ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
CyberPK has joined #linux-sunxi
<CyberPK> Hi, i'm a newbye and i'm tryng to add bluetooth support to my generic AllWinner A13
<CyberPK> i've managed to downlad, compile and successfully load some modules
<CyberPK> as bluetooth.ko, btusb.ko, rfcomm.ko
<CyberPK> but i don't how to activate the bluetooth on my rom.
<CyberPK> i think i should add also bluetoothd binaries and libblue...so but i don't know where to get them
<CyberPK> can someone help me?
<rz2k> its 3AM in europe
<rz2k> who you want to see here :p
<krimpsok> I am here o/
<ssvb> just us chickens :)
<CyberPK> ahahaha i know :D here are 2:41 AM
<CyberPK> maybe someone can help, who know
<CyberPK> :D
<ssvb> what kind of bluetooth device do you have?
<CyberPK> a generic csr
<CyberPK> 0a12:001
<ssvb> are you using linux or android?
<CyberPK> android
<ssvb> oh, I'm not familiar with it :(
<CyberPK> it's ok
<CyberPK> thank you
<CyberPK> :)
<CyberPK> maybe, you know some resource where i can read and learn?
ganbold has quit [Remote host closed the connection]
<krimpsok> So you compiled the modules and insmod them but no dice when turning it on in the settings screen or?
tinti_ has joined #linux-sunxi
ganbold has joined #linux-sunxi
<CyberPK> i can't enable it by the settings because there is nowhere an option to do it in my tablet
<krimpsok> Is it running a vendor specific rom or something like Cyanogenmod?
CyberPK_ has joined #linux-sunxi
<CyberPK_> but i'm sure i need somethink else that i don't know
<CyberPK_> it running a vendor specific rom
<CyberPK_> softwiner nuclear evb
<CyberPK_> SoftwinerEvb 4.1.1
<krimpsok> I'll do some digging for you but can't promise anything :)
CyberPK has quit [Ping timeout: 245 seconds]
<CyberPK_> thank you very much
<CyberPK_> :)
<rz2k> you should head to the android porting mailing list
<CyberPK_> i'm searching on internet too, but i'm a newbie and i can't find anything useful
<CyberPK_> android porting mailing list?
<CyberPK_> where?
<CyberPK_> thank you :)
drachensun has quit [Ping timeout: 258 seconds]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
CountryGeek has joined #linux-sunxi
bsdfox_ has quit [Read error: Connection reset by peer]
CyberPK has joined #linux-sunxi
CyberPK_ has quit [Ping timeout: 245 seconds]
krimpsok has quit [Quit: Ik ga weg]
<hno> slapin_n1, I lurk around, but haven't had my eyes on IRD tonight. Maybe tomorrow.
tinti has quit [Read error: Connection timed out]
tinti has joined #linux-sunxi
CyberPK has quit [Quit: Page closed]
w00tc0d3 has quit [Quit: No Ping reply in 180 seconds.]
w00tc0d3 has joined #linux-sunxi
rz2k has quit []
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
el has joined #linux-sunxi
<el> when I'm trying to sync cubie_android_ics, I'm getting error : "fatal: Authentication failed"
<el> can anyone help me?
shineworld has joined #linux-sunxi
wowon has quit [Quit: Leaving]
rellla has joined #linux-sunxi
ZaEarl has quit [Ping timeout: 245 seconds]
BJFreeman has quit [Quit: had a good time]
rellla2 has joined #linux-sunxi
rellla has quit [Ping timeout: 256 seconds]
rellla2 is now known as rellla
n01|work|away is now known as n01
<oliv3r> lkcl: ideally wait until we have a blob that works for sun[4567]i before geting onto their case. also leaves us room to examine that one
<oliv3r> lkcl: worst things come, they get pissy, and don't release any libcedar anymore; unlikly but could happen :)
<oliv3r> lkcl: also i'd think git://github.com/linux-sunxi/cedarx-libs.git is the most recent one
FunkyPenguin has joined #linux-sunxi
<FunkyPenguin> has anyone been able to build a kernel for A20?
<FunkyPenguin> i dont see a defconfig in the git tree
paulk-desktop has joined #linux-sunxi
<oliv3r> FunkyPenguin: nobody has hadware yet
<oliv3r> FunkyPenguin: This is a lie actually, hipboi has some A20 chips now that he got on a cubieboard. He has gotten our current BSP working I think, or atleast tested that the old kernel works on the new hardware
<oliv3r> but that's all very experimental
<oliv3r> ssvb: ping
tzafrir has quit [Ping timeout: 264 seconds]
<oliv3r> Turl: am I a manual spammer?! :p
<oliv3r> i think so, but I bet that's the wrong answer ;)
<lkcl> oliv3r: ta
<FunkyPenguin> oliv3r: i've got an A20 box (actually got 4 of them)
<FunkyPenguin> would it be possible to get the config from /proc on android?
<oliv3r> lkcl: i'm documenting some findings on the wiki, i'll let you know when i'm done for the day
<oliv3r> lkcl: note, that this is me dabbling with things i have little understanding, so forgive me on that :p
<oliv3r> FunkyPenguin: if the stock kernel is compiled a such
<oliv3r> FunkyPenguin: chances are, your A10 kernel wil run just fine, or near fine.
<oliv3r> FunkyPenguin: WHERE DID YOU GET THOSE! :p we want some too :)
<FunkyPenguin> oliv3r: so build against sun4i?
<oliv3r> FunkyPenguin: they are shipping allready, awesome
<oliv3r> FunkyPenguin: how much did you pay for it?
