<mnemoc>
"NOD" claims it was trivial to find it googling.... but my G fu seems to be in a very different league of his
<oliv3r>
hehe
<oliv3r>
well you'd cover your tracks
<oliv3r>
we need some form of opensource wikileaks :p
<oliv3r>
wikileaks for datasheets and manuals :)
<mnemoc>
dataleaks.org :p
<oliv3r>
(haven't opened it yet) but it's not the same that is currently on the wiki for the A10, right?
<oliv3r>
hehe, taken :( 'sheetleaks.org'
<mnemoc>
that's the "public" ~80p datasheet
<oliv3r>
kk, just checking
<mnemoc>
this one is ~400p
<oliv3r>
sheetleaks.org does sound very ... diareah like though
<mnemoc>
:D
<oliv3r>
doesn't seem to be taken though :)
<cheng>
thanks mnemoc for the manual. This will come handy one day.
<oliv3r>
With all the heat assange et al have been getting over sites like wikileaks though; I wonder what chances would be of getting in trouble running 'sheetleaks.org'
<oliv3r>
oh even hipboi commented it being authentic :D
<oliv3r>
'when i WAS at allwinner' I guess he's no longer?
<mnemoc>
oliv3r: he left a week ago
<mnemoc>
oliv3r: "announced" it on u-boot's ML
merbanan has joined #arm-netbook
<oliv3r>
I need to be on more ML's
<oliv3r>
he should have leaked some code >: )
<oliv3r>
codeleak.orrg
<oliv3r>
:D
<mnemoc>
problem of code leaks is that they destroy any RE effort
<mnemoc>
legally speaking
<oliv3r>
If I could sustain my spending habbits running a site like those, i would. some risk with future employement though :(
<cde>
well leaked code is useful but dangerous, ideally you want a clean room process
<oliv3r>
code -> 'datasheet' -> cleancode :D
<mnemoc>
cde: would this leaked "user manual" harm a clean room process?
<cde>
it depends on your country. general ideas and algorithms can't be copyrighted (at least in EU) so unless the datasheet contains the actual implementation you should be safe
<cde>
"For the avoidance of doubt, it has to be made clear that only the expression of a computer program is protected and that ideas and principles which underlie any element of a program, including those which underlie its interfaces, are not protected by copyright under this Directive."
<mnemoc>
\o/
<RaYmAn>
so cleanroom is ok I guess - you could probably still get sued for using leaked datasheets, but code is fine? :)
<cde>
and the Directive overrides any EULA, which is nice. basically you have the right to RE as long as you don't publish the disassembled/decompiled code
<mnemoc>
neat
<cde>
well you can only get sued if you were under NDA, which is not the case here
<cde>
since the datasheet is now public, anyone not under NDA can use it without fear of prosecution
<jinzo>
(at least in EU)
<oliv3r>
so what about devs writing code in the EU from these sheets
<oliv3r>
and then using the code, in let ssay, the USA
<RaYmAn>
cde: isn't it akin to book piracy essentially?
<cde>
I don't know. I'm not a lawyer :)
<RaYmAn>
I mean, allwinner still has copyright on the datasheets even if it's leaked
<jinzo>
probably nothing if you just use it.
<RaYmAn>
(the actual text, not the ideas)
<jinzo>
there's an interesting story flowing around on the first versions of PGP - that they couldn't legally export it from the USA (being cryptography) they printed the code into books and exported that and then scanned it back in
<oliv3r>
so allwinner could use copyright of the document to get it 'unpublished'
<oliv3r>
but you know how that goes, once its out there, it's out
<cde>
I guess so
<oliv3r>
and as said, one could always wikify the content into a 'cleanroom datasheet' like thing
<mnemoc>
oliv3r: linux-sunxi.org is waiting :)
<oliv3r>
haha
<oliv3r>
so much to do, so little time :(
<mnemoc>
+1
<oliv3r>
if someone would hire me fulltime to do this kind of work, i'd love to
<mnemoc>
me too
<cat1>
+1 :)
<jinzo>
hm, looks like A10 has some hardware crypto functions - that's nice.
<lundman>
rot13 AND xor
merbanan has quit [Ping timeout: 246 seconds]
<oliv3r>
I wonder how the A15 will look like. Simple drag/drop swap of the Arm core, some small fixes, or a complete revamp of the entire SoC (invalidating this datasheet completly)
<mnemoc>
we have the small datasheet of sun3i, sun4i and sun5i
<mnemoc>
almost everything outside the ARM core remains identical
<oliv3r>
so this sheet will be still very interesting when sun5i is released
<mnemoc>
same disp, same nand, same mmc, same ethernet, same sata, ...
