<hno>
z4k4ri4, many USB cables have quite poor cable quality making them bad as power cables.
<hno>
too much resistance making the voltage drop too much
<hno>
mnemoc, the shame is on the Nokia board of directors that hired him.
<hno>
and the major shareholders who accepted it.
<mnemoc>
yes. it's primarily their fault
rellla has joined #linux-sunxi
<hno>
but it's quite worrying from an industry perspective that MS now get Nokia as a weapon.
<z4k4ri4>
Yes so many bad USB cable and USB power supply, I'm thinking just buy PC PSU and some converter for Cubie power port and micro USB for my raspi.
<z4k4ri4>
Anybody knows is there some PC power jack to USB converter?
<tkoskine>
hno: But Microsoft didn't get the ownership of patents from Nokia, they only licensed those.
<oliv3r>
z4k4ri4: if your cable is to long, the voltage will start to drop too much and it won't be enough power
<z4k4ri4>
My cable is just 1M+
<oliv3r>
z4k4ri4: as hno said, cheap USB cables tend to be crap for power :)
<oliv3r>
z4k4ri4: and length seriously matters in that regard
<oliv3r>
tkoskine: now they own those patents
<oliv3r>
and got a cheap phone company
<oliv3r>
nomatter how crap surface and WP are; now they can manufacture then cheap and evelop them in house as MS hardware
<z4k4ri4>
Well this is not some cheap USB cable. I bought USB cable with "gold" connector
<oliv3r>
i can sprinkle fairy dust on it, it won't make the gauge and internal resistance of the wire better ;)
<oliv3r>
z4k4ri4: also gold isn't all that usefull in digital transmission
<oliv3r>
it was very usefull for analog of course
<z4k4ri4>
Yeah, I know :)
<tkoskine>
oliv3r: No they don't. Microsoft bought only Nokia mobile phone business excluding the patents.
<oliv3r>
i have USB harddisks, that work fine on a single cable when connected directly to the PC; but fail to work even with double connections when i use them with 1m cables
<tkoskine>
(At least that is what Finnish newspapers say.)
<oliv3r>
tkoskine: you are right
<oliv3r>
tkoskine: Microsoft will pay EUR 3.79 billion to purchase substantially all of Nokia's Devices & Services business, and EUR 1.65 billion to license Nokia's patents, for a total transaction price of EUR 5.44 billion in cash.
<arokux1>
panda84kde, it was just a guess, I don't know what the problem really is, but is solution requires manual intervention to confirm that you're are a "nice" person
<panda84kde>
arokux1: thanks. Should I ping someone?
<arokux1>
I pinged mnemoc already
<z4k4ri4>
Guys, what is the good microSD card for cubieboard?
<rah>
wuh?
<rah>
MS bought Nokia?
shineworld has joined #linux-sunxi
<oliv3r>
rah: gifted rather
<panda84kde>
rah: wasn't that already clear when Elop joined Nokia?
<oliv3r>
z4k4ri4: many :p whaat specifically you after?
<rah>
panda84kde: err.. no
<z4k4ri4>
Just some for installing ubuntu or other linux on my cubie
<z4k4ri4>
some good
<z4k4ri4>
Is sandisk good?
<oliv3r>
as long as it's not a fake it'll be fine
<oliv3r>
i use old 512 MB sdcards :p
<oliv3r>
the mmc controller isn't super fast anyway
<rah>
"Today’s agreement will accelerate the momentum of Nokia’s devices and services"
<rah>
lol
<rah>
it's true though
<oliv3r>
downward?
<rah>
it'll accelerate their momentum to the grave
boycottg00gle has joined #linux-sunxi
CoSM0 has quit []
popolon has joined #linux-sunxi
kaspter has quit [Ping timeout: 246 seconds]
<hramrach>
they used windows already, anyway
arokux has joined #linux-sunxi
<shineworld>
In past Nokia was best phone hardware maker but now ... there are so much valid competitors ...
arokux1 has quit [Ping timeout: 256 seconds]
<hramrach>
I guess they gave up on trying so it was good deal to sell off thje phone division
<eagles0513875>
hey all :)
<eagles0513875>
how is everyone
<hramrach>
They used to be good before phones became 'smart'
<hramrach>
then they did the wrong choice of 'smart' os. More than once. dead
mouchon has quit [Read error: Connection reset by peer]
<jelly-home>
that's what killed them, MS just came in to poke on the rotting remains
shineworld has quit [Remote host closed the connection]
<hramrach>
so MS will accealerate their phones one way o another but they don't really care anymore. and MS gets a platform of their own to play with Win8. Not much success pushing Windows on third party phines so far
shineworld has joined #linux-sunxi
<hramrach>
phones
z4k4ri4 has left #linux-sunxi [#linux-sunxi]
* hramrach
cherishes his 'dumb' phone
<hramrach>
so reliable! so long battery life!
mouchon has joined #linux-sunxi
<shineworld>
I'm on N4 ... battery life is my pain
cajg has joined #linux-sunxi
rellla2 has joined #linux-sunxi
<mouchon>
hi
<mouchon>
simple noob question, i would like to change the i2c-sunxi to give busses frequency as param to the module. What's the best option ? add entry in flex or pass it as module argument ?
arokux1 has joined #linux-sunxi
tinti has joined #linux-sunxi
arokux has quit [Ping timeout: 260 seconds]
<hramrach>
mouchon: if the frequency is basically fixed based on hte device connected it's probably good idea to put in fex
<mouchon>
hramrach: The a20 olimex uext1 and uext2 connector have i2c-1 and 2 wired and you can put any i2c device you want, so you may have different frequency
ykchavan has quit [Quit: Leaving]
sud0x3 has joined #linux-sunxi
<hramrach>
mouchon: are uexts hotpluggable?