<FunkyPenguin> yeah, got mine a just over a week ago
<FunkyPenguin> tbh i have no idea how much it cost, work paid for it :D
<oliv3r> that's your best bet
<oliv3r> that's the 'stock 3.3' kernel
<oliv3r> FunkyPenguin: can you try try a few things for research sake? :)
<oliv3r> http://linux-sunxi.org/Submitting_Boards for example would be awesome (IF a10mem info even works, but we assume it does)
<FunkyPenguin> oliv3r: great, thanks for that i checked out linux-sunxi master so i'll pull that tree now
<FunkyPenguin> oliv3r: i can try and research, depending on workload
<oliv3r> also could you extract dd nanda.tar.gz and send it to me (oliver@linux-sunxi.org) so I can put it up on github?
<oliv3r> https://github.com/linux-sunxi/sunxi-bin-archive has some other dumps and since oyur the first, yours is most interesting for now :)
<oliv3r> FunkyPenguin: linux-sunxi master is where the fun should be eventually, but we might have to port a few things first. It is assumed for now though, that sun4i works quite well on sun6i
<oliv3r> sun7i* (sun6i = a31)
<FunkyPenguin> ok let's see what i can get
<FunkyPenguin> oliv3r: thanks
<mripard_> oliv3r: the sun6i and sun7i use a different interrupt controller
<mripard_> so I'm not sure you can just put a sun4i kernel on sun7i and expect it to work
<oliv3r> mripard_: now we are mixing 3 things up :p
<oliv3r> mripard_: i Made the mistake of calling sun7i, sun6i. I hopefully corrected myself
<oliv3r> mripard_: you then said sun6i and sun7i use different interrupt controllers
<mripard_> yes
<mripard_> different from sun4i
<mripard_> but sun6i and sun7i share the same one
<oliv3r> so sun7i uses sun6i interrupt controller?
<oliv3r> but otherwise sun7i is more closly related to sun4i
<oliv3r> interesting
<oliv3r> mripard_: can we use dt to 'load' the interrupt controller?
<mripard_> well, both sun6i and sun7i no longer use the Allwinner custom interrupt controller, but the ARM one, so it's pretty standard
<mripard_> and yes, you can get that from the DT
<mripard_> I'm not saying it's hard
<mripard_> I'm just saying it won't just work
<oliv3r> mripard_: ah, okay, i thought it was a 'new' custom aw IC
<oliv3r> mripard_: but that does mean some porting for stage 3.0 and stage 3.4 will be required for FunkyPenguin and he can't run our code just yet
n01 has quit [*.net *.split]
WarheadsSE has quit [*.net *.split]
n01 has joined #linux-sunxi
WarheadsSE has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
<oliv3r> mnemoc: ping
_BJFreeman is now known as BJFreeman
<oliv3r> wtf? this looks exrmpely ugly to me, right? https://github.com/huceke/libcedar/blob/master/src/CedarDecoder.cpp#L40
<jelly-home> oliv3r: sure, you could replace it with just the base name without the directories
<jelly-home> oliv3r: as dlopen(3) says, If filename contains a slash ("/"), then it is interpreted as a (relative or absolute) pathname. Otherwise, the dynamic linker searches for the library as follows [...]
<mnemoc> oliv3r:
tzafrir has joined #linux-sunxi
<oliv3r> mnemoc: what IS the latest cedarX code, it's in our github right? empat0 forked it from our github and possibly submitted his changes (if there are any)
<ssvb> oliv3r: hi, but I'm leaving in a few minutes, will be back in a few hours
<oliv3r> ssvb: lol okay, i'll have my Q's ready in a few hours :p
<ssvb> but using android libs is not going to be easy, if even possible (via lowlevel libve api)
<oliv3r> ssvb: oh no, i was gonna ask real low level noob questions :)
<oliv3r> ssvb: but go, don't let me hold you up :)
<mnemoc> oliv3r: afaik linux-sunxi/cedarx-libs has the latest
<rellla> oliv3r: do you know, why gimli, empat0 and willswang did their own cedarx implementations? would it be worth to bundle it in a way? what's the cleanest way to base on?
<oliv3r> rellla: that's what i was gonna go ask next :p but from what I saw:
<oliv3r> rellla: they only use libvecodec.so and didn't use all the files that come with it
<oliv3r> rellla: empat0 i think did use ours and probably has the closest implementation of it
<oliv3r> rellla: willis used libvecodec.so and baked some stuff around to integrate it with vlc
<oliv3r> gimili just did some magic voodoo :p around it, but is for xbmc decoder only
<oliv3r> so to summarize, empat0 did his changes upstream (linux-sunxi) and integrated it with xbmc; gimili uses libvecodec.so and did the integration to xbmc in secret , hence the duplication; willis integrated libvecdoec.so to vlc
<rellla> sure? for me it looks like empat0 bases directly on the original libvecore.so and (only) uses the "orig" api and modules (vbv, fbm ...)
<rellla> while gimli and wills did a kind of wrapper libcedarx.so for the libvecore.so and build their own api to use in xbmc and vlc.
