<Turl>
45 * IRQF_SHARED - allow sharing the irq among several devices
<arokux2>
Turl, everybody else does the same..
<Turl>
I dunno, just looking at stuff that looks suspicious
<arokux2>
thanks Turl
naobsd has quit [Quit: Page closed]
<arokux2>
Turl, the code is hooked in differently, though
<arokux2>
but I doubt this is a problem....
<arokux2>
Turl, hm........ I just remembered of jukivili 's patch
<arokux2>
Turl, you know, unaligned mem access
<arokux2>
Turl, maybe aw's code works but then fails because of unaligne access
<arokux2>
Turl, no...........
<arokux2>
Turl, (I've applied a fix suggested by him, but the same problem..)
<libv>
ssvb: running against X effectively kills the job interleaving
<libv>
ssvb: i hit 60% with no job handling when i brought up q3a
<libv>
ssvb: running the same code in egl in an X11 window, versus running the equivalent limare code really is 2/3rds of the performance
<libv>
in quite a few tests
<libv>
and i am consistently 6% faster than the binary driver against X11
<ssvb>
libv: limare equivalent, but implemented with X11 would be to just just kick DRI2SwapBuffers instead of FBIOPAN_DISPLAY without ever waiting for any confirmations from the X server
<arokux2>
Turl, any ideas what I should do next?
<libv>
i guess so, i have not looked into that path yet, i just know that the hardware has that much more to give
<arokux2>
Turl, I do not understand why my own driver fails differently, could it be normal...?
<Turl>
dunno
<ssvb>
libv: the problem with the current mali blob (and partially with dri2) is that we only have 2 buffers, and the right way to use them correctly is to wait after swap before we are allowed to safely write to the other buffer
<ssvb>
libv: there is exactly the same problem with the framebuffer mali driver, if it is used "correctly"
<libv>
ssvb: we can still run the vertex shader while waiting
<libv>
waiting with the fragment until the buffer is available
<libv>
so yes, you are right, we will not get 100% hw usage
<libv>
but we will get between 80 and 90
<ssvb>
libv: if using sufficiently large number of buffers and cycling through them, we would never have to wait
<libv>
ssvb: my current mali scheduling code goes part of the way.
<libv>
it runs the 2 processor asynchronously to begin with
<libv>
but it starts building up the third frame already
<libv>
so the oldest is rendering out to the freshly released buffer
<libv>
the middle on is churning through the vertices and generating the polygon lists
<libv>
and the final one is collecting draws already
<arokux2>
Turl, good night
<arokux2>
good night gus
<arokux2>
guys*
<ssvb>
libv: let's say we have 10 buffers, after you swap the first buffer (push it to the display server), you can instantly write to the next one, or even to another one after it (knowing that they have been likely freed long ago)
<ssvb>
libv: you can only catch up with the non-freed buffers only if having the framerate higher than the monitor refresh rate
<ssvb>
libv: but this needs throtting anyway, so the performance is not a problem
<libv>
having too many buffers will kill any feeling of attachment to input
<ssvb>
hmm, it shouldn't affect input, only extra memory is wasted
<ssvb>
or yes, it can, but this is tunable
<ssvb>
depends on how we drop frames when rendering too fast
egbert has quit [Ping timeout: 264 seconds]
<ssvb>
currently with the sunxi display controller, the overlay re-configuration or FBIOPAN_DISPLAY only happens on vertical blanking
egbert has joined #linux-sunxi
<ssvb>
so if you push new buffers too fast, the some of them will be automagically skipped
<ssvb>
*then
<libv>
let's first get a proper mesa driver up and running, and then worry about the performance details
<libv>
the large lines about getting good performance out of this engine are known
<ssvb>
well, I mean that just increasing the number of used buffers allows not to bother about scheduling vertex shaders or doing anything complicated
<libv>
memory is always at a premium
<ssvb>
true
<libv>
and randomly mapping and unmapping buffers is quite the overhead
<libv>
when i stopped re-mapping frame local memory i gained quite a few %
<ssvb>
mapping and unmapping buffers is totally unnecessary, this will kill performance for sure
<libv>
in any case, details.
<ssvb>
bad idea to begin with
<libv>
which we will solve.
<libv>
but not before i figured out how to convince mesa to handle fbos
<ssvb>
mali binary driver does not do mapping/unmapping buffers in normal conditions, it just gets 2 buffers and is happily swapping between them
<ssvb>
works fine as long as the x server and the client have the same idea about which one of them is front and which one is back
<ssvb>
the only real problem is that we want at least 3 buffers to avoid extra waiting
<libv>
again. later
<libv>
this can be played with endlessly. later.
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
<ssvb>
there is no need to play, this is kinda solved long ago
<ssvb>
just needs the open source client side code :)
Gerwin_J has quit [Quit: Gerwin_J]
CoSM0 has quit [Read error: Connection reset by peer]
CoSM0 has joined #linux-sunxi
CoSM0 has quit [Read error: Connection reset by peer]
<arokux2>
Turl, ?
CoSM0_ has joined #linux-sunxi
<Turl>
arokux2: ?
<arokux2>
Turl, chinese code now fails same way as mine
<steev>
i have it checked out here, but only as part of the sunxi-bsp
<steev>
though i have it for building the script.bin file and that's all i know to use it for (i don't boot my cb2 via fel or whatever)
<steev>
cuz linux, and i don't feel like figuring out all the stuff needed to get livesuite running on something other than debian or fedora
<Cubear>
I'm gonna try using the softfloat compiler
rellla2 has joined #linux-sunxi
rellla has quit [Ping timeout: 264 seconds]
<oliv3r>
Cubear: sunxi-tools isn't that big, so you could just crosscompile each binary manually and ignore the make file
forcev is now known as FunkyPenguin
FunkyPenguin has quit [Changing host]
FunkyPenguin has joined #linux-sunxi
<steev>
Cubear: changing to a softfloat compiler won't change anything, but as oliv3r suggests, compiling manually passing cc as the cross compiler might work
<steev>
but you'll need libusb and such cross compiled as well
<Cubear>
I'll try. The softfloat packages seem broken anyway, some dependency unavailable
n01 has joined #linux-sunxi
<mnemoc>
sunxi-tools' makefile takes CC iirc
<Cubear>
shall I edit the CC line from gcc to arm-linux-gnueabihf-gcc?
<mnemoc>
you can just pass it on command line
<mnemoc>
make CC=arm-linux-gnueabihf-gcc
<Cubear>
yep, now it complains about libusb
<Cubear>
/usr/lib/gcc/arm-linux-gnueabihf/4.4.6/../../../../arm-linux-gnueabihf/bin/ld: skipping incompatible /usr/lib/i386-linux-gnu/libusb-1.0.so when searching for -lusb-1.0
<mnemoc>
what tool you want? libusb is only needed for fex
<mnemoc>
fel
<Cubear>
I was hoping to compile all tools and bundle them somehow. But I really need just the fex tools.
<mnemoc>
you can, for example, make CC=arm-linux-gnueabihf-gcc bin2fex fex2bin
<Cubear>
yes, I'll do that now
<mnemoc>
for fel you need to install libusb's dev package
<Cubear>
./fexc: ELF 32-bit LSB executable, ARM, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.26, BuildID[sha1]=0x0efd3849c8999aa10ab3b7aaf80fc5aa77eff7fb, not stripped
<Cubear>
seels to work
<mnemoc>
=)
<Cubear>
hello nand-part
<mnemoc>
we should autotoolize that package
<Cubear>
alright, thanks for the assistance
<mnemoc>
yw
<n01>
'monin
<mnemoc>
moin
<Cubear>
btw I have another question
<Cubear>
I have decided to use the USB0 OTG port for a wifi device. Don't worry, I've tested it and it seems to function properly
<Cubear>
the only problem is that it doesn't get detected at boot
<Cubear>
I have to unplug it and replug it in order for it to work
<Cubear>
is there a way to make the OS detect the device at USB0 at boot?