<hramrach>
if not then you have only one device connected at any time and fex makes more sense.
hipboi has quit [Quit: Leaving]
<wigyori>
morning
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<oliv3r>
lo
<andoma>
hi
<mouchon>
hramrach: yes uexts for me are hotpluggable
<mouchon>
may be other opiton is to let it be configurable via ioctl
rellla2 has quit [Remote host closed the connection]
<Turl>
panda84kde: what's your wiki user?
hansg has joined #linux-sunxi
<panda84kde>
Turl: panda84
<Turl>
panda84kde: you should be able to spam now :)
<oliv3r>
rellla: wow, how many weeks did you have it? :)
<rellla>
oliv3r: around 2 or 3;)
<oliv3r>
mripard: did you see see my paste last night?
<mripard>
hmmm, no :)
<rellla>
oliv3r: i'm diff'ing sun7i and sun4i kernel module atm...
<oliv3r>
mripard: a20 prints a crashdump; same kernel works fine on a10
<oliv3r>
rellla: wingrime is doing unification etc so ping him to get the very best
<rellla>
ok
shineworld has quit [Remote host closed the connection]
shineworld has joined #linux-sunxi
\\Mr_C\\ has quit []
<oliv3r>
mripard: yay! I got an ack!
Ehsand has quit [Remote host closed the connection]
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<shineworld>
very dummy question about uEnv.txt: but if in kernel Load & Entry point address are 0x40008000 I need to use same value on boot_mmc=fatload mmc 0 0x43000000 ${fexfile}; fatload mmc 0 0x48000000 ${kernel}; bootm 0x48000000 or leave these ?
<oliv3r>
leave that as you have
<oliv3r>
that's good
\\Mr_C\\ has joined #linux-sunxi
<shineworld>
I'm upset... why my kernel freeze at first line...
<shineworld>
I'm using the kernel 3.3.0 out from linchee build of olimax sdk
<shineworld>
without any changes
vinifr has joined #linux-sunxi
<mripard>
oliv3r: you'll have to be more specific than that :)
<_enrico_>
maybe arch id is different from your uboot
<vinifr>
is it better to generate thread patch with git-send-email or git-format-patch?
<rellla>
<oliv3r>rellla: wingrime is doing unification etc so ping him to get the very best
<mripard>
oliv3r: again, I'll need more infos.
<mripard>
what's your config, which commit have you been using, on which board, did you make any modification on it, etc.
<vinifr>
Turl, and how to generate patchset each being reply to previous patch?
<vinifr>
Turl, i tried 'git format-patch --thread=shallow', but does not work
geecko has quit [Quit: Quitte]
<Turl>
vinifr: shallow means everything is a reply to cover letter or first patch if cover letter is absent
<oliv3r>
mripard: multi7 with only allwinner enabled and debugging (earlyprintk) + my driver (but that crash happens long before mine gets probed and kernel boots and runs fine; i tested it on cb1 and cb2; where cb2 crashes and same kernel runs fine on cb1
<oliv3r>
mripard: i used your a20-clocks tree as head; but have since rebased it on 3.11 (but haven't tested that)
ganbold_ has joined #linux-sunxi
<wingrime>
rellla: pong
<rellla>
wingrime: are you doing some sun7i/sun4i cedar unification atm?
<wingrime>
rellla: cedar same
<wingrime>
rellla: I not see much diffs
<rellla>
yes not much, but a few
<wingrime>
rellla: Now i trying finish disp patch for full dynamic unification
<wingrime>
rellla: actualy on a13 cedar down clocked
<wingrime>
rellla: on a20 up clocked
<rellla>
so is there a branch with a working sun7i cedar kernel module somewhere? or is it enough to build the sun4i one in stage/sunxi-3.4?
vinifr has quit [Quit: Saindo]
<hramrach>
shineworld: did you get the thing working?
<hramrach>
that's probably the arch id problem or script.bin problem
<hramrach>
depending on the content of hte first line
<hramrach>
I have an image for cubieboard but might not work with
<hramrach>
olimex board
<shineworld>
hramrach, I've get script.bin from nand image so I guess is right
<shineworld>
hramrach, I've removed the __machid=0xf35 from my uEnv.txt and it's the same
<hramrach>
shineworld: not necessarily
<hramrach>
I got script.bin from nand image first and the kernel would not boot from nand with that
<hramrach>
the kernel in nand might take som shortcuts not looking in some sections and the memory parameters might be unset
<shineworld>
ah .... where I can get a valid script.bin (fex) for lichee 3.3.0 present in olimex sdk 1.03 ?
