ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
tinti has quit [Remote host closed the connection]
vinifm has joined #linux-sunxi
<vinifm> Someone here is doing a pwd driver?
<vinifm> *pwm driver
vinifm has quit [Quit: Saindo]
LoCoZeNoz_ZUE has joined #linux-sunxi
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
BJfreeman has quit [Ping timeout: 248 seconds]
_BJfreeman has joined #linux-sunxi
_BJfreeman is now known as BJfreeman
rm has quit [Ping timeout: 245 seconds]
rm has joined #linux-sunxi
<hramrach_> hello
<hramrach_> hmm, rella not here
<hramrach_> anybody was able to reproduce the issue wit non-decodable file?
MadSpark has joined #linux-sunxi
root_ has joined #linux-sunxi
w00tc0d3 has quit [Write error: Broken pipe]
BJfreeman has quit [Quit: had a good time]
w00tc0d3 has joined #linux-sunxi
w00tc0d3 has joined #linux-sunxi
mturquette has quit [Read error: Operation timed out]
rellla has joined #linux-sunxi
awafaa_ is now known as awafaa
<rellla> hramrach_: ^
<oliv3r> morni'
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
hansg has joined #linux-sunxi
w00tc0d3 has quit [Quit: http://quassel-irc.org - Chat comfortabel. Waar dan ook.]
dragonn has joined #linux-sunxi
rellla has joined #linux-sunxi
root_ has quit [Ping timeout: 252 seconds]
<FunkyPenguin> moin
<gzamboni> morning
<FunkyPenguin> what's the correct defconfig to use for an A20? sun7ismp_defconfig or sun7ismp_min_defconfig
<gzamboni> dunno, min should mean minimal configurations to boot
<FunkyPenguin> ok, thanks
<hramrach_> mornin
<gzamboni> i am having some strange ttl issues when booting from a nfs share.
<hramrach_> rellla: https://github.com/hramrach/video-samples/blob/master/cedarx-no-output.mkv?raw=true this filr does not play but seems limited to libhybris
<gzamboni> i can't login, when i type root and enter, its like if i have given 2 enters with blank passwords
<hramrach_> so possible a wrapper issue rather than cedar issue
<rellla> hramrach_: yeah. it played with libhybris - but without video, only sound. the video layer doesn't com up
<gzamboni> others characters behave strangly like the back space: ^H
<hramrach_> gzamboni: try different application for serial connection
<rellla> btw. i closed your Xserver issue ;)
<gzamboni> i am using minicom, whats strange is that is well configured and it works perfectly when booting up from the sd card
<hramrach_> yes, did not get to it, sorry
<hramrach_> gzamboni: possibly you may need to set correct terminal tupe in inittabl or wherever the getty is spawned
<hramrach_> *type
root_ has joined #linux-sunxi
<gzamboni> in my inittab it is like this: T0:2345:respawn:/sbin/getty -L ttyS0 115200 linux
<gzamboni> im using rella's debian fs
dragonn has quit [Ping timeout: 260 seconds]
<rellla> gzamboni: puuh, that old ;)
root_ has quit [Client Quit]
<gzamboni> should do a newer one with debian final 7
mab_ has joined #linux-sunxi
<hramrach_> gzamboni: minicom is not linux
<rellla> gzamboni: good idea.
<gzamboni> can i use getty to connect to ?
<hramrach_> you set terminal type to linux. minicom is not linux console. stuff breaks
<gzamboni> yeah, i see that, do u know any terminal of type linux ?
<hramrach_> look minicom docs I guess
<gzamboni> or what kind of terminal should i set to make it work correctly on minicom ?
<gzamboni> ok
<FunkyPenguin> um, where do i get the rootfs.cpio.gz that it complains about when i try and build - http://paste.opensuse.org/60690623
<hno> gzamboni, vt100 is always a safe bet.
<gzamboni> in the minicom configs i can set it to VT102 or ANSI
<hramrach_> you probably need to set enter and backspace to do the right thing in minicom, too
<hno> also I generally prefer screen over minicom for console usage.
<hramrach_> or cu
<hno> screen /dev/ttyUSB0 115200
<hramrach_> anything but minicom, really
<gzamboni> :) thank you guys
<hno> minicom is nice for calling BBS:es using a modem, but haven't connected a modem to my computer in 15+ years..
<hramrach_> hence it's over-engeneered for the simple task of using a serial console and stuff gets in your way
<gzamboni> screen looks nice :)
<oliv3r> i'm measuring some pins on the PHY on my scope, and notice both TXC and RXC are only 250 kHz (without lan connected). it should be 25 MHz (100mbit) or 2.5 MHz (10mbit), so kinda doubting that is right
<oliv3r> does the a10 clock drive the emac?
<oliv3r> er phy?
<oliv3r> 231 kHz*
<hramrach_> maybe you should measure with cable connected
<oliv3r> same :p
<hramrach_> weird
<hramrach_> but if it transmits data that looks like measurement or evaluation error
<oliv3r> phy doesn't work at all :p
<oliv3r> lan is dead
<oliv3r> driver crashes, etc
<hramrach_> hmm
<oliv3r> Management clock runs nicely at 48 MHz
<hramrach_> I don't have a scope so can't tell what's on my phy
<oliv3r> which is WAY to high
<hramrach_> hmm
<mab_> are there any news about brightness pwm getting to work? Last time I tried was 2 month ago (A13, sunxi 3.4.43, debian)
<oliv3r> This pin provides a clock synchronous to MDIO, which may be asynchronous
<oliv3r> to the transmit TXC and receive RXC clocks. The clock rate can be up to
<oliv3r> 2.5MH
<oliv3r> mab_: there's a new PWM driver for 3.4, should do brightness too
<oliv3r> for 3.9 i've nearly finished a dt-based PWM driver, which should control backlight
<mab_> oliv3r: is it in the testing branch on github sunxi?
<oliv3r> the 3.4 is on the ML
<oliv3r> the 3.9 isn't done yet nor pushed
<mab_> thanks, will try
<oliv3r> we could do with a BROM dump of the a13 ...
<oliv3r> drachensun: can you extract the BROM from an A31 for us? it would go nicely on hno's Allwinner-info github
<gzamboni> strange, now when i type root as username it jumps directly to incorrect password, and if i try another unexistent user it ask me for password
<gzamboni> maybe be its and file permission thing. i noticed for the ssh i had to modify some file permissions
<gzamboni> this pwm fs driver is awesome, havent tried yet but i didnt know we could control pwm easily like that
rz2k has joined #linux-sunxi
<oliv3r> hno: i was looking at the cubie schamtic, and something interesting on page2, the SYS section has 'EFUSE_VDDQ'
<oliv3r> which on cubie is tied to GND
<oliv3r> so it's probably programming language
dragonn has joined #linux-sunxi
<oliv3r> programming voltage*
<oliv3r> i hope that the next cubie PCB brings out that pin too :)
<rellla> ssvb: the high level api (is it called high level?) to use libve hasn't changed much in newer cedarx.
<gzamboni> ok, its working now, booting from nfs share, it will avoid me from changing all the time the sdcard
<hno> oliv3r, aha!
<hno> as usual with Allwinner the correct details are in schematics (or code) an not datasheets...
<hno> T9 is listed as GND of type Input in the data sheet... other GND pins are listed as of type GND.
<oliv3r> the efuses might need 12V to program
<oliv3r> maybe 5V
<oliv3r> not quite unusual for fuses to require special programming voltage
<oliv3r> hno: what are the chances that Emac-MDC (Mii clock) is wrongly clocked? My scope says 48 MHz during u-boot
<hno> Not sure. Have never looked at that part.
hipboi|cubie has quit [Read error: Connection reset by peer]
<hno> oliv3r, looks like it's set up the same as in kernel.
wingrime has joined #linux-sunxi
<oliv3r> so programming isn't really possible then for mere mortals
<hno> seems so, and is supported by the fact that the mele does not seem to respond to programming requests.
<oliv3r> and that pin is probably not reachable, not connectable
<hno> oliv3r, specifications for the PHY says: MDC: This pin provides a clock synchronous to MDIO, which may be asynchronous to the transmit TXC and receive RXC clocks. The clock rate can be up to 2.5MHz.
<oliv3r> yeah, i pasted that earlier
<oliv3r> and 'up to 2.5 MHz' is what worries me
<oliv3r> cause the A10, generates a full 48 MHz on it
<oliv3r> so i probably best start looking at the clock register
<hno> oliv3r, it's very hard to do anything on an BGA, even less on pins which is routed to GND. You need to desolder the BGA, rework the mainboard and reflow the BGA again.
<oliv3r> yeah, i don't think that's quite possible
<oliv3r> unless it's brought out somewhere, before going to GND of course
<oliv3r> which is highly unlikly
<oliv3r> but not entirely unreasonable, if mele wants to reprogram their chips in-product or somethign
<oliv3r> CCM_AHB_GATE_EMAC we only have that :(
<oliv3r> nothing related to MDC or how EMAC configures the mdc :S
<oliv3r> and of course, no docu
<oliv3r> anybody that can probe PIN25 on the RTL PHY
hipboi|cubie has joined #linux-sunxi
<oliv3r> hey hipboi
<oliv3r> hipboi|cubie: question; The PHY doc says, MDC pin should not exceed 2.5 MHz (Mii clock). My scope measures 48 MHz!