<rellla> also gimli did a rewrite of fbm and vbv
eebrah has quit [Ping timeout: 264 seconds]
<oliv3r> rellla: exactly
<oliv3r> emapt0 used 'ours' and pushed changes into our github
vicenteH has joined #linux-sunxi
<rellla> wills bundled vbv_ ve_ and fbm_ modules in his libcedarx.c
<oliv3r> rellla: I didn't do very close inspection :p
<oliv3r> I was checking out our git and did update our wiki so far on the android lib
<oliv3r> still working on it ;)
<rellla> oliv3r: don't do closer inspections. just RE it ;P
<oliv3r> rellla: lol, not technical enough Im affraid :p but i'll will put more info as I find it on the wiki
<oliv3r> step 1 in RE is tearing it appart and documenting things anytway :p
<rellla> pinging iainb could be useful, to ask why he was stopping his work
BJFreeman has quit [Ping timeout: 264 seconds]
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJFreeman
<oliv3r> rellla: yeah, ask him
<oliv3r> no update for 9 months :(
<oliv3r> rellla: hmm, i do notice that ianb's RE of libcedarx (open_cdxalloc) is mostly android stuff
<oliv3r> I see adapter/avheap/ where alloc is for android, so it's quite possible that's not even needed any longer for regular linux
<rellla> oliv3r: yes, iirc alloc isn't needed for linux anymore.
paulk-desktop has quit [Remote host closed the connection]
paulk-desktop has joined #linux-sunxi
ganbold_ has joined #linux-sunxi
awafaa has joined #linux-sunxi
<lunra> Has anyone got the iptables LED interface thingy to work? I think it's pretty cool, but I can't make it work properly, led-delay keeps getting set to zero, even though I'm telling it to be 1000ms. The command I'm using is {iptables -A INPUT -p tcp -j LED --led-delay 1000 --led-trigger-id blinky} and iptables -L gives {LED tcp -- anywhere anywhere led-trigger-id:"blinky" led-delay:0ms
<lunra> }
<lunra> The consequence is that it's very hard to see when a packet arrives because the LED is on for the time it takes for the system to set the pin and then unset it
<lunra> I'm on a cubieboard, but this is a software problem from what I can tell.
<oliv3r> iptables led interface?
<lunra> yep, iptables can set an LED whenever a packet is matched
<oliv3r> never seen that :) I bet openwrt implements that too, you could see how they do it
<lunra> Thanks for the pointer
<oliv3r> otherwise, might be a bug in the led driver of course
<lunra> Hm, I'm doing it correctly
<lunra> Yeah. Looks like a bug in iptables or the LED driver (more likely the former). That's the problem with niche features, they break often because noone really cares.
<oliv3r> or knows about
<oliv3r> :p
<lunra> true
<oliv3r> mnemoc: was it you that mentioned that cedar access registers directly? I'm pretty sure you're right. MACC_REGS_BASE is in libcedar objects, which is the 'media ACCelerator.
<lunra> I'll leave it running anyway, and if I somehow get a SYN flood on my external SSH port, well then I'll know because my room will suddenly be green (cubieboard LEDs are really bright!)
<lunra> UGH
<lunra> still can't get SL going
<lunra> I can create a user for you on the opensim server if you want to play with that, else I'll just pack it up
<lunra> ah wrong channel
<lunra> (sorry folks)
<ssvb> oliv3r: I'm back :)
<ssvb> about the libvecore library, one of the challenges is that the ABI seems to be unstable
<libv> i really should also mask out libve in my highlight...
<ssvb> libv: what kind of highlight?
<libv> libvecore triggers my libv highlight
<ssvb> :)
<oliv3r> ssvb: right now, i'm only documenting little things I find
<oliv3r> ssvb: what my simple noob question was about
<oliv3r> we currently have libvecore.a and libvecore.so
<oliv3r> using ar we can extract libvecore.a, and get a list of object files, which where built using C files, nothing ispecial here
<oliv3r> we can use readelf to find out what C files libvecore.so was built from and atleast compare that (did that, see wiki) :p
<oliv3r> we can also use readelf -s to find function names etc among other thigns
<oliv3r> with libvecore.a, or rather, the seperated modules, we know exactly which functions/objects etc belong to which objectfile and thus which C file
<oliv3r> Now I can do readelf on libvecore.so but that just returns a messy bit.
<oliv3r> I get all functions etc naturally, but besdies prefixes, makes it hard to compare
<oliv3r> libv: nah, we like when you lurk/respond :)
<rellla> oliv3r: http://linux-sunxi.org/Lib*ecore/linux-armhf_functions has everything twice ;)
<oliv3r> rellla: ouch
<oliv3r> rellla: oh that's quite possible, I haven't filtered it/did anything nice with it
<rellla> just mentioned, that a diff doesn't make much sense ;P
<oliv3r> rellla: that's how it jus tis
<oliv3r> rellla: arm-pc-linux-gnueabi-readelf -W -s libvecore.so | grep FUNC | awk '{ print $8 }' | sort
<oliv3r> try it :)
<oliv3r> is there a quick and easy way to remove dupes?
<oliv3r> rellla: but that does explain why the list was twice as long :p
<oliv3r> rellla: gonna try running that through calc then :p
<oliv3r> rellla: btw, I diffed a grep FILE :)
<oliv3r> there, removed dupes :)
<oliv3r> libre calc ftw
torindel has joined #linux-sunxi
<oliv3r> ssvb: which bits did you find that could be frome ffmpeg; i have looked at a few things, but found nothing yet. That said, I find it extremly hard to imagine they'd written all this themselves
<CountryGeek> oliv3r: sorry to be so late to the game, but ... sort | uniq will remove the dups
<oliv3r> uniq! duh, well i used calc, but i'll use that next time!
<ssvb> oliv3r: for example 'ff_mpeg4_DCtab_chrom'
<oliv3r> ssvb: ah yes, the ff_ prefixes you mentioned
<oliv3r> i was dabbling around, but hadn't found that specific one
<oliv3r> ssvb: how much do you know of gcc? What is crtstuff.c for in gcc?