<hramrach>
depends on the board, obviously
<shineworld>
cubieboard2
<hramrach>
maybe sunxi-boards has some
wingrime_ has joined #linux-sunxi
<shineworld>
I will try ... however I'm trying with another kernel just to check if my uEnv.txt is fine: shine@shine:~/android/build/jwrdegoede/linux-sunxi$ git checkout -b cubieboard2 origin/sunxi-3.3-cubieboard2
<hramrach>
I did not test using that hwpack but at least the script.bin should work
<shineworld>
perfect... time to download... for me Android on SD is a MUST
<hramrach>
if you get it working write up somewhere what you did
<shineworld>
with CB1 all work fine but with CB2 is another history at moment
<shineworld>
SURE, I've an account in linux-sunxi
<hramrach>
I cannot get android working on cubieboard A10 either
<hramrach>
sure, you can download some obsolete half-broken image
<hramrach>
but for me 'working' means recent kernel
<hramrach>
and for that I would have to build my own image which somehow did not work out
<mripard>
oliv3r: I doubt it's multi_v7
<mripard>
since multi_v7 have SMP enabled and with 4 CPUs max
<oliv3r>
mripard: well the v7 defconfig; not sure the exact name
<mripard>
and your log says your max is 1
<wingrime_>
hno: do you know, I tryed to use hdmi to dvi cable and this not not works
<wingrime_>
hno: there any solution?
<hramrach>
I use hdmi to dvi cable all the time
<shineworld>
Actually I'm running android (openbox) with 3.0.x kernel... I've tried to move to 3.4 but there was always something not working fine
<hramrach>
well, it does not work with android.
<hramrach>
I tried the openbox thing and it boots and then immediately complains that launcher is locked up or w/e
<shineworld>
I though often to move from Android but is too much happy for me write an app which works fine in any phone, cubieboard or PC with xasm simulator
<hramrach>
and input oes not work
<hramrach>
it canot be built with 3.4 kernel because the compiler in openbox cannot build the kernel
iamfrankenstein has quit [Ping timeout: 245 seconds]
<oliv3r>
mripard: and I swear it's multi_v7
<hramrach>
I tried hacking the build system and set different compiler for the kernel
<wingrime_>
hramrach: somthing wrong
<hramrach>
we are dealing with software. there is always something wrong
<shineworld>
Actually 10 cubies are running 24/24 7/7 in different industrial owens :)
<shineworld>
Android powered obviously
<wingrime_>
hramrach: also, you mean that hdmi-to-dvi not workrs with stock kernel?
<oliv3r>
mripard: but kernel config sholdn't make it crash, should it? so even if I accidentally somehow disabled SMP
<hramrach>
if you are willing to pick a display that is compatible with the broken HDMI driver or give up graphics output then the android would work
<hramrach>
wingrime_: the stocke kernel has broken HDMI driver which dan only drive a TV. if you have a DVI TV then the hdmi to dvi cable would work ;-)
<shineworld>
I've used LVDS display for them so nothing HDMI
<oliv3r>
mripard: i'll retest tonight; i'm now just realzing i was having issues t afirst with my driver not being recognized or something, turns out, i had the wrong DTB (name matched, content very likly wrong)
<oliv3r>
mripard: I dont' think i rechecked that error with the finall kernel + final dtb
<oliv3r>
mripard: so let me double check that tonight
<hramrach>
LVDS might work too. the android images were made for tablets after all
<shineworld>
yeah... are simple to configure and require few modifies on android sources to works fine (openbox is tailored to TV)
<shineworld>
hramrach, the problems isn't with kernel, I'm using original from olimex-sdk-1.03 but u-boot from github.com/linux-sunxi/u-boot
<shineworld>
hramrach, using your all works fine
<shineworld>
I can't use olimex-sdk-1.03 u-boot because don't have the spl part
<shineworld>
so I ask to you : but is linux-sunxi / u-boot ready to be used with cubieboard2 ?
<shineworld>
I've build it with: make CROSS_COMPILE=arm-linux-gnueabihf- Cubieboard2
<hramrach>
shineworld: I use hno's u-boot in that image
<hramrach>
bu the changes should be in sunxi u-boot by now
<hramrach>
also linux-sunxi 3.4 stage branch boots on cubieboard2
<hramrach>
most stuff should work
<hramrach>
Linux cubieboard2 3.4.43+ #33 SMP PREEMPT Tue Sep 3 09:49:49 CEST 2013 armv7l GNU/Linux
<shineworld>
uhm.. I've tried many and many times... I will check better next days ... at moment I will continue to solve script and android related questions
<hramrach>
one problem is the kernel in that 3.3 iamge uses old AW mach id but sunxi u-boot uses new mach id
<hramrach>
but you should see that immediately - it will complain ht eid is wrong
<hramrach>
other problem is when you get broken kernel/script.bin
ibrah2 has joined #linux-sunxi
rellla has joined #linux-sunxi
gnufan has quit [Ping timeout: 240 seconds]
sud0x3 has quit [Ping timeout: 240 seconds]
<shineworld>
it don't compain ht eid because stay frozen at Loding kernel message
<shineworld>
seem isn't able to load kernel
<Turl>
shineworld: did you enable earlyprintk?