<hno> oliv3r, the clock is set in mac_mcfg, but no info on how the divisors works.
<oliv3r> i'll play with my scope :p
<oliv3r> can't make it worse
<oliv3r> mac_mcfg, in the net driver I assume?
<hno> oliv3r, Can you try a little experiment for me? Boot from nand and ask the allwinner bootloader to jump to fel by pressing 2 as boot1 starts up. Then use fel to load u-boot.bin into SDRAM and start it. See if that makes a difference.
<oliv3r> sure, can do
<oliv3r> so boot allwinner nand, and interrupt boot
<hno> yes, just keep sending the digit 2 while boot1 starts.
<oliv3r> ohh, press button as boot1 starts
<oliv3r> ohh
<oliv3r> ok
<oliv3r> [ 0.243] Jump to fel
<hno> that
<hno> that's it!
<oliv3r> lsusb to confirm?
<hno> yes. Now load u-boot.bin using fel.
<oliv3r> er, never done that :p
<oliv3r> not sure if i have everything setup
<oliv3r> need to find usb cable first
<oliv3r> lsusb doesn't see it
<hno> are you sure? Check your cable...
<oliv3r> its power via
<oliv3r> but my monitor is its hub
<oliv3r> i'm not at the u-boot console
<oliv3r> let me boot full android
<hno> should work fine via hubs.
<hno> your computer is connected to the monitor USB hub?
<oliv3r> let me use a diff cable
<oliv3r> android is booted
<hno> oliv3r, where did you measure 48 MHz?
<oliv3r> on the MDC pin, pin 25
<oliv3r> buttom of the PCB
<oliv3r> ok lsusb still show snothing, even with different cable, but without hub
<oliv3r> let me grab my tablet
<oliv3r> oh, no console, nvm
<oliv3r> [ 3.740650] android_usb: already disabled
<oliv3r> [128460.706677] scsi 13:0:0:0: Direct-Access USB 2.0 USB Flash Driver 0100 PQ: 0 ANSI: 2
<oliv3r> did have that at some point
<hno> ?
<oliv3r> [128459.703757] usb 1-4: New USB device found, idVendor=18d1, idProduct=0003
<oliv3r> that's some android device
<oliv3r> so has to be the cubie :p
<oliv3r> can't get it to re-appear though :S
<hno> so something fishy about your usb somewhere..
<oliv3r> well i do see my mice, scope when i power it on etc
<oliv3r> ttyUSB0 also works
<oliv3r> but yeah, USB on my motherboard is iffy. sometimes (1-10 with latest bios, before 1-3) it doesn't boot, have to power off/on again. I found out,that without USB hardware connected, it always booted
<oliv3r> also, 64bit linux wouldn't see/work with USB hardware
<oliv3r> with the latest bios, it does work pretty well
<hno> odd
<hno> Ok, i'll try to measure it here.
<oliv3r> i'll grab my tablet, and see if it works in fel mode (via button) there
<oliv3r> tablet also isn't see ?!
<hno> I suspect bad cable.
<oliv3r> but it powers it fine :S
<hno> different pins
<oliv3r> i'll try my other cable
<oliv3r> diff cable, diff port, same thing
<oliv3r> let me try adb
<oliv3r> fuck me.
<oliv3r> i do lsusb as root it does show
<oliv3r> bad cable.
<hno> Often helps to tighten the mUSB contact a little by slightly compressing the metal.
<hno> on the bottom side
<mnemoc> hno: hi, I found a bug on the soc-detect/sun5i code which is now fixed. it would be awesome if you can give it another try (it's also rebased so better reset the old local branch)
<mnemoc> also changed the way SUNXI_VER_ things are composed, I hope it's more useful now
<hno> amery/wip/sunxi-3.4/soc-detect?
<mnemoc> yes
<oliv3r> phone doesn't work either, bad cable
<oliv3r> crap
<oliv3r> Bus 004 Device 027: ID 1f3a:efe8
<oliv3r> jump to fel works :)
<hno> :)
<hno> Now from sunxi-tools:
<hno> fel write 0x4a000000 $uboot
<oliv3r> gotta sync
<hno> fel exe 0x4a000000
<oliv3r> $uboot being my u-boot binary
<hno> yes
<oliv3r> works
<oliv3r> so how is that different from your u-boot boot?
<hno> This uses Allwinner clock initialization.
<oliv3r> ahh, ok
<oliv3r> still, nice feature
<oliv3r> so, with boot0/1 init; we can boot straight to u-boot via fel; when done from BROM, we need u-boot-felboot
<oliv3r> to init clocks
<oliv3r> i documented whatever gets setup before 'jmp to fel' gets done on the BROM wiki page
<hno> yes
<hno> and also SDRAM etc.
<oliv3r> ah of course
<hno> mnemoc, <6>sunxi: Allwinner A13 revision A (AW1625/sun5i) detected.
<mnemoc> \o/
<mnemoc> hno: btw, the method also works on A31
<oliv3r> md.l 0x01c0b07c 1
<oliv3r> can a few people run that on a few boards (that have working MAC_PHY)
<oliv3r> i have 0x000000
<hno> oliv3r, have you started the ethernet in u-boot?
<oliv3r> how do I 'start' it? i done mii device emac; mii dump
<oliv3r> i'm running the boot0/1 -> fel u-boot version
<mnemoc> hansg: around? can you try wip/sunxi-3.4/soc-detect from my github on your A10S please?
<mnemoc> hansg: it's based upon stage/sunxi-3.4 so it includes your latests commits too
<oliv3r> mnemoc: i need to rebase that and re-submit my fixes on top of those?
<hansg> mnemoc, what is your github? I guessed: https://github.com/mnemoc but that is 404
<mnemoc> hansg: amery
<mnemoc> oliv3r: yes, please
<mnemoc> hansg: beware, it's FAT. so better `git remote add`
<hansg> mnemoc, yeah I always use git remote add
<mnemoc> )
<mnemoc> :)
<hansg> mnemoc, I'm busy with $dayjob atm. I'll try to make some time for this tonight
<hno> oliv3r, I also get 0, but it works.
<hansg> mnemoc, cool that you're working on this btw, I had sorting this out on my own todo :)
<oliv3r> hno: btw, when i probed the MII data pin, i did get somewhat different result on mii dump; 0xfff instead of 0xffff, probabyl due to capacitance etc and values being 'wrong'
<oliv3r> hno: hmm, so that's not gonna be helpfull
<oliv3r> hno: what DOES mii dump show for you?
<mnemoc> hansg: awesome. thank you. if this works on yours we can use sunxi_is_sun5i() and sunxi_is_a10s()/_a13()/_a10()/_a31() stuff on the unifications
<hno> oliv3r, depends on what I dump. It takes two arguments.
<hansg> mnemoc, yeah. Eventually I would like to see things unified sofar that we can have one kernel for both sun4i and sun5i
<hno> oliv3r, "mii info" is better.
<mnemoc> hansg: yes. that's my goal too. maybe even farther into single-sunxi considering cortex-a7 is also armv7
<mnemoc> hno's chip-id (sunNi family) detecton is simple enough to be run very early
<mnemoc> and works on A31 too
<hno> and works on A10 with a blank SID...
<mnemoc> yeah :)
<mnemoc> A31 has blank SID too
<hansg> note: got to get back to work, see you all later
<mnemoc> cu
fredy has quit [Ping timeout: 245 seconds]
vinifm has joined #linux-sunxi
<oliv3r> hno: mii ifno gives zero info
fredy has joined #linux-sunxi
<oliv3r> hno: if i do mii dump without arguments it gives me a list, -- PHY control register --
<oliv3r> https://github.com/hno/u-boot/blob/sunxi-current/drivers/net/sunxi_wemac.c#L331 magic values to set up the clock where simply copied
<oliv3r> but why don't I see 0xd << 2 in that register ...
<vinifm> there is no CONFIG_PROC_KCORE in sun4i_defconfig, could i create and set CONFIG_PROC_KCORE=y ?
<hno> oliv3r yes those are copied from the kernel driver.
<hno> but... I think that is only called when first using the ethernet.
<hno> oddly the mii command works regardless.
<hno> and the register shows 0 even after initializing the network.
<hno> no it doesn't
<mnemoc> vinifm: sure
<hno> 01c0b07c: 00000034 4...
<mnemoc> vinifm: the defconfig is only a "known to work" starting point
<vinifm> but: Depends on: PROC_FS [=y] && MMU [=y] && !ARM [=y]
<hno> oliv3r, "ping 1.1.1.1" to initialize. Then check the clock.
<vinifm> means I can not set it on ARM?
<ssvb> rellla: we don't want any CedarX API/ABI backwards compatibility breaking changes at all
<hno> Iguess MII clock setup should move to sunxi_wemac_initialize so it's set when the hardware is probed, not on first use. It's amasing it works without it.
<ssvb> rellla: any minor change is a potentially hard to debug issue
<rellla> ssvb: the changes are between libve speaking with the adapters (FBM+VBV), not on the toplevel how to use libve
<rellla> so it should be at least application independent imo.