<ssvb> oliv3r: run 'objdump -D libvecore.so', search for "<ff_mpeg4_DCtab_chrom>:" chunk and compare with http://ffmpeg.org/doxygen/trunk/mpeg4data_8h_source.html#l00041
<ssvb> oliv3r: that's not code, but an array of constants, though it is very suspicious they decided to name it exactly in the same way as in ffmpeg if it is just a coincidence
<oliv3r> ssvb: interestting, I used 'arm-pc-linux-gnueabi-readelf -W -s libvecore.so | grep FUNC | awk '{ print $8 }' | sort | uniq | grep ff_' and didn't find that one
<oliv3r> ssvb: ahh, yeah i only focused on function names for now
<oliv3r> well if the use a table, which is factual data? is that still a violation of the GPL?
<oliv3r> ssvb: I do think they may have renamed functions, just in case, to obfuscate things?
<ssvb> imho it's enough to be a circumstantial evidence, somebody might do a thorough investigation if he wants to
<oliv3r> ssvb: well was digging through some random function names, jpeg, mjpeg etc etc
<oliv3r> i then did some of those functions against google/git grep ffmpeg.git
<oliv3r> ssvb: but so far found not a signle funciton name dupe
<oliv3r> that doesn't mean anything of course
<ssvb> what about 'ff_huff_build_tree'?
<bfree> CountryGeek: I'm later to the game ;) sort -u
<oliv3r> also, ff_mpeg4_DC* appears to be 'some' function? in the objdump, but is a const table in the ffmpeg source
* CountryGeek nods
<oliv3r> bfree: double ouch; if only I knew :p
<ssvb> oliv3r: it's not a function, objdump is just a bit dumb, look at the byte values representing these "instructions"
vicenteH has quit [Ping timeout: 248 seconds]
<oliv3r> ssvb: i'm no asm guru :p
<oliv3r> ssvb: I a readelf on the android objects, 136: 00000000 636 FUNC GLOBAL DEFAULT 52 ff_huff_build_tree
<oliv3r> 136: 00000000 636 FUNC GLOBAL DEFAULT 52 ff_huff_build_tree
<oliv3r> ssvb: so IF they renamed everything, they may have forgotten that one, or copy pasted that one
<ssvb> oliv3r: in any case, I'm not in favour of any GPL enforcement actions
<ssvb> that's a way to mutually assured destruction
<ssvb> the community still depends on the documentation, kernel sources and a bit of support from allwinner whatever insignificant it is at this point
<ssvb> I'm not going to provide any more comments on this matter :)
<oliv3r> ssvb: haha, i'm not gonna do anything either, but lkcl might pursuide them to release the cedarX sources :)
<oliv3r> ssvb: heck I even mentioned to lkcl above in the log, that he best wait until some new blobs are released at the very least ;)
ZaEarl has joined #linux-sunxi
<oliv3r> ssvb: anyhow, if they have in fact use ffmpeg (huffman.c has atleast some bits) it could ease RE-ing a lot imo
<ssvb> that's for sure
<oliv3r> that's all I was looking at actually O: )
<oliv3r> was seeing how hard it was to do the jpeg decoder (and encoder) via cedar
<ssvb> oliv3r: FYI, libjpeg-turbo should already do a pretty good job
<ssvb> enough for realtime mjpeg encoding/decoding
<oliv3r> ssvb: but that's besides the exersize isn't it :p
<oliv3r> jpeg is very simple compared to all the other audio and video codecs
<ssvb> true
<oliv3r> jpeg, mjpeg, mp3 (even if possible) then video i'd reccon
<oliv3r> god so much to do, so little time and I get distracted by so many things that I'd want to do
<oliv3r> I wonder how openMAX IL and DL work together. I bet you don't put entire codecs into the DL, just some generic functions you can accelerate i'd hope (dct for example) which can be used by a pletora of codecs
<oliv3r> if only we had docs and register info for the Media Accelerator
<oliv3r> ssvb: jpeg_dec is done with libjpeg (maybe even turbe)
<oliv3r> turbo* allthough I serisouly doubt that
<oliv3r> ssvb: 100% certain they're using libjpeg for it; they took the functions they needed, copy pasted them into a single file and added their own 'hal'; (libjpeg is public domain afaik so no license violation)
<oliv3r> though I do think you have to add something to the README or some such
eebrah has joined #linux-sunxi
shineworld has quit [Quit: Leaving]
<BJFreeman> the repository is to big to download with git clone, is there another way to get a image of the respository
<oliv3r> BJFreeman: which repository?
<BJFreeman> linux-sunxi\
<oliv3r> this should get you started
<BJFreeman> thanks
<mnemoc> the not __git version might be more useful for him
<mnemoc> even if sightly larger
<oliv3r> the non-git version has .git removed?
<mnemoc> it has both, .git and files
<oliv3r> ah and __git is only the .git bit? :) e.g. 'bare' iirc
<mnemoc> yes, and that can confuse newbies
<oliv3r> ah
<oliv3r> BJFreeman: yes, you may want dl the other linux-sunxi archive from the dump ;)
<oliv3r> using
<oliv3r> the OpenMAX logos and trademarks under a royalty-free license. The
<oliv3r> 10.000 $ for Khronos members, and 2.500 $
<oliv3r> procedure requires a fee, that is
<oliv3r> for non-members
<BJFreeman> ok I have the uboot
<oliv3r> so why would I want to be a member again?
<BJFreeman> but only have 512 mem
<mnemoc> oliv3r: royalty-free for those who pay the license :p
<oliv3r> i ment more, that it's cheaper for non-members to get a sticker :p
<mnemoc> *g*
<oliv3r> :(
<mnemoc> first read 10k vs 25k
<oliv3r> but it's 10k vs 2.5k :(
<oliv3r> sotp confusing me!
