<rz2k>
gpu supports opengl/gles/cl; still no idea how it does the gl one, glxinfo reports vivante as gl client and it does 250fps in first test of glamrk2 (the standard one, not gles) with 40% load on cpu.
<bsdfox>
wow
<rz2k>
best part of it, vpu and gpu are completely separated, Jas posted a vid with movie and glmark2 running at the same time without lags
<bsdfox>
just put a little active cooling around it and you're golden
<rz2k>
I use it without the case, so it goes to max pretty fast without anything on SoC
<bsdfox>
does it boot from sd like the a10?
<bsdfox>
or is linux on the flash
<rz2k>
it has two sd slos
<rz2k>
one internal and one external
<rz2k>
it doesnt have the NAND
<bsdfox>
no nand?
<bsdfox>
ahh
<rz2k>
yeah
<rz2k>
and standard sdcard is class4, so many guys are complaining on standard android about performance
<rz2k>
under heavy mmc load
<rz2k>
so you need to replace it asap, I was forced to do it because after couple of dd's shipped sdcard just died
<rz2k>
:)
<ssvb>
this sounds like a good discussion topic for #arm-netbook :)
<rz2k>
ssvb: yeah, how is your g2d stuff going?
<rz2k>
saw your commit
<rz2k>
is it worth trying?
<ssvb>
yes, it works
<ssvb>
but needs a kernel patch, as usual :)
<rz2k>
cool, thanks for your work. this is the one of many g2d applications, right?
<rz2k>
I mean copyNtoN
<ssvb>
yes, it affects scrolling in xterm, geany, gedit, ...
<ssvb>
but unfortunately does not help firefox
<ssvb>
we need somebody to build firefox with gles layers and webgl support :)
<ssvb>
I just need to also add XV acceleration (should not be too difficult) and push a new release out
<hramrach>
shouldn't firefox use gtk>cairo>w/e
<hramrach>
would be too easy I guess
<ssvb>
firefox uses a lot of tricks and ugly hacks to avoid doing things which tend to be slow
<ssvb>
I also need to write a blog about EXA and why it is not good enough for us
<hramrach>
exa is bad .
<hramrach>
there are newer accel stuffs
<hramrach>
it's just nobody wrote cool new accel for most drivers
<hramrach>
well, it turns out whatver tricks firefox use they are slow and ugly in the end
<ssvb>
well, nowadays they focus on D2D and GLES acceleration, so legacy software rendering and X11 code paths suffer
<hramrach>
like nobody uses X11
<Turl>
what are they using for fx for android?
<Turl>
hramrach: well, most use D2D
<bsdfox>
hmm I wonder if chrome/chromium is available for arm now
<hramrach>
why would not it
<bsdfox>
there weren't any packages in ubuntu/linaro last I checked
<ssvb>
AFAIK firefox on Android uses GLES for compositing layers (scrolling, zooming, ...) and software rendering for the content of these layers
<Turl>
considering there's a full *chrome*book I'd be surprised it it weren't running on arm :)
<bsdfox>
and I think it would take a long time to compile on my 512mb mele :)
<bsdfox>
Turl, that was kinda my logic :)
<hramrach>
apt-get install chromium
zub has quit [Ping timeout: 252 seconds]
rz2k has quit []
h36sa has joined #linux-sunxi
calris has quit [Ping timeout: 245 seconds]
calris has joined #linux-sunxi
zub has joined #linux-sunxi
<WarheadsSE>
I realize i should know this by now, but is there a sitting copy of the most recent build of the cubieboard uboot somewhere
egbert has joined #linux-sunxi
egbert_ has quit [Read error: Operation timed out]
<WarheadsSE>
nvm, ill make one ;P
ganbold_ has quit [Remote host closed the connection]
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
ganbold has joined #linux-sunxi
orly_owl has quit [Ping timeout: 272 seconds]
uminded has joined #linux-sunxi
h36sa has quit [Ping timeout: 245 seconds]
<uminded>
http://packages.linux-sunxi.org/debian/ just saw this link around the interwebs, is the project planning on hosting a full debian armel/armhf distro?
h36sa has joined #linux-sunxi
<drachensun>
has anyone tried to the use the u-boot lichee branch for booting from nand?
orly_owl has joined #linux-sunxi
bfree_ has quit [Remote host closed the connection]
<shineworld>
interesting ... actually I've board by some days so I've just compiled original AndroidTV and changed some android sources to meet my needs
<Epsylon3>
but first error all people do: kernel is not made in kernel directory
<Epsylon3>
in android repos
<shineworld>
I'm happy to have some new interesting forks on project
<Epsylon3>
but in root folder
<shineworld>
uhm... only a base question for you : do you have a script to help in store android to SD and run it on SD ?
<shineworld>
I'm just begin to try it but is a very lost of time at moment
<Epsylon3>
E/audio_hw_primary( 76): Unable to find 'Playback MIXPAS' mixer control
<shineworld>
because I've to understand u-boot and related
<Epsylon3>
rmm, whats the new name for that :/
<Epsylon3>
ADC Input Mux ... Ldac Left Mixer
<Epsylon3>
well this evening... need to go at work...
<shineworld>
uhm main branch is ICS or ICS-OPENTV ? I need to git .....
<Epsylon3>
there is no uboot stuff in android tree
<Epsylon3>
ics-opentv is the original one
<Epsylon3>
or almost
<Epsylon3>
ics, current branch i work on
<shineworld>
ok, but to run it on sd ... BROM -> SPL -> UBOOT -> ?
<oliv3r>
there! Done with FEX_Guide wikification! just needs to be tagged what section is required by what (uboot, spl, boot0/1, kernel) and then the kernel bits, once in the FDT can be moved to a seperate document
focus has quit [Remote host closed the connection]
<mnemoc>
oliv3r: congrats!
<n01>
\0/
<oliv3r>
shineworld: what is opentv; because it is not wikipedia.com/opentv which is some propriatery framework crap
<mnemoc>
oliv3r: it's just allwinner android for large displays....
<oliv3r>
ok, they really should not call it opentv
<mnemoc>
they call it "open" because G doesn't have to authorize the name usage
Guest94292 has quit [Remote host closed the connection]
<oliv3r>
it's like saying, 'oh look, we made this new xml standard for our documentation, we call it openxml'
<shineworld>
Not probabily is the AllWinner ics distro (based on AOSP) for some OpenTV gadget powered by A10 and pre-build and stored in cubieboard nand
<oliv3r>
yeah, they should have atleast googled opentv before using it :p
<mnemoc>
that's asking too much
<oliv3r>
mnemoc: i'm sorry, I appologize
<oliv3r>
i hate though, how the word 'open' has been raped and abused so badly (just like opentv (the real one) and openxml etc)
<mnemoc>
how you dare to expect people to *research* before naming things?!
<oliv3r>
Lets call it OpenCRAP and make people think it's really open!
<shineworld>
oliv3r, is a very elementary distro, very few installed apps, but had already gapps on it (breaking some google rules ? )...
<oliv3r>
LibreTV :p
focus has joined #linux-sunxi
<shineworld>
however I've build (fixing something) it and works
<rellla>
oliv3r: aren't you fex'o'phobic now? thanks a lot!
<oliv3r>
what I don't understand about CM (this will be a little rant) they support many architectures and handsets, why does everybody have to recompile the same stuff over. Why isn't there a CM10.2 package, with all the libre parts in 1 image, and then 'hwpacks' for specific boards. I guess you'd need a few of each due to different architecture optimizations, but with armv6, armv7 you'd cover a lot
<mnemoc>
shineworld: can you see what's there that isn't on our more "official" trees?
