<hramrach>
but for now usable static defconfig should do
<ssvb>
mnemoc: sorry for dropping the ball, I have no problems using local defconfig patches, so improving default defconfigs was kinda nice to have but definitely not critical
<mnemoc>
I use the defconfigs for test building after merging commits. and also share pre-builts
<hramrach>
I just use local config
<hramrach>
because the defconfigs had no drivers
<hramrach>
also on 3.4 teh usb configuration is somehow broken
<mnemoc>
so it's more useful if those tests/pre-builts are more complete
<hramrach>
so I added some workarounds but not really awesome
<hramrach>
if somebody knows how the kconfig is supposed to work maybe there is a better fix to be had
<ssvb>
can we check which kernel modules are shipped with the popular desktop linux distros and just try to provide a similar set?
<hramrach>
basically the usb configuration does not stick if you do savedefconfig ; cp ; defconfig
<hramrach>
ssvb: you know any popular distro for devboards?
<ssvb>
hramrach: who cares about devboards? we have a full fledged desktop computer with everything including SATA hard drive, ethernet, hdmi, ... (for cubieboard and tv boxes)
<hramrach>
yes but no PCI but tons of weird pins with i2c, spi ..
<ssvb>
at least everything usb related that is provided by x86 desktop distros could be enabled
<hramrach>
everything USB related can be enabled, yes
<wingrime>
ssvb: hramrach: how can I measue IOWAIT?
<hramrach>
I have no idea
<wingrime>
ssvb: look like android works slow becose nand have "wait cycles"
<mnemoc>
if the cores of the things are enabled, people can build the .ko of their desired driver and reuse the core
<wingrime>
I mean not interpurtable
<hramrach>
and the wait is not preemptible?
<hramrach>
mnemoc: well, no. they would have to have the exact same source and build the whole kernel, anyway
<hramrach>
afaik you cannot build a module without building the kernel
<wingrime>
hramrach: simple wait
<wingrime>
hramrach: with while
<hramrach>
busy wait
<hramrach>
that really kills performance on single core
<mnemoc>
hramrach: if usbserial is enabled in the defconfig, and so in the prebuilt kernel. they can compile only the wanted .ko
<mnemoc>
and drop it in
<mnemoc>
for example.
<hramrach>
after building the whole kernel they can just drop in the .ko
<hramrach>
but by then the whole kernel is built, anyway
<wingrime>
hramrach: I prefer fix it in a first prio
<ssvb>
mnemoc: I'm almost sure nobody will do that, and the people who are skilled enough will just crosscompile their own kernel instead of using the hwpack
<hramrach>
busy wait is the first impl when prototyping a driver, typically
<hramrach>
it's very precise, too
<wingrime>
how make it interpurable
<wingrime>
I know only wait_event
<hramrach>
you just use some sleep function
<hramrach>
but you must know that the hardware can tolerate late finish of the wait
<wingrime>
only problem is,that if I fix this I gen some performace downgrade
<hramrach>
that is if you wait longer nothing more happens
<mnemoc>
ssvb: good point
<wingrime>
but get perfect responce
<hramrach>
perfect response is waht most people perceive as performance
<hramrach>
you can make a switch, possibly
<wingrime>
do you know best way for replace while
<hramrach>
I am not driver developer
<hramrach>
I would look at other driver or grep for sleep/delay
<techn__>
wingrime: include schedule request into that while
<hramrach>
we have CONFIG_SND_SUN4I_SOC_CODEC on 3.4 and CONFIG_SND_SUN5I_SOC_CODEC on 3.0 :s
<mnemoc>
I haven't finished backporting that unification to stage/3.0 yet
<hramrach>
so that's expected
<mnemoc>
will be "solved" tonight
<mnemoc>
and will merge the rest of hansg/olvier sound changes
techn__ has quit [Read error: Connection reset by peer]
techn_ has joined #linux-sunxi
n01 is now known as n01|work|afk
<wingrime>
mnemoc: why do you still not accept cedarx unification
simosx has joined #linux-sunxi
ganbold___ has quit [Remote host closed the connection]
ganbold__ has joined #linux-sunxi
<wingrime>
I want cm to test on a13
<techn_>
Turl: how about that pull request?
paulk-desktop has joined #linux-sunxi
torqu3e has quit [Quit: torqu3e]
ssvb has quit [Quit: Leaving]
<drachensun>
This is kind of a generic driver question, why would a driver never be probed?
eebrah has joined #linux-sunxi
Quarx has quit []
n01 has joined #linux-sunxi
Dave77 has quit []
eebrah_ has joined #linux-sunxi
<mnemoc>
drachensun: you mean modprobed by udev automatically? because the driver doesn't provide the right alias usually
<drachensun>
ah ok
<wingrime>
ssvb: are you tested dma patches ?
<mnemoc>
wingrime: two reasons. 1) still haven't had time to finish merging hansg's sound unification, (some aren't trivial to backport to 3.0) 2) haven't seen feedback of people about it's safety
<wingrime>
mnemoc: you can look at diff
<wingrime>
only clocks "slighly different"
<wingrime>
tecnh_: what with android ?
<mnemoc>
wingrime: yes, but time availabilty is VERY scarse. working over 14h/d on finishing $work$ for a trade show
<mnemoc>
wingrime: will try after merging the sound stuff
eebrah_ has quit [Ping timeout: 255 seconds]
<wingrime>
mnemoc: I still have some dma optimize/cleanup patches in my repo and I still wait benches for others
<wingrime>
mnemoc: and now I trying avoid "busy wait" I think it a great problem
<wingrime>
mnemoc: but I still noob with profiling
<mnemoc>
do your clean ups rely in something still unmerged?