<oliv3r> stop*
eebrah has quit [Ping timeout: 245 seconds]
Undertasker has joined #linux-sunxi
<hno> oliv3r, isn't that 10K unlimited license, or 2.5K per product?
<hno> BJFreeman, which u-boot? If you are staring at the output fron the u-boot Allwinner ships in NAND then don't worry. If you are staring at output from an SD card u-boot SPL boot then you have the wrong SPL.
<ssvb> oliv3r, rellla: a bit of progress with libhybris wrapper for android libs - https://github.com/ssvb/libhybris/commit/09d999011b5b646cf216c4aa2f62f7497e024f40
<mnemoc> sadly android libs are armel...
<ssvb> I would say the chances of success are not too bad
<mnemoc> are these libs better than linux's?
<ssvb> mnemoc: if I understand it correctly, libhybris takes care of armel/armhf compatibility
<mnemoc> ssvb: yes, but we lose neon
<ssvb> mnemoc: I still don't know android cedarx libraries are better, first I need to make them work correctly with VLC :)
<ssvb> mnemoc: we don't lose anything
<mnemoc> fair enough ;-)
<ssvb> for armel compatibility, have a look at FP_ATTRIB define in https://github.com/libhybris/libhybris/blob/master/hybris/glesv2/glesv2.c
ZaEarl has quit [Ping timeout: 245 seconds]
<ssvb> the functions are called via these thunks, and they provide calling conventions conversion
<mnemoc> :)
<ssvb> that said, cedarx/libvecore even does not use any floating point arguments, so it should be a non-issue
<mnemoc> can we use libhybris to use "system" ffmpeg instead of the gpl violating copy bundled in the .so?
<ssvb> I don't think so
<mnemoc> I mean, hacking that into the same glue. of LD_PRELOAD is better?
<mnemoc> ok
<ssvb> cedarx is likely using a heavily modified ffmpeg (with significant parts of decoding pipleine offloaded to hardware)
<mnemoc> good point
<ssvb> libhybris is a framework which creates an environment for transparently loading and using android libraries
<mnemoc> it's an awesome idea
<ssvb> so it would look like having android /system directory with some android blobs and a libvecore.so in /usr/lib, which is essentially a wrapper emulating native libvecore.so
<ssvb> but using libhybris and android blobs under the hood
<ssvb> mnemoc: it's not my idea, all the credit belongs to vgrade and Stskeeps :)
<mnemoc> i mean, libhybris itself. like wine
<mnemoc> accepting the crapness of the world instead of fighting against it
ganbold_ has quit [Ping timeout: 276 seconds]
<mnemoc> and for us it's a decent way of using newer/better mali/cedarx libs
ZaEarl has joined #linux-sunxi
<mnemoc> is there anything in the kernel capable of "faking" the register area so we can watch what the libs do when accessing "raw" memory?
lkcl has quit [Ping timeout: 264 seconds]
<ssvb> mnemoc: https://github.com/iainb/CedarXWrapper tries to do this in userland
<ssvb> surely one can also easily make a kernel based solution
ganbold_ has joined #linux-sunxi
<ssvb> ARM has emulation hooks for unaligned memory reads/writes, so all the code is already there
<mnemoc> nice
<ssvb> better ask libv, reverse engineering folks know a lot of tricks :)
<mnemoc> afaik he doesn't touch the kernel
<mnemoc> only LD_PRELOAD
<ssvb> any solution is fine, as long as it does the job
focus has joined #linux-sunxi
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
n01 is now known as n01|away
<techn_> ssvb: oliv3r: disassebler show that it uses atleast part of this file http://en.verysource.com/code/4495555_1/vc1decpic.c.html
<libv> ssvb: there was someone who already was capturing register accesses
<libv> ian ...
<ssvb> libv: yes, I know - https://github.com/iainb/CedarXWrapper
<libv> ah, right
<libv> wow, 9m ago already
<libv> i pointed him in the right direction back then
<ssvb> we are yet to find out if it is the right direction
<techn_> ssvb: oliv3r: also part of these files could be used http://iphome.hhi.de/suehring/tml/doc/lenc/html/index.html
<ssvb> I would say, there is a chance that Allwinner A10 might have another dedicated ARM core which is executing codecs code from SRAM
<libv> ssvb: the description there sounds sane to me
<ssvb> it is a good start for sure
<techn_> ssvb: Imho they have ripped quite a lot from those reference decoders
<ssvb> techn_: I don't want to know all these details
<techn_> :)
<techn_> but seems that blob is combination of many licenses
<oliv3r> mnemoc: should be quite easy in kernel mode. though I don't have any real knowledge, what the IOCTL does, is a function that calls readl/writel, so you can do plenty with those arguments passed to the ioctl, before handing it over to readl and writel
<oliv3r> mnemoc: also the raw access wasn't there in sun4i_cedar.c; i only see it in sun7i
<ssvb> oliv3r: the memory area with cedarx hardware registers is mapped as uncached and banged from the userland
<oliv3r> mnemoc: what reference of ffmpeg do you speak of btw? All I noticed so far, was functions probably copy pasted/renamed; i don't think it's the full ffmpeg
<ssvb> oliv3r: if sun7i has special ioctls for this, then it is now a bit easier to trace
<ssvb> oliv3r: if it was just a full unmodified ffmpeg, then we would have no need for this blob in the first place ;)
<techn_> as I said earlier it contains atleast parts of ffmpeg, reference h264 encoder, reference vc1 decoder, ..