<oliv3r>
rellla: i just thought it needed some cleanup and addition of parameters :)
ibrah has joined #linux-sunxi
<oliv3r>
what bums me is that most of those parameters may go away sometime when it's moved to FDT and the driver is cleaned up
<oliv3r>
mnemoc: btw, that ssd driver I asked about yesterday, I started it but boy is it really messy
<mnemoc>
oliv3r: :)
<shineworld>
mnemoc, ? I don't understand ...
<shineworld>
my english is so limited sometime
<oliv3r>
mnemoc: also, I uploaded the datasheet for it, and it explains all i2c addresses and their meanings etc, so a proper datasheet. Now the first step (after cleaning and submitting) will be to put those magic values behind defines, but then, do we want some form of documentation for each define in the .h; or just refer to the datasheet?
<mnemoc>
shineworld: there is this .repo-less opentv thing used by hipboi you fixed. and there are real git repositories maintained
<mnemoc>
shineworld: it would be good to know if there is anything useful. and integrate it in the repos
<oliv3r>
rellla: also, those chinese translated garbage is now readable english (well I tried to :p) just need libv to go over the disp section to verifiy/correct those as that was the 'hardest' section in understanding what stuff ment
<shineworld>
ah so there is a repo for AllWinner OpenTV .... that sound good
<shineworld>
have you the git address for it ? I could add my efforts to community
<rellla>
oliv3r: don't change my nice overwhelming <tables> !! ;)
<oliv3r>
rellla: I think I actually copied them (and improved where possible)
<oliv3r>
i was lazy/sloppy om some minor thhingies
<shineworld>
ok so is the official fork where to works ... I will check differences when at home
<mnemoc>
shineworld: currently it insists in per-device device trees, but we should also migrate it to a multi-device like Epsylon3's
<mnemoc>
shineworld: Turl maintains device/allwinner/common, which needs to be forked and adjusted to be a multi-device with device/allwinner
<shineworld>
I'm used with multi-device way because in my repo I've to support a lot of armv6 phones brands
<mnemoc>
awesome
<rellla>
oliv3r: i only put the in-fex-docu regarding disp-options in imo readable <table>-format. nothing changed on the basics so far.
<shineworld>
to simplify I've a device/get-semc.sh /device/get-cooper .... so main tree is common and one can choose what platform add
<oliv3r>
rellla: well look at what I did then; i doubt I changed much then, I did try to translate some of the chinglish
<mnemoc>
shineworld: having the multidevice thing, my idea was to improve sunxi-bsp to be able to build images to all devices in sunxi-boards reducing build duplication
<shineworld>
yes, is a good idea
<shineworld>
at TeamFun we do same thing (or similar)
<rellla>
chinglish is nice, as long as you can do google-translate \o/
<mnemoc>
i don't know android, so any help to improve the bsp and the local manifest would be greatly appreciated
<shineworld>
just time to practice a little with it
<mnemoc>
:)
* rellla
wishes to get some news on cedarx, so this week may become a sunxi-positive, since libv/cwabbots statements look very promising...
<shineworld>
I don't know so much the sunxi family
<shineworld>
I'm feeling a little lost when I walk in wiki and related documents (at moment)
<mnemoc>
also, as we plan to deprecate 3.0 (and forward port sun[67]i from 3.3 to sunxi-3.4) we need help testing and fixing sunxi-3.4 for android usage
<shineworld>
cya... meeting time
<mnemoc>
shineworld: don't hesitate to ask any doubt you have
<shineworld>
ok ... I need to go ... boss call
<mnemoc>
shineworld: but key is, there are thousands of allwinner based devices out there, but behind the hoods the only device-specific detail is the .fex file and TS driver
<oliv3r>
but isn't a ASOP build the same? I mean, there's many 'hwpacks' out there, but nearly all binaries are the same on identical architectures (armv7-hf)
<oliv3r>
i find the whole android world to be _extremly_ messy
<oliv3r>
and i don't think it should or needs to be
<mnemoc>
android actually is not hf
<oliv3r>
so what if we would build it as such? wouldn't the compiler atleast be able to optimize some of it? even if just a small fraction?
<mnemoc>
also, even if userspace is mostly cross platform they tend to need per-device kernels. we don't
<oliv3r>
or is dalvic not hf capable at all
<oliv3r>
per device or per soc?
<mnemoc>
them, per device
<mnemoc>
we have script.bin ;-)
<oliv3r>
well FDT and pinmux is hoping to change that
<oliv3r>
they will have at some point FDT though :p
<mnemoc>
5.0 will be 3.8 based
<oliv3r>
does AW we have mainlined some stuff for 3.8?
<mnemoc>
but very few socs are FDT compatible as of 3.8
<mnemoc>
oliv3r: they are working on 3.3
<oliv3r>
If not, I'm curious what kind of wtf reaction it will be when they try to port their changes to 3.8
<mnemoc>
oliv3r: they will just hack their own stuff in
<mnemoc>
and screw fdt and pinmux
<oliv3r>
mnemoc: most probably they will nuke our stuff, and continue on their changes
<mnemoc>
their 3.3 tree has their own implementation of a continuous memory allocator
<mnemoc>
oliv3r: absolutely
<oliv3r>
stupid stupid stupid really, they should have worked with the community when the first silicon was being designed (as Greg basically says they should)
<mnemoc>
otoh it's possible they hope we to mainline the enough stuff before they need to care about android 5.0
<oliv3r>
get specs + code out as early as possible; oh well; android 6 or whatever will use 3.12 or whatever will be interesting
<oliv3r>
depends on how much they know we do
<mnemoc>
they know
<oliv3r>
:S
<mnemoc>
allwinner devs are CC/ed by mripard
<oliv3r>
oh really; ok then they should know
<mnemoc>
and they contact him for... guidance
<oliv3r>
really? heh
<oliv3r>
didn't know that bit :)
<mnemoc>
also, the rename to reumilla in their copyrights seems to have been a pro-leak decision
<oliv3r>
ok, technical question; if I use bit-fields ina struct, why do I have to declare them as unsigned, how can a single bit be 'signed' ever. I guess multple bits could have 1 sign bit
<oliv3r>
mnemoc: pro? post-leak?
<oliv3r>
or pre-leak?
<mnemoc>
pro, favour
<oliv3r>
I read the whole reuumilla and newbietech or whatever discussion
<oliv3r>
but didn't quite understand it, why rename?
<mnemoc>
in that way the leaked code lets people use the devices but at the same time "no one" can blame allwinner itself. it's reuumilla
<oliv3r>
i mean, we are crediting allwinner with their code when submitting?
<oliv3r>
ahh
<mnemoc>
chinese are VERY fearful
<oliv3r>
i guess
<mnemoc>
they don't want credit. they want to 1) not be called for free support. 2) not to be sued
<oliv3r>
so what about the allwinner headers we add? are they not 'scared' of those?
<mnemoc>
our allwinner headers are pre-reuumilla
<ganbold>
mnemoc: is there any datasheet for wemac of a10?