<Cubear>
Using either of the normal USB HOST ports works fine
<mnemoc>
you might want to change the usb0 to host mode in your script.bin
<Cubear>
yeah, I'll do that next, now that I have the fex2bin and bin2fex.
panda84kde has joined #linux-sunxi
<Cubear>
nope, the compiled fexc not working on cb1
<Cubear>
~/sunxi-tools# ls
<Cubear>
bin2fex fex2bin fexc nand-part
<Cubear>
~/sunxi-tools# ./fexc
<mnemoc>
the first two are symlinks
<Cubear>
-bash: ./fexc: No such file or directory
<steev>
is it marked as executable?
<mnemoc>
if it's cross compiled it won't run
<mnemoc>
unless you run it on the right machine
<steev>
mnemoc: it should
<Cubear>
it's green in ls
<steev>
Cubear: ls -lah
<steev>
green doesn't tell us anything
<Cubear>
-rwxr-xr-x
<mnemoc>
steev: i mean, he seems to have compiled it in a "PC"
<Cubear>
green = executable afaik
<steev>
mnemoc: yes, but the point of cross compiling is to run it on the target system
<mnemoc>
steev: but i think he is running it in his "PC"
<steev>
"not working on cb1"
<Cubear>
I upoaded it to cb1
<Cubear>
if I try to run it on PC I get this: /lib/ld-linux.so.3: No such file or directory
<Cubear>
but I didn't expect it to run on the PC so...
<mnemoc>
ok
<oliv3r>
file fexc
<oliv3r>
:)
<Cubear>
./fexc: ELF 32-bit LSB executable, ARM, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.26, BuildID[sha1]=0x0efd3849c8999aa10ab3b7aaf80fc5aa77eff7fb, not stripped
Soru has quit [Read error: Connection reset by peer]
<n01>
wow
<n01>
uname -a?
<Cubear>
Linux cubie 3.4.61-custom+ #4 PREEMPT Mon Sep 16 20:49:43 CEST 2013 armv7l GNU/Linux
<oliv3r>
Cubear: ./fexc
<oliv3r>
:p
<oliv3r>
Cubear: is the partition mounted exec
<Cubear>
/dev/root on / type ext4 (rw,noatime,errors=remount-ro,commit=60,data=ordered)
<jelly-home>
Cubear: execve() failing with ENOFILE usually means there's no dynamic library loader for that ELF binary (ie. a 32bit binary on a pure 64bit system)
<oliv3r>
Cubear: copy it to /bin just in case
<oliv3r>
or listen to jelly home; he's much smarter
<jelly-home>
(or armel/softfloat binary on a system without softfloat glibc)
<n01>
find . -name ldconfig*
<oliv3r>
hf/sf is a good point
<Cubear>
-bash: /bin/fexc: No such file or directory
<Cubear>
are you saying I used the wrong compiler options when cross compiling?
<oliv3r>
Cubear: possibly or the wrong compiler
<oliv3r>
arm-gcc --version
<Cubear>
hmm this might be it
<Cubear>
I compiled the kernel on a different PC
<oliv3r>
doesn't matter
<oliv3r>
a hf compiler should be able to proudce proper softfloats
<oliv3r>
and fexc is int only anyway so shouldn't matter much anyhow
<tgaz>
zumbi: do you know what's broken about it? out of curiosity
<atsampson>
Cubear: it would be helpful if you could put the broken executable somewhere so we could look at it...
tinti has joined #linux-sunxi
rellla2 has quit [Ping timeout: 248 seconds]
<zumbi>
tgaz: yes, it is using old linker path
rellla has joined #linux-sunxi
<zumbi>
Cubear: try to rebuild ubuntu cross toolchain packages, that should work fine, as it supports multi-arch
<zumbi>
tgaz: emdebian cross toolchain is compiled with explicit headers and libs found under /usr/$triplet/{lib,include}, deprecated upstream long time ago
<tgaz>
zumbi: but there doesn't appear to be any rpath in there, and the referenced files seem to be found. (i still wonder about that ld-linux.so.3 path that seems to be rewritten)
<tgaz>
but, i'll remember your explanation for future problems. thanks.
SEnthoven has joined #linux-sunxi
<zumbi>
tgaz: maybe I am wrong... hold on
<zumbi>
binary codes /lib/ld-linux.so.3 => /lib/ld-linux-armhf.so.3, while, it really is ld-linux-armhf.so.3 -> arm-linux-gnueabihf/ld-2.13.so
<zumbi>
well, I guess that should work.. maybe ldconfig needs to be run.. now I am starting to guess
<lkcl_>
Cubear: what i've done is just "apt-get install gcc-4.7" on the EOMA68-A20 then "apt-get install git" and then checked sunxi-tools out.... absolutely no problems.
<tgaz>
zumbi: i don't get the magic "/lib/ld-linux.so.3 => /lib/ld-linux-armhf.so.3" rewrite ldd seems to say
<tgaz>
is it just falling back to *-armhf* if it doesn't find the original path?
<tgaz>
how hard can it be just executing some CPU instructions ;)
<tgaz>
oh. ldd -v. never knew about that.
<SEnthoven>
Hi, maybe one of you guys can help me out. I've made a board with an Allwinner A10 and 4x Hynix H5TQ2G83BFR DDR3 SDRAM chips. The processor starts but the DATA_TRAINING sequence for the ram fails. Any ideas?
<Cubear>
zumbi, I'm currently on an old Centrino laptop with debian wheezy using the emdebian toolchain. When I get back home in a few hours I can use my stationary PC with ubuntu 13.04 and its toolchain.
<Cubear>
I'll let it rest until then.
<arokux1>
uh, so much stuff here already.
Soru has joined #linux-sunxi
<arokux1>
Turl, yep, saw it already and replied. he's based his work on my old patches, which I changed.. not sure I did the right thing by changing commits instead of commiting fixes. how do you think? but then.. he should have checked the IRC for WIP.
<arokux1>
SEnthoven, hi! what board? :) we do kernel here, you might have better luck in #arm-netbook
<tgaz>
(probably some dynamic linking gone wrong due to incompatible toolchain)
<Turl>
arokux1: fixup patches are nice to see what changed, but otherwise updating commits is fine
<arokux1>
Turl, the thing is.. I was preparing the commits for mainline. but I could have squashed them just before submitting.
<SEnthoven>
arokux1: its my own board. yeah, the problem is a bit below the kernel. Its somewhere between ram config parameters and the hardware
<atsampson>
Cubear: you'll be pleased to hear it fails the same way for me (I think with the same Debian image you were using) ;)
<atsampson>
Cubear: it's because the executable is built expecting /lib/ld-linux.so.3 to exist, and it doesn't
<Turl>
mripard, oliv3r I wonder if Greg forgot about SID
<atsampson>
Cubear: on the other hand, you can invoke ld.so by hand to run it quite happily: do "/lib/ld-linux-armhf.so.3 ./fexc"
<oliv3r>
Turl: he'd push it with rc1
<Turl>
oliv3r: yeah but we're @rc2 :)
<n01>
my wdt is in :)
<Turl>
n01: indeed it is :p
* n01
think oliv3r is gonna kill himself now
<Turl>
lol
<Turl>
n01: I think he already has kernel patches :p
SEnthoven has left #linux-sunxi [#linux-sunxi]
<n01>
oliv3r: please don't do it, we need you
<atsampson>
Cubear: (or do "ln -s /lib/ld-linux-armhf.so.3 /lib/ld-linux.so.3" so it can find ld.so where it's expecting it, and where most systems have it)
\\Mr_C\\ has quit []
<Cubear>
atsampson, w00t, thanks for pointing that out
<zumbi>
but.. that's a workaround
<atsampson>
well, the alternative is hacking the GCC spec file so it produces executables that know where Debian puts ld.so, which is a bit more involved
<zumbi>
or get a properly configured cross tool
<zumbi>
but, yeah, for a quick fix, might do :)
<granite_crusher>
arokux1: hi, I tested minimally usb on A13 yesterday, but have other work to do :/
<arokux1>
granite_crusher, that is absolutely no problem, just be persistent :)
<arokux1>
granite_crusher, do you have some findings? :)
<granite_crusher>
arokux1: well I couldn't get rootfs boot from usb... got kernel panic. I am not good at that... however I tried to check usb things with that cpio you sent me, it is minimal, so I could test storage USB
<granite_crusher>
I dmesg'ed all USB peripherals I have
<arokux1>
granite_crusher, so storage works for you too?
