit [Quit: Page closed]
Black_Horseman has joined #linux-sunxi
popolon has joined #linux-sunxi
<atsampso1> for the benefit of anybody searching the logs for the problem I was having earlier...
<atsampso1> if you're trying to bring Android up on a sunxi machine, and adb logcat is showing "eglInitialize(0x1) failed (EGL_BAD_ALLOC)" "E/SurfaceFlinger( 2195): couldn't find an EGLConfig matching the screen format"...
<atsampso1> ... then it's because your mali proprietary libs don't match what the kernel's mali module expects
<atsampso1> drivers/gpu/mali/mali/include/linux/mali/mali_utgard_uk_types.h has the version number from the kernel module in, and you can find the appropriate libs in sunxi-mali
<atsampso1> message ends. now to get the touchscreen working...
Paullo612 has quit [Ping timeout: 250 seconds]
<atsampso1> woohoo, and that's it working -- everything built from source except the mali blobs -- so I'm prepared to consider that a successful evening, and the replicant deblobbing can wait for another day
<arokux2> atsampso1, good job!