<wingrime>
It must make kernel work quicker when you write/read something from nand
shineworld has quit [Quit: Leaving]
<n01>
mripard_: when I try to build uImage I get multiple (or no) load addresses. Even if I specify LOADADDR=0x48000000 I'm not able to boot (It is stuck after Starting kernel ...
<n01>
any hints?
<techn_>
wingrime, you need your board specific layer
hansg has quit [Remote host closed the connection]
<oliv3r>
did sun4i do hdmi 1.4?
<mnemoc>
1.3 iirc
<mnemoc>
oliv3r: mr oliverlist, you forgot to subscribe
<oliv3r>
mnemoc: i forgot to subscribe to?
<mnemoc>
sunxi's ML
<oliv3r>
so if A20 has hdmi 1.4 (brief says so too) It should have a 'data' channel for ethernet (will check pinouts)
<mnemoc>
but I just whitelisted your oliverlist@
<oliv3r>
mnemoc: it should be oliver+list@
<mnemoc>
no +
<oliv3r>
i get tons of mailsf from the list
<oliv3r>
well I have oliver@ and I have oliverlist@ as two seperate mailboxes. I have set up oliver+list as an alias for oliverlist :)
<oliv3r>
everything else (oliver+rest) goes to oliver; its just so i can seperate mailing lists from everything else
<oliv3r>
i see what you mean now
<oliv3r>
i used the wrong identity to mail
<oliv3r>
i should remove that one actually
<oliv3r>
done on my laptop :p (I have way to many identities to manage)
<oliv3r>
but feel free to un-whitelist it
<oliv3r>
i should be punished :)
<n01>
Turl: can you pastebin your cmd script for uboot?
<oliv3r>
oh wow the A20 schematic was last update april 1st; so it's quite recent :)
<oliv3r>
mnemoc: have you looked at the PIO schematic bit?
<oliv3r>
mnemoc: i'm trying to do A20/PIO
<oliv3r>
but get stuck on the first bit allready :)
<oliv3r>
litterally
<Turl>
n01: I used uEnv.txt, you'll have to wait a bit as I don't have my cubie handy now
<n01>
no prob :)
<n01>
I'm always around
<Turl>
it was nothing too fancy though, just a couple of lines to download the stuff to ram and then the bootm call
<mnemoc>
oliv3r: no, haven't looked
<Turl>
you'll need a tftp server on your pc, you can set that up meanwhile :)
<n01>
I'm using mmc so I don't need a tftp server :) anyway loading the kernel @ 0x40008000 I got something like "undefined instruction
<n01>
after starting the kernel
<oliv3r>
mnemoc: you did A10 PIO right? Where did you learn about the functions bound to a pin. I can see that some pins have more functions , but it doesn't say which mux is what, though it looks like the same right to left on what we have
<n01>
if I put the kernel @ 0x48000000 still got nothing
<n01>
(no undefined instruction but no messages on console)
<n01>
still investigating though
<oliv3r>
i really should use netboot for testing! i keep putting it all on the SD card over and over; good call
<oliv3r>
n01: wasn't it at 0x40000800?
<oliv3r>
or something closer to that
<n01>
ok we have 3 different addresses now :D
<mripard_>
4800000 seems way too high, probably out of the memory space
<n01>
23:30 < mripard_> scripts/mkuboot.sh -A arm -O linux -T kernel -C none -a 0x40008000 -e 0x40008000 -n 'Allwinner A1X' -d zImage_dt uImage.dtb
<Turl>
oh ok
<mripard_>
what toolchain are you using?
<n01>
gcc-linaro-arm-linux-gnueabihf
<mripard_>
yeah, don't
<n01>
O_o oh
<Turl>
not kernel related, but I've had issues with linaro toolchain generating crap when used on buildroot with cortex-A8 flag
<mripard_>
linaro toolchains are soooooooooo buggy
<n01>
ok, codesourcery? crosstool-ng? gcc-arm? :)
<Turl>
codesorcery/mentor one has been behaving ok so far
<mripard_>
codesourcery is a good choice yes
<n01>
:D ok I'll switch, thank you guys
<zumbi>
mripard_: ooi, why are linaro tools buggy?
<Turl>
mripard_: have you ever seen a sigill on init? that's what linaro got for me
<vinifm>
and debian toolchains?
<mripard_>
don't ask me. ask the linaro toolchain guys
<mripard_>
but I've seen things like Turl speaks about on a regular basis with linaro toolchains
<oliv3r>
mnemoc: yeah cubieboard schmatic has GPL_CLK on PIO, its now TWI3_SCK; i think they had problems with GPS, as most datasheet references are removed aswell; so they probably replaced it with twi3
<zumbi>
vinifm: also debian toolchain is buggy?
<oliv3r>
gentoo's crossdev builds a good one :D
<vinifm>
no, i guess
<mripard_>
vinifm: I don't know, I don't use debian :S
<oliv3r>
but tbh, if linaro toolchain is buggy, they should be able to fix it quite easily
<zumbi>
mripard_: I have been using the ubuntu toolchain, based on linaro, it contains codesourcey patches, worked fine so far for me
<oliv3r>
fedora probably has arm toolchian too?
<vinifm>
i uses and i like
<vinifm>
I have not had problems
<zumbi>
i was wondering if there was some known big bug, or if is just words gone with the wind
<mripard_>
zumbi: yes, I use the one in ubuntu, and never had any problem with this one
<zumbi>
that one is debian+linaro based
<mripard_>
I know.
<zumbi>
linaro did a pass over codesourcery patches and integrate them in their tools as well
<oliv3r>
mnemoc: everything is marked as I/O; i'm sure that's not true (I think i saw some input only, output only things somewhere) how is it defined in the 3rd colomn?