<mnemoc>
sun5i = A12/A13, already in use broadly
<mnemoc>
sun6i = next thing
<oliv3r>
same ethernet, aww; i hope they fix their mac issues
<oliv3r>
ah, i ment sun6i sorry
<oliv3r>
but makes perfect sense of course, least amount of work for a new more powerfull soc
<jinzo>
and another good reason why they could do more on open source/mainlining
<mnemoc>
they seem to not see any value on letting people outside their walls to do any system software on their platform
<akaizen>
Regarding the RE: you can have someone disassemble / look @ code and draw up a spec sheet
<akaizen>
Hand that off to another team which implements it in a clean room
<mnemoc>
akaizen: my wondering was about the impact of having leaked datasheets
<akaizen>
As long as the two groups dont communicate can there is no breach.. IANAL but thats what I know of some companies / groups doing for certain projects (one off the top of my head is Broadcom GPU implementation for RaspberriPi)
<mnemoc>
can those who implement the "clean room" code read the datasheet?
<mnemoc>
can they prove they didn't?
<akaizen>
That im not sure about
<akaizen>
Because on the hardware side it implementation specific (per the chip)
<akaizen>
If you could make an argument that the specific registers and addresses / what not are similar to function names
<mnemoc>
when you RE a closed chip, it's always "implementation specific"...
<akaizen>
then you could have whoever draws up the spec sheet put in the address / info from data sheet
<akaizen>
then hand that off to clean room team
<akaizen>
From the Oracle JAVA case
<akaizen>
they said that the function names are essentialy to make it binary compatible but did not consistiute copyright since it was necessary but not ciritical to the purpose of the function
<akaizen>
like i said if you could make that argument for the chip: i.e. instructions are necssary but trivial and not tied to implementation
<mnemoc>
in our case it's about registers, not functions
<jinzo>
yes that's all dandy and cool, but this project lacks people/coders as it is...
<akaizen>
Yes, so make that analogy
<akaizen>
I think it could be done
<akaizen>
Well im a systems engineer... had some time to work on this and was working on re implementiong libve into a proper driver
<mnemoc>
the documentation about the "mixing processor" is specially nice documented
<RaYmAn>
must be license from someone else
<RaYmAn>
:P
<mnemoc>
:D
<jinzo>
:D
<akaizen>
just got into it with my mk802, so not sure where the most needed parts are.. there is no organization to this project...
<akaizen>
well not that is apparent to me :)
<akaizen>
Seems spread across a couple wikis, mailing lists and sporadic irc logs
<jinzo>
akaizen, indeed.
<jinzo>
but that's quite usual and it's improving
<akaizen>
I'd be happy to setup a mailing list and a wiki where we can all collaborate, shouldnt take more than an hour
<jinzo>
the rhombus-tech wiki and linux-sunxi.org are the way to go imo.
<akaizen>
ahh cool, didnt know about the google gorup
<jinzo>
ofcourse - but creating a new one won't help anyone
<akaizen>
Well, I was thinknig make a new one put in all the info from various sources and in the process clean it up
<mnemoc>
imnsho RT's wiki should be about RT goals, i.e. EOMA68 cards and eoma68 i/o boards
<akaizen>
basically a git pull and merge :)
<mnemoc>
not about other A10-based devices
<mnemoc>
or generic A10 stuff
<jinzo>
mnemoc, indeed - but it unfortunately hosts quite some other info too.
<mnemoc>
jinzo: indeed
<akaizen>
there are like 4 posts on the linux-sunxi mailing list...
<mnemoc>
akaizen: yes, it's very new
<jinzo>
akaizen, the arm-netbook one in the topic is quite active - but again mixed bag of cats
<akaizen>
yup that one is good :)
<mnemoc>
yes, that's mostly what we (ab)use.
<akaizen>
thats where i found out about empat zero libve and that the libcedar and included files work
<akaizen>
but some specific codes init_data is missing which is neede to decode them
<mnemoc>
this channel wasn't supposed to be about A10 either...
<akaizen>
like MPEG4 that empat_zero mentioned was unable to get working
<akaizen>
Do you guys have a status as to what is done, being worked on / needed?
<akaizen>
i can clean up the wiki before i go to sleep
<akaizen>
Is the wiki not open for editing?
