Turl changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask and wait! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi | #blameOliver
Turl has joined #linux-sunxi
Turl has quit [Changing host]
Turl has joined #linux-sunxi
Turl has quit [Client Quit]
Turl has joined #linux-sunxi
Tsvetan2 has joined #linux-sunxi
Tsvetan has quit [Ping timeout: 245 seconds]
Turl has quit [Remote host closed the connection]
Turl has joined #linux-sunxi
eebrah_ has quit [Read error: Connection reset by peer]
dalee has quit [Ping timeout: 250 seconds]
lump has quit [Ping timeout: 246 seconds]
awatch has quit [Ping timeout: 248 seconds]
eebrah_ has joined #linux-sunxi
<wens> arokux: the A20 doesn't have the pins to do full GMII
<wens> arokux: it can do MII or RGMII
pfdm has joined #linux-sunxi
<wens> arokux: i like the structure of stmmac, looks extensible, maybe make stmmac-sunxi, like dw_mmc -> dw_mmc-socfpga
<wens> arokux: if i can get it to work with gmac, i'll pull sunxi bits into an extension driver
hramrach has joined #linux-sunxi
<wens> wonder how i should do the gmac clock. the bits actually depend on whether MII or RGMII is used. and i don't think it's gated.
froy has joined #linux-sunxi
wingrime has joined #linux-sunxi
eebrah_ has quit [Ping timeout: 244 seconds]
guizamboni has quit [Quit: Leaving]
wingrime has quit [Ping timeout: 244 seconds]
popolon has quit [Quit: Quitte]
Seppoz_nowork has quit [Ping timeout: 264 seconds]
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
egbert has quit [Disconnected by services]
egbert_ has joined #linux-sunxi
Sonicadvance1 has quit [Remote host closed the connection]
Sonicadvance1 has joined #linux-sunxi
forcev has quit [Ping timeout: 240 seconds]
FunkyPenguin has joined #linux-sunxi
fredy has quit [Excess Flood]
fredy has joined #linux-sunxi
awafaa has quit [Ping timeout: 264 seconds]
awafaa has joined #linux-sunxi
pfdm has quit [Ping timeout: 250 seconds]
eebrah_ has joined #linux-sunxi
eebrah_ has quit [Read error: Connection reset by peer]
eebrah_ has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
eebrah_ has quit [Ping timeout: 248 seconds]
eebrah_ has joined #linux-sunxi
wolfy has joined #linux-sunxi
tzafrir has quit [Ping timeout: 245 seconds]
eebrah_ has quit [Read error: Connection reset by peer]
eebrah_ has joined #linux-sunxi
eebrah_ has quit [Read error: Connection reset by peer]
eebrah_ has joined #linux-sunxi
HeHoPMaJIeH has joined #linux-sunxi
eebrah_ has quit [Read error: Connection reset by peer]
eebrah_ has joined #linux-sunxi
Quarx has joined #linux-sunxi
atiti has joined #linux-sunxi
Dapsaille_Sleep is now known as Dapsaille
<Dapsaille> o/, morning
huosen is now known as Amitabha
rellla has joined #linux-sunxi
ZetaNeta has quit [Ping timeout: 264 seconds]
BluesBoy has joined #linux-sunxi
ZetaNeta has joined #linux-sunxi
<plaes> \o
<mnemoc> moin
<oliv3r> morni'
<oliv3r> arokux: what's your switch say? :)
<oliv3r> wens: i don't think that's true, EMAC was MII only (no RMII i think) GMAC uses the same pins, so is (G)MII compatible
<oliv3r> the PHY! only has the RGMII pins
<oliv3r> wens: actually, stmmac -> designware; disignware < sun/st
<oliv3r> wens: see how i2c was done by maxime
hramrach has quit [Remote host closed the connection]
deasy has quit [Ping timeout: 244 seconds]
hramrach has joined #linux-sunxi
Quarx has quit [Ping timeout: 246 seconds]
<wens> oliv3r: GMII has 8 pins for tx/rx each, MII has only 4 each, A20 does not have tx4-7 rx4-7 connected
<wens> oliv3r: not following # stmmac -> designware; disignware < sun/st