<granite_crusher>
storage USB works, I mounted it
<granite_crusher>
(I didnt knew vfat needs some language enabled in kernel..)
<arokux1>
granite_crusher, very nice. once you have time we can debug booting from rootfs (it should work) -- on some evening
<granite_crusher>
my full speed (
<arokux1>
granite_crusher, yeah.. that was a problem for me too with vfat. it was a while till I found all needed configs.
<arokux1>
granite_crusher, so did you post you dmesges somewhere?
<granite_crusher>
my full speed (USB 1.1) M-audio Fast track pro got recognised in dmesg, however to test it fully I need full rootfs with alsa and jack fully working
<andoma>
anyone know if there is some kind of performance counters in the sunxi devices? so I can measure DRAM bus utilization, etc
Soru has quit [Read error: Connection reset by peer]
<granite_crusher>
arokux1:what so special about dmesg?
<granite_crusher>
it looks same as on pc
<granite_crusher>
arokux1: or You need a proof?
<arokux1>
granite_crusher, just interesting. maybe there were some useful messages. sometime you won't notice if you do not know what to look for.
<granite_crusher>
I probabbly didn't saved it (Ishould check if I loged putty)
<Turl>
granite_crusher: full speed worked for you? hm
<arokux1>
granite_crusher, but most probably nothing interesting there anymore..
<Turl>
granite_crusher: mine just did nothing and killed the usb port until I rebooted
<arokux1>
be back in 5 min
<Turl>
andoma: there's the normal cpu perf counters, but I haven't seen anything more advanced than that
<andoma>
Turl: ah ok
<granite_crusher>
arokux1, Turl: I have to warn about one thing on olimex-a13 boards: USB 5V is enabled allways by 3.3V applied directly on switch (logic 1), while pin CPU USB_EN is open with jumpper
<granite_crusher>
so this part about enabling usb in dts of olimex is not functional, until somebody would resolder the connection
soul has joined #linux-sunxi
<arokux1>
granite_crusher, hm.. "this part" - which part? if it didn't work then usb storage wouldn't work ...
<granite_crusher>
I mean it can just work by me, because 5V is always enabled
<arokux1>
granite_crusher, I see, it seems to be working for everybody else so far. some ppl told there could be problems but I haven't yet find users with them.
<arokux1>
granite_crusher, "USB 5V is enabled allways by 3.3V applied directly on switch.... " -- where do you know that?
<granite_crusher>
arokux1: from schematics
<oliv3r>
open hardware ftw
<granite_crusher>
there is usch 5 pin chip which should listen for signal from allwiner,pin PG11 aka USBH_EN, but line is cutted with jumper(open)
<granite_crusher>
arokux: and instead it is enabling 5V by applying 3.3 to that small chip
<arokux1>
granite_crusher, mm.. the power is enabled by GPIO outputting 1, Turl said GPIO pins can only output 3.3V, so I think this is standard to do 5V power supply like this?
<granite_crusher>
arokux1: so I can't test that part, unless I would unsolder somthings and solder other things (olimex leaved such place for)
<granite_crusher>
arokux1: GPIO can output 0 or 1 it simply doesnt matter, because that 0 or 1 never reach the usb hub
<granite_crusher>
instead 3.3 is soldered directly
<granite_crusher>
so power is enabled allways
<arokux1>
granite_crusher, oh, so you are saying that the regulator which i have in dt does actually nothing?
steev has joined #linux-sunxi
<granite_crusher>
arokux1: I don't know, 1. I could messure with voltmeter if it changes, or I could resolder stuff (not today) to do reall tests
<arokux1>
granite_crusher, no need to resolder anything. you could just comment out one line and recompile the kernel and see if storage works.
<granite_crusher>
arokux1: it is patch? or one line? :)
<arokux1>
granite_crusher, the line you need is at the top, the link includes an anchor at the and: #L2R164 so this is a link to a line.
<arokux1>
granite_crusher, you do not have wifi on board, right?
<oliv3r>
question: moving my dev tree's (that tend to be pushed to github anyway) to a raid0 (from a raid5 or 6) a good idea? I do think the disks (which potentially is a 5400 -> 7200 move iirc)
<oliv3r>
personally i find it good plan; dataloss isn't a huge issue, everything mirrord etc
<granite_crusher>
arokux1: no, but there are place for it 4 pins intentionally (USB) so I can easiely measure if there is 5V (as I sad it is allways even if there is no USB plugged)
<granite_crusher>
arokux1: olimex sells few versions generally the same; mine is without nand and wifi
<arokux1>
granite_crusher, ok, I see, no need to measure just now. you can try uncomment that line and see if storage still works. next: you need to setup rootfs on your usb-stick I think. having full fledged distro with package management will make things much easier.
\\Mr_C\\ has joined #linux-sunxi
<arokux1>
granite_crusher, hm.. have you actually got usb2ethernet working?!
<granite_crusher>
arokux1: do you mean this line? ret = regulator_enable(sunxi_ehci->vbus_reg);
<arokux1>
granite_crusher, yes, this is the line you need to comment out.
<oliv3r>
nvm the disks are really old n slow :p
<granite_crusher>
arokux1: I will try to do that at evening, now I have some science and academic buisness to do, see you
<arokux1>
granite_crusher, alright.
granite_crusher has quit [Quit: Page closed]
<Cubear>
*bump*
<Cubear>
I'm just reporting that the earlier issue I asked about about wifi device on USB0 not being detected at boot
hipboi has quit [Quit: Leaving]
<Cubear>
after editing and compiling the fex file it works fine
<Cubear>
gg
<oliv3r>
Cubear: \o/
<arokux1>
Cubear, what board do you have, btw?
<Cubear>
cubieboard A10
<Cubear>
I'm using wifi IEEE802.11 n on USB0 OTG port
<arokux1>
Cubear, what kernel are you running?
<Cubear>
3.4.61
<arokux1>
Cubear, do you have interest in testing mainline kernel?
<Cubear>
I might, the one I currently have is from linux-sunxi github repo
<Cubear>
needs to be stable for my project
<Cubear>
Oh, I also have a Cubieboard A20, but haven't touched it yet
<arokux1>
Cubear, sure. once you have time to play, you can come along and ask.
<Cubear>
I'll stick around in this channel
<arokux1>
Cubear, cool. currently I'm working on USB host (not OTG) for mainline. it works partially - storage works but wifi module fails. you could test with your external wifi or other usb devices, once you have some time.
kaspter has quit [Remote host closed the connection]
<Cubear>
yeah, I have a bunch of wifi devices lying around, used to do wardriving
<steev>
arokux1: oh that's what you need tested? i have shit tons of usb wifi devices
<Cubear>
Alfa AWUS036H ftw!
<steev>
that's actually how i use my cb2
<steev>
or at least
<steev>
i would if i could actually ssh into it
<steev>
dunno why it's the only device on the network i can't ssh into
<steev>
at least, on wifi
<Cubear>
speaking of networking, has anyone managed to setup a working L2 wlan0 <--> eth0 bridge on cubieboard?
<Cubear>
i.e. use cubie as a wifi access point
<Cubear>
without NAT
<Cubear>
I can set up the AP on my laptop using the USB wifi dongle and hostapd. It works fine, and I can access the internet from a client device.
<Cubear>
but when doing the same setup on cubie using the same wifi dongle it fails
<Cubear>
I've done some packet logging and it would appear that non-broadcast traffic cannot pass from eth0 side to clients on wlan0.
rz2k has joined #linux-sunxi
FDCX_ has quit [Remote host closed the connection]
<arokux1>
ideas the missing dma support can cause that -- but still not sure.
<arokux1>
steev, yes, it would be great if we could identify working usb devices. maybe it could give us a clue what is wrong. current state of affairs is the following: there are usb EHCI patches against sunxi-next. usb storage works, but usb wifi fails. I have a 230-line piece of code with which wifi works in the sunxi-3.4. *same* code won't work in -next, so now I actually suspect something is wrong somewhere else. specifically there've been
<arokux1>
Turl, maybe you could pick up bamvor 's work on the PLL6 clocks for sun6i and add it to the patches you'll submit?