<shineworld>
in uEnv.txt ?
<shineworld>
sorry I'm dummy in some things
<Turl>
you need to pass earlyprintk on cmdline
<shineworld>
I will add
<Turl>
and enable DEBUG_LL and EARLY_PRINTK on kernel config
ibrah2 is now known as eebrah|noPC
<shineworld>
I will do :) thanks for suggestions
sud0x3 has joined #linux-sunxi
<_enrico_>
shineworld: as i said, did you check that mach-id in kernel and u-boot match?
<shineworld>
when I finish to fix adroid needed file I will check :) just some time
<shineworld>
actually I get:
<shineworld>
shine@shine:~$ adb shell
<shineworld>
- exec '/system/bin/sh' failed: No such file or directory (2) -
<Turl>
shineworld: sounds like /system didn't get mounted
<shineworld>
yeah...I'm checking init.rc
<shineworld>
Turl, you are right: [ 41.724075] init: cannot find '/system/bin/sh', disabling 'console'
<boycottg00gle>
config created by make VERSION=3.4.3 kernel-3.4.3-armv7hl-sun7i.config -f Makefile.config looks good for olinuxino a20 - will use that one
<boycottg00gle>
will it be possible to have one kernel for all allwinner boards?
<boycottg00gle>
that is the idea behind the fex stuff - right?
<Turl>
for all boards with one specific SoC
<boycottg00gle>
sorry for producing so much noise but i am really noob here
<boycottg00gle>
Turl: so there will be a kernel for a10, a13, a20 ?
<Turl>
there's work towards making it one kernel for a bunch of SoCs
<Turl>
but today, it's one kernel for SoC/family (sun4i, sun5i etc)
<boycottg00gle>
ah, where come thos sunXi names from?
<Turl>
oliv3r: yay, greg will take sid for 3.12 :)
rellla has quit [Ping timeout: 264 seconds]
ZetaNeta has quit [Quit: I need no food or water. I am feed with the magic emerge leaves behind it after re-emerging the world]
ZetaNeta has joined #linux-sunxi
eebrah|noPC has quit [Ping timeout: 260 seconds]
ibrah2 has joined #linux-sunxi
Seppoz has joined #linux-sunxi
massi has quit [Remote host closed the connection]
tinti has quit [Ping timeout: 245 seconds]
<shineworld>
Turl, do you know what is databk partition in 4.2.2 ?
<ykchavan>
inserted some printf in fel.c and found that somehow 'libusb_bulk_transfer' is being called with unusually long length of 33488. Usually, it is 16,32...
<shineworld>
I'm used to common 10 partitions
<Turl>
shineworld: nope, how big is it?
<Turl>
shineworld: sounds like "data backup" going by the name only
<Turl>
ykchavan: gdb is your friend :)
<ykchavan>
:-)
<shineworld>
I don't know, I'm moving init.sun7i.rc from block (nand) to uSD
<shineworld>
and I've got:
<shineworld>
#try to format /databk
<shineworld>
wait /dev/block/databk
<shineworld>
setupfs /dev/block/databk
<shineworld>
mount ext4 /dev/block/databk /databk wait noatime nosuid nodev barrier=0
<shineworld>
umount /databk
<shineworld>
rmdir /databk
<shineworld>
never listen before in any system
<Turl>
shineworld: is '/dev/block/databk' used anywhere else?
<shineworld>
I need to grep ... a sec
<Turl>
shineworld: it might be 'data backing' too
<shineworld>
usually there is a "recovery" partition for backup system... .... I will google a little
<shineworld>
is used by /system/bin/preinstall-sh
<Turl>
can you paste it?
<Turl>
probably a way to preinstall bloatware? :)
<shineworld>
probably
<shineworld>
I will try to ingnore for a while
tinti has joined #linux-sunxi
tinti has quit [Changing host]
tinti has joined #linux-sunxi
<hno>
ykchavan, make sure you use the home button and not the recovery button.
<ykchavan>
hno, I am working on tablet which has 'FEL/reset' button.
<hno>
wingrime, I have never succeeded in using hdmi->dvi cable, but have not tried in a long long time. Don't usually run anything with other than serial output.
<ykchavan>
I can confirm that tablet is in FEL by using livesuit.
<ykchavan>
My A10 tablet does not have home or recovery button. All are software buttons.
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
<hno>
ykchavan, you MUST use the soft home & power button to enter fel mode for being able to dump boot0 & script.bin. You can not use the FEL/reset button.
tinti has quit [Ping timeout: 264 seconds]
<hno>
ykchavan, you don't have any hard buttons other than power button? All touch screen buttons?
<hno>
any volume button?
<ykchavan>
I have hard buttons 'reset' 'power' 'vol+'(also menu) 'vol-'(also esc)
<hno>
then use vol+ instead of home.
<ykchavan>
ok.
<hno>
you can have OTG cable connected the whole time if you prefer.