<ssvb> rellla: haven't you shown some sort of constants reshuffling in an enum earlier?
<rellla> ssvb: yes, but armel, armhf and the new libcedarv.h is the same
<rellla> seems that the android on was an older one
<ssvb> I guess this might be just pure luck :)
<ssvb> it does not seem like they have ever considered keeping stable API, let alone ABI compatibility
<hno> oliv3r, fixed the clock setup now.
<hno> oliv3r, I bet all values in your mii dump is 1. (ffff). This is "no response"
<ssvb> rellla: libve core also does not change "much", but that's enough to be a PITA
LoCoZeNoz_ZUE has quit [Ping timeout: 260 seconds]
<mnemoc> hno: btw, did you get my mail? tried the ssh/sudo already? :)
<hno> yes. not tried yet
LoCoZeNoz_ZUE has joined #linux-sunxi
<mnemoc> ok
<hno> mnemoc, works fine.
<hno> oh, lxc containers. nice.
<mnemoc> :)
<mnemoc> hno: also, I've enable the >5TB traffic at 100Mbps. we are around 3TB this month, thanks to the cubieboard images....
<hno> but can'l login to the builder.
<mnemoc> uhm
<mnemoc> you are ipv6, right?
<hno> ipv4 at the moment.
<mnemoc> ipv4 188.40.48.2 2201 should do
<mnemoc> or it's an auth problem?
paulk-desktop has joined #linux-sunxi
<mnemoc> uhm
<mnemoc> no .ssh there :(
<mnemoc> 1m
<mnemoc> hno: try now
<hno> better
<hno> password set eveywhere now I think.
<mnemoc> hope that machine helps you. beside bw, it's seriously underused
<hno> what did you use to set up the lxc containers?
<mnemoc> hno: manually.... (and the stock lxc package)
<mnemoc> had to made an script for the networking. stock things don't play way with bridges :\
<mnemoc> hno: see the ./firewall.sh in /root
<mnemoc> i think it's pretty readable
<hno> tried to use some of the templates included in lxc some time ago but did not have much success, it even crashed my host root fs for me..
<mnemoc> ouch
<hno> yes. but need to give it another try shortly.
<hno> need to migrate some old vserver containers.
<mnemoc> lxc is far from ready to replace a vserver
<mnemoc> meh. getting a maps.google.com "upgrade" while using the phone as modem isn't nice...
rellla has quit [Remote host closed the connection]
dragonn has quit [Quit: Wychodzi]
dragonn has joined #linux-sunxi
mab_ has quit [Ping timeout: 252 seconds]
mab_ has joined #linux-sunxi
Yaku321 has joined #linux-sunxi
<vinifm> do you guys debug kernel drivers?
<FunkyPenguin> oliv3r: what is needed to build an a20 kernel? i've tried both sun7ismp_defconfig & sun7ismp_min_defconfig
unioah has joined #linux-sunxi
<mnemoc> FunkyPenguin: have A20 hardware with uart?
<FunkyPenguin> mnemoc: i need to improve my soldering skills considerably if i want a working board with uart
<mnemoc> :(
<FunkyPenguin> im just trying to cross build the kernel first, but that fails
<mnemoc> what kernel?
<mnemoc> meh. out of battery. back in 15m
Yaku321 has quit [Disconnected by services]
Yaku-noob has joined #linux-sunxi
Yaku321 has joined #linux-sunxi
Yaku-noob has quit [Ping timeout: 252 seconds]
<hramrach_> oliv3r: 01c23800: 16236750 80758352 52574848 08025081 Pg#.R.u.HHWR.P..
<hramrach_> result of sun4i#md.l 0x01c23800 4
PiyushVerma has joined #linux-sunxi
<PiyushVerma> ssvb: we just decode with ffmpeg and it take that much cpu
<PiyushVerma> that's a unbelivable fact for AAC Codec
<oliv3r> FunkyPenguin: i have no idea; don't have a a20 yet :(
<ssvb> try to profile it and see where the cpu usage is spent
<oliv3r> FunkyPenguin: nobody does
wingrime has quit [Read error: Connection reset by peer]
<oliv3r> hramrach_: awesome, thanks
<oliv3r> i'll add it to the wiki; that a mele?
<ssvb> PiyushVerma: if you are using some mainstream armv7 linux distro, the chances are high that ffmpeg is compiled with NEON support disabled
<FunkyPenguin> oliv3r: heh, ok thanks
<PiyushVerma> ssvb: I compile seperately with neon opt but not diff
<FunkyPenguin> oliv3r: when trying the min defconfig i get http://paste.opensuse.org/60690623
<PiyushVerma> might be ffmpeg internal codec not optimised for that
<ssvb> PiyushVerma: as I said, just profile it and you will see what's wrong with it
<oliv3r> FunkyPenguin: so far I know, that A20 is very similar to A10, but with different interrupt controlelr
<FunkyPenguin> what provides the rootfs.cpio.gz?
<oliv3r> you :p
<PiyushVerma> ssvb: right I will try that with strace
<oliv3r> FunkyPenguin: do you have console +u-boot access?
<ssvb> PiyushVerma: no, strace is not a good tool, you need 'perf'
<FunkyPenguin> ah, i thought it was auti generated
<FunkyPenguin> s/auti/auto
<FunkyPenguin> oliv3r: not yet
<oliv3r> FunkyPenguin: ohh you are building lichee branch; the 'SDK'. Yeah, not sure if we have anybody able ot build that yet :)
<ssvb> PiyushVerma: what is your distro?
<FunkyPenguin> im trying to build the kernel first before moving on to getting things running :)
<PiyushVerma> I just build rootfs with buildroot
<oliv3r> FunkyPenguin: there should, in thoery, be a defconfig that works. I suppose best way to get one, is to get it from a running kernel, /proc/config.gz
<PiyushVerma> ssvb : May be I should switch to debianhf
<FunkyPenguin> oliv3r: ok thanks, i'll try that. problem is it doesn like hdmi-dvi on my monitor and ai dont have a tv to hand to try
<ssvb> PiyushVerma: if you can build perf with buildroot, then it should be also fine
<PiyushVerma> ssvb: right I will check that
<PiyushVerma> ssvb: thankyou very much I will go through that
<mnemoc> get an uart please
<ssvb> PiyushVerma: also as I said before, VLC is wasting a lot of time *resampling* audio with some poorly optimized code (can be easily seen with perf)
<ssvb> PiyushVerma: audio decoding is not a bottleneck by itself
<PiyushVerma> ssvb: well I am not using VLC
<PiyushVerma> ssvb: I am directly using ffmpeg which purely decode and play audio without resampling and it take that much cpu
<ssvb> whatever you are using may have a similar problem
<PiyushVerma> where it take approx 20% for flac codec
<PiyushVerma> which say it's codec specific
<PiyushVerma> right
<PiyushVerma> let's first profile it
<oliv3r> FunkyPenguin: well, none of the developers have any a20 hardware yet
<PiyushVerma> oliv3r: I have a20 hardware and also have HDMI Monitor :)
<oliv3r> FunkyPenguin: but a uart on your end would help ;) helps us looking at a few things :)
<oliv3r> i'm not envious at all FunkyPenguin ... not at all.
<mnemoc> without uart access we can't support A20
mab_ has quit [Ping timeout: 256 seconds]
<ssvb> PiyushVerma: that's interesting, what kind of A20 hardware?
<ssvb> PiyushVerma: could you please run https://github.com/ssvb/tinymembench on it?
<PiyushVerma> ssvb: Let me take picture and put on web
<FunkyPenguin> oliv3r / mnemoc: yeah, im trying to convince colleagues to do the soldering for me
<oliv3r> no header
<ssvb> PiyushVerma: cool, it's always fun to see some nice pictures :)
unioah has quit [Quit: Leaving.]
<PiyushVerma> ssvb: I have 5 pcs board
<PiyushVerma> may be can pass if some one can utilise
<ssvb> I think many of the developers have submitted a request at http://cubieboard.org/cubieboardii-developer-request/
<mnemoc> but they haven't been shipped afaik
vinifm has quit [Quit: Saindo]
LoCoZeNoz_ZUE has quit [Ping timeout: 248 seconds]
<PiyushVerma> ssvb: I have add u in google plus and uploaded photo
tinti has joined #linux-sunxi
<PiyushVerma> got direct link :)
<mnemoc> what device is that?
<mnemoc> can you make a page on the wiki for it?
<mnemoc> those 4 pins aren't uart0?
<PiyushVerma> yes I solded that
<PiyushVerma> and that work
<mnemoc> \o/
<mnemoc> can you go to u-boot and get us some register dumps?
ozonys has joined #linux-sunxi
<hramrach_> oliv3r: that's cubieboard
<ozonys> PiyushVerma: how to get one ?
<oliv3r> hno: they are all 1; and after writing 034 to that register, I get a much nicer 833.3 kHz
<PiyushVerma> mnemoc : need to connected that board. working on other. By the way how to dump register
<mnemoc> PiyushVerma: md.l command on u-boot
<mnemoc> or devmem2 from a running os with /dev/mem
<PiyushVerma> ozonys: I can available to developer for 35$
<mnemoc> PiyushVerma: where????