<oliv3r>
arokux1: what specifically is he sharing? cause I think we have it allready
<mnemoc>
uhm.... Turl: oliv3r: please let me know when 3.10.12-r1 (backporting) is done to make a version jump =)
<arokux1>
oliv3r, lichee-3.3 and android for A31 (sun6i)
<oliv3r>
ahh the old a31 sdk
<oliv3r>
i think we have that too
<oliv3r>
i'm sure we allready have that
<Cubear>
arokux1, to list a few: poor signal, power issues, most devices I got didn't have an option for an external antenna, dropping connections/packet loss, etc
<mnemoc>
google has continued working in android-3.10 .... /me smells android 5.0
<arokux1>
mnemoc, I thought google publishes sources some time after the release of android?
<Turl>
mnemoc: I haven't backported anything yet, shame on me :)
Soru has joined #linux-sunxi
<mnemoc>
arokux1: android yes. kernel no
<mnemoc>
their kernel tree moves freely
<mnemoc>
and they are working in 3.10 since some weeks ago
<arokux1>
oliv3r, you are right. is there any newer sdk for A31?
<oliv3r>
arokux1: i don't know what kernel is in the newer sdk; but i'm expecting a new sdk drop with a51
<oliv3r>
Turl: have you commited your shiney new sunxi_defconfig?
<mnemoc>
oliv3r: officia name?
<arokux1>
mnemoc, will android 4.4 (aka kitkat) be still based on old kernel?
<arokux1>
oliv3r, ask you source :p
atiti has joined #linux-sunxi
<mnemoc>
i think it's still 3.4. to reduce vendor nightmares
atiti has quit [Remote host closed the connection]
<Turl>
oliv3r: no :p I didn't even write it
<arokux1>
mnemoc, or put it other way. android-4.4 != android-5, right?
<mnemoc>
sure
<mnemoc>
4.x vs 5.x
<arokux1>
Google announced Android 4.4 KitKat , The release had long been expected to be numbered 5.0 and called 'Key Lime Pie'.
<Turl>
oliv3r: if you backport sth please make a wiki page and list the stuff :)
<mripard>
again, android is a userspace, it doesn't require any particular kernel version
* Turl
still doesn't know what kitkats taste like
<mripard>
Turl: hmmm, why ?
<mnemoc>
arokux1: they will make the major version jump when there is something important changed in the platform
<mripard>
the sid isn't merged?
boycottg00gle has joined #linux-sunxi
<mripard>
Turl: you don't ? :(
<mnemoc>
check the kernel versions used i the previous N.x
<Turl>
mripard: I don't see it on torvalds/master
<mnemoc>
for the same N, kernel remains
<Turl>
mripard: no, nestle doesn't sell them in here :|
<n01>
Turl: sad life
<n01>
I hope you still have nutella
<Turl>
we got mantecol to beat all your kitkats though :)
<mripard>
Turl: hmmm, if we have to send something to argentina, I might placed some in it then :)
<Turl>
n01: we do, but it's a 'rare' product
<Turl>
n01: never bought a jar of it either tbh
<Turl>
mripard: :)
atiti has joined #linux-sunxi
<mnemoc>
mantecol beats kitkat 100 times
<oliv3r>
i luv kitkat
<oliv3r>
kitkat white, kitkat chunky
<oliv3r>
mmmm kitkat
<arokux1>
mripard, while you are here.. I'd like to remember you to cc linux-sunxi.
<oliv3r>
android always makes me hungry :(
<mnemoc>
you don't know mantecol :p
<Turl>
mnemoc: +1
<oliv3r>
Turl: didn't write it? you where hacking it just yesterday!
<Turl>
oliv3r: I wasn't
<n01>
linux-sunxi exchange? like on reddit :)
<Turl>
oliv3r: I was just critizising the one mripard sent upstream :p
FDCX has joined #linux-sunxi
<arokux1>
mripard, also small update on usb part: my hard work yesterday has given indications that my driver might do everything correctly and the problem is elsewhere. not 100% sure, though.
lzg_ has quit [Quit: Ex-Chat]
<mripard>
arokux1: ok. while you are here, I still don't get why you want to impose this :)
<mripard>
arokux1: cool :)
<arokux1>
mripard, it is much easier to follow everything sunxi related...
<mripard>
Turl: yeah, you like to critizise, but I don't have much things to critisize from you lately :)
<mripard>
arokux1: you won't have all of the patches anyway. I don't even have all of them
<Turl>
mripard: :) you should get some mail one of these days
<mripard>
if you want to be aware of all the changes, subscribe to lkml.
<arokux1>
mripard, mainline work is posted against all the possible mailing lists, so difficult to follow all..
<Turl>
maybe later today, who knows :)
<mripard>
then follow lkml
<Turl>
mripard: I never cc lkml :p
<arokux1>
mripard, is it difficult to add yet another CC, I cannot understand your arguments...?
<mripard>
Turl: bad bad bad :)
<Turl>
mripard: I sent you a tested-by and a comment last night
<oliv3r>
Turl: ohh one actually went upstream; good then you can submit that ;)
<Turl>
oliv3r: I'm waiting for v2 :p
<oliv3r>
mantecol looks very very yummie, sounds like a menthol cigarette
<mnemoc>
it's dry fat :p
<mnemoc>
but amazing taste
<oliv3r>
dry fat?
<oliv3r>
oh thought it was nougat wit pralinee
<oliv3r>
covered in white choco
<oliv3r>
hmm, kitkat <3
<Turl>
it's peanut
<oliv3r>
ohh snickers on steroids
<oliv3r>
Turl: possible title for wiki page?
<oliv3r>
stable_kernel_backport or something?
<mripard>
arokux1: my point is that you won't be able to see *all* the patches anyway.
<oliv3r>
i only backported SID btw :p
gzamboni has quit [Ping timeout: 264 seconds]
soldoKyn has quit [Quit: Konversation terminated!]
<mripard>
and I find it hard to believe that you want some patches to sort some Kconfig options alphabetically
<oliv3r>
Turl: but it'll only have 'sid' on it :p
<oliv3r>
fine
<oliv3r>
done :D
rz2k has quit [Write error: Broken pipe]
<Turl>
mripard: I'm surprised nobody came to blame you for making unnecesary churn on the kernel :p
<arokux1>
mripard, Kconfig isn't so interesting, but other things are. ok. if you say it is not feasible to cc linux-sunxi, then can you please give a recommendation of how to follow sunxi related patches against mainline? one possibility would be to subscribe to lkml and add filter on "sunxi", also I'd need to subscribe to the mailing lists Turl send patches too.. is there any better option?
Cubear has quit [Quit: Leaving]
<Turl>
arokux1: I'd say lakml has all the things
<oliv3r>
crapster
<arokux1>
Turl, not your stuff :p
<oliv3r>
how long is the PHY/mdio timeout probe?
<oliv3r>
oh nvm
<oliv3r>
i think console should be just after
<oliv3r>
Turl: have you testedy our initramfs v3?
<Turl>
oliv3r: yeah I used it for quite a while, why?
ganbold_ has joined #linux-sunxi
<arokux1>
oliv3r, is that initramfs v3 for our stepping stone 3.10?
<steev>
arokux1: tbh, usb isn't my forte, i can plug in devices and tell you, yes or no, but i *hate* the way it's done in mainline now - mostly because it's taken so fucking long and still not up to part for what i need to re-add efikamx support back
<steev>
s/part/par/
<mripard>
Turl: unneccessary?
<steev>
speaking of usb wifi... what's the deal with 3 different broadcom wifi drivers?
<mripard>
Turl: arokux1: lakml mostly have everything yes.
<Turl>
mripard: often when people sort stuff these people come up out of nowhere :)
<n01>
steev: some from bcm, some open, some RE
<arokux1>
steev, yes, sticking the (EHCI) devices in and telling if they work would be enough already. you could stick OHCI-only too, just to see what happens. "the way it's done" -- what exactly?