<ykchavan>
I see
_enrico_ has quit [Quit: Bye]
<hno>
ykchavan, did it enter FEL mode?
<ykchavan>
I am struggling
<ykchavan>
It is booting regularly.
<ykchavan>
hno, looks like it worked.
<ykchavan>
Let me add this in wiki.
<ykchavan>
Are these 2 different EFL modes?
arokux has joined #linux-sunxi
<boycottg00gle>
drivers/misc/pwm-sunxi.c:51:24: fatal error: pwm-sunxi.h: No such file or directory
tinti has joined #linux-sunxi
<hno>
ykchavan, there is many ways to enter FEL mode, and differs in how much of the board have been initialized before FEL is active.
<hno>
the reset/recovery button is most low leve, enters FEL mode before anything is initialized.
<hno>
the button method is soft, where the bootloader enters fel mode after initializing everything.
<ykchavan>
Adding this to wiki.
<hno>
bu that time the bootloader and script.bin have both been loaded into memory and is why it is possible to read them using fel.
<hno>
good. Added a little info about vol+/vol- after you asked.
leowt has joined #linux-sunxi
<ykchavan>
My tablet shows different light indicator for both of these FEL mode.
<ykchavan>
blue for reset fel. pink for home fel.
<ykchavan>
So, they are definitely different.
<oliv3r>
Turl: whoot finally! victory :)
<ykchavan>
hno, I mounted nanda , copied script.bin and copied script.bin via FEL mode. They both differ.
<ykchavan>
Is this expected?
<atsampson>
is the one you got from FEL mode actually sensible and not just uninitialised memory? (can bin2fex read it?)
<ykchavan>
Let me get fex from both and compare.
<ykchavan>
script.bin copied via FEL is corrupt.
<ykchavan>
boot1.header can be read via bootinfo
<ykchavan>
and working script.bin got by mounting nanda. So I got both files anyway.
sud0x3 has quit [Ping timeout: 260 seconds]
sanka has joined #linux-sunxi
<oliv3r>
mripard: happens with current 3.11 aswell; fresh dtb, no modification besides my driver :)
<oliv3r>
double che king my .config now
<oliv3r>
mripard: SMP is auto selected by ARCH_MULTI_V7
<oliv3r>
ohh smp is not on! i found the setting, but /SMP said it was on
<oliv3r>
strange
<oliv3r>
and I deffo didn't touch that setting
<ykchavan>
hno, with fel command, I tried to dump same number of bytes that my script.bin(copied via mounting nanda) has. starting from 0x43000000. Still it did not work.
<ykchavan>
Have you seen any case where script.bin starting from another memory location?
<hno>
ykchavan, it's always at same location. But you you are happy with mounting nanda and getting the file with adb then that works too.
<ykchavan>
by the way, script.bin is readable in hex editor :-)
<ykchavan>
I guess only padding is done to create this bin from fex.
<hno>
is there any differences if you convert them to fex?
<ykchavan>
yes
<hno>
what differencs?
<ykchavan>
script.bin via fel gives corrupted output.
<Turl>
oliv3r: maybe ARCH_SUNXI doesn't depend on SMP, and if you disabled all the other ones, it went off
<oliv3r>
that's quite likly
<oliv3r>
there's a flag, boot SMP kernel on uniprocessors
<oliv3r>
so should be save i suppose
<hno>
ykchavan, that's odd.
<Turl>
oliv3r: SMP_ON_UP?
<oliv3r>
i guess
<oliv3r>
i just saw it
<oliv3r>
yep that fixed it
<oliv3r>
mripard: is the kernel to panic though if i choose non-SMP mode? i would expect it not to be the case
<oliv3r>
Turl: [ 0.000000] Linux version 3.11.0 (oliver@riley) (gcc version 4.6.3 (Gentoo 4.6.3 p1.13, pie-0.5.2) ) #11 SMP PREEMPT Tue Sep 3 19:28:22 CEST 2013
<oliv3r>
so cool :)
<Turl>
oliv3r: I don't see it panicing on your log
<ykchavan>
I would like to get memory dump and search for the script.bin file origin. Can you suggest me limit from which memory location to which?
<hramrach>
I don't even know how to test host mode
<hramrach>
is it supposed to work automagically or do you need to frob something to switch?
<hno>
oliv3r, oh, Linux USB stack almost as bad as the PS3 USB stack was when it got hacked..
<hramrach>
alos I don't have the correct cable to connect another device to OTG
<arokux>
hramrach, I have all the needed equipement and will test it soon
<hno>
hramrach, to test host mode you need to configure the port as host, and suitable cable.
<hramrach>
configure in what way?
<hno>
in your script.fex.
BJfreeman has quit [Quit: had a good time]
<hno>
or by building the module as host-only (then ignores port type set in script.fex I think)
<hramrach>
it builds as module?
<hno>
yes if you ask it to. but no difference is built in to kernel.
<hramrach>
and if it's not broken. Last time I tried it did not build as module
<hramrach>
but need the cable first. I think I saw some odd USB bits somewhere
<hno>
not too hard to make your own cable.