<mnemoc>
they changed to reuumilla immediately after we asked them to GPL our sources
<mnemoc>
ganbold: dream on
<oliv3r>
ganbold: there's 'some' e-mac doc in the A10/A13 datasheet, but not much
<mnemoc>
ganbold: but please review the lichee/a10-dev and lichee-3.3/a20-dev branches in my github for possible fixes and improvements regarding wemac
<mnemoc>
funny thing is that the wemac driver was submited as a davicom
<mnemoc>
to mainline
<mnemoc>
and the only thing we know is that they used dm9000 driver as template to write their own
<mnemoc>
anyhow, nice to see more code getting mainlined :)
<oliv3r>
aye :) the more the better
<oliv3r>
and it's emac! not wemac (well in script bin and in their datasheet it is refered to as emac, dunno where the whole W came from
<mnemoc>
.ko name
<oliv3r>
*facepalm*
<mnemoc>
sun6i's is called gmac in their tree
<oliv3r>
Gigabit mac :D
<ganbold>
mnemoc: I'm trying to write wemac driver for freebsd, but not sure what I'm doing :)
<mnemoc>
exactly
<oliv3r>
i wonder how much it is based on the dm9000
<mnemoc>
ganbold: look at the code submited to lkml as ... reference
<mnemoc>
oliv3r: there is 0 similarity in the registers
<mripard>
mnemoc: oliv3r: feel free to comment on our patches.
<mnemoc>
mripard: nah, I'm happy with it that way
<oliv3r>
mripard: que?
<mnemoc>
the davicom thing
<oliv3r>
oh, i don't know anything about the emac, just that the nameing gets mangled and is inconcistant
<mnemoc>
back then Tom acked there was a connection between that IP and davicom
<oliv3r>
maybe we should rename it to aw-emac :D
<mnemoc>
sunxi :)
<oliv3r>
yes! duh, sorry
<oliv3r>
sunxi-emac
<mnemoc>
mripard: I just commented it as an example of how much we need to "guess" because of allwinner's secrecy
<mnemoc>
mripard: i doubt allwinner made their own ethernet IP, but I'm unsure they actually paid for the license of whatever ethernet IP they are using
<mnemoc>
mripard: and that can explain why they refuse to document it
bsdfox has quit [Read error: Connection reset by peer]
<mripard>
mnemoc: probably
<oliv3r>
how about (when I have some time) rename it all to kernel/drivers/net/ethernet/sunxi/sunxi.c (don't think it needs to be called sunxi-emac'
<oliv3r>
though it would help if it has a name
<mnemoc>
oliv3r: we already have gmac codes...
<mnemoc>
oliv3r: as mripard said, reply on lkml ;-)
ibrah_ has joined #linux-sunxi
ibrah has quit [Ping timeout: 252 seconds]
<ganbold>
mnemoc: do you have link to this mail on lkml?
<oliv3r>
before starting a discussion on lkml; I think we should be atleast in agreeance here :p so it's been pushed as ethernet/davicom/wemac; so either rename it to ethernet/sunxi/dm9000? sunxi/emac? davicom/sunxi?
<mnemoc>
dm9000 is not
<mnemoc>
it's *possible* it's davicom IP, but it's not dm9000
<oliv3r>
just spouting out some names before replying with a patch :p
<oliv3r>
but I think having sunxi as a directory would be fair, even if it is heavily or remotly based on davicom IP
<oliv3r>
they changed it enough for it to become their own I guess?
<mripard>
sunxi is not an option, it's not a vendor
<oliv3r>
otherwise, davicom/sunxi :)
<oliv3r>
then it's davicom/sunxi
<mnemoc>
allwinner ?
<oliv3r>
i don't think we can do allwinner/sunxi?
<mripard>
if you want that logic, then allwinner/emac and gmac
<oliv3r>
or allwinner/emac
<mnemoc>
allwinner/sunxi-{emac,gmac} ?
<oliv3r>
I guess that's best then :)
<mnemoc>
to avoid poluting the .ko space
<mnemoc>
I mean, emac/gmac are just too weakly named
<mnemoc>
so a prefix would be nice
<oliv3r>
that's the whole reason why i'm bringing it up :)
<mripard>
ok, I'm fine with this
<oliv3r>
ok i'll write a patch today/tomorrow to rename it from davicom/wemac, to allwinner/emac and allwinner/gmac
<mnemoc>
no prefix?
<oliv3r>
yes sorry
<oliv3r>
sunxi-emac/gmac sorry
<oliv3r>
and i'll post it to lkml
<mripard>
it's not merged yet
<oliv3r>
where is your git tree and where shall I post it
<mripard>
so just suggest the new name
<mripard>
in a mail
<mripard>
i'll do the patch
<oliv3r>
to sunxi ML?
<mripard>
I have some changes to make anyway
<mripard>
no, to LKML
<oliv3r>
not linux-arm-kernel? ok to lkml
<mripard>
well, both LAKML, LKML and netdev are recipient to this patchc
<mripard>
so just reply to all :)
<mnemoc>
oliv3r: [reply-all] for the win
<oliv3r>
not subscribed to any of those :p
<mripard>
me neither
<mripard>
doesn't matter :)
<oliv3r>
nor do I have those mails in my inbox to reply to
<oliv3r>
i'll find it ;)
<mripard>
that's fine then, I'll do it :)
<mnemoc>
oliv3r: and subscribe, so you can reply next time ;-)
<oliv3r>
more ML spam *sigh* :p
<oliv3r>
could have gotten the mail to reply
<oliv3r>
mripard: sorry :(
<mnemoc>
oliv3r: you can reply from the gmane url I pasted above
<oliv3r>
i'll go finish cleaning up the touchscreen driver :)
<oliv3r>
i opend the infradead one :p
<mnemoc>
infradead didn't even care to reply when I asked them to host a list for us :<
<oliv3r>
:(
<oliv3r>
ok so gmane, i have the site open, i went to 0/5 (The coverpage instead of the 1/5 patch). how do I reply all via their webinterface
<oliv3r>
do I need to somehow log in?
<oliv3r>
ive read tons via gmane (old stuff etc mostly) never 'did' anything with it
<oliv3r>
didn't even know you could
<oliv3r>
'follow up'!
<mnemoc>
oliv3r: look for "follow up" on the [actions] menu
<mnemoc>
meh... need more coffee. I've been late on every answer today
<oliv3r>
sent :)
<oliv3r>
that's pretty cool, didn't know that was posisble
<mnemoc>
it's not the most awesome example of modern web design, but it does it's job awesomely
<oliv3r>
Dear Oliver, Please refer to the attached ZET6221 spec. Any question, please feel free to contact me. Moble:+8613538asdf
<oliv3r>
not perfect, but a start
jochensp has quit [Ping timeout: 260 seconds]
<oliv3r>
it's 2mb, looking at it now
<mnemoc>
specs?!
<mnemoc>
:D
<mnemoc>
oliv3r: so, [[ZET6221]] + attachment ? :)
<oliv3r>
well
<oliv3r>
let me just say I got a boner from this
<oliv3r>
:p jk
* mnemoc
steps back
<oliv3r>
mnemoc: check the two pdf's I uploaded yesterday/today
<oliv3r>
maybe you can move them to dl.linux-sunxi.org/datasheets ?
ibrah_ has quit [Read error: No route to host]
ibrah has joined #linux-sunxi
<oliv3r>
last modification date 22- may 2012, initial version 26-feb 2011, so it's been kept updated too
<oliv3r>
where is wingrme when you want to make him happy
ibrah has quit [Ping timeout: 272 seconds]
<oliv3r>
anybody? all ran away scared?