<PiyushVerma> I am in Taiwan
<PiyushVerma> so 35$ + Shipping cost
<mnemoc> can you please make a wiki page for it?
<oliv3r> hramrach_: 1024 kiB?
<PiyushVerma> yes
<PiyushVerma> actually board not all ready. you can see some jump wires
<PiyushVerma> nead sdcard
<ozonys> ok
<mnemoc> PiyushVerma: missing sata, that's a pity
<PiyushVerma> right
<mnemoc> but still, please document your board on our wiki :)
<oliv3r> md.l 0x01c23800 4
<oliv3r> can you run that frmo a u-boot prompt?
<oliv3r> looks like a evaluation board
<oliv3r> DEMO_V2.0
<mnemoc> evaluation boards are more feature complete. that's the rpototype for a product
<PiyushVerma> Evaluation board comes in 1000$ :)
<PiyushVerma> mnemoc: you are right
<PiyushVerma> mnemoc: prepairing this board in china
<mnemoc> PiyushVerma: if you add good hacking info you are welcomed to add commercial info on the page too ;-)
<PiyushVerma> mnemoc : I will check
<PiyushVerma> any way I can also get hackberry only in 35$ for quantity order
<mnemoc> we have enough A10 hardware. problem is hackable A20 and A31 with uart atm
<PiyushVerma> ok
<oliv3r> A20
<oliv3r> i want a20 only :)
<oliv3r> so mnemoc since you never answerd my question; fosdem 2014; sunxi-devroom
<mnemoc> oliv3r: I can't go
<oliv3r> i meant, if I would go
<mnemoc> but looks like a great plan
dwilkins has quit [Ping timeout: 248 seconds]
<oliv3r> well i was worrying, what would we talk about for 60 minutes
<mnemoc> libv or hno can advise you
<mnemoc> they have experience talking about things
<oliv3r> libv allready has
dwilkins has joined #linux-sunxi
<oliv3r> well i'll see what I think
<oliv3r> gonna see if its worth going out for just the saturday
<mnemoc> oliv3r: *you* in particular don't have any excuse to not be there during the whole thing
<oliv3r> i'm poor
<mnemoc> it's near and it's free
<oliv3r> and have someone at home to think about
<mnemoc> and you have time for saving
<mnemoc> saving 200E is enough for you to be there the whole thing, so start saving :)
<oliv3r> :S
<oliv3r> you have no clue how poor i am!
<mnemoc> how many months up front?
<oliv3r> well i guess i could save for it :p
<oliv3r> maybe share a car with others going
<mnemoc> :)
dwilkins has quit [Ping timeout: 256 seconds]
<mnemoc> and share a room on a youth hostel reserved in advance
<oliv3r> stop trying in talking me into it; figure out a) purpouse of the talk; content of the talk :D
dwilkins has joined #linux-sunxi
<mnemoc> oliv3r: talk details aren't relevant at this point :)
<oliv3r> recruitment for devs? :)
<drachensun> FunkyPenguin: I think if you use the allwinner build script, build.sh it will compile. That worked for the A31 when a standard build would fail
<drachensun> I can get an A20 tablet with a solderable header
<drachensun> if you guys are interested
<drachensun> I've got 2 sitting on my desk but I've been focused on the A31 since I got the devkit
dwilkins has quit [Read error: Connection reset by peer]
<oliv3r> dwilkins come back
<oliv3r> drachensun: oh! a20 is good
<oliv3r> md.l 0x01c23800 4
<oliv3r> md.l 0x01c00024 1
dwilkins has joined #linux-sunxi
<oliv3r> dwilkins: can you link that patch that you want accepted agian?
<PiyushVerma> drachensun: where from u got A31 Devkit
<oliv3r> I just talked to the other of said patch (i think) who just says 'it never will be applied'
<drachensun> PiyushVerma: I bought it from wits-tech, http://www.wits-tech.com/pagesen/
<PiyushVerma> drachensun : understand how much ?
<drachensun> $1600
<drachensun> USD
<drachensun> mostly I wanted all the docs and build setup
<PiyushVerma> drachensun : ok, are you planing to develop own hardware can u share the cost and make somthing useful for both of us ?
<drachensun> but it came with less documentation than I had hoped
<drachensun> I hadn't planned on designing anything custom yet
<PiyushVerma> darchensun: I went to witsteck and see this devkit but they unable to give me gpu driver support for linux
<drachensun> I might be in a few months though
<drachensun> yeah, that is the still the case
<drachensun> the framebuffer console doesn't even work in Linux
<drachensun> but it will boot Linux
<PiyushVerma> aah pity
<mnemoc> drachensun: we added fbcon for a1x
<mnemoc> actually techn_ did
<PiyushVerma> drachensun: what's u planing to do with devkit ?
<PiyushVerma> as I ask early about dev ....
<drachensun> mnemoc: yeah, I looked back at the patches, thats when I decided to try and merge the sun6i core into the driver fixes rather than the other way around
<oliv3r> did you get more docs then what we have available?
w00tc0d3 has joined #linux-sunxi
<mnemoc> drachensun: it's easier when you see things commit by commit than bulk from a raw SDK
<drachensun> mnemoc: Yeah, of course they delete the history from what they sent me, but you have that one with the history
<drachensun> I got it to compile though, so I'll start testing it in about an hour here
<oliv3r> yeah, we cheat ;)
<drachensun> oliv3r: I don't think so, Luke said he was already posting the schematics and such but I dont know if I actually saw them on his site
<mnemoc> i'll try to reconstruct import/lichee-3.3/{a20,a31}-dev on top of sunxi-3.4
<mnemoc> commit by commit
<drachensun> I do have a lot of little step by step guides but most seem to be things we knew, like how to customized the fex file
<oliv3r> :S
<oliv3r> do you have to solder the uart pins to your a20 hardare? or dot hey have headers
<oliv3r> in those docs, is there information on how to program the Security ID? (SID)
<drachensun> I haven't seen anything about the security ID
<drachensun> yeah, the A20 requires soldering
<drachensun> its got 3 test points
<drachensun> that are labeled for once
<drachensun> right opposite the micro USB
<drachensun> I haven't soldered it up to test them yet though
<hramrach_> oliv3r: 1024 of what? it has 1G ram.
<oliv3r> 1024 MB :p
hansg has quit [Quit: Leaving]
<drachensun> ok, anyone that wants an A20 tablet that has a solderable UART header, please let me know today. I am putting in a new order to my supplier and so will get them added
<drachensun> I can ship pretty much anywhere, you can check the shipping cost at http://www.pengpod.com/pengstore just select a PengPod 700, put it in the cart and check the shipping to you
<drachensun> They are 1G ram, 8G flash and 7", otherwise very simliar to the PengPod 700, I think its a 1024x600 screen, so a little better
<drachensun> They will be $80 plus shipping
<oliv3r> that IPS screen?
<oliv3r> dthat's better then 109$ that shows now?
paulk__ has joined #linux-sunxi
<drachensun> this is a special deal for active developers
<oliv3r> ah kewl :)
<oliv3r> i trying to get a quota for shipping to netherlands
<oliv3r> holy shit, 37 USD for DHL express world wide
<oliv3r> ouch
paulk__ has quit [Client Quit]
<drachensun> yeah, they are going to whack you for VAT and brokerage fees too most likely
<oliv3r> ouch
<drachensun> I've noticed when people ship internationally from Europe its a lot cheaper than they charge us here in the US to send it out
<oliv3r> usually, when i buy from DS or ebay; i don't get vat nor customs :)
<mnemoc> USPS <3
<oliv3r> i've worked for the USPS when I was 18 :D
<mnemoc> last year?
<drachensun> we used to use them, about 1 in 3 to Canada either got lost and pilfered
<oliv3r> quite a few years ago
<mnemoc> drachensun: that's .ca post fault, not usps
<drachensun> that was before we shipped everywhere else so I'm assuming is better to other places
<drachensun> yeah, it must be since everyone seems to praise it
<mnemoc> if both national post services are reliable, no courier can beat them
<mnemoc> if one is meh. forget about the package
<oliv3r> my father sends me packages every so often, using USP
<oliv3r> USPS
<drachensun> they want $40 for a box this size as well but I guess you would avoid duties and brokerage
<oliv3r> wats brokerage?
<mnemoc> normal usps to .es takes less than 5 working days, no taxes, and delivered to your door :)
<drachensun> well they have one option for $29
<mnemoc> oliv3r: their kind service of dealing with the customs officer for you
<drachensun> what DHL charges for collecting your duties and getting it through customs
<oliv3r> oh even with USPS i've had a customs bill
<mnemoc> in .es at least they FORCE the customs to tax you, so they can charge you extra
<oliv3r> 25 or so euro
<drachensun> oh really? a lot of international customs say with USPS they don't have to pay duties on their end
<mnemoc> DHL/UPS/FedEx I mean
<oliv3r> but that was a big box, 700 euro's worth of parts :)
<mnemoc> drachensun: depends on the whimp of their national post service
<mnemoc> drachensun: in .es I've received up to 400 EUR packages via normal post without taxing
<mnemoc> .uk and .de are specially annoying
<drachensun> ok, so there is some threshold under which they dont bother with in Spain
<oliv3r> pure luck
<oliv3r> here, officially, the break point is 45 Euro
<mnemoc> spaniards don't stress themselves with reviewing small packages, regardless the amount on the invoice
<oliv3r> and then, only IF the customs officer checks it
<drachensun> In the US under $1500 doesn't require a formal entry and virtually nothing in the US has a duty (at least that I would buy), so I almost never get a bill
<drachensun> but yeah, it sucks to pay a DHL $30 entry fee for $0 in duties
<oliv3r> 40*
<oliv3r> so is the screen IPS?