<mnemoc>
akaizen: register-only
<mnemoc>
akaizen: too many spam boots and lack of people to keep them under control
<akaizen>
gotcha
<akaizen>
So what parts are needed / cirtical? From my understanding: We got uboot, need some kernel stuff, working on VPU, GPU
<akaizen>
I'm approaching from xbmc / media center path so im definitely missing some other stuff
<mnemoc>
vpu libs are "debian/ubuntu armel" only
<mnemoc>
i.e armv4
<akaizen>
yea serious RE on libve.so to get hard float
<mnemoc>
GPU wise we have "ubuntu armhf" libs for r2p4,r3p0 and r2p1
<akaizen>
according to the wiki some is 'trivial'
<mnemoc>
akaizen: according tot he ffmpeg hacker who added that to the wiki
<mnemoc>
trivial is relative ;-)
<akaizen>
hehe yea
<mnemoc>
my cubieboard arrived \o/
<akaizen>
nice
<akaizen>
so many nice arm boards
<jinzo>
great :D
<akaizen>
the rk3306 (ug802) looks nice too
<akaizen>
damn these companies, cant they abide by GPL
<akaizen>
i want a ARM FARM
<cheng>
so soon hardware engineer will be out of job
<jinzo>
bah - rockchip isn't exactly known for it's quality/power
<RaYmAn>
mnemoc: pics or it didn't happen
<RaYmAn>
:P
<mnemoc>
RaYmAn: :D
<oliv3r>
cheng: why is that? Hardware engineers design beautiful hardware, write proper specs and let the software community do the rest. They will of course need help there :)
<oliv3r>
hardware engineers are not supposed to write software anyway, they tend to not be able to do this, properly
<cheng>
with so many board coming out, soon people just buy these board and develop on top of that.
<akaizen>
Rockchip = bad? rk3306 is dual-core a9
<akaizen>
dunno about their VPU but has quad core mali400
<oliv3r>
i think rockchip is one of the bigger GPL violators, no
<akaizen>
supposedly lower heat under load too
<jinzo>
akaizen, maybe they improved - but in the past their offerings were utter shit :)
<oliv3r>
cheng: well with regards to hardware engineers, (and not chip designers), if the for example allwinner had mainlined everything from before the start, then all the engineer has to do, design a board around the A10 do some minor software setup (pin mappings etc) and done.
<akaizen>
jinzo: hehe, lets hope for the better. but i do hear they are one of the worst gpl violators so i will boycott for now
<akaizen>
cubie board looks awesome tho
<jinzo>
yeah - and incredibly cheap
<cheng>
how about just work on a peripheral board and jump (wires) from these boards, skip the design part around a10
<oliv3r>
I kinda want one now, just to play with
<oliv3r>
:p
rvalles has quit [Ping timeout: 260 seconds]
rvalles has joined #arm-netbook
cheng has left #arm-netbook ["Leaving"]
ZeroJ has quit [Ping timeout: 246 seconds]
merbanan has joined #arm-netbook
<akaizen>
Cleaned up the main page a little bit
<mnemoc>
akaizen: don't be shy, go on :)
<akaizen>
hehe
<cat1>
mnemoc: do we reserve fb block here arch/arm/mach-sun4i/core.c?
<mnemoc>
cat1: i would prefer to find how to read the script.bin at that stage
<VarmVaffel>
what happens if you try to write to a register than can only be read?
<cat1>
mnemoc: what prevents us from doing this reservation later, let say from here drivers/video/sunxi/disp/dev_fb.c?
<VarmVaffel>
will the CPU restart, or an interrupt triggered or something?
<mnemoc>
cat1: in previous experiments it didn't work.... but please, try
<cat1>
mnemoc: ok, will try. how would script.bin help?
<mnemoc>
cat1: script.bin tells you if the fb will be enabled by the driver or not
<mnemoc>
cat1: which is evil, but that's how life is at this moment
itamarjp has joined #arm-netbook
<mnemoc>
cat1: so we try to only reserve it only if it will be used
<mnemoc>
cat1: which halts 3.4+ :<
<cat1>
mnemoc: right, but it is driver's business to know.
<cat1>
mnemoc: i reworked 3.6 and halts one too :|
<mnemoc>
cat1: problem is we still need a fixed address
<mnemoc>
cat1: as it's hardcoded in the mali userspace driver
<cat1>
mnemoc: ah, and it might not be available when modprobing disp and co..
<cat1>
mnemoc: tricky..
<mnemoc>
exactly
<akaizen>
Cant we use mmap syscall with MAP_FIXED to reserver adter the fact?