<mnemoc>
first the boner, then your desire of making wingrme happy, ... what do you expect?
<oliv3r>
:p
<oliv3r>
ok the datasheet is pretty detailed
<oliv3r>
I think it's exactly what wingrme needs
<oliv3r>
even has a small reference C program
<oliv3r>
:)
<oliv3r>
full protocol description
<mnemoc>
nice
<oliv3r>
yep it's the real deal
<oliv3r>
it's uploaded to the wiki in any case :)
<oliv3r>
and that within 12 hours
<oliv3r>
how cool is that
<mnemoc>
:)
<oliv3r>
i'll put my e-mail somewhere on the wiki, as a 'contacting manufactures for datasheets' or something
<oliv3r>
i guess if you copy/paste the reference code, you have 90% of the driver allready
<oliv3r>
looks like it's linux reference code, but can't be sure
<oliv3r>
datasheet even explains how the boatloader works and how to update its flashrom
<oliv3r>
or download the flashrom from the IC i think
<oliv3r>
so it can be read, dissasmbled etc
ibrah has joined #linux-sunxi
<ganbold>
looks like mripard's wemac change is much better than one in linux-sunxi
<oliv3r>
in kernel structure, u8 or uint8_t?
uminded_ has joined #linux-sunxi
techn__ has joined #linux-sunxi
techn_ has quit [Read error: Connection reset by peer]
uminded has quit [Ping timeout: 255 seconds]
hansg has quit [Ping timeout: 260 seconds]
shalem has joined #linux-sunxi
ibrah has quit [Ping timeout: 276 seconds]
<mnemoc>
oliv3r: u8
<mnemoc>
oliv3r: uint8_t on userspace
<oliv3r>
ok, i'll do a 2nd patch to clean that up
<oliv3r>
and kernel to userspace is __u
<oliv3r>
I did read that part
<oliv3r>
but in this driver, there should be hardly any if at all 'to userspace' bit
<oliv3r>
coordinates should stay internal and go to the input layer and that probably does the conversion
rellla2 has joined #linux-sunxi
rellla has quit [Ping timeout: 245 seconds]
<mripard>
ganbold: stefanro and myself spent quite some time on cleaning it up...
uminded_ has quit [Quit: Leaving]
ibrah has joined #linux-sunxi
<ganbold>
mripard: do you have it in your github?
<mripard>
yes
<oliv3r>
ok, how does this, or will it work; currently drivers check the script.bin for certain parameters. Right here, I have a touch panel driver, that specifically checks for ctp7_name, so a driver is bound to it's number in script.bin? how will this handled in FDT later? Will this be prettier? because this looks extremly ugly. How would we want to deal with this meanwhile? rename/change it to ctp_chipname_name?
ibrah has quit [Ping timeout: 264 seconds]
<Turl>
hi mripard
<oliv3r>
turl! ii wanted to ask you something
<oliv3r>
you are doing android images right? they based on ASOP or on Cyanogenmod?
<Turl>
oliv3r: ask away :)
<oliv3r>
couldn't find much on the wiki
<Turl>
CM
<oliv3r>
do you have any info on installing your CM on a A10 device on the wiki? most android stuff goes about fully compiling it etc
<Turl>
mripard: I saw you sent a new iteration of the 8250 patch, but included an incorrect IS_ERR on the uartclk if; read the conversation with Russell on that patch
<Turl>
oliv3r: yeah there's a page about installing CM on zatab iirc
<mripard>
Turl: erf...
<mripard>
I have archived the wrong mail, and so I just saw the first one...
<mripard>
I guess we will have a v5
<Turl>
:)
<Turl>
can't you just send a replacement for the bad one? it's isolated on a file
<oliv3r>
do other drivers really have a '*_write_reg' for each i2c register that's being changed? Don't they usally have some array to send all init data via a for loop or something? this is getting tedious :p
Skaag has quit [Read error: Connection reset by peer]
<slapin_nb>
oliv3r: hi, I remembered something among these months, something about wiki...
<oliv3r>
lol, i made many many contributions
<oliv3r>
if you haven't found the ( . ) (. ) yet
<oliv3r>
you are not searching enough!
rz2k has quit [Read error: Connection reset by peer]
rz2k has joined #linux-sunxi
<slapin_nb>
oliv3r: hmmmmm
<calris>
Is sunxi armv7a-hf?
<mnemoc>
sun4i and sun5i are cortex-a8, sun6i and sun7i are cortex-a7. all these qualify as armv7a, but the hf thing depends on how you compiled your userspace
gzamboni has joined #linux-sunxi
<calris>
Ino userspace - U-Boot
<mnemoc>
u-boot is float-free
<mnemoc>
so if your toolchain is hf or not is not relevant
<calris>
Thanks
drachensun has quit [Ping timeout: 256 seconds]
<oliv3r>
where's wingrme today? he usually is so active
<oliv3r>
god, I hate the fact, that the kernel + checkpatch want single statement blocks without curly bracers
<oliv3r>
curly bracers cost 1 extra line and are well worth it
<oliv3r>
not using them causes bugs, proven fact, some get caught in reviews, some do not
<Turl>
oliv3r: they only cause bugs if you've been programming on python before C
drachensun has joined #linux-sunxi
<oliv3r>
Turl: lies :p i'
<oliv3r>
i've never touched python and made mistakes
<oliv3r>
especially when late, tired and in a hurry :p
<mnemoc>
my only problem with linux coding style are the f* 80 columns :p these days you have have side by side two terminals of at least 100 chars each
<oliv3r>
i know exactly how bracers work around single statements, it's just bound for problems :) 'one true brace' rule fixes that, and the kernel should adapt to it;
<mnemoc>
and those 20 char have big impact in readability when you have to break a line
<oliv3r>
and yes, 80 collumns, wtf. we're not in 1965 anymore
<oliv3r>
it would almost incourage you to use short cryptic names
<oliv3r>
who writes code on 80 colulmn terminals anymore
<mnemoc>
well... for looos with long count var names are much harder to read than i=0
<mnemoc>
above 100, code lines are indeed unreadable
<mnemoc>
but 80 is way too tight
<mnemoc>
sometimes, when passing structs as arguments to a namespaced function, you can hardly fit the first argument in the first line
<oliv3r>
yep
<oliv3r>
i fully wholehardedly agree
<oliv3r>
and even then '81 lines, no nonono, bad programmer!'