<drachensun> I'm not sure, let me see if I can find something on it
<oliv3r> wmy current tablet, is so bad; if i put the birghtness down, i can't view it from any angle properly, colors get bad really fast. int he 'front' its good
_BJfreeman has joined #linux-sunxi
<drachensun> they aren't selling this normally yet, so I dont have to regular write up
<oliv3r> i do find the 80 a really good deal; but with 40 shipping ontop, it's iffy :(
_BJfreeman is now known as BJfreeman
mturquette has joined #linux-sunxi
mturquette has quit [Changing host]
mturquette has joined #linux-sunxi
<Turl> drachensun: no USPS/EMS? :P
<Turl> Express worldwide $48.59
<Turl> drachensun: (I wasn't going to buy it though, just interested in checking shipping charges)
<mnemoc> for us used to buy from .cn/.tw shipping prices from other countries are always absurdly bloated
<Turl> mnemoc: it's not just that, but with DHL I'd also have to add a 300km bus trip, DHL 'handling' charges ($$$$, I still don't know what they handled though if I have to go and do the customs papers myself) plus the actual customs tax
* mnemoc wonders if the .cn gov subsidizes CNP/HKP or it's just that they are far more efficient
<mnemoc> Turl: DHL/UPS/FedEx-only on a shop means immediately closing the window and deleting the bookmark
<mnemoc> (for me)
<mnemoc> they are thieves
<Turl> mnemoc: pretty much, yeah
<Turl> EMS is way better if you need the speed
<drachensun> Turl, where are you? I know we used DHL in Argentina once and it was a big mess like you describe
<drachensun> We dont have EMS here, that I know of
<mnemoc> EMS is the express service of USPS
<Turl> drachensun: EMS is USPS express
<mnemoc> EE......US
<drachensun> ah ok
<mnemoc> vs the usual RR......US
<drachensun> yeah when I order aliexpress ems I get it via the post office, in 3-4 weeks
<Turl> drachensun: usually 5-7 days here
<mnemoc> same to .es
<mnemoc> 1-1.5w
<mnemoc> due to working day issues
<Turl> the only "problem" with EMS is that if it is taxable, they will tax it
<drachensun> If I could figure out a provider for USPS that did the customs forms electronically I might start offering it
<mnemoc> registered airmail <3
<Turl> mnemoc: :)
<mnemoc> yes, takes a month from .cn/.hk ... but smooth and untaxed
<mnemoc> <1w from .us
<Turl> the breaking point here is 25USD, pretty low when compared to other countries
<drachensun> yeah, HKP, however they do it is amazing, I wish we had it here
<mnemoc> after packages leave .cn they take the same to arrive to .es than after leaving .hk, the main difference is that hkp ships things next day, and cnp can have packages lingering in their offices for 1-2 weeks easily
<drachensun> but I assume that cheap, 4 week service is just, label it for the far sides post office and throw it in a container, so really its the cost of the other countries post office plus pennies
<drachensun> I don't know of a way to do that from the US to anywhere else though
<mnemoc> get an "office" in .hk and ship from there ;-)
<mnemoc> miniand does it that way
mab has joined #linux-sunxi
<ssvb> hramrach_: can you check what VE frequency is requested for your problematic videos here: https://github.com/linux-sunxi/linux-sunxi/blob/sunxi-v3.4.43-r0/drivers/media/video/sun4i/sun4i_cedar.c#L614 ?
<drachensun> really
<drachensun> well that actually would make since
<ssvb> hramrach_: the kernel code is rounding the requested frequency down, for example the Sintel trailer asks 180MHz, but actually gets 160MHz
<mnemoc> offshore hsbc account in .hk, easy paypal, easy payments to mainland
<drachensun> I looked at setting up registered foreign corporation office in China and it was beyond my current means. I forgot Hong Kong has its own different set of laws though
<mnemoc> and an ape doing the posting
mab has quit [Client Quit]
mab has joined #linux-sunxi
<mnemoc> exactly, .hk != .cn
<mnemoc> not here, not there. just benefits
<ssvb> hramrach_: it might be a bit tough if some video needs 230MHz for decoding but also merely gets 160MHz :)
<ssvb> mnemoc: what is the latest code drop from allwinner?
<mnemoc> not sure
<mnemoc> been very disconnected the last... months
<mnemoc> ssvb: the last leak with history I know of are the -dev branches in my github
<mnemoc> (commits from march 1st 2013)
<mnemoc> but I think there are newer history-less SDK dumps
<techn_> mripard: Great. :) :* It seems to actually be a Mentor Graphics Inventra USB Controller (musb), that already support for it in mainline kernel, so it only needs a thin layer to adapt it to sunxi. Moreover, it's already supporting the PIO mode, so we could avoid relying on DMA to merge it. (Thanks to Arnd Bergmann for noticing
<mnemoc> :o
<Turl> techn_: that means we get usb for free? :D
<Turl> techn_: your message got truncated by freenode I think btw :P
<ssvb> mnemoc: thanks, IOCTL_SET_VE_FREQ still looks kinda shitty, but at least they are trying to use clk_set_rate with better granularity - http://git.rhombus-tech.net/linux?p=linux.git;a=blob;f=drivers/media/video/sun6i/sun6i_cedar.c;h=ec8bd60e91f1bd8d495849f0edb0b8afbd0eacdf;hb=refs/heads/allwinner-sunxi-a31#l672
<mnemoc> ssraw import or has history?
<mnemoc> I'm rebasing a20-dev upon 3.4 currently, a31 next
<mnemoc> (reference-3.4 first, not sunxi-3.4)
<ssvb> do you have any hardware to test it already?
<mnemoc> nope
<mnemoc> just git/C dancing
<ssvb> heh, shooting in the dark
<mnemoc> as usual
mab has quit [Ping timeout: 248 seconds]
<ssvb> PiyushVerma: did increasing CedarX VE clock frequency help?
<mnemoc> pretty much sun6i at this point. they branched a20 later
<hno> drachensun, you generally need to sign an NDA to get access to more documentation like the user guide etc, if there exists an user guide for A31..
* mnemoc wonders what A31 docu mripard got
<hno> User guide have not been included in any devboard shipmends for any AW CPU as far as I know, just the briefs + SDK + some PCB layout stuff.
<mnemoc> yes, but he received the A13 user guide without buying anything
<mnemoc> publicly avaialble in dl.linux-sunxi.org these days
<mnemoc> so *maybe* he can/did get A31 too
<hno> Yes, and they know it's leaked so considered public.
<mnemoc> afaik they authorized the distribution of that one
<hno> before or after it was published on linux-sunxi.org?
<mnemoc> before
<mnemoc> mripard gave it to me after AW authorized it
<hno> good.
<hno> wonder if there is updated A10 user gude as well. The A13 one is more complete..
<hno> filling in several blanks on the A10.
<Turl> hno: the latest I've seen had the holes too
<mnemoc> they probably NDAs signed forbidding them to document those holes
mdfe has joined #linux-sunxi
wingrime has joined #linux-sunxi
<hno> mnemoc, does not explain A13 manual filling ins several holes.
Yaku321 has quit []
<mnemoc> .oO
<mnemoc> maybe(tm) they got authorized after writting the A10's :p
<Turl> or maybe they got one guy to write it and he just took the A10, s/A10/A13/ and as he got bored then, starting documenting other hw blocks :)
<mnemoc> :)
<drachensun> ok it started to boot
<drachensun> I got Error: unrecognized/unsupported machine ID (r1 = 0x00000f34)
<mnemoc> backport of who knows what for who knows why
nove has joined #linux-sunxi
<techn_> mnemoc: 3.4 was first kernel to support a7/a15?
<mnemoc> that would explain the bulk backport to 3.3
<mnemoc> but doesn't explain *why oh why* 3.3
<mnemoc> and their backport includes typo fixes to things mainline still has in 3.4.43....
<mnemoc> so it's from >3.4
mab has joined #linux-sunxi
<techn_> line 740
<techn_> 3.3 seems to have somekind a15 support.. maybe little support for a7
<mnemoc> hope to push this rebased commit soon for others to review. I'm already dizzy
<techn_> mnemoc: how much there is patches to kernel parts?
LoCoZeNoz_ZUE has joined #linux-sunxi
<mnemoc> techn_: git diff amery/import/lichee-3.3/base..amery/import/lichee-3.3/a20-dev (and same for a31-dev)
<mnemoc> that base is mainline + android, and both -dev are alrady rebased upon it
<mnemoc> so that diff is the sun[67]i-only part
<mnemoc> http://sprunge.us/LZZc a20-dev --stat
<wingrime> mnemoc: you have new leak?