Gerwin_J has joined #linux-sunxi
<Turl>
bbl
<steev>
n01: yes, but, why - if a driver doesn't work, one should *probably* use backports, OR, the defconfig shouldn't enable 2 different ones, OR remove the exports from the RE/bcm ones that aren't from the kernel
<drachensun>
turl: arokux1: I wasn't told of any restrictions on the SDK when is provided and at least one other reseller has already posted it all, several months ago
<steev>
arokux1: it's a long story, and i don't want to get into it, but basically, the chipidea stuff (which IS a good thing) is what had held us back from mainlining EfikaMX support when I was at genesi. we were pushing mainline support for a very long time, and all of a sudden the maintainers of the imx stuff were like, nope, not gonna accept til <arbitrary
<steev>
thing happens> - and we got sick of waiting since the goal posts kept moving, and now the efikamx is considered ancient
<arokux1>
thanks drachensun
<n01>
steev: I think that not every driver works well with every card
<steev>
n01: you still can't export things multiple times, and the defconfig enables at least 2 of them
<arokux1>
steev, I see. let me know you have some time for trying usb devices you have.
<steev>
arokux1: just build the 3.10 stuff?
<oliv3r>
Turl: i keep failing to get a serial console with v3; using the same config settings i've always used, multi_v7_defconfig + debug + SD
<oliv3r>
SID*
<steev>
oliv3r: modify the inittab?
<oliv3r>
arokux1: no, but good point, a nice default working initramfs for testing purpouses would be nice to have
<steev>
isn't that what the miniroot thing was supposed to be about?
<steev>
arokux1: i'm *very* familiar with u-boot and kernel loading and such
<arokux1>
steev, re: why not bootz the zImage?: dunno, it was like this when I started hacking,..
<oliv3r>
steev: because the older versions of u-boot didn't liek zImage, but turned out to be ar ather simple change, so now they support all, we never went back
<oliv3r>
so u-boot CAN boot it all now, but updating muscle memory and wiki's takes time :)
<oliv3r>
and 'why not' :p
<steev>
oliv3r: extra step, shave some time off yo
<arokux1>
steev, can we get some advantages over uImage?
<steev>
at least, for why not
<n01>
steev: also with zImage u need fdt right?
<arokux1>
steev, I can update wiki etc, if "bootz the zImage" has some advantages.
<arokux1>
n01, fdt is needed anyway, or?
<oliv3r>
steev: extra step for the compiler?
<oliv3r>
steev: shave like 0.1 ms off
<steev>
damn it guys
<n01>
with zImage don't you have to decompress the kernel also?
<steev>
stop saying my name so much
<n01>
lol
<steev>
it sets off the notification to my phone
merbanan has quit [Ping timeout: 245 seconds]
<steev>
which causes it to vibrate in my pocket
<steev>
but anyway
<steev>
you only need fdt if you have a kernel with fdt. uImage does some stuff
<steev>
like crc and whatnot, iirc
<steev>
but, zImage seems to do that too these days
<steev>
at least, my sabresd bitches about the 3.12 kernel i've been trying to build for it failing the crc check
<n01>
we agree on that, but it turns out that we have fdt
<oliv3r>
steev: hi
* arokux1
subscribed to lkml and wonders how long will it take for the first e-mail to appear
<steev>
I would never sub to lkml. If it's important people will CC me
<oliv3r>
lkml is way to busy
<n01>
arokux1: better lakml
<arokux1>
steev, I want to follow sunxi stuff. mripard says it is the best way to do it...
<arokux1>
n01, I'm there already..
<n01>
arokux1: lakml != lkml
<n01>
hu ok
<arokux1>
here it is! first e-mail
<oliv3r>
*checking*
<oliv3r>
did you cc sunxi?
eebrah has quit [Ping timeout: 246 seconds]
<arokux1>
oliv3r, ? I've just got first e-mail from lkml, not actually send smth :)
<arokux1>
sent*
<steev>
Yeah. Laml is what I sub to, and even then idgaf about 98% of the emails
<arokux1>
steev, I put a filter to archive everything, unless there is "sunxi" in it.
<oliv3r>
Turl: also, when trying your v2 cpio, (and v3 i'm sure too) I keep getting wrong ramdisk image format :S
<steev>
You may want to rethink that, maybe sun*i
<oliv3r>
so i'll just build it in again ;)
<oliv3r>
arokux1: ohh; yeah lkml is like 300+ mails per day
<arokux1>
steev, gmail isn't so smart
<oliv3r>
ifndef sunxi rm -r mail :)
Guest67810 has joined #linux-sunxi
naobsd has joined #linux-sunxi
atiti has quit [Ping timeout: 252 seconds]
<oliv3r>
i frackin' don't understand this shit anymore
<arokux1>
steev, rephrasing... no need to switch to zImage, since no advantages?
<oliv3r>
a few weeks ago i was building kernels, testing them like crazy
<oliv3r>
now i can't even boot one?
<arokux1>
oliv3r, they change, you know.
<steev>
Tool chain change?
<oliv3r>
lol
<oliv3r>
arokux1: if it werne't for the stupid fact, that i'm trying to boot the same kernel, from the same branch i used; i haven't changed anything on this branch since it was accepted on mainline :)
<oliv3r>
steev: nope same compiler
<oliv3r>
now i'm getting: Error: unrecognized/unsupported machine ID (r1 = 0x00001008).
<oliv3r>
oh wait
<steev>
Pass machid
<arokux1>
oliv3r, hm... i know the reason!
Guest67810 is now known as eebrah_
<steev>
Or remove it
<oliv3r>
nope, ARCH_SUNXI is set
<arokux1>
oliv3r, you have built-in initramfs, right?!
<oliv3r>
ah, good point
<oliv3r>
relocation bug
<arokux1>
oliv3r, what point?
<arokux1>
oliv3r, yep.. fdt_high...
<oliv3r>
different bootloader introduced the relocation point
<oliv3r>
bug
<oliv3r>
but
<oliv3r>
the thing is
<arokux1>
hno, fix it please!! ^
<oliv3r>
i always booted this cpio :)
<oliv3r>
arokux1: hno is a little busy the next 2 months
<arokux1>
oliv3r, I know, nevertheless :p
<oliv3r>
i don't understand what changed though; the rootfs is the same size as before ..
<arokux1>
oliv3r, try to set and see if it helps: env set fdt_high ffffffff
<oliv3r>
Oh maybe i accidentaly disabled compression
<arokux1>
oliv3r, if yes, you'll think why
<oliv3r>
finally, console :D
<oliv3r>
htings work agian :p
<oliv3r>
with rootfs v1
<arokux1>
oliv3r, fdt_high helped?
<n01>
hoo hno what was the big news you are workin' on?
<n01>
is it still a secret?
<arokux1>
n01, backread he unveiled some details
<n01>
do not have bnc (yet)
<arokux1>
n01, check topic for logs
<n01>
-.-
<arokux1>
n01, where do i find the meaning of all those smileys? :)
<steev>
arokux1: so i should use multi_v7_defconfig with your kernel?
<arokux1>
steev, you can use what ever you want actually just be sure to enable everything needed for sunxi and enable EHCI HCD + sunXi EHCI.
<arokux1>
steev, multi has a lot of things you do not need i.e. all kind of different arm socs
<steev>
arokux1: i prefer to start from a defconfig, since a defconfig should have everything enabled to at least get the device to some sort of prompt
<arokux1>
steev, yep, then go for multi.
<steev>
it's compiling now
<arokux1>
steev, what is your -j? :)
<steev>
arokux1: -j8
<steev>
it will take about 10 minutes
<steev>
i could do 16 but i don't really notice that much of a difference
<arokux1>
steev, far more less. several minutes for -j5
<steev>
i need to get a new laptop
<arokux1>
steev, ah, laptop, then maybe 10..
tinti has quit [Ping timeout: 240 seconds]
<steev>
let me kill the chromium compile
<steev>
this laptop is a couple years old, once i move, i think i'll get a new one, but this one was free, so i can't really complain
<steev>
real2m59.247s
<steev>
okay so it MAY be less than 10 minutes
<steev>
hno: btw, what's with the cb2 not actually powering off on... uh, halt/poweroff
<arokux1>
steev: no support for AXP in the mainline?
<steev>
arokux1: eh?