<akaizen>
that way kernel wont use it for something else if uboot is allocating it and not telling kernel
<akaizen>
*** his book that he recommends is good about that stuff
<akaizen>
tl;dr check out the comic on his page
<akaizen>
Alright guys, great talking to you. I'm going to add some stuff from the rhombus-tech page to the wiki tomorrow and clean it up
<akaizen>
good night
<cat1>
akaizen: gn!
<cat1>
mnemoc: we do not want to use kernel command line to pass information related to g2d and fb reservation, right? at least it would be less painful workaround.
<mnemoc>
cat1: linux-sunxi@googlegroups.com :)
<cat1>
mnemoc: you mean i should start discussion there ;)
Mazon has quit [Read error: Connection reset by peer]
Mazon has joined #arm-netbook
<oliv3r>
good afternoon
<mnemoc>
arete74: wow, who was the lucky?
jquip has joined #arm-netbook
<VarmVaffel>
what would make an ARM to go into Thumb mode?
<RaYmAn>
blx? :P
<RaYmAn>
or bx
<VarmVaffel>
oh ok
<RaYmAn>
(with suitable flags)
<VarmVaffel>
is that the only reason? :P
<RaYmAn>
your question is rather...unspecific
<VarmVaffel>
I bet it is
<VarmVaffel>
I have no clue what I'm doing
<RaYmAn>
compilers will do it automatically if you tell it your code should be thumb
<RaYmAn>
the processor shouldn't switch to/from thumb without being explicitly told to
<VarmVaffel>
right
<VarmVaffel>
I just had a problem with this, I compiled the code like normal
<VarmVaffel>
downloaded it to the CPU
<VarmVaffel>
and then I got an "Invalid instruction" error
<VarmVaffel>
did a bit of googling and found out it had been switched into Thumb mode for some reason
<VarmVaffel>
I didn't actually do anything to make it switch, so I'm just wondering what might cause it to suddenly do that
<jquip>
Hallo there! u-boot load from the mmc card is becoming painful... :( why do I have to switch on the device 3-4 times to get linux on? :/ will chained bootloader work the first-time I power on the device?
<jquip>
srry for butting in like that ;)
<VarmVaffel>
This is IRC, it's basicly designed for butting in :P
<jquip>
haha :) yeah..
<jquip>
this bootlader is driving me mad.... it takes *indefinite number of tries to get the a10 machine on...
<lundman>
pull power twice here
<jquip>
mostly in the range of 3-4 times... and I have to press the power on button until the screen shows some backlight, and then I know its gonna boot..
<VarmVaffel>
if you stick around in here someone might answer
<RaYmAn>
jquip: have you tried with different cards?
<RaYmAn>
It's the bootrom that handles reading u-boot off SD, so it might well be less compatible than e.g. linux. (though, it could well be something like card detect misbehaving physically)
<jquip>
card detect aye?? enlighten me further dear sir..
<RaYmAn>
we've seen some devices where the micrsd card detect functionality was..unstable
<RaYmAn>
so it would only detect a card inserted sometimes
<RaYmAn>
and hence only boot from card sometimes
<RaYmAn>
depending on OS setup, it might well still show up if e.g. it uses another method for card detection
<jquip>
aah... "some devices" implies a particular model? as in "all the devices of that model"??
<RaYmAn>
there are so many different a10 tablets/devices/etc out there
<RaYmAn>
most of them not with the best build quality ;)
<RaYmAn>
so I'm suggesting a hardware error on your specific tablet, not necessarily the entire range of devices
jquip has quit [Ping timeout: 245 seconds]
cheng has joined #arm-netbook
Yakuzzi has joined #arm-netbook
jlj has joined #arm-netbook
Yakuzzi has quit [Remote host closed the connection]
Yakuzzi has joined #arm-netbook
rvalles has quit [Remote host closed the connection]
rvalles has joined #arm-netbook
<oliv3r>
if someone's bored ... I just 'filled in' linux-sunxi.org/A10/registers
<oliv3r>
if someone could do a quick proof-read-check review?
<oliv3r>
:)
<mnemoc>
oliv3r: isn't it a bit too verbose to do A10/Foo Controller instead of just A10/Foo ?
<oliv3r>
Well there is Foo Controller and Bar Controller :p
<mnemoc>
in that context by Foo we can only be refereing to the Foo controller
<mnemoc>
and by Bar, the Bar controller
<oliv3r>
unless you mean having a Foo#Controller and a Bar#Controller
<mnemoc>
what might be in Foo which isn't about the Foo Controller?