<oliv3r>
make it a 95 warning, 100 error limit
<oliv3r>
where 95 is acceptable
<oliv3r>
or something :)
ZaEarl has joined #linux-sunxi
<mnemoc>
specially considering all the elite of kernel devs use fancy macbooks and now this chromebook pixel thing
<mnemoc>
with a ton of pixels wide
<mnemoc>
so they can probably fit 4 80char terminals side by side
<mnemoc>
and so with a 100char hardlimit they would still be able to fit 3
<Turl>
oliv3r: I write code on a 80x24 term most of the time
<mnemoc>
uh
<mnemoc>
my usual term for programming is 110x58
<oliv3r>
i stretch my term to something comfterable
<Turl>
mnemoc: too big, requires moving eyes :P
<oliv3r>
that said, at home i have 2x 24 1920x1200 and at work 1 of those + a 21" 1600x1200
<oliv3r>
80 vs 100 lines is a minor difference in characters, but yet gives a lot more freedom
<Turl>
oliv3r: I work with 2x24 1080p
<oliv3r>
the 80 collumn limiti s purely arbetrary from the old days of 80 collumns terminals
<mnemoc>
in 1024x768 times you hardly can put to >80 terms sides by side
<mnemoc>
but no one has that resolution these days
<Turl>
I was using 1366x768@16" up to not so long ago
<Turl>
it's an incredibly common res for notebooks
<mnemoc>
I'm using 1366x768@11.6" right now
<Turl>
@11 ouch
<mnemoc>
and I work have 110x58 side by side
<mnemoc>
the left column is split to allow a 110x14 term for irssi :p
<Turl>
I don't usually work with stuff side by side
<Turl>
when I need that kind of stuff I use geany :P
<mnemoc>
i use one for vi and the other for commands
<Turl>
I have geany set up with limit columns line at the 72th column or so
<Turl>
so if I hit it I know I don't have much more to go until it's 80
<oliv3r>
:231
ganbold_ has joined #linux-sunxi
<mnemoc>
oliv3r: do you number the irc lines? :p
<oliv3r>
wrong window :(
<oliv3r>
but good catch ;)
<oliv3r>
mnemoc: i'm almost done finishing up with the formatting issues, and want to push that before doing anything else (it could use some love)
<oliv3r>
i pushed his patch into my repo so i could format patch (with his credits) it and then format-patch my patch after it
<oliv3r>
or do you prefer pulling the e-mail manually?
<mnemoc>
nah
<mnemoc>
preserving authoriship is good
<oliv3r>
i'll git email both patches then
<mnemoc>
thanks
wingrime has joined #linux-sunxi
<oliv3r>
there. he. is.
<oliv3r>
i got you a present
<oliv3r>
guess he's not here afterall :(
ganbold_ has quit [Ping timeout: 264 seconds]
ganbold_ has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
ganbold_ has joined #linux-sunxi
<oliv3r>
oh god, not this again 'Alignment should match open parenthesis'
<mnemoc>
vi does that automatically for me
<oliv3r>
it does for me too
<oliv3r>
but checkpatch doesn't agree
<oliv3r>
actually, vim put it on 3 tabs, checkpatch wanted it at 2 tabs
<oliv3r>
because the tab ended just at the ( above
<mnemoc>
2 tabs + 7 spaces maybe
<oliv3r>
nah, it was perfect
<oliv3r>
i have visual tabs
ganbold_ has quit [Remote host closed the connection]
<oliv3r>
'avoid line continuations in quoted strings' fu.
<oliv3r>
'line is to long, make shorter'
<oliv3r>
but don't split it fool!
<oliv3r>
:389q
<oliv3r>
crap can't fix even fix this one :)
ganbold_ has joined #linux-sunxi
<oliv3r>
4 tabs is to long, it'll be more then 80, but 0 tabs, 1 tab 2 and 3 tabs all give me the 'to long :S
<Turl>
can't you split the string?
<Turl>
"hello this is the string.\n"
<Turl>
"this is some more of it :P"
<oliv3r>
'don't cut continued strings'
<Turl>
\n? :)
<oliv3r>
or some bla bla, so i made it two printk's to make it shut up :p
<oliv3r>
the code needs to be refactored anyway, it's horribly ugly
<oliv3r>
looks like it got copied from a nother horribly driver and just adjusted
<oliv3r>
which brings me to my next point, do we have a '
<oliv3r>
'pretty done driver' that can server as a template?
<n01>
:2
<n01>
lol
<oliv3r>
for touchscreen or similar enough to be useable
<Turl>
oliv3r: grab one that's recent and that's in mainline :D
<oliv3r>
well those sometimes are even a little messy :p
<oliv3r>
but that's probably the best starting point
ganbold_ has quit [Remote host closed the connection]
ganbold_ has joined #linux-sunxi
jochensp has joined #linux-sunxi
<oliv3r>
crap, so first it tells me to replace printk with printk(KERN_something
<oliv3r>
i change it all to KERN_DEBUG; now it tells me to use netdev_dbg or dev_dbg
<oliv3r>
fuuckk you
vicenteH has joined #linux-sunxi
<vicenteH>
hello
<oliv3r>
Hi
<vicenteH>
I am trying to flash my A13 allwiner tablet,
<vicenteH>
using livesuit for linux
<vicenteH>
but getting this error:
<vicenteH>
Dev Plugin The Device Path is: /dev/aw_efex0
<vicenteH>
pcall FAILED Tools_ii.lua:2845: attempt to index global 'hwSCanData' (a nil value) 2 275
<vicenteH>
can anyone give me any clue?
<oliv3r>
flashinng from linux is very much untested
<oliv3r>
as far as I know, you are the first one to even try :p
<vicenteH>
wow
<vicenteH>
unfortunately I get similar errors from windows xp and 7
<oliv3r>
we do have an updated USB driver on github that you should use
<oliv3r>
as the one that comes with the package won't compile
<slapin_nb>
anybody remembers of some cheap DDR3 memory found in sticks or tablets? also some part number for NAND...
* slapin_nb
buys parts for prototypes
<oliv3r>
oh protoypes!
<oliv3r>
i just did a git checkout sunxi-3.4; how do i load the defconfig using sunxi-bsp? (sunxi-3.0 worked without problems
<mnemoc>
oliv3r: go inside linux-sunxi/ and switch branch
<wingrime>
mnemoc: I send reload fix for 3.4
<wingrime>
oliv3r: this doc have reference to special tuning utility for firmware
<wingrime>
oliv3r: It very intersting to contol sensitivy
<Dreadlish>
wingrime: found
<Dreadlish>
wingrime: and its workin ;)
<Dreadlish>
wingrime: so, thanks very much
<oliv3r>
wingrime: you like it? :)
<oliv3r>
mnemoc: i switched the branch, but make comes up with lots of questions about .config
<wingrime>
oliv3r: it have c code examplrd
<oliv3r>
like Whether to enable certain new errate etc
<oliv3r>
wingrime: i think tohse 'examples' are very very very closely related to their own binary only driver
<wingrime>
oliv3r: example for firmare update
<oliv3r>
wingrime: it looked like linux code
<wingrime>
oliv3r: same I see in IDA
<oliv3r>
wingrime: upload and download, so you could download it, RE it and re-upload a newer one I think
<oliv3r>
wingrime: IDA?
<Dreadlish>
oliv3r: kind of RE ide
<wingrime>
oliv3r: Interactive Disassembler Pro
<Dreadlish>
very helpful.
<wingrime>
oliv3r: Interactive Disassembler Pro = IDA
<wingrime>
oliv3r: best soft for reverce-engenering
<Dreadlish>
omg... verbosity of rtl8188eu driver is very scary.