<steev>
not talking about mainline
<arokux1>
steev: ok
<steev>
not powering off on halt is a u-boot thing
Gerwin_J has quit [Quit: Gerwin_J]
<arokux1>
steev: so U-Boot lives even after kernel has started after shuts down?
<steev>
arokux1: u-boot is the same as a bios, for lack of a better term
<steev>
whoops
<n01>
libv ssvb can I ask you something in pvt about drm/kms/fb? (you are the graphic layer guru :)
<steev>
forgot to add the dtb
<steev>
3571712 bytes read in 165 ms (20.6 MiB/s) <-- zImage 4466032 bytes read in 205 ms (20.8 MiB/s) <-- uImage
<ssvb>
n01: I'm not a guru, just trying to apply common sense where appropriate :)
<n01>
because you are parsing the header
hipboi has joined #linux-sunxi
<steev>
n01: correct
<steev>
ssvb: pfft, you're modest as fuck too
tinti has joined #linux-sunxi
<steev>
arokux1: this is supposed to boot all the way? it stops here at rtc
<steev>
[ 0.892850] drivers/rtc/hctosys.c: unable to open rtc device (rtc0)
<n01>
ok I'll ask here, f* the OT :) I have this board with renesas CPU an a video display controller called vdc5. We are not interested in X, just plain fb (we will work with openVG). Now, it is better to write the driver on fbdev or trying DRM/KMS?
<arokux1>
steev: which board are you on?
<steev>
cubieboard2
<n01>
ssvb ^^
<arokux1>
steev: my tree worked for Turl
<arokux1>
he has cb2, too
<steev>
n01: drm/kms
<steev>
arokux1: hm
* steev
pokes Turl
<n01>
steev: it looks painful
<steev>
n01: i don't touch graphics, too much maths
<n01>
ok, so why drm/kms?
<arokux1>
steev: so you just taken multi config?
<steev>
arokux1: i'll make menuconfig and... see if i can figure out what's missing
<arokux1>
steev: what about rootfs?
<steev>
arokux1: yeah. i'm using same rootfs
<steev>
one momento
<arokux1>
steev: where is it? built into the image?
<arokux1>
warzon: I'm not sure about the absence of hf at your toolchain prefix too.
<warzon>
i'm sure it can't be the case
<arokux1>
Tsvetan: there? it is probably a good idea to use git tags in your blog posts.
<warzon>
arokux1: i think that new wiki will help
<Tsvetan>
arokux1: sure we do
<Tsvetan>
DIT: After some discussion with Henrik on irc channel he made git TAGs on uboot and kernel which snapshot the GitHub repositories to match what I wrote above, this means even there are new changes to the GitHub repositories, if you follow the steps above and something is broken you can do after the # git clone …
<Tsvetan>
for u-boot
<Tsvetan>
for kernel
<Tsvetan>
# git checkout v2013.07-rc1-sunxi
<Tsvetan>
# git checkout sunxi-v3.0.76-r0
<Tsvetan>
these commands will point GitHub to same snapshot we used in the posting to build uboot and kernel
<Tsvetan>
Mali has developer forums where bugs can be reported. However, our experience when reporting bugs was “yes, we know this is broken” followed by a lack of replies. Even Google seems to be working around Mali issues in Chromium, suggesting that even they probably cannot get ARM to fix their bugs or implement missing features advertised as supported.
<Tsvetan>
duh :-)
<arokux1>
uhm, so even competition will not force these guys to fix their bugs?!
<Tsvetan>
in 2015 Samsung will have same technology as Intel
<Tsvetan>
who could imagine this is possible few years ago?
<geecko>
Tsvetan, with a toy user interface on top ;)
<Tsvetan>
geecko without this end user product with higher margin Samsung would not have the money to invest in this technology
agath_at_work has joined #linux-sunxi
<agath_at_work>
Hello! I am trying to use the watchdog with the Allwinner A20 (cubieboard2) but even if I periodically kick it, it restarts the system as if it didn't receive the keepalive.
<agath_at_work>
I tried also through systemd, the result is the same. Is there a known bug?
<arokux1>
agath_at_work: (just curious: what kernel is it?)
geecko has quit [Quit: Quitte]
<n01>
agath_at_work: wdt in known to work, how are you kicking it?
<arokux1>
n01: I wonder why its name is wdt?
<arokux1>
watchdog timer -- I see.
<n01>
yep
CaCtus492 has quit [Ping timeout: 248 seconds]
CaCtus491 has joined #linux-sunxi
<n01>
agath_at_work: btw, which kernel are you using?
<arokux1>
warzon: did it help?
<agath_at_work>
it's a 3.3.0 (Arch Linux distribution)
<warzon>
yes, thanks a lot
<warzon>
thanks a lot for help!
<agath_at_work>
I tried to kick it manually and through systemd
<agath_at_work>
it's the first time I work with an hardware watchdog, I read on a doc that /dev/watchdog starts working when you open a file handle, and then you can throw in any character (any but V) to kick it... so I tried with cat >/dev/watchdog
<agath_at_work>
however I tried to enable the watchdog in systemd, and the result is the same...
<arokux1>
agath_at_work: are you sure you want to use this old kernel?
<n01>
agath_at_work: using a wdt in that way is tricky. I used to test my driver using the command "watchdog" that you can compile with busybox
<n01>
are you using busybox?
<agath_at_work>
I am using the Arch Linux distro on a Cubieboard2....
<agath_at_work>
I may try to compile a new kernel if the problem may be that
<n01>
dunno why arch linux uses a kernel so weird
<arokux1>
they find a difficult to keep pace with us
<n01>
arokux1: we are not reliable
<arokux1>
n01: reliable?
<n01>
agath_at_work: try to use the watchdog command to kick the wdt
<agath_at_work>
ok I will try with it
<n01>
arokux1: they need a stable kernel, mainline kernel is not stable at all
<WarheadsSE>
you know, you all could just ping me.
<WarheadsSE>
I can't follow the pile all by myself, occasionally i need someone to rell me when you have an architecture merged into the main sunxi branches
<arokux1>
WarheadsSE: btw, are there instructions how to build an arch rootfs from scratch?
<n01>
arokux1: why don't you use buildroot?
<WarheadsSE>
arokux1: build from scracth... ?
<arokux1>
n01: why should I? I want full fledged distro. and Arch is the best distro out there...
<arokux1>
WarheadsSE: yep, what scripts do you use to create you rootfs.tar.gz?
<n01>
lol I know, I'm an arch user myself
<arokux1>
n01: nice to meet you :)
<WarheadsSE>
pacstrap ?
<WarheadsSE>
same way you do with an x86, just we do those little steps for you
ganbold_ has quit [Remote host closed the connection]
<WarheadsSE>
arokux1: seriously
<WarheadsSE>
pacstrap if you want to make a fresh rootfs yourself
<WarheadsSE>
boot an image
<WarheadsSE>
then strap
<arokux1>
WarheadsSE: I understand that, but you surely have you scripts that do everything, I want just to put several additional packages and some config.
<WarheadsSE>
nope
<WarheadsSE>
we really just pacstrap && install packages
<WarheadsSE>
well, and enable dhcp & ssh
<WarheadsSE>
there are no scripts mde for that
<arokux1>
WarheadsSE: and then you put kernel&friends manually?
massi has quit [Quit: Sto andando via]
ykchavan has quit [Ping timeout: 240 seconds]
<WarheadsSE>
mount, place
<WarheadsSE>
the kernel package hands "kernel and friends"
<arokux1>
WarheadsSE: so you actually pacstrap on those tiny arm boxes?!
<WarheadsSE>
yes
<WarheadsSE>
its not that hard.
hipboi has quit [Quit: Leaving]
<arokux1>
thanks WarheadsSE now I understood everything. I think I'll take time and update your kernel pckgs
<arokux1>
WarheadsSE: do you have any objections if I use the latest sunxi-3.4 tag?
<arokux1>
it is supposed to be stable by our standards...
ZetaNeta has joined #linux-sunxi
<arokux1>
WarheadsSE: tag - sunxi-v3.4.61-r0
book` has quit [Ping timeout: 245 seconds]
<WarheadsSE>
none, test it
<arokux1>
WarheadsSE: alright!
atiti has joined #linux-sunxi
<WarheadsSE>
the build system will puke error messages directly at _you_ if it has issues building
panda84kde has quit [Quit: Konversation terminated!]