<oliv3r>
I dind't interpret the document, only wrote it down :p
<oliv3r>
anyway how much different IS a dram controller from an sram controller
<mnemoc>
dive in and find out ;-)
<oliv3r>
haha
<oliv3r>
lack of time says no
<oliv3r>
i have to go home now :)
<oliv3r>
Feel free to improve my work :D
<oliv3r>
(i notice i forgot to add some space between foo and the number
<mnemoc>
oliv3r: you took it over from me, now it's all yours ;-)
neal__ has joined #arm-netbook
drachensun has joined #arm-netbook
drachensun has quit [Client Quit]
neal__ has quit [Client Quit]
drachensun has joined #arm-netbook
ssspiff has quit [Remote host closed the connection]
cheng has left #arm-netbook ["Leaving"]
jlj has quit [Ping timeout: 240 seconds]
jlj has joined #arm-netbook
rellla has quit [Quit: rellla]
eflatun has joined #arm-netbook
rellla has joined #arm-netbook
rellla has quit [Client Quit]
jlj has quit [Ping timeout: 246 seconds]
<techn>
RaYmAn: jquip's problem could be that power regulator problem
<techn>
atleast I'm wondering that is the problem with my mid_a10
<techn>
but I cannot verify that without serial cable
<techn>
btw. has anyone tried that buf2 patch with r3p0/1?
ceo16 has quit [Ping timeout: 240 seconds]
rellla has joined #arm-netbook
ceo16 has joined #arm-netbook
eflatun has quit [Ping timeout: 246 seconds]
ssspiff has joined #arm-netbook
ssspiff has quit [Changing host]
ssspiff has joined #arm-netbook
<RaYmAn>
techn: also a possibility :) there's plenty of potential issues to choose from :P
<techn>
yep :)
<techn>
.. I'm going to try now buf2 with r3p0.. thumbs up ;)
<RaYmAn>
cool
<RaYmAn>
hope it works
<RaYmAn>
did you attempt to get egl working with fbdev and r3p0 btw? or anyone else?
<techn>
RaYmAn: I got it working with r2p4.. but partially
<RaYmAn>
on x11?
<techn>
yep
gimli has joined #arm-netbook
<techn>
no luck with r3p0 :(
rellla1 has joined #arm-netbook
robws has quit [Ping timeout: 244 seconds]
robws has joined #arm-netbook
<techn>
forgot r3p0 x11 drivers
<mnemoc>
doh
rellla1 has quit [Quit: rellla1]
j1nx_ has joined #arm-netbook
eflatun has joined #arm-netbook
<rm>
nice
<rm>
at least this new MK802 booted from the get go
<rm>
mnemoc, # CONFIG_SUNXI_IGNORE_ATAG_MEM is not set
<rm>
and I do not see the full 1GB
<rm>
should I try "=y" here?
<mnemoc>
rm: from where are you booting?
<mnemoc>
not sure what's the state of hno's u-boot with regard to 1GB boards
<rm>
what do you mean from where :S
<mnemoc>
CONFIG_SUNXI_IGNORE_ATAG_MEM falls back to allwinner's hack to read memory size
<mnemoc>
uSD or NAND
<rm>
SD
<rm>
what's "ATAG MEM"?
gimli_ has joined #arm-netbook
<mnemoc>
how bootloader passes the memory info to linux
<mnemoc>
you can also try passing two mem=s
<rm>
okay, so I get it
<rm>
CONFIG_SUNXI_IGNORE_ATAG_MEM=y means to ignore what bootloader (u-boot) does, and try the Allwinner hack
<mnemoc>
yes
<rm>
thanks
<mnemoc>
i should really rename that CONFIG_...
<rm>
also interestingly this mk802 advertised as 512M on the product page
<rm>
but the actual device has that DDR3/1G sticker on it
gimli has quit [Ping timeout: 246 seconds]
<jelly-home>
ordered a $56 mk802 device, wonder what they left out to make it cheaper
eflatun_ has joined #arm-netbook
<Yakuzzi>
they have to compete with the ug802 so they have to be cheaper to still sell
<rm>
did not check what android says. Hmm.
<rm>
jelly-home, they are sub-$50 now
gimli_ has quit [Ping timeout: 246 seconds]
<rm>
the other day I found an A13 tablet for $50
<rm>
if you leave out the capacitive touchscreen, battery, case, etc, etc
<rm>
naturally an MK802's BOM is perhaps $15
<rm>
if that
eflatun has quit [Ping timeout: 268 seconds]
<jelly-home>
I don't want an A13 yet. Trying to evaluate and test as a replacement STB
<jelly-home>
nod
eflatun_ has quit [Remote host closed the connection]
eflatun_ has joined #arm-netbook
<Yakuzzi>
did anyone see this somewhere else then alibaba yet ?