<mnemoc>
oliv3r: rm build/*-linux/.conf
<mnemoc>
oliv3r: rm build/*-linux/.config
<oliv3r>
ah, ok; thne yes, I belive that those samples where pulled directly from their code
<wingrime>
Dreadish: I have patches not sended to stop dmesg spam
<oliv3r>
mnemoc: i did make mrproper
<oliv3r>
but i'll do that
<mnemoc>
oliv3r: it's built within build/*-linux/, not withint the source tree
<wingrime>
oliv3r: actualy I look like we depend in firmware config
<Dreadlish>
wingrime: okay, its isnt annoying enough ;d
<oliv3r>
mnemoc: ah ok
<oliv3r>
wingrime: how did you build your driver without firmware? looks like initially you either have to 'download' it from your chip or cut it from the binary blob
<oliv3r>
unless you RE it :p
<oliv3r>
wingrime: well i hope the datasheet is helpfull :)
<wingrime>
oliv3r: actualy ALL TS IC are simply some MCU
<wingrime>
oliv3r: actualy it have flash ram
<wingrime>
oliv3r: and updating firmware non nessesory at boot at all
<oliv3r>
wingrime: ok good, well before you mess with it, remember to download it atleast once ;0
<wingrime>
oliv3r: with dataschet a can write firmware update
<wingrime>
oliv3r: and strip firmware from original ko
<Dreadlish>
with datasheet you can do rainbows with that ic ;d
<oliv3r>
im supprised they released the datasheet so easily
<oliv3r>
but it's g ood one :)
<oliv3r>
mnemoc: works
<wingrime>
Dreadish: I can do rainbow when I understad what architicture have inner TS conntroler mcu
<wingrime>
Dreadish: it can be 8051
<oliv3r>
i think it's 8051
<wingrime>
Dreadish: or simular
<oliv3r>
the datasheet says it is
<Dreadlish>
;)
<oliv3r>
i think they call it x51 what I quickly saw
<wingrime>
So than I can do some intersing stuff with IDA
<Dreadlish>
so, 8051 are available afaik
<Epsylon3>
"mnemoc: oliv3r: it's built within build/*-linux/, not withint the source tree"
<wingrime>
IDA can 8051
<Epsylon3>
in your bsp ?
<Epsylon3>
or in repo... sorry i just came back
<Epsylon3>
for the device tree, its because ive a lot a repos
<wingrime>
Epsylon3: can you test 3.4 for android ?
<Epsylon3>
that will take some time but yep
<Epsylon3>
git ?
<wingrime>
Epsylon3: we want depricate 3.0 soon As possible , so hepl us
<Epsylon3>
yep, why not
<mnemoc>
oliv3r: easy reason. they *want* to sell their chips
<mnemoc>
Epsylon3: sunxi-bsp
<Epsylon3>
oh, the official one :)
plan_b has quit [Quit: plan_b]
<Epsylon3>
ive fetched it already
<Epsylon3>
will be faster
<mnemoc>
Epsylon3: as it's intended for multi-device we make one build/foo-linux for each foo_defconfig
plan_b has joined #linux-sunxi
<mnemoc>
Epsylon3: android wise the idea is to be able to build one CM image for each supported device
<Epsylon3>
in fact its easy to include a kernel in a android build
<Epsylon3>
the reverse is more difficult
<mnemoc>
Epsylon3: and if the device is new, use a template to create the corresponding devices/allwinner/mydevice within a multi-device devices/allwinner
<Epsylon3>
if you dont understand the lunch system
<Epsylon3>
to set env
<mnemoc>
Epsylon3: [HELP WANTED]
<Epsylon3>
on ?
<mnemoc>
android (CM) support in sunxi-bsp :)
<Epsylon3>
you just have to put the zImage as "kernel" in the device tree
<Epsylon3>
to build the bootimage
<oliv3r>
mnemoc: true, maybe the domain threw them off, though i was very clear in my mail; anyway, its all good :)
<Epsylon3>
and disable a variable in Boardconfig to skip building the kernel
<Epsylon3>
only modules are dependant to that
<Epsylon3>
nothing else...
<mnemoc>
cool
<Epsylon3>
nothing use kernel headers
<mnemoc>
and modules?
<mnemoc>
.ko
<Epsylon3>
modules are just dependant to the kernel
<Epsylon3>
same modver / build config i mean
<Epsylon3>
modules are in system/lib/modules on android
<Epsylon3>
but...
<Epsylon3>
ive seen also some builds with modules in rootfs
<Epsylon3>
(boot.img)
<Epsylon3>
but its not possible for a tunable device
<Epsylon3>
to much modules
<mnemoc>
can I tell android from where to get them (with a var) instead of copying the files for each device
<Epsylon3>
too
<Epsylon3>
sure
<Epsylon3>
check the PRODUCT_COPY_FILES stuff
<Epsylon3>
source, destination
<mnemoc>
awesome
<oliv3r>
mach/irqs.h is only for arm, is it?
<Epsylon3>
mnemoc: do you have a repo ?
<Epsylon3>
aosp or tarball ?
<Epsylon3>
i mean the build folders etc
<mnemoc>
Epsylon3: currently we use cm's official + a local manifest poiting to Turl's stuff
<Epsylon3>
ok perfect
<Epsylon3>
i know all of this ;p
<oliv3r>
if you can document that on the wiki too ;)
<Epsylon3>
its more difficult to detect changes in the open tv tarball
<mnemoc>
Epsylon3: forget about that opentv thing :p
<Epsylon3>
ouch
<Epsylon3>
cm wiki is for that
<Epsylon3>
there is a lot to tell
<Epsylon3>
i do that since years
<mnemoc>
and help us getting sunxi-bsp's `make android` working :p
<Epsylon3>
a still learn everyday (weeks now :)
<mnemoc>
the only Android-frienly dev we have is Turl :<
<Epsylon3>
he have more time than i have
<Epsylon3>
currently
<Epsylon3>
new job
<mnemoc>
ok
<Epsylon3>
has*
<mnemoc>
np
<mnemoc>
at least now I have an idea about how to make sunxi-bsp's `make android` work the way I wanted it
<Epsylon3>
i guess the interresting part for you in the file build/core/tasks/kernel.mk
<wingrime>
oliv3r: most strange
<wingrime>
oliv3r: it have pressure
<Epsylon3>
depends on what do you want to do
<wingrime>
but I have only 1 fixed pressue on real device
<hramrach>
Turl: what's wrong with the nand partitions?
<hramrach>
it's not clear from the paste
<wingrime>
hramrach: check gadget reload patvh I made
<hramrach>
I wrote some of the patches but they are generally noop
<hramrach>
yeah, looking for another patch somebody posted here
<Epsylon3>
i guess the kernel i need to check is stage/sunxi-3.4 ?
<mnemoc>
Epsylon3: that's the test branch, it's rebased
<hramrach>
thanks for the patch
<Epsylon3>
yes ? and its this one ?
<Epsylon3>
git pull --rebase :)
<Epsylon3>
fix all
<oliv3r>
how do I change the config with sunxi-bsp, i can't edit the .config :p and make menuconfig probably doesn't work?
<mnemoc>
oliv3r: `make linux-config`
<oliv3r>
ah ok
<mnemoc>
that will trigger menuconfig on the right dir
<oliv3r>
do i add the CROSS_COMPILE parameter?
<mnemoc>
only if it's different than the one set in Makefile
<oliv3r>
then i have to :)
<mnemoc>
we use linaro's there iirc
<Epsylon3>
tekkuli, wingrime, amery...
<mnemoc>
tekkuli = techn
<wingrime>
oliv3r: I hopeed translate coordinats on TS IC but no way....