<ssvb> "copy/pasted renamed" is in other words http://en.wikipedia.org/wiki/Derivative_work
rz2k has joined #linux-sunxi
<oliv3r> ssvb: i don't understand the code there (for sun4i :( )
<oliv3r> techn_: ah, they used reference codecs; well they deffo used libjpeg for the jpeg bits, ff for huffman stuff
<oliv3r> ssvb: yep, which means lkcl could in theory atleast poke them 'give us cedarX codes, your secret is out'
<oliv3r> well one can dream
<oliv3r> ssvb: how does that code work? the sun4i bit
<oliv3r> haven't found the mpeg2 bits, but i'm sure it's also a reference decoder then?
<ssvb> oliv3r: for accessing the hardware registers from the userland you do something like this - https://github.com/willswang/libcedarx/blob/master/libcedarx.c#L264
<ssvb> oliv3r: mmap returns you a nice pointer to the memory buffer with the mapped hardware registers, and then this address is provided to the blob via "ve_get_reg_base_addr()" function
<ssvb> oliv3r: on the kernel side it is the code that I linked earlier - https://github.com/linux-sunxi/linux-sunxi/blob/sunxi-v3.4.29-r1/drivers/media/video/sun4i/sun4i_cedar.c#L899
<ssvb> oliv3r: is it clear now? sorry if my explanations are crappy
n01 has joined #linux-sunxi
<oliv3r> techn_: ssvb mnemoc : -rw-r--r-- 1 1000 1000 2064 May 3 10:24 h264_register.o
<oliv3r> that file is purely responsible for direct register access for the h264 codec
<oliv3r> not sure if you somehow CAN wrap that
<oliv3r> anyhow, how does AW deal with different kernel/different libvecore.* mismatch? (ioctl for register access, vs the 'sun4i' way0
<ssvb> I guess sun7i is just going to have an updated libvecore userland blob
<rm> hmmm last I heard the A10S still had issues even booting up
<rm> with u-boot
<rm> not via android-then-reboot
<oliv3r> on the ML is saw its nearly solved
<oliv3r> ssvb: yeah, but it'll mean 2 blobs to rule half :)
<ssvb> oliv3r: as I mentioned earlier, the libve ABI seems to be unstable and has already been changing from release to release, so it might be in fact more than 2 blobs :)
vinifm has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
ganbold_ has joined #linux-sunxi
sanka has joined #linux-sunxi
<techn_> rm: pretty expencive nand :p
<oliv3r> ssvb: i know, i'm just curious how they will handle that
drachensun has joined #linux-sunxi
<ssvb> yay! libhybris cedarx libvecore seems to work fine :)
<ssvb> and android libs are indeed not affected by weighted P-frame prediction bug
<jelly-home> yo dawg, lemme make a wrapper for your wrapper so you can wrap libcalls while you wrap libcalls
* ssvb needs to clean up the code and provide some instructions
<drachensun> lol
<drachensun> in all seriousness though, pretty exciting stuff, I am trying to copy vgrade example work right now
ZaEarl has quit [Read error: No route to host]
tinti has quit [Ping timeout: 256 seconds]
tinti has joined #linux-sunxi
<oliv3r> atleast that will lower the dependancy on AW to supply timly real linux decoders
<oliv3r> other hand, it'll do the same what wine does, AW won't see reason to release normal linux decoers
<oliv3r> libvecore*
<ssvb> now we have a clear confirmation that the linux blob is buggy
<ssvb> libhybris is likely to have an extra memory footprint (I have not measured this yet), this could be a good justification to ask for an updated linux blob
lkcl has joined #linux-sunxi
shineworld has joined #linux-sunxi
fra79Wii has joined #linux-sunxi
<fra79Wii> Does someone using the touchscreen experience weird behaviour at low battery?
ZaEarl__ has joined #linux-sunxi
<fra79Wii> It was really bad with the stock firmware (3.0.8) now it's quite fine, but I have to turn brightness to the minim an sf clock back the cpu to 1GHz to have it almost usable
<fra79Wii> After the 15-10% of battery left
eebrah has joined #linux-sunxi
eebrah has quit [Ping timeout: 245 seconds]
sanka has quit [Quit: Leaving]
eebrah has joined #linux-sunxi
<Undertasker> bad voltage converter design
<Undertasker> Additional capacitors will maybe solve this
<ssvb> Turl: what license applies when redistributing compiled android binaries?
<ssvb> I would like to provide a tarball with libc.so libcorkscrew.so libcutils.so libdl.so libdsyscalls.so libgccdemangle.so liblog.so libm.so libstdc++.so libutils.so libz.so Android binaries
<ssvb> building the whole Android themselves is not for the faint hearted end users :)
<oliv3r> ssvb: I suppose if your intentions are clea[rn] it should be fine, make sure you have a README with a link to your site that has full building instructions?
<oliv3r> if you add git hashes it should be better
<ssvb> I don't know much about android, but this beast took half a day to download and several hours to build
tinti has quit [Remote host closed the connection]
<ssvb> I would assume a lot of people might prefer a small tarball with a few binaries (as an alternative to just giving up) :)
rellla has joined #linux-sunxi
fra79Wii has left #linux-sunxi ["Messaggio"]
fra79Wii has joined #linux-sunxi
<ssvb> oliv3r: hmm, actually just mounting /dev/nandd as /system and using a copy of android from nand also works :)
CyberPK has joined #linux-sunxi
fra79Wii has quit [Remote host closed the connection]
fra79Wii has joined #linux-sunxi
<CyberPK> hi all
<CyberPK> can someone help me to enable bluetooth on A13 android tablet?