<mnemoc> and a31's http://sprunge.us/WiEA
sanka has joined #linux-sunxi
<mnemoc> wingrime: no. I'm rebasing the impot/lichee-3.3/a20-dev upon the 3.4 base of our sunxi-3.4
<techn_> mnemoc: I ment to between lichee-3.3 and android-3.4 :p
<techn_> But I can diff my self :)
<mnemoc> that base includes android-3.3
<Turl> sound/soc/sun7i/spdif.rar | Bin 0 -> 365402 bytes
<Turl> .rar? :)
<mnemoc> *g*
<mnemoc> after this pass is finished, next will be unremoving stuff
<mnemoc> that blob was probably introduced after deleting .c
mab has quit [Ping timeout: 260 seconds]
<mnemoc> arch/arm/mach-sun6i/pm/standby/gen_check_code.exe | Bin 0 -> 163924 bytes <--- nice one too
Tartarus has quit [Excess Flood]
Tartarus has joined #linux-sunxi
<drachensun> well, if I turn of SMP
<drachensun> and almost all the drivers
<drachensun> it boots the A31
<mnemoc> *g*
<mnemoc> it's pretty much a very early a31 support
<mnemoc> VERY early
<mnemoc> just for curiousity
<drachensun> with how the build system works its sort of a pain to swap in kernels but I can try it
<drachensun> is that your applying the kernels from the A20 branch back to the stable?
<mnemoc> almost
<drachensun> applying the commits I mean, ugh
<mnemoc> I'm applying them over the not-sunxi base of sunxi-3.4
<mnemoc> that way I don't have to fight merging sunxi drivers (yet)
<drachensun> I gotcha
<mnemoc> only the difference in mainline+android bases
<mnemoc> 364 commits to go, so don't expect anything fancy :p
<mnemoc> :)
<rz2k> that exact commit
<rz2k> never thought I find it
<mnemoc> those will be gone on pass2
<mnemoc> noise removal
<mnemoc> i'm probably bad on the head because I kind of enjoy cleaning legacy junk
<drachensun> what exactly are they up to there?
<mnemoc> drachensun: removing legal responsibility
<mnemoc> it's allwinner written backward
<mnemoc> and upside down
<mnemoc> they have files owned by s/allwinner/newbie/
<drachensun> I see
<mnemoc> weird chinese management ideas
<wingrime> mnemoc: don't forget that you take a responsibility not removing legasy )))
<wingrime> (crap unused code
vinifm has joined #linux-sunxi
<mnemoc> wingrime: "documentation" :p
<mnemoc> wingrime: give me functional code that provides the same knowledge, then remove the old junk
<wingrime> I waitinh new leaks with docs
<wingrime> it looks thay use some 3rd nand IP vendor
<wingrime> most "cool leak" will be verilog files
<wingrime> it can exmplain everything without any docs, and also we can emulate it on Icarus Verilog
<vinifm> i got vlc error: http://pastebin.com/Uw5PZX1Q
<vinifm> sunxi-3.0.62 had no this error
<vinifm> sunxi-3.0.62 did not have this error
<mnemoc> and what are you trying now?
<vinifm> to play same video.mp4(720p)
<mnemoc> what kernel
<vinifm> i mean, I'm doing the same test
<vinifm> before sunxi-3.0.62, now sunxi-3.0.76
<mnemoc> if you found a regression you need to narrow the range of commits between "works" and "doesn't work"
<mnemoc> test the last .62 tag please
<hno> Nice, been hunting a fel boot issue for last days and in the end it turned out there was no bug, just need to wait a little for the DRAM initialization to complete before trying next fel command.
<mnemoc> vinifm: sunxi-v3.0.62-r2
<mnemoc> hno: :D
<hno> apparently A13 takes longer to initialize DRAM than A10.
<mnemoc> the a20 8250 code has loops trying commands up to 1000 times
<hno> serial driver?
<mnemoc> yes
<hno> that's normal Allwinner busy-waits I think.
<wingrime> look like workaround
<hno> where is it?
<mnemoc> append ?w=1
mdfe has quit [Disconnected by services]
<techn_> vinifm: there is two dma patches.. you could try without them?
mdfe_ has joined #linux-sunxi
<vinifm> how do I do that? :)
<mnemoc> techn_: if he is using .76, he already has those
<mnemoc> doh. "without them"
<techn_> havent tried git-revert :/
<mnemoc> it makes a commit for the reverse patch
<hno> mnemoc, right. A normal allwinner busy-wait. It needs to wait for the UART to finish any pending transmission before DLAB can be set.
<mnemoc> something similar can be happening with your a13 dram init
<mnemoc> it's "busy" doing something else
<vinifm> git revert commit=?
<mnemoc> yes, in reverse order
<mnemoc> git revert $hash
<hno> mnemoc, it wasn't the DRAM init that failed. It was the shell script that was too impatient.
<mnemoc> :D
<hno> Fix: sleep 1 after fel exec of the dram code.
<mnemoc> *g*
<hno> Had me deeply worried for a while that there was issues in the DRAM code.. ahttps://github.com/linux-sunxi/sunxi-tools/commit/e42ec9b7ad88a2f1743103c1735382858bd968b7
<mnemoc> looks fair :)
<mnemoc> :o
<hno> really should go into u-boot tree as it's all u-boot code + stubs..
<mnemoc> sure
<mnemoc> does sunxi-current support usb?
<hno> it's been one of those odd progams that just gets code cut out from it the more it's worked on.
<hno> no.
<hno> should be easy to add the host controller I think, but the OTG controller needs to be documented a bit first.
<mnemoc> i believe we should point toward uboot/fastboot for reflashing
<mnemoc> either with the installed uboot or one pushed via fel
<hno> there is no fastboot in u-boot.
<mnemoc> ours or mainline?
<hno> neither.
<mnemoc> what other standard-ish flashing methods exist?
<mnemoc> or all ad-hoc
<hno> There is a full mtd subsystem in u-boot for flashing.
<mnemoc> ah, cool
<mnemoc> `mtd` -> usb -> uboot -> mtd-sunxi ?
<mnemoc> i mean, the goal
<hno> have not yet integrated the sunxi mtd driver. Needs more work.
<mnemoc> right, but what's the ultimate goal regarding flashing?
<hno> fel -> boot linux -> do whatever you want.
<mnemoc> so a linux installer
<hno> yes.
<mnemoc> i hoped for a more 'raw' method for repartitioning and writting/reading partitions over usb
<hno> but it will be possible to flash from u-boot as well if you prefer.
<mnemoc> that way all the fanciness remains on the "PC"
<hno> You can still do that with a linux installer.
<mnemoc> sure
<hno> but with a linux installer you get many more options on how the install should be done.
<mnemoc> tue
<mnemoc> true
<hno> And the exact same installer can also be booted using sdcard.
<mnemoc> and u-boot doesn't get fatter
<hno> it doesn't anyway.
<mnemoc> like for adding usb support
<mnemoc> or "fastboot" protocol or fs creation
<hno> usb gadget support is something I like to add regardless, but for providing u-boot console without need of an uart.
shineworld has joined #linux-sunxi
mashku has joined #linux-sunxi
<mashku> Hi
<hno> hi mashku
<mashku> will mk802 also run gk802 ?
<mashku> quad-core i.MX 6 SOC
<hno> there is too many variants of both.
<mashku> oh
<mashku> I was wondering if I can run Debian on it
<mnemoc> mashku: this channel is about Allwinner SoCs, not freescale's ;-)
<mnemoc> but yes, you can run debian on it
<mashku> mnemoc, I also have Cubiebox ;)
<hno> mashku, so ask about things relevant to Cubiebox :)
<mashku> but I have already Debian up and running nicely n Cubie
<mashku> :)
<mashku> and I want the same on this Freescale gadget
<mashku> mnemoc, do you place to release Cubie with four cores ? ;>
<mashku> plan
<mnemoc> i'm just a user
<mnemoc> i'm not involved at all in cubietech
<mnemoc> but afaik they plan a a31-based "miniserver"
<mnemoc> which is quad a7
<mashku> oh
mab has joined #linux-sunxi
<mashku> I use cubie for server, to host my home website
wingrime has quit [Ping timeout: 264 seconds]
<rm> mashku, go to #imx6-dongle
<rm> for the gk802
<rm> and yes there's debian for it released a couple of days ago
<mashku> I saw this website
<mashku> but its without accelerated Xserver
<mashku> I belive?
<rm> I don't know
<mashku> oh by the way.. how to connect to it via serial port?
<rm> ask further questions on #imx6-dongle
<mnemoc> mashku: wrong channel
<mashku> ok
<ssvb> mashku: do you want xserver acceleration to make your home webserver serve pages faster? ;)
<mashku> nah!!
<mashku> I wanted to play movies on one of them:)
<mashku> cubie is fine for web server
<hramrach_> ssvb: how do I check the frequency?