<Epsylon3>
amery i ve seen him previously with Turl in another chan
<Epsylon3>
almery
<mnemoc>
that's /me
<Epsylon3>
oh
<Epsylon3>
forgot that lol
<Epsylon3>
that was on atrix channel
<Epsylon3>
when they got their zatab
<oliv3r>
mnemoc: yeah, making a crosscompiler is relaly easy in gentoo :)
<mnemoc>
that's not me then
<mnemoc>
Epsylon3: on IRC i've used only this nick for over 15y already
<Epsylon3>
right, so i mismatch
<Epsylon3>
or i dont remember your nick ;)
<mnemoc>
on github and email I use my name, alejandro mery -> amery@
<Epsylon3>
ok, for the info im Tanguy Pruvot
<mnemoc>
nice to meet you :)
<Epsylon3>
ive seen some of your commits in Quarx repos
<mnemoc>
uh
<Epsylon3>
he make a cm rom too for a 7" tablet
<Epsylon3>
ainol novo7
<oliv3r>
mnemoc: ssd2533 patches actually build against 3.4; so while i don't know if the driver actually works, atleast the patch applies cleanly
<Epsylon3>
ive one too, but a different touchscreen
<mnemoc>
TS tends to be the only important difference between sunxi tablets
<Dreadlish>
yup
<mnemoc>
oliv3r: happy enough with what you posted?
<Epsylon3>
or just different version
<oliv3r>
so how to handle that in fex files? if the only difference i the touschreen?
<Dreadlish>
someone was doing jb based on that what he had on novo7
<Dreadlish>
and the only difference was ts ;d
<oliv3r>
mnemoc: it applies and builds, i'll test it, so I hope it works; i'll clean it up after that
<Epsylon3>
not sure its the chip itself of the rom in the chip
<Epsylon3>
or*
<oliv3r>
the datasheet has define names for all the addresses so wanna use those instead of the magic values
<mnemoc>
Dreadlish: we don't have android libs for jb
<Dreadlish>
mnemoc: i know
<Dreadlish>
just sayin
<oliv3r>
mnemoc: i haven't updated defconfig though;
<Epsylon3>
but results were not as fast as 4.3 for the full rom
<Epsylon3>
4.4
<oliv3r>
mnemoc: for make linux-config to work, you have to have built your kernel atleast once to pull in the defconfig?
Dave77 has quit []
<mnemoc>
oliv3r: it shouldn't be needed
<mnemoc>
oliv3r: `make` should do it's magic
<oliv3r>
/bin/sh: line 0: cd: /silo/build/sunxi-bsp/build/sun4i_defconfig-linux: No such file or directory
<mnemoc>
meh. counts as bug
<oliv3r>
i did rm -r sun4i defconfig, so i understand it can't find it
<oliv3r>
(i need to rebuild the entire kernel without preempt, hansg's fedora 18 kernel is without preempt)
<mnemoc>
Epsylon3: but compiling for.... armv5? instead of armv7a-hf
<Epsylon3>
#include "f_fs.c" was not in 3.0.62
<Epsylon3>
the include in android.c
<oliv3r>
make without linux-config works nnormally
<mnemoc>
`make` is `make hwpack`
<Epsylon3>
i will turn it off
<mnemoc>
Epsylon3: btw, gadget support in 3.4 still experimental
<mnemoc>
wingrime is working on it
<oliv3r>
well make hwpack works, make linux-config does not work (without sun4i dirrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr in build
<wingrime>
Dreadish: and thats why this is still ~alpha~
<wingrime>
Dreadish: but wait, now with datascheet I make it work defenetky
<wingrime>
Dreadish: try reload .ko
<Dreadlish>
ok
<Dreadlish>
reloaded and still the same
<wingrime>
reboot
<Dreadlish>
done at least 2 times ;d
<Dreadlish>
but wait
<wingrime>
Dreadlish: it strange
<Dreadlish>
yup, i know
<wingrime>
Dreadlish: if you touch I2c configs in fex///
<Dreadlish>
no, i haven't
<wingrime>
realy first time see that
<Dreadlish>
hmm
<wingrime>
I had other problem
<wingrime>
but with datacheet I know where
<Dreadlish>
if bin2fex work properly i could get some info from another fex that was working
<wingrime>
Dreadlish: last time you report as All worked?
<wingrime>
Dreadlish: what did you canged?
<Dreadlish>
wow
<Dreadlish>
wingrime: i changed fex to get usb working
<Dreadlish>
but i have solution i think ;d
<Dreadlish>
yup
<Dreadlish>
i have it
<wingrime>
?
ibrah has quit [Read error: No route to host]
<Dreadlish>
wait a sec
<oliv3r>
holy shit why is my 8192cu.ko from 3.4 8MiB whereas hansg's 3.4 is only 750kb
shineworld has joined #linux-sunxi
<wingrime>
oliv3r: this shitty ts complitly ignores i2c adresses
<wingrime>
oliv3r: only raw bytes
<Dreadlish>
wait
<Dreadlish>
not
<Dreadlish>
it doesnt work
<oliv3r>
wingrime: what do you mean?
<wingrime>
oliv3r: usualy i2c have registers
<wingrime>
oliv3r: usualy i2c with read/write
<wingrime>
oliv3r: you send commands for read / write
<wingrime>
oliv3r: this only return raw data after request
ibrah has joined #linux-sunxi
<mnemoc>
indeed
<oliv3r>
wingrime: do you think you can work with it? or is it almost impossilbe?
<mnemoc>
err, wrong scroll
<oliv3r>
wingrime: I guess a firmware has to be re-written?
<oliv3r>
mnemoc: patch works
<wingrime>
oliv3r: It realy not simple
<wingrime>
oliv3r: yep it wotk (for me now) but buggy
<Dreadlish>
wingrime: i put old fex and its still the same
<Dreadlish>
what i changed? i just reinstalled rootfs
<Dreadlish>
rest is the same
<mnemoc>
oliv3r: great. pushed
<wingrime>
oli3r: it funny , I don't even need for ask to data
<wingrime>
oli3r: it funny , after irq it will send data without any request
ibrah has quit [Ping timeout: 256 seconds]
<Dreadlish>
funny.
<wingrime>
oli3r: and defenetly funny that IC will send unknown byte count
<n01>
Is the guy working on MTD here in chan?
<shineworld>
Hi all, I a question: why MAC change every time I power on the board ?
<Dreadlish>
had you set it in .fex?
<oliv3r>
wingrime: sounds like that chip will take a lot to support :(
<shineworld>
me ? but MAC should be a registered part of component ....
<Dreadlish>
shineworld: should, but isn't
<oliv3r>
shineworld: i think you can set/override it in the fex
<Dreadlish>
shineworld: its in fex.
<shineworld>
so I've to buy MAC values from corporate ?
<oliv3r>
so if you don't 'set it in the fex, it'll be random memory values :)
<shineworld>
I can't sell objects in world with "random" mac... I don't thing is legal
<oliv3r>
shineworld: one very good question, if I download the latest firmware for my tablet, a livesuit image, it can't have my unique mac
<shineworld>
*think
<oliv3r>
so it shomehow works properly somewhere
<oliv3r>
i'll check what my tablet does irght now
<oliv3r>
you should check what is in your fex
<shineworld>
ok, so I can set a valid MAC in script.bin (fex) file
<wingrime>
oliv3r; now I know how make it stable work
<shineworld>
just to know ;) normally a CPU manufacturer buy a stock of MAC for its factory and put in its chips but perhaps this isn't our case ;)
<shineworld>
because is a penal case to use random MAC
<wingrime>
shineworl: everey a10 have unical number
<wingrime>
can be readen from reg
<shineworld>
ah ok, so mac can "linked" to it to avoid conflicts but there is a part of MAC related to manufacturer ... uhm I will search for some dev evb (for example could be first part of mac used by vmware interfaces)
<shineworld>
thanks for suggestion of cpu reg
<shineworld>
just to avoid to put manually any number
<mnemoc>
A10-based devices with ethernet usually get the mac address set via u-boot env
<mnemoc>
that's why the mele have uart pins
<shineworld>
ok !!! thanks
<mnemoc>
doing so via script.bin doesn't scale, but you'll need a custom image per unit. but u-boot's env is usually left untouched
<mnemoc>
s/but/because/
rellla has joined #linux-sunxi
<techn__>
anyone u-boot guru? :)
<techn__>
I have situation where u-boot seems to load correctly but somewhy it wont find "boota".. I'm suspecting that it can't load "env" partition, which supposed to have u-boot helpers
<techn__>
so how I can load it manually?