<CyberPK> i've already compiled and loaded successully the modules bluetooth.ko and btusb.ko
<CyberPK> but i can't go further because i don't know what else is needed and where to get without recompile the entire kernel
<CyberPK> :(
<vinifm> cyberpk, first check the list of compatible devices
<vinifm> in btusb.c
<CyberPK> ok.
<CyberPK> i think my device is compatibile
<CyberPK> is a generic csr
<rellla> ssvb: nice to read about libhybris progress!
<ssvb> rellla: these are the needed changes for libcedarx wrapper - https://github.com/ssvb/libcedarx/commit/e15e6ab4a167e243cf5254d2adc32f70b3b25381
<ssvb> something similar may be also required for XBMC
<rellla> do you plan to wikify the whole thing? i'll look into xbmc the next days.
<vinifm> ssvb, in cross compile, '--with-tuning=ARCH', ARCH=?
<vinifm> cross compile vlc
eebrah has quit [Ping timeout: 245 seconds]
<ssvb> rellla: yes, I will provide complete instructions at least for CedarX VLC tomorrow
<rellla> yeah. then i'll try to understanf it and port it to xbmc ;)
<ssvb> vinifm: sorry, I have not tried cross compiling vlc yet
<vinifm> ssvb, I'm trying to a long time since that day :(
CyberPK has quit [Ping timeout: 245 seconds]
<vinifm> it looks like it will never end
fra79Wii has quit [Read error: Connection reset by peer]
fra79Wii has joined #linux-sunxi
eebrah has joined #linux-sunxi
<ssvb> rellla: great :)
<rellla> so in our cedar source, it is already included.
<rellla> thats what i meant, with "there are too many different versions or implementations on the way"
<rellla> don't know whats the newest or more complete or which one belongs to which blob...
<Turl> ssvb: they're under apache2 iirc, so that's it
shineworld has quit [Quit: Leaving]
<drachensun> vgrade or any one else familiar with mer, should I just use xorg-launch-helper to launch X under mer?
FunkyPenguin has quit [Ping timeout: 245 seconds]
<Turl> ssvb: I'd worry more about redistributing cedar tbh :)
<ssvb> Turl: thanks
<ssvb> Turl: I'm not going to redistribute cedar :)
<rellla> ssvb: iirc willswang first was using the armel blob for his libcedarx, and in armel-libve.h they aren't included.
FunkyPenguin has joined #linux-sunxi
<rellla> so he probably missed updating his files
<ssvb> rellla: I spotted 3 types of ABI for cedar blobs, but fortunately the blobs from jellybean have the same ABI as the buggy armhf linux library currently in use
<rellla> good for us
<fra79Wii> Damn android irc has lost the connection and the log while interesting topics are on plate...
<fra79Wii> Since you were talking about it I've seen the libvecore.a and libcedarx for android...
<fra79Wii> For what purpose would be use that lib in android... I'm struggling to port cedarX to 4.2...I have to know all the possible use of cedarx..
<fra79Wii> Available code...
n01 has quit [Quit: leaving]
FunkyPenguin has quit [Ping timeout: 264 seconds]
FunkyPenguin has joined #linux-sunxi
<oliv3r> fra79Wii: all you can do is pray
<fra79Wii> Yes..that allwinner release the damn sdk...
<fra79Wii> I manage to make camera almost work with older blob
<fra79Wii> But it's an awfull hack
<oliv3r> fra79Wii: offer them lots of money for cedarX sources ;)
CyberPK has joined #linux-sunxi
<drachensun> you would be surprised how hard it is to get them to take your money sometimes
<rellla> oliv3r: there is only one guy, that knows that they may have sources at least. so first find him :p
<fra79Wii> I was thinking to use that lib-cedarx for android into the android ffmpeg..somehow...
<fra79Wii> I was so glad one year ago when I saw beginning of rev-engineering..
<CyberPK> vinifm: No it isnt't!
FunkyPenguin has quit [Ping timeout: 258 seconds]
<CyberPK> my device is a generic csr 0a12:0001
<oliv3r> rellla: yeah, 1 dev, who probably doesnt care anymore :)
<fra79Wii> rella: who is the guy??
<rellla> *PROBLEM*
<vinifm> CyberPK, it is in black_list :(
<CyberPK> why???? :(
<CyberPK> what it mean that is in blacklist? :(
<rellla> i'll check my mail tomorrow, maybe allwinner answered \o/ <- joke
<vinifm> which is not compatible
<vinifm> / CSR BlueCore devices { USB_DEVICE(0x0a12, 0x0001), .driver_info = BTUSB_CSR },
FunkyPenguin has joined #linux-sunxi
<rellla> fra79wii: don't know. mnemoc said (was it mnemoc?) they had/have exactly 1 guy who has access to cedarx sources. and that one wasn't the best friend of tom ;)
<fra79Wii> I mean cedarX is present in all allwinner chip. They have only to win on rekesing the sources...
<rellla> they probably don't want to let other people dig into the sources ... btw. what i a license?
<fra79Wii> A10 is almost compleyely open...if you compare to raspberry pi it's better, it would be even better with an open vpu
<rellla> fullack
<rellla> tell it allwinner!
<Undertasker> They don't care, becasue their main business are cheap android devices, not linux
<fra79Wii> ...waiting for lima to be ready... The vpu would be the only thing remaing close...
<oliv3r> A10 is still miles better then Rpi
<fra79Wii> Undertasker: but even in Android the openess is a good thing.