<hramrach>
you can't make your own for the price of pre-made
<hramrach>
the connectors are aeach alone more expensive than the whole cable
<arokux>
hramrach, not if you have old usb mouse (a tip given to me by hno)
<arokux>
(and if you do not need this mouse anymore :) )
<hramrach>
you still need the other end
<hramrach>
and hte cable has no connector found on old mouse
<arokux>
the other end is inside the mouse
<arokux>
(i'm talking about mele, its usb otg is on the board and has special connector
<arokux>
)
<hramrach>
my OTG port is plain mini-usb
<hramrach>
so you can use something like mini-A male - A female to make it look like host connector
<arokux>
ah, ok
leowt has quit [Quit: leowt]
robb83 has joined #linux-sunxi
<robb83>
hi
ibrah2 has quit [Remote host closed the connection]
<oliv3r>
i got a back of 10 connectors from e-bay for various purpouses
atiti has joined #linux-sunxi
ibrah2 has joined #linux-sunxi
ibrah2 has quit [Ping timeout: 256 seconds]
ibrah has joined #linux-sunxi
atiti has quit [Read error: Operation timed out]
<buZz>
oliv3r: i bet massages dont come cheap now
<oliv3r>
heh
<boycottg00gle>
debugfs is cool
<boycottg00gle>
unfortunately i still get "[ 1465.503202] HDMI not connected" when in fact it is connected
ykchavan has quit [Quit: Leaving]
shineworld has quit [Quit: Leaving]
wingrime_ has joined #linux-sunxi
<wingrime_>
ssvb: ping
<ssvb>
hi wingrime
<oliv3r>
i'm tired :(
<oliv3r>
i wanted to code some, but too tired :S
<wingrime_>
ssvb: I suppricely can't get cedar output on hdmi output but with rgb display all works fine, also same for a13 - can't request scalar, so question, how do you fixed it last time for XV?
<ssvb>
wingrime: a13 hardware has only one scaler, a10 has two scalers
<wingrime_>
ssvb: yes, but how xv works now on a13?
<ssvb>
wingrime: if you enable the 'scaler_mode' for the framebuffer in fex file, it take up one scaler
n01 has joined #linux-sunxi
<wingrime_>
ssvb: can we disable such in kernel?
<ssvb>
ah, that's probably because the xorg driver reserves the scaled layer for itself
<mturquette>
Turl: very belated pong
<mturquette>
(due to US holiday weekend)
<ssvb>
wingrime: so there is no scaled layer left for vdpau
<Turl>
mturquette: :)
<Turl>
mturquette: I wanted to ask you about automagic reparenting
<ssvb>
wingrime: I need to change it to request the scaled layer only when it's really in use
<Turl>
mturquette: is it supported so far?
<wingrime_>
ssvb: but hdmi request one?
<boycottg00gle>
i suspect my hdmi hardware is broken - any tips to check?
<mturquette>
Turl: i merged James Hogan's remuxing series for 3.12. It will be in the pull requeust I send to Linus this week
<wingrime_>
ssvb: fb0 realy need one or not?
<oliv3r>
boycottg00gle: i never gotten my hDMI to work on my tablet
<ssvb>
wingrime: that's bad, try to change fb0_scaler_mode_enable and fb1_scaler_mode_enable to 0
<boycottg00gle>
oliv3r: ah ok - for me it worked at some time sometimes and somewhow (picture was cut off) - but then it stopped working
<wingrime_>
ssvb: I see, maybe we need force kernel not trust this ?
<hno>
hramrach, just massacre two USB cables with suitable contacts.
<ssvb>
wingrime: you can also use https://github.com/hglm/a10disp to change the scaler mode of the primary framebuffer layer at runtime
* boycottg00gle
thinks he killed his hdmi using a10disp
<wingrime_>
ssvb: you can't do it directly from xv86-video-sunxi?
<Turl>
mturquette: I'll need to check it out then, it'd be of use to configure our mmc clock
<wingrime_>
ssvb: also I , thinking about how add kms / dri on top currect driver..
<boycottg00gle>
oliv3r: at least you don't ignore me
<wingrime_>
ssvb: if you can't we can't change it with /dev/disp I can try add new IOCTL for this
<wingrime_>
*can change
<boycottg00gle>
thought this /dev/disp will be removed?
<ssvb>
wingrime: the ioctls for toggling the scaler mode already exist
<oliv3r>
boycottg00gle: nobody is ignoring you; but i'm affraid nobody knows the answer :(
<oliv3r>
boycottg00gle: i responded to let you know that we don't ignore you; but ... don't know :)
<wingrime_>
ssvb: so, you can simply add some options to xconfig or force it
<boycottg00gle>
oliv3r: thanks
<oliv3r>
boycottg00gle: but ssvb knows everything about a10disp and hdmi and stuff ;)
<ssvb>
wingrime: yes, basically borrow some bits from a10disp and disable the scaler mode for the primary framebuffer layer
<boycottg00gle>
ssvb: if i connect hdmi shouldn't the HPD register be != 0 ? if it stays at 0 is it safe to assume a hardware problem?
ykchavan has joined #linux-sunxi
<boycottg00gle>
ssvb: or is that complete nonsense and there may be many other reasons?