<arokux2>
steev, will you have a chance to test some usb stuff?
<steev>
at some point
<steev>
although, if you were to say, hook me up with an image so i don't have to dick around with an initramfs myself.....
<steev>
but i'm planning on going to bed here at some point soon
<steev>
just waiting on a reply from my "boss" but i don't think i will very soon, since he's on his way to derbycon
deasy has joined #linux-sunxi
<arokux2>
steev, (sorry do not know your timezone) you mean if I could provide you a kernel image with initramfs built-in?
Sonicadvance1 has joined #linux-sunxi
<Sonicadvance1>
what
<arokux2>
steev, the problem is I do not know which usb devices you have... you'd need to enable them in the kernel config. but I can provide you initramfs based on buildroot.
<binaryferret>
Is it ok to ask u-boot questions here? or does that have its own chan?
<arokux2>
binaryferret, sunxi related - yes, and yes u-boot has its own chan
<steev>
arokux2: the initramfs i can just shove in the kernel?
<arokux2>
steev, yep
<steev>
arokux2: honestly, if it's a usb wifi device... enable it, i have it ;)
<steev>
except the bcm crap
<binaryferret>
It's sunxi related. I have a cheapo chinese a13 tablet. I'm wondering which config to use when calling make. What information from the tablet would be most useful?
<arokux2>
binaryferret, did you do any search on our wiki?
<arokux2>
steev, do not understand you last msg
<steev>
arokux2: i meant i dont have any usb broadcom cards
<binaryferret>
arokux2: Yes, it's been the primary source of my information, very useful. There are a number of a13 choices, along with sun5i
<binaryferret>
I have not looked at that page arokux2. Will read now. Thank y ou.
<steev>
i've got rt2500, rt73, rt2800, rt3070 (pretty much same as a 2800), ath9k_htc, carl9170, p54, rtl8187, and a few others
<steev>
actually, i don't think i have an orinoco usb
<arokux2>
steev, so you want me to enable them in the kernel and provide you an image?
<steev>
but yeah, anywho
<steev>
nah
<steev>
just gimme a rootfs i can build into the kernel
<steev>
i'll probably want to make a few changes to the kernel config itself
<arokux2>
steev, do you need firmware for those?
<steev>
mmm, yeah
<steev>
some of them at least - ath9k's definitely, same with rt2x00
<binaryferret>
I've already retrieved the script.bin along with converted it back to a .fex file. Will try and use that useful resource to get bootinfo. Thanks arokux2
<arokux2>
binaryferret, cool. please improve wiki if you do some additional finding.
<arokux2>
steev, do you know some good place where I can upload a cpio?
<binaryferret>
Okdokey. I've been trying to just get linux running on it, but had mixed results. So i thought I'd take a few steps back and make sure that I have the right u-boot built.
<arokux2>
fill the option General Setup / Initramfs source file(s) with a path to this cpio
<arokux2>
if initramfs is built into the uImage, you just use "-" as a placeholder for its address. Taking all this into account your bootm might look like
<arokux2>
bootm 0x46000000 - 0x49000000
<arokux2>
steev, ^
<steev>
yah
<oliv3r>
Turl: how can I see the diff between v1, v2 and v3 of yer rootfs?
<oliv3r>
datestamp of /root/.bashrc is identical
<arokux2>
oliv3r, may I ask what is this rootfs for?
<arokux2>
naobsd, hi, you there? I have a problem with my USB driver. storage works, but USB WiFi won't. can you try if some USB WiFi module works in *BSD you are using? I do not know if *BSDs have support for wifi..
<oliv3r>
arokux2: still backreading; but 3.4 kernel's module acceptace is quite relaxed. So as long as it's not wrong, it can get it
<oliv3r>
in*
<arokux2>
oliv3r, module acceptance?
<arokux2>
WarheadsSE, what is buildarch=4? should I touch it?
<steev>
arokux2: oh, for the record, i'm in texas, but i work from home and work whatever hours i want, since i was up all night working on stuff, i go to bed soon
<arokux2>
steev, ok :)
<steev>
well, within reason
<steev>
currently working with Pwnie Express, so i do try to be around whenever they are in the office
<arokux2>
WarheadsSE, how do you think should I update .config there are *lots* of changes, should I go one by one?
tinti has quit [Ping timeout: 240 seconds]
<WarheadsSE>
arokux2: did you read the readme I told you to read?
<WarheadsSE>
apparently not
<WarheadsSE>
steev: nice
<WarheadsSE>
is that what you wanted _that_ for?
<arokux2>
WarheadsSE, oops, sorry, didn't thought there is some info related to it.
hugobo has quit [Quit: Page closed]
<WarheadsSE>
hence why i told you to read it
<WarheadsSE>
and I would expect you to update the config since it is a version jump.
hugobo_ has joined #linux-sunxi
<arokux2>
WarheadsSE, it is a HUGE leap. :)
<WarheadsSE>
even more impetus.
<WarheadsSE>
you ask me so many obvious questions :P
<hugobo_>
On A13 CPU, when booting from internal NAND flash, at which address the BROM looks for a boot0 program header?
<hugobo_>
(I am assuming that boot0 is NOT part of the BROM)
<arokux2>
hugobo_, consult our wiki
<arokux2>
WarheadsSE, I do not know exactly how should I approach this problem. it is a jump from 3.0 to 3.4. I can update sunxi related config but there sure were some upstream changes... what do you think is the best way to update config?
<WarheadsSE>
arokux2: there aren't that many non-core modules that can "break like crazy"
<WarheadsSE>
just make oldconfig
<aep>
arokux2: if i intentionally skip a <> part, same error. i guess its not more fine grained than "parse error"
on1x has quit [Ping timeout: 245 seconds]
<hugobo_>
My assumption is that boot0 is located BEFORE the 4 MBR copies...
on1x has joined #linux-sunxi
<arokux2>
hugobo_, why do you care?
<hugobo_>
Because I want to use the internal NAND to store my U-Boot, kernel and filesystem on it!
<hugobo_>
But I don't want all that android crap :)
<hugobo_>
and boot0 that calls boot1 that calls U-boot SPL that calls U-boot :)
<hugobo_>
It seems much simpler on the SD card, BROM calls U-boot SPL which calls U-Boot...
<aep>
arokux2: actually, if i intentionally break the syntax, its slightly different. with valid syntax it says : invalid gpio pin : -1
FR^2 has quit [Quit: Connection reset by peer]
<arokux2>
hugobo_, hno must know the answer, but he's not here
<hugobo_>
arokux2: Ok, I am starting to have a good understanding of the boot process, it just seems less complicated than I originally tought, but I find that the documentation is lacking about using a custom layout for the internal NAND flash. Anyway, thank you
<hugobo_>
I will try to ask hno
<hugobo_>
when he is here
<arokux2>
hugobo_, hno is busy these days at work :(
<arokux2>
hugobo_, please improve our wiki, if you know more than is available there.
<hugobo_>
Ok, then I think I will need a little more reverse engineering then!
eebrah_ is now known as eebrah
<aep>
bah found it. a random character slipped into the [gpio_para] section header
naobsd has quit [Quit: Page closed]
<aep>
and the error is totally unrelated. its printed anyway, even with a valid gpio config
<arokux2>
aep, hm.. maybe you could fix it and submit a patch?
<aep>
hm yeah, lets check
<binaryferret>
arokux2: Once I have the script.fex etc what should I be looking for to determine which u-boot config to use? there are a few a13 ones a13_mid and sun5i
<binaryferret>
Oops sorry was just afk. Will try and get the bootinfo, having issues building the sunxi tools atm. Just having a play
vinifr has joined #linux-sunxi
vinifr has quit [Client Quit]
<binaryferret>
I have script.bin/fex and a bunch of other info from the board. Going to try the other options for grabbing the boot info as that link gives three different paths.
<arokux2>
wow, drachensun cool. Ubuntu Touch is freely available?