<oliv3r> you can boot and use A10 an A10 fully (except for GPU and VPU) and with limare/lima even WITH gpu
<oliv3r> rpi you cant even boot gpl
<oliv3r> vpu will be RE-able easly
<oliv3r> just requires a lot of knowledge and work
<Undertasker> Allwinners customers also don't care, because they only want a working Android, so why should Allwinner invest any time?
CyberPK has quit [Ping timeout: 245 seconds]
<oliv3r> well working android still requires a linux GPL kernel :)
<Undertasker> As if they would care about this in China
<oliv3r> AW actually is reasonably GPL compliant
<oliv3r> they even leaked their entire git tree + history
<fra79Wii> Sllwinner tabs are not so cheep anymore though... All the hw have gone cheap now
<oliv3r> A20 will be nice
<fra79Wii> Allwinner..sorry...the complete openess would attract more people to trust and buy...
CyberPK has joined #linux-sunxi
<Undertasker> Always the same game. When all the information is leaked or reverse engineered, that hardware is obsolete
<fra79Wii> oliv3r: yes I hope would be possible to swap the SoC. I'm looking forward for A20 for months
<Undertasker> cubieboards should be available soon
<oliv3r> fra79Wii: FunkyPenguin actually has 4 A20 based boards :)
<oliv3r> Undertasker: yes, but in this case, the A20 is very similar to the A10
<oliv3r> and the code has been leaked for those aswell
<oliv3r> so once we have hardware, we'll have the same level of support for A20, that we have for A10 now
<oliv3r> and A20 isn't even easly avialable
<oliv3r> A40 will hopefully be also quite similar *pray*
<fra79Wii> Wow..would be quad A7 with mali?
<oliv3r> fra79Wii: we are hoping :p, praying even
<oliv3r> we guess it may be single or dual A7, quad A16
FunkyPenguin has quit [Ping timeout: 252 seconds]
FunkyPenguin has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
BJFreeman has quit [Ping timeout: 252 seconds]
FunkyPenguin has quit [Ping timeout: 256 seconds]
FunkyPenguin has joined #linux-sunxi
fra79Wii has quit [Ping timeout: 255 seconds]
<oliv3r> hmm, interesting, looks like all libs put their data directly in the sram for processing
vinifm has quit [Quit: Saindo]
fra79Wii has joined #linux-sunxi
<fra79Wii> But A7 is more performing than A8?
<fra79Wii> More power efficient for sure...but st same clock are almost the same?
<fra79Wii> I thought for some reason that cortex A7X2 was slightly better/if not the same than cortex A8
fra79Wii has quit [Quit: AndroIRC - Android IRC Client ( http://www.androirc.com )]
fra79Wii has joined #linux-sunxi
fra79Wii has left #linux-sunxi [#linux-sunxi]
rellla has quit [Remote host closed the connection]
fra79Wii has joined #linux-sunxi
_BJFreeman is now known as BJFreeman
paulk-desktop has quit [Quit: Ex-Chat]
<fra79Wii> For some reason I though dual a7 was almost the same than cortex A8
<fra79Wii> Performance whise
FunkyPenguin has quit [Ping timeout: 264 seconds]
FunkyPenguin has joined #linux-sunxi
<Turl> fra79Wii: from what I heard A7 is similar to A9 perf wise
FunkyPenguin has quit [Ping timeout: 248 seconds]
<fra79Wii> Not bad then...less power drain, more horse
<fra79Wii> Sw video playback is not terrible though in android A10
FunkyPenguin has joined #linux-sunxi
Undertasker has left #linux-sunxi [#linux-sunxi]
<fra79Wii> I guess with more CPU (eg A20) it might be possible to live w/o cedarX
<fra79Wii> Turl: how is the zatab/allwinner-dev-team port to jb 4.2 going
FunkyPenguin has quit [Ping timeout: 276 seconds]
<Turl> fra79Wii: I haven't worked on it for a long time
<fra79Wii> I've used a lot your code.. I was wondering if you also noticed an substantial improvement of performance using 4.2
<Turl> fra79Wii: but it was pretty much all working except camera and cedar
<fra79Wii> I got camera almost working (I can see myself but not snapshots)
FunkyPenguin has joined #linux-sunxi
<Turl> I didn't bother with camera much as the hal links with cedar
eebrah has quit [Read error: Connection timed out]
<fra79Wii> I'm trying to use the sw implementation of androidx86
<fra79Wii> Of camera...
<fra79Wii> I just have to work on the camera.cfg parsing
eebrah has joined #linux-sunxi
<fra79Wii> 'Just' as far as I know now
CyberPK has quit [Quit: Page closed]
FunkyPenguin has quit [Excess Flood]
FunkyPenguin has joined #linux-sunxi
<fra79Wii> I have to go... Gave a nice week end bye!!
<fra79Wii> Have..
fra79Wii has left #linux-sunxi ["Messaggio"]
<oliv3r> looks like android 2.3.4 was used/compiled against to get those blobs
<oliv3r> no wonder they are so shit and have a hard time working with 4*
FunkyPenguin has quit [Ping timeout: 276 seconds]
FunkyPenguin has joined #linux-sunxi
eebrah has quit [Read error: Connection timed out]
FunkyPenguin has quit [Ping timeout: 268 seconds]
Guest11682 has joined #linux-sunxi
FunkyPenguin has joined #linux-sunxi
FunkyPenguin has quit [Ping timeout: 252 seconds]
FunkyPenguin has joined #linux-sunxi
FunkyPenguin has quit [Ping timeout: 256 seconds]
Guest11682 has quit [Read error: Connection timed out]
Guest11682 has joined #linux-sunxi
FunkyPenguin has joined #linux-sunxi
FunkyPenguin has quit [Excess Flood]