<ssvb>
boycottg00gle: for the assistance with the hdmi stuff better ask hansg
wingrime_ has quit [Quit: leaving]
<boycottg00gle>
ssvb: thanks
wingrime_ has joined #linux-sunxi
<boycottg00gle>
this hdmi stuff makes me crazy - if it wouldn't have worked at all, ok - but it worked somehow and i can't reproduce it... have to get a break and will get a new board this week and try again - thanks again - see you
<wingrime_>
ssvb: yes, looks you was right , that was fex problem,
<wingrime_>
but now I see more one bug
boycottg00gle has quit [Remote host closed the connection]
<wingrime_>
ssvb: cedar output shakes
<Turl>
mturquette: so I need determine_rate then?
<ssvb>
wingrime: what kind of hardware is that and what are you using for output?
<wingrime_>
ssvb: hdmi
<wingrime_>
ssvb: hdmi-to-dvi
<ssvb>
wingrime: is it a13? and what is the screen resolution and refresh rate?
<wingrime_>
ssvb: a10
<ssvb>
wingrime: the screen starts shaking if there is apparently not enough memory bandwidth for scanout
<wingrime_>
ssvb: 1920x
<wingrime_>
1080 60 Hz
notmart has quit [Quit: notmart terminated!]
<ssvb>
wingrime: yeah, that's a common problematic scenario, reducing the refresh rate to 50hz (if your monitor supports this) can make it a bit better
<wingrime_>
ssvb: this is fixable ?
<ssvb>
wingrime: also changing the color depth to 16bpp and clocking the memory higher helps
<ssvb>
wingrime: and disabling the colorkey puts less stress on the memory
ibrah has quit [Remote host closed the connection]
<wingrime_>
ssvb: jemk's code also non optimal,
<wingrime_>
ssvb: it change all layer params for every call
<wingrime_>
ssvb: not simply move layer
<ssvb>
wingrime: excessive re-configuration ioctls should not be a big problem
<wingrime_>
ssvb: where bottle neck ?
<ssvb>
wingrime: the problem is that the 1920x1080 framebuffer *and* your 1920x1080 video layer are both read 60 times per second to send the pixels to hdmi
<wingrime_>
ssvb: now I test 720p only sintel
<wingrime_>
ssvb: if I downscale it a bit, it works good
<ssvb>
wingrime: it's a simple math, the display controller needs to read the framebuffer part overlapped by the video to check for the colorkey (if you are scaling the video to full screen, that's 1920x1080) *and* also the buffer with the video frame
<ssvb>
wingrime: for the framebuffer it's 32 bits per pixel, and for the decoded video buffer it's 12 bits per pixel
<ssvb>
wingrime: add everything together, and you get the required memory bandwidth
<wingrime_>
ssvb: can we just speedup disp using pll or some divider?
<oliv3r>
overlcock disp!
<wingrime_>
oliv3r: AW claim that we can 4k with a31
<ssvb>
wingrime: nobody knows yet, somebody has to investigate and fix this disp issue :)
<oliv3r>
better disp :D
<ssvb>
wingrime: maybe yes, if you equip the board with a very fast memory
<ssvb>
wingrime: what is you memory clock speed now?
<wingrime_>
ssvb: 408
<wingrime>
rellla: actualy on a13 cedar down clocked
<ssvb>
wingrime: not just for fixing this screen shaking issue, but also for general performance
<oliv3r>
on cubie1 you can even try 512 MHz
<oliv3r>
tom said it ran stable
<oliv3r>
but i have my doubts :)
<ssvb>
oliv3r: I overclocked the memory to 504 in CB2 just for a test, it booted, but failed to work reliable
ykchavan has quit [Quit: Leaving]
<wingrime_>
ssvb: are you touched cas?
<wingrime_>
ssvb: it must be increased
<oliv3r>
ssvb: but with so many unknowns in the memory controller; it may be tweaked to do more, the memory chips aren't the issue are they
<ssvb>
oliv3r: also Tsvetan2 says that a certain percentage of A20 chips just can't work with the memory clocked at higher than 400mhz
<ssvb>
oliv3r: it was something like 80% of A20 chips were fine with 480mhz, but the remaining 20% failed even at something closer to 400
<wingrime_>
ssvb: also, I just checked blob, it have _MUCH_ artifacts on sintel 720p
<wingrime_>
ssvb: but it not shakes at least
<ssvb>
wingrime: does the blob powered video player enable the colorkey?
<wingrime_>
ssvb: artifaces looks like broken weght-prediction
<oliv3r>
ssvb: we should build some mini sd image, that people can download and burn for a10 and a20 that tests stability; like memtestx86, but then ssvbtestxarm :)
<ssvb>
wingrime: yeah, the linux blob is broken, that's why we used the android blobs via libhybris
<wingrime_>
ssvb: it must not reconfigure layer every frame
<wingrime_>
ssvb: only move window
<ssvb>
wingrime: maybe
<arokux>
anybody willing to test some usb code in mainline?