<oliv3r>
ah, so i have to keep my u-boot enviroment safe, as that stores my mac, so how odes livesuit deal with that?
<mnemoc>
oliv3r: when making the livesuit image you tell what partitions to touch
<oliv3r>
and they leave the uboot; ok i'll see if I can back that up :)
<oliv3r>
when I build with V=99; i get arm-pc-linux-gnueabi-ld: drivers/gpu/mali/mali/linux/mali_osk_atomics.o: invalid string offset 4293984260 >= 213 for section `.strtab'
<oliv3r>
what causes that?
<mnemoc>
no ida, but re-running make solves it
<hramrach>
wingrime: still panics
<mnemoc>
i also don't know what mali insists in rebuilding itself every time
<oliv3r>
ok, strange
<oliv3r>
it did
<hramrach>
that patch looks like it adds bits that might be usful but the issue that makes it panic is not fixed
<oliv3r>
but without V=99
<hramrach>
Isuspect there is sime state left in the otg driver which should be cleared on gadget unlaod but is not
<hramrach>
it crashes in some list function dereferencing null
<techn__>
mnemoc: does some of lichee-dev branches support boota and a13?
<techn__>
and nand
<wingrime>
hramrach: when ?
<wingrime>
hramrach: strange realy
<wingrime>
hramrach: are you realy aplied patch
<wingrime>
hramrach: I defenetly reload driver 5 times without any probles
<mnemoc>
techn__: no idea what "boota" is
<mnemoc>
techn__: but we have lichee/a13-dev in my github
<techn__>
mnemoc: command which reads android bootimg format :/
<hramrach>
wingrime: it crashes in connecting cable
<mnemoc>
techn__: ah, so you mean u-boot
<hramrach>
did not try just realoading
<techn__>
mnemoc: yes
<wingrime>
hramrach: I don't saw
<mnemoc>
techn__: iirc lichee-dev has sun5i support
<wingrime>
hramrach: I you use host and gadget at same time ?
<techn__>
great
<wingrime>
hramrach: I can't reproduce
<wingrime>
hramrach: You have find fix own-way
<wingrime>
hramrach: whait....
<wingrime>
hramrach: are you made insmod ?
<wingrime>
before cable connection
plan_b has quit [Quit: plan_b]
<mripard>
Turl: yes, we can
<mripard>
send him a mail
<mripard>
I'll try to find him somewhere on IRC tomorrow
<oliv3r>
mnemoc: how do make the kernel compile verbose? ( i need to pass -DDEBUG to my module)
<hramrach>
wingrime: I set up g_ether for the first time, ping from the other machine. remove module, insert module, remove cable, insert cable
<hramrach>
I still get the messages about taken endpoints
<hramrach>
well, and it says git is up to date and copied the kernel once more to make sure
<hramrach>
this is not modular code so that should make it
<hramrach>
I tried looking at the gadget system and don't have any idea how it fits together
<hramrach>
would have to look at it when there is more time
<wingrime>
hramrach: now I finishing TS driver
<hramrach>
oh, and I don't have any idea if I use host and gadget at teh same time
<hramrach>
I have mouse connected to the other usb port but that's supposed to be another controller completely
<hramrach>
I have OTG configured but only connect A10 <-> pc
<wingrime>
hramrach test more
<wingrime>
hramrac:g_mulit
<wingrime>
and "android"
<techn__>
does lichee-dev support UART on sdcard slot?
<hramrach>
where is 'android' ?
<wingrime>
hramrach: andorid can be used when gadges is static non module
<wingrime>
hramrach: set any gadget to statuc
<wingrime>
hramrach: and you will see selector
<techn__>
lichee/lichee-dev-ICS :/
<hramrach>
that will work, presumably
<techn__>
should I use that with cwm10?
<hramrach>
since it is the unload code that is broken
n01 has quit [Quit: leaving]
mdfe has joined #linux-sunxi
<hramrach>
g_multi cannot be inserted
<mnemoc>
techn__: in u-boot-sunxi's lichee-dev there are two sun4i boards and two sun5i boards. one for mmc as mmc, another for mmc as uart/jtag
<hramrach>
no such device
<mnemoc>
techn__: but i've never used any of this myself. it's just... theory
ibrah has joined #linux-sunxi
<techn__>
mnemoc: I tried sun5i_sdcard earlier.. but it freezed on followin row "MMC: ÿ"
<mnemoc>
ow
<wingrime>
hramrach: my last patch actualy fix for reload
<wingrime>
hramrach: my last patch actualy fix for "rmmod"
<techn__>
mnemoc: but it seems that none of default sun5i u-boots wont support boota-method
<hramrach>
well, rmmod never crashed for me
<hramrach>
but reconnecting the cable did and still does
<techn__>
so lichee-dev seems to be the way.. lichee/lichee-dev-ICS wont support sun5i :/
<hramrach>
reconnecting cable after rmmod/modprobe
<wingrime>
hramrach: last patch should fix endpoint leak,
<wingrime>
check this
<hramrach>
the one that just wen in 3.4?
<hramrach>
went
<wingrime>
only for 3.4
<hramrach>
yes
<wingrime>
and already in stage
<hramrach>
I pulled from stage
<hramrach>
endpoints still leak
<wingrime>
strange... i rmmod/insmod many times without problems
<hramrach>
the patch does not printk anything special so can't tell from the log but should be applied
<wingrime>
hramrach: also, are you tested cable on 3.0 ?
<hramrach>
only 3.4
<wingrime>
it looks for me 3.4 host work bad as on 3.0
wingrime has quit [Ping timeout: 264 seconds]
ibrah has quit [Ping timeout: 240 seconds]
ibrah has joined #linux-sunxi
<hramrach>
well, the in-built g_ether works
<hramrach>
and you cannot unload so does not crash
<hramrach>
maybe netconsole time?
shineworld has quit [Remote host closed the connection]
<oliv3r>
frack; i can enable and see the debug strings from pr_debug in debufs/s control file, but nothing gets printed in my dmesg. (I enabled it with module <modulename> +p) where does debug output go?
<oliv3r>
cause my wifi doesn't even work anymore with my self-built kernel :)
<oliv3r>
maybe i should swap the entire kernel and all modules :)
<oliv3r>
oh probably should check my script.bin too :S
<oliv3r>
I guess I should start with making one specifically for my tablet :)
mdfe has quit [Remote host closed the connection]
ibrah has quit [Ping timeout: 256 seconds]
paulk-desktop has quit [Quit: Ex-Chat]
chujalt has joined #linux-sunxi
<chujalt>
Hola, saludos
<hramrach>
hmmm, so netconsole odes not work
<hramrach>
it oes not even bring up the gadget interface
<hramrach>
and endpoints are already taken even with built-in gadget so that particular thing is probably a red herring
<hramrach>
where do you ask about netconsole not working :s