<hramrach_> anyway, it plays with native libve\
<hramrach_> hmm, looks like I need to get new kernel
<hramrach_> still at 3.4.29
dragonn has quit [Quit: Wychodzi]
dragonn has joined #linux-sunxi
<ssvb> hramrach_: I mean the files which play choppy
<vinifm> the same error occurred only one time, using sunxi-v3.0.62-r2
<mnemoc> vinifm: try to catch wingrime. he did the dma unification
<ssvb> hramrach_: as for https://github.com/hramrach/video-samples/commit/51fe08b21efd3922c3d4101cb327759cb8d78c48 - I get "Error: Decode failed(-5), try to reset decoder!" with VLC and libhybris wrapper
<ssvb> hramrach_: which means VRESULT_ERR_UNSUPPORTED = -5, //* failure caused by not supported stream content;
<ssvb> hramrach_: to rule out the possibility of a regression in android blob, probably it's a good idea to also try it in the native android player
<techn_> we could have too old kernel module for cedarx
<techn_> I think I saw some patches somewhere :/
<ssvb> techn_: at least the ioctl setting the VE clock frequency seems to be wrong
<hramrach_> ssvb: I don't have android
<hramrach_> need more sd cards I guess
<ssvb> hramrach_: ok, I'll try to test it myself in a few minutes (I have android in NAND)
<hramrach_> I have Debian there because I tested nand installation ;-)
<ssvb> :)
dragonn has quit [Quit: Wychodzi]
dragonn has joined #linux-sunxi
<hramrach_> printk building ..
<ssvb> techn_: thanks, looks somewhat less broken
<techn_> but that's sun5i.. hope it works with sun4i too :)
<mnemoc> note the a10s hacks don't necesarily work on a13
<mnemoc> that's one of the reasons for the soc-detect thingy
<mnemoc> they have separated mantainers for each soc
<ssvb> it also looks like they bump the default VE clock frequency from 160MHz to 240MHz in this patch
<ssvb> or maybe not (there are no references to 160MHz there), still it feels like 240MHz should be faster :)
<hramrach_> I get IOCTL_SET_VE_FREQ set ve freq 240, ve clock 960000000/4
<hramrach_> the initial VE clock is 720 and the zeroes and the rounding presumably assumes that but there is code for reading the freq form somewhere for that calculation
mab has quit [Ping timeout: 276 seconds]
<hramrach_> hmm, you get the exact clock for 240 requested and 960 the VE freq
<hramrach_> but does the lib ever request anything but 240?
<ssvb> hramrach_: hmm, was 240 in the "arg_rate" variable?
<hramrach_> yes
<ssvb> hramrach_: I get 180 for the Sintel trailer
<hramrach_> and 960 in the variable that gets divided
<hramrach_> and it's initially 720
<hramrach_> which would give 180 for /4
<hramrach_> but I have 960 which gives exactly 240 for /4
<ssvb> which they claim to be 160 in the comments
<ssvb> but 240 in "arg_rate" variable probably indicates that libvecore thinks that this particular video really needs higher VE clock frequency
<hramrach_> I get
<ssvb> have you also tried other videos?
<hramrach_> I get 240 req for sintel too
<ssvb> weird
<hramrach_> I am runnig vlc
sameerynho has joined #linux-sunxi
<ssvb> maybe bump it to 320 and check if this makes any difference?
<ssvb> hopefully 320 is within valid range and cedarx will not fry
<hramrach_> this is quite a mess
<oliv3r> hno: btw, do you want me to rebase and resubmit my rename patch for u-boot?
<ssvb> btw, I tested cedarx-no-output.mkv in native android - does not seem to work (shows no preview, immediately jumps to the next video if I attempt to play it)
<hramrach_> on L1030 pll4clk_rate is set
<hramrach_> a bit higher for you withjout the debug prints
<vinifm> there is another error, but do not know if it's bad code: http://pastebin.com/NrxBq2Ts
<hramrach_> hmm, so it's broken with android libve and neither lib works completely then
<ssvb> maybe I could try older android image, but not sure if I want to reflash the device right now
<ssvb> pick your poison, they seem to fix and break things arbitrarily :)
<hramrach_> anyway the comment about 160 make it clear that the freq is supposed to be 960 and hte 720 initialized at the top of the file is bogus
<hramrach_> so the difference in requested frequency is probably in the libve
paulk-desktop has quit [Quit: Ex-Chat]
<hramrach_> yes, the android libve requests 180
<techn_> ## Checking Image at 4c000000 ...
<techn_> Unknown image format!
<ssvb> hramrach_: so in fact the linux blob was clocking VE at 240MHz, but the android blob asks 180MHz and gets 160MHz
<sameerynho> does any one works on IR driver ?
<ssvb> hramrach_: seems like bumping it at least to 240MHz should be safe
<techn_> sameerynho: no. but someone send some IR driver to ML
<hramrach_> yes, don't have any video sdample at hand that makes it ask different clock
<mnemoc> the same guy who made the gpiolib driver iirc
<hramrach_> asking 240 gives you the second highest clock
<mnemoc> but never sent the v2
<hramrach_> fastest is /3 but the blob does not request that
<sameerynho> techn_: hmm, does that work good ?
<techn_> sameerynho: dunno
<ssvb> hramrach_: now we only need to confirm if increasing the VE clock speed helps with the choppy videos or not
<sameerynho> techn_: i want to work on A10 supports on kernel, i want a field to zoom on
<hramrach_> the one that causes fpe in vlc is choppy at 240
<sameerynho> techn_: i was thinking about IR driver
<techn_> sameerynho: great :)
<mnemoc> sameerynho: there is a sun4i-ir driver already
<hramrach_> could hack kernel to bump hte freq to /3 and see if that make a difference
<sameerynho> mnemoc: so we have to update the todo list
<hramrach_> but it's _really_ bad so don't have much hope
bsdfox_ has quit [Read error: Connection reset by peer]
<mnemoc> sameerynho: ?
<ssvb> hramrach_: yes, I would search for a different problematic sample
<mnemoc> sameerynho: ok. the *legacy* tree has an IR driver, mainline doesn't
<mnemoc> legacy = script.bin-based, 3.0 or 3.4.
<sameerynho> mnemoc: so we have to make changes and make it ready for mainline
<mnemoc> yes
<techn_> mainline is already accepting sunxi patches :)
<ssvb> hramrach_: in fact it was PiyushVerma who originally reported the slow playback problem (fast in native android, slow in linux with libhybris), so he must have some good sample video to test
<sameerynho> mnemoc: nice, does sunxi repo contains any branch the synced with mainline
<mnemoc> sunxi-next
<hramrach_> I only observe slow playback with everything
<hramrach_> on the video that causes fpe in cedar and pretty much anything with HD res and much movement
<sameerynho> mnemoc: is there any document about sunxi repo workflow ?
<ssvb> hramrach_: hmm, so does increasing VE clock frequency help with these HD resolution videos with much movement?
<mnemoc> sameerynho: sunxi-next is special, it's rebased often.
<hramrach_> I am not sure the encode is good
<hramrach_> never seen it play back smoothly on anything
<sameerynho> mnemoc: so sunxi don't use pull requests ?
<mnemoc> sameerynho: patches are sent to the mailing list
<mnemoc> sameerynho: first to the sunxi ML for review by other sunxi devs
<sameerynho> mnemoc: so which branch should i push my commits ?
<hramrach_> but on x86 it decodes in software which is not expected to work and on sunxi it decodes on cedar which is not flawless either
<mnemoc> sameerynho: no
<sameerynho> mnemoc: thanks
bsdfox has joined #linux-sunxi
bsdfox has quit [Changing host]
bsdfox has joined #linux-sunxi
<hramrach_> any good way to acelerate h264 playback?
<mnemoc> sameerynho: that would bypass most of the devs. just like with mainline, or arm-soc, the sunxi repo only takes things posted on the ML
<mnemoc> for proper community discussion
<hramrach_> the multithreaded ffmpeg is promised for ages but not hitting any actual distro afaik
<sameerynho> mnemoc: thanks man , i'll join the ML
<techn_> sameerynho: code is your documentation.. and there is couple datasheets.. http://dl.linux-sunxi.org/
mdfe__ has joined #linux-sunxi
<mnemoc> techn_: he means documentation about the development workflow, not hardware
<techn_> oh.. missed the last word ;)
<mnemoc> :)
<sameerynho> mnemoc: techn_: thanks guys
<oliv3r> sameerynho: note though, that IR needs certain clocks, that aren't in mainline yet; Turl is working on the clocks
<techn_> mnemoc: does dev@linux-sunxi.org work yet reliable?