<ssvb>
wingrime: we should try to better integrate it with the xorg driver anyway
<wingrime_>
ssvb: it all can be done using drm
<ssvb>
wingrime: yes, it can, and it also can be done without drm
<ssvb>
wingrime: with our current software stack, doing it without drm is a bit easier
<hramrach>
with drm you can technically get non-X programs use the same infrastructure
<mturquette>
Turl: correct.
<wingrime_>
ssvb: cedar have tons of registers
<Turl>
mturquette: is there any special reason why?
<wingrime_>
ssvb: maymust be configured for every frame
<wingrime_>
*many
<mturquette>
Turl: easier to introduce a new callback that combines set_rate/set_parent than to modify all the existing set_rate callbacks out there
<ssvb>
hramrach: that's assuming that we have a properly working drm kernel driver
<ssvb>
hramrach: which is a non-trivial task by itself
<Turl>
mturquette: I was thinking the framework could've just called round_rate with all the different parent clocks and then choose the nearest one
<ssvb>
hramrach: it all depends on whether we want everything to work on the 3.4 kernel right now, or have it added in the mainline kernel in some (maybe distant) future
<wingrime_>
ssvb: drm can be added on top current disp
<hramrach>
I guess at this point it makes seanse to start moving towars mainline. it has enough drivers to give somewhat usable system
<ssvb>
wingrime: in theory yes, but it is already a bit convoluted and may become really unmaintainable if you make it support yet another api
Black_Horseman has joined #linux-sunxi
<libv>
can we have a proper working drm driver?
<libv>
there is no driver out there for which display, 3d, 2d and media are completely separate blocks.
<libv>
you either have everything mashed together (and controlled by one big entity in hw), or you have kms drivers
<jelly-home>
gpgpu/cuda seems decently separated even if one blob rules the kernel side of the things
<jelly-home>
at least from my limited userspace perspective
wingrime has quit [Ping timeout: 245 seconds]
<arokux>
Turl, would you like to test some usb code for me? :)
<arokux>
Turl, (on cb)
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
<wingrime_>
2048x1536@60fps
<wingrime_>
ssvb: maximum supported by a20 docs
<wingrime_>
ssvb: for rgb
akaizen_ has joined #linux-sunxi
akaizen has quit [Read error: Connection reset by peer]
<arokux>
anybody want to sell cb2?
<wingrime_>
arokux: it's so difficult to by it ?
<wingrime_>
*bye
<wingrime_>
*Buy
<arokux>
maybe i could get it cheaper :)
<wingrime_>
ssvb: according manual ,hdmi module gets 24 bus, VS and HS signals, so, looks like problems can be when hs and vs are too late
ganbold_ has quit [Remote host closed the connection]
Seppoz has quit [Remote host closed the connection]
Seppoz has joined #linux-sunxi
BJfreeman has quit [Quit: had a good time]
<mturquette>
Turl: .round_rate can suggest a better rate with with &best_parent_rate, something similar needed to be done for &best_parent (or whatever)
<mturquette>
Turl: the function signature of .round_rate didn't allow for it to pass back the best parent
<mturquette>
Turl: and introducing a new callback was easier than updating all of the existing users of .round_rate and hoping that there would be no regressions
_enrico_ has quit [Quit: Bye]
arokux has quit [Remote host closed the connection]
vinifr has joined #linux-sunxi
sanka has quit [Remote host closed the connection]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
xtofury has joined #linux-sunxi
<Turl>
mturquette: oh, that's why it's a pointer :) I never fully understood that bit
<vinifr>
Turl, i did a test with my email, but does work. should work with email list?
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<Turl>
vinifr: I don't think you need --thread on send-email if you're using properly threaded patches made by format-patch
<mturquette>
Turl: yeah, it's not the most beautiful api but it is a bit more powerful than some of the legacy clock framework implementations
<Turl>
mturquette: having it on-framework would've been nice though - I'd just have to enable a flag and it'd work
<mturquette>
Turl: that is another change: if you use .determine_rate then the remuxing behavior happens by default
<mturquette>
Turl: you only set a flag if you do NOT want to automagically switch parents
<mturquette>
so that inversion in logic (compared to CLK_SET_RATE_PARENT for parent rate propagation) was another reason that I wanted to have a new callback
<Turl>
mturquette: the thing is I'm using a composite clock
<mturquette>
Turl: interesting. i haven't seen a lot of adoption of the composite clock type
<mturquette>
Turl: and that clock type is really working within some tight constraints
<mturquette>
i'm loathe to make changes to it for supporting .determine_rate versus .round_rate
<Turl>
mturquette: well, I stuffed a lot of stuff behind composite clocks
<Turl>
probably something like this will be enough http://git.linaro.org/gitweb?p=people/mturquette/linux.git;a=blobdiff;f=drivers/clk/clk.c;h=bc020372106ba939191f3c4e0130e81df650b3bf;hp=f33f1ccf1b2f22ba51823aaa5f0395a14b84ad39;hb=e366fdd72529c545ccf327569ee250c1673be221;hpb=819c1de344c5b8350bffd35be9a0fa74541292d3
<Turl>
now the problem is how to get it into the composite clock
<Turl>
as mux has it, but it could also be on the 'rate' component