<drachensun>
IANAL
<drachensun>
:-)
<drachensun>
All the software is GPL I dont really see how it could be a problem
<drachensun>
but they say "free for non commercial use"
<arokux2>
drachensun, no, I mean if it can be downloaded already, I thought they wanted to publish it later.
n01 has quit [Read error: Connection reset by peer]
<drachensun>
well I don't plan to sell it, just maybe make it available for download after purchase in a known good configuration
<mnemoc>
arokux2: in 3.4 we don't want to make intrusive style changes as they just conflict with the goal of unification and integration with newer source drops from allwinner (more difficult to see the real diff). 3.10 otoh is expected to take already unified 3.4 drivers, DTS them, migrate them to common frameworks, clean the style, and iteratibely reach mainline quality and get submited to the gods
<arokux2>
mnemoc, I see. my question was about the quality of the patches. should we be strict on that?
<drachensun>
my xchat client can't see certain letters
<mnemoc>
arokux2: I just arrived "home".... haven't looked at the ML yet
<drachensun>
unless they are in the chat window, bizarre
<mnemoc>
arokux2: new code needs to be strict regarding quality
<arokux2>
drachensun, send me some of them, if I can see.
<drachensun>
some of what?
<mnemoc>
arokux2: even in 3.0. any line which gets modified or added needs to comply. but the lines without code changes should not be "fixed"
<mnemoc>
because that adds noise to the diff
<mnemoc>
and makes comparison harder
<mnemoc>
arokux2: which patch in particular are you talking about?
<ykchavan>
binaryferret, matching script.bin should be big task.
<mnemoc>
thunderbird sloooooowly getting the new mails
<binaryferret>
ykchavan: Just in case you're interested. I noticed my system was using certain binary files (as, ld) from /usr/local/bin and these were older versions
<binaryferret>
of the ones located in /usr/bin
<binaryferret>
so i just copied them over. Fixed the issue and was able to build sunxi-tools
<mnemoc>
arokux2: thanks! 3528 mails to go...
<arokux2>
mnemoc, are you really sure you want to read them? I think ppl do not care if they haven't got a reply for a month already :p
<ykchavan>
binaryferret, nice to know
<arokux2>
mnemoc, mark them all read and start some real work.
<mnemoc>
arokux2: I need to find pending issues and outstnding patches....
<arokux2>
mnemoc, they will find you! do not worry. if there is a bug or unapplied patch affected person will complain again!
<mnemoc>
arokux2: true
<arokux2>
mnemoc, so do it now! "Mark all as read" and liberate yourself.
<arokux2>
:)
<mnemoc>
:p
<Turl>
oliv3r: ping :p
<Turl>
drachensun: ping
rz2k has joined #linux-sunxi
<binaryferret>
some of the board config names have things like _fel, _sdcon _uart1 _1gb after them. What do they denote?
<binaryferret>
that's from u-boot sunxi boards
<binaryferret>
_mid is another
<binaryferret>
says fel is for usb booting in the wiki doc couldn't find a reference for the others.
<Turl>
binaryferret: _FEL is to boot via FEL
<Turl>
binaryferret: _sdcon is for uart via sd, but iirc it's only on the ancient lichee uboot?
<Turl>
binaryferret: _uart1 must be to output on uart1 ;)
<Turl>
binaryferret: _1gb is for the _1gb ram version of whatever
<Turl>
mid means media internet device
<Turl>
it's part of the thing's name
<binaryferret>
Thanks Turl.
<arokux2>
binaryferret, maybe you could dump this info on the wiki?
<binaryferret>
Is it? doh, I could only find info on _FEL. although I did skim read it
<binaryferret>
will have another look.
<binaryferret>
Actually I need to sleep, I keep vanishing to play with sunxi tools etc and raise my head and another few hours have past. Thanks for the help today chaps.
ykchavan has quit [Quit: Leaving]
pitillo has joined #linux-sunxi
<nedko>
rz2k: how well mtd driver works?
<rz2k>
nedko: hi, well... not really well :p
<rz2k>
it works
<rz2k>
but there are problems with ECC
<rz2k>
UBIFS goes nuts sometimes
Tartarus_ has joined #linux-sunxi
<rz2k>
use the block emulation layer if you want to run the board 24/7
<rz2k>
also, all the info about MTD is at mailing list
<rz2k>
I've posted all the results around
<nedko>
rz2k: the mtd device exposes the whole nand device, correct? including the part where boot0 and boot1 are usually written?
<nedko>
yes i read some of the nand1k block device code
<rz2k>
there's also a problem with last blocks
<rz2k>
for some reason, bad block detection works sort of bad
<rz2k>
and doesnt detect badblocks at the end of nand
<rz2k>
I had proven that if I enable last 8 blocks - UBIFS dies after each reboot, if I remove that last blocks - everything works (didnt test for 1K reboot cycles like Yuq did)
<arokux2>
rz2k, why it is so mysterious? no doc?
<arokux2>
mnemoc, where should I send my key?
<arokux2>
Turl, his personal e-mail? ^
<rz2k>
arokux2: 1) no docs except the block emulation layer driver 2) nand controller itself is sort of weird
<arokux2>
rz2k, this controller is on SoC, right?
<rz2k>
yes
<rz2k>
for example, you can only read data in 1K blocks
<rz2k>
if you request block less or more, DMA will never finish
<rz2k>
or you can read by one byte
<rz2k>
so, to read 8192+640 nand
<rz2k>
you need 9 dma operations
<Turl>
arokux2: /msg it to him? :)
<rz2k>
i dont mention weird address system and conversions around
<arokux2>
rz2k, and so the aw's code has all this weirdness?
<rz2k>
yes
<Turl>
arokux2: remember he just needs the public part (id_rsa.pub usually)
<Turl>
it's a text file you can open and copy
<rz2k>
MTD driver is 1:1 to the nand_phys_read and nand_phys_write files found at sunxi_nand/nfc
<arokux2>
Turl, for some reason I've thought it is more than that, but yes, only several hundred symbols :)
<arokux2>
rz2k, did you take a look if things improved for A31?
<rz2k>
all allwinner devices have the same nand controller
<rz2k>
iirc, even sun3i has exactly same register layout
<rz2k>
just the emulation layer of older version
<rz2k>
sun67i have emulation layer of newer version that sun45i, so you need nand-part of newer version
<rz2k>
s/that/than/
<arokux2>
rz2k, this emulation is done in hw? (I'm not sure I understand..)
<rz2k>
no
<rz2k>
emulation layer is /drivers/block/sunxi_nand/ driver
<rz2k>
that exposes you /dev/nand device in block form
<arokux2>
rz2k, ah, ok.
<arokux2>
rz2k, what is the nature of the nand controller then? character device?
<rz2k>
special MTD device, check /drivers/mtd. there is a framework for working with NAND flashes. to this framework you can connect your NAND controller driver that will expose the nand controller to the framework. On the filesystem side - you can select out of special MTD filesystems
<rz2k>
jffs, jffs2, yffs, UBIFS
<rz2k>
they all have their own tools to work with the mtd devices
<arokux2>
rz2k, hm, interesting. why do you think allwinner did not go for a mtd driver then?
<rz2k>
because it is harder to implement
<rz2k>
and also because in sun3i times when allwinner wrote this driver
<rz2k>
well...
<arokux2>
rz2k, but an mtd driver does have advantages as I read on wiki " there is detection of bad blocks and of wear leveling."
<arokux2>
rz2k, is that correct?
<rz2k>
mtd in 2.6.3x that was recommended by google for android 1.x, mtd there had huge gap between the NAND flashes around the market and the ones mtd did actually support
<rz2k>
yes it has those
<rz2k>
wear leveling is done by your mtd filesystem like UBIFS
<rz2k>
on the block emulation layer in /drivers/sunxi_nand/ wear leveling is done kernel side
<rz2k>
i have no idea how they store the information needed for it to work well enough
<rz2k>
or if it does actually work there
<rz2k>
as I said, there is something wrong with nand controller itself in terms of bad block detection, mtd framework has its own implementation of badblock scanning, BUT, for some reason, it does not work well enough to detect everything, like I said - bad blocks at the end of nand for some reason are detected as good
<rz2k>
check the links I posted to nedko, they have some explanations by Yuq and hno