<mnemoc> techn_: i think so, but it's a humble alias
<mnemoc> eventually I'll install a proper postfix on the sunxi server and then mailman
<sameerynho> oliv3r: Clocks ?
mdfe_ has quit [Ping timeout: 260 seconds]
<mnemoc> but not there yet
<mnemoc> s/alias/forward/
<techn_> I tried that when it was announched.. it didn't work then :/
<mnemoc> some domains don't allow forwards
<mnemoc> fancy anti-spam systems
Tartarus has quit [Excess Flood]
<mnemoc> so G might trash your mails if they don't come from the right server
Tartarus has joined #linux-sunxi
<oliv3r> sameerynho: clocks :)
<mnemoc> little things that control timing
<oliv3r> currently we only have OSC24, OSC32k and PLL1
<oliv3r> if you want something easier; there's wiki that needs lots of work done, mostly on the memory map
<oliv3r> or of course, sun3.4 driver unification
<oliv3r> not to mention, sun[67] driver porting
nove has quit [Quit: nove]
<sameerynho> oliv3r: what do you mean by unification ? i'm not good in english
<oliv3r> currently, there's several drivers, for the exact same hardware
<oliv3r> for example, sun4i-snd and sun5i-snd
<oliv3r> those 2 drivers are 99.9% identical
<oliv3r> and can be unified to sunxi-snd
<sameerynho> oliv3r: aha
<oliv3r> that one has been done actually ;)
<sameerynho> oliv3r: nice its good for start
<mnemoc> oliv3r: but has the wrong name :(
<sameerynho> oliv3r: :D so give me an undone one
<oliv3r> mnemoc: i do that with my next rebase fix patch
<mnemoc> oliv3r: cool
<oliv3r> sameerynho: mnemoc knows better what needs to be done
<mnemoc> mnemoc needs to go to sleep :(
<oliv3r> what driver still needs unification?
<sameerynho> mnemoc: sorry for bothering you, but is there any todo to give me for starting ?
<mnemoc> but search for drivers suinx sunNi instead of sunxi in our tree
<mnemoc> using*
<mnemoc> there are a bunch
<oliv3r> git grep sun4i
<mnemoc> uhm... more of git ls-files drivers | grep sun4i
<sameerynho> thanks guys
<oliv3r> my git-mojo is weak
<mnemoc> find drivers | grep sun4i :p
<oliv3r> sameerynho: also, try setting up the sunxi-bsp (see github) and try to build the kernel :)
<mnemoc> and sun5i
<mnemoc> good night
<oliv3r> nn mn
<sameerynho> oliv3r: unfortunately i can't use bsp because of my poor internet connection
<sameerynho> mnemoc: good night
<oliv3r> sameerynho: the bsp is reallys mall
<oliv3r> kernel + u-boot are huge :)
<sameerynho> oliv3r: i see, i had a huge trouble with download them :D
<oliv3r> git clone was giving you trouble?
<sameerynho> oliv3r: yeah specially on kernel tree
<oliv3r> well you'll need to compile to test your patches :p
<mnemoc> I have some xz'ed git clones in dl.
<mnemoc> as starting point
<oliv3r> dl.linux-sunxi.org/users/*
<sameerynho> mnemoc: i made them by my own on my server
<oliv3r> sameerynho: what hardware do you have
<sameerynho> god that was a pain in the a**
<mnemoc> __git is the .git only
<mnemoc> the other also have the files
<sameerynho> oliv3r: A10
<mnemoc> good night for real now
<sameerynho> mnemoc: ;)
<oliv3r> cubie? tablet
<sameerynho> oliv3r: cubie
<oliv3r> good :)
<sameerynho> oliv3r: i'll have to go to sleep too, but is it ok to talk to you tomorrow about the TODO ?
<oliv3r> but eventually you'll need git too
dragonn has quit [Ping timeout: 248 seconds]
<oliv3r> sure
<sameerynho> oliv3r: thanks and good night my friend
<oliv3r> nn
sameerynho has quit [Quit: Leaving]
rz2k has quit [Read error: Connection reset by peer]
<rm> hm, on the GK802 u-boot loads just a regular zImage
<rm> which can even be on Ext4 and be a symlink to a real file
<rm> someone remind me why we still bother with FAT partitions and uImages
<mnemoc> rm: our u-boot supports ext34 too
<rm> and uImage vs zImage?
<mnemoc> no idea
* mnemoc gone
<oliv3r> mnemoc: you still here?
<oliv3r> there's a reason for fat rm
<oliv3r> our u-boot can also be loaded from the fat its on!
<oliv3r> but probably compatibilty reasons with stock u-boot, fat nanda etc
shineworld has quit [Remote host closed the connection]
LoCoZeNoz_ZUE has quit [Quit: Leaving]
sanka has quit [Quit: Leaving]
vinifm has quit [Quit: Saindo]
<Turl> oliv3r: I hope IR clocks are ready for 3.11
hglm has joined #linux-sunxi
<hno> and felboot source continues to shrink.. looks like it will just be a linker script when integrated in u-boot. Only two stubs left from being built with the full SPL objects.
<Turl> hno: great
<hno> rm, standard u-boot really wants uImage format for kernel & ramdisks.
<hno> but there is no need for fat.
tinti has quit [Read error: Connection reset by peer]
<Turl> techn_: ping
<oliv3r> hno you are the best :D
<hno> rm, ok, found how to enable zImage support. Was a couple of undocumented config defines. Please test if it works in sunxi-current.
mashku has quit [Quit: Leaving]
<Turl> oliv3r: what device are you hacking on these days? cb?
<oliv3r> Turl: now that i have one yeah; but i allready broke the PHY :p
<oliv3r> well me, or it was broken; but only got to work a few times :(
<oliv3r> or, the 48 MHz frequency on the MDC broke it; though hno fixed that in sunxi-next
<oliv3r> _might_ need a fix in regular emac too; but haven't found it
<Turl> MDC?
<oliv3r> Media II clock
<oliv3r> MII clock
<Turl> oh
<oliv3r> pin 25 on the phy
<oliv3r> before emac is init; it outputs some random 48 MHz clock there
<Turl> doesn't work on uboot either oliv3r?
<oliv3r> nope
<oliv3r> datasheet says, max. 2.5 MHz on that clock pin
<oliv3r> after we init it, it outputs 833.3 MHz
<Turl> that's one hell of an overclock :)
<Turl> oliv3r: what clock is it on CCM?
<oliv3r> well i think that's default for everybody
<oliv3r> the EMAC is on AHB
<Turl> ahb can't work at 800Mhz
<oliv3r> 833.3 kHz
<oliv3r> my bad
<oliv3r> what's the default AHB freq?
<Turl> it depends on the cpu freq
<oliv3r> during u-boot
<Turl> uboot bumps it to 1Ghz and 3:2:2 iirc
<oliv3r> 3:2:2?
<oliv3r> you mean devidor
<Turl> AXI, AHB, APB divs indeed
<oliv3r> they have shared dividors? (I forgotten long long how it all goes together)
<oliv3r> which is stupid, since I wrote the CCM page :S
<oliv3r> they are sperate!
<Turl> oliv3r: yeah but they're chained
<Turl> AHB is axi/n
<oliv3r> i'm browsing, see if i see anything
<oliv3r> you use the wiki page at all?
<Turl> I used it when I started
<Turl> nowadays I pull the user manual pdf when I need to check something mostly
<oliv3r> oh :(
<oliv3r> well the wiki is based on usermanual, headers and code
<oliv3r> so _should_ be more extensive
<Turl> oliv3r: yes, 3, 2, 2 there
<oliv3r> i haven't put the a13 usermanual next to it
<oliv3r> also, small errors in the usermanual, are fixed in the wiki
<oliv3r> so wiki should be better :(
<Turl> oliv3r: the pdf is local though :P
<Turl> oliv3r: http://linux-sunxi.org/A10/CCM#CCM_PLL5_CFG spot the mistake :)
<oliv3r> but I don't wanna have it written for nothing! :p
<oliv3r> M1 is dupe
<oliv3r> comment for M is wrong
<Turl> hadn't noticed the M comment heh :)
<oliv3r> but CCM_PLL5 doesn't even list M1
<oliv3r> in the math bit
<Turl> oliv3r: yeah I still dunno what M1 is :)
<oliv3r> so might be an error in one of the sheets etc
<Turl> I asked mripard, he was gonna see if he could find out
<oliv3r> oh AND i used the 'datasheet' also for making that page!
<Turl> hno: do you know what M1 is? you should've played with PLL5 a bit, it's the DDR clock
<oliv3r> we should add 'download as pdf' button
<Turl> that would be cool
<oliv3r> maybe M and M1 are 'combined'
<oliv3r> spent way to much time on that page, think 2 1/2 if not 3 days
<Turl> oliv3r: I'm thinking one output does /M/P and the other does /M1/P or something
<oliv3r> if you have 2 chips? or 2 banks?
<oliv3r> fixed the wiki :p
<Turl> there's one output to ddr and another to 'others' according to the pdf
<oliv3r> yeah, CCM_PLL5 does list that
<oliv3r> but 'others' is 24 MHz * N * K / P
<oliv3r> no M there :)
<Turl> oliv3r: and no P on ddr
<oliv3r> so that theory doesn't hold :p
<oliv3r> let me do a quick code browse
<Turl> hno is always afk when this topic comes up :P
<oliv3r> lol
<Turl> no M1 there
<oliv3r> yeah
<oliv3r> but does exist :)
LoCoZeNoz_ZUE has joined #linux-sunxi
<oliv3r> those two references are the only ones I can find with git grep FactorM1
<oliv3r> so unless they use it via some other name/trick; that's it
<oliv3r> that one looks even better; single page, 'book' etc