<wens> oliv3r: A31 has all the pins for GMII
<oliv3r> wens: ah, i stand corrected, so GMAC supports MII, RMII, RGMII, GMII, but a20 lacks pins so doesn't do GMII
<oliv3r> i can agree with that ;)
<wens> :)
<oliv3r> we should rename stmmac to designware (as u-boot does)
<oliv3r> then add support for both stm and sunxi, just like mripard did for the marvel i2c controller
eebrah_ has quit [Ping timeout: 272 seconds]
<oliv3r> wens: i thought GMII and MII had the same number of pins, hense the confusing
deasy has joined #linux-sunxi
deasy has quit [Read error: Connection reset by peer]
_massi_ has joined #linux-sunxi
<mnemoc> the EMAC is available on two sets of pins
<oliv3r> on a20 yeah :D
<oliv3r> why can't I ever design a board :)
<mnemoc> focus has some nice docu about learning to do PCBs
<mnemoc> using oss obviusly
<oliv3r> i've done a few gEDA designs (basic ones)
<mnemoc> for using a SoM you only need to wire connectors
<oliv3r> but when you start to talk about DDR timings, antenna design etc, random decoupling caps etc you loose me
<oliv3r> sweet Ian got ahci to work, gotta diff his branch with mine and see what he changed
<mnemoc> the SoMs come with the DDR timings problem solved ;-)
<oliv3r> true
<oliv3r> it's like EOMA68 but diff
<oliv3r> :p
<mnemoc> each module has it's interface
notmart has joined #linux-sunxi
ZetaNeta has quit [Ping timeout: 245 seconds]
<mnemoc> olimex made some SoMs too
ZetaNeta has joined #linux-sunxi
<mnemoc> or look at the first cubie baseboard or the DAC thing... just turning extension headers into something useful
maz_ has joined #linux-sunxi
<mnemoc> sure on both there is a huge greed factor, but the design is very very simple
<oliv3r> yeah
<oliv3r> i must say I do appreciate the CT design
<oliv3r> it would fit nicely into a settopbox
<oliv3r> or PC box
<mnemoc> I want it on a VESA mountable case...
<mnemoc> for POS and kiosks
<oliv3r> yeah i've seen the DAC module; the price was outragous though; 300 USD
<oliv3r> mnemoc: you gonna open a kiosk? :p
<mnemoc> nah, but it's the perfect computer to hook behind an old TFT and make it productive
deasy has joined #linux-sunxi
<oliv3r> true
<oliv3r> but it'll need touchscreen :)
<mnemoc> office desktops, POSs, info terminals, etc etc
<mnemoc> sure
<mnemoc> but not always
<oliv3r> but yeah i can see potential for many such situations
<oliv3r> but seriously 300 EUro for a dac
<oliv3r> crazy
<oliv3r> unicorns pissed over
<plaes> o_O
popolon has joined #linux-sunxi
Seppoz has joined #linux-sunxi
Quarx has joined #linux-sunxi
maz_ has quit [Ping timeout: 272 seconds]
vanner has joined #linux-sunxi
rellla2 has joined #linux-sunxi
<mnemoc> oliv3r: regarding the gmac/mii thing, please look at the plat/script.h API
<oliv3r> mnemoc: no full link :(
<oliv3r> mnemoc: but what specificall do you mean?
<mnemoc> using that to query script.bin instead of aw's API
<mnemoc> well... wrong branch, but the file is the same
boycottg00gle has joined #linux-sunxi
<arokux> oliv3r: I'm connected directly to 100Mb router
<arokux> oliv3r: this is what I was able to get out of gmac under u-boot adding some additional bit poking
<arokux> oliv3r: however dhcp command failed...
<arokux> so I still wonder if it works.
<mnemoc> (in linux it does)
<arokux> maybe I should try designware first
<arokux> mnemoc: yep
<mnemoc> but random mac address is annoying :p
<arokux> wens: here is my update, take a look at previous messages ---^
<arokux> mnemoc: be glad it works :p
<mnemoc> I am!
<mnemoc> haven't needed to remove the uSD in a long time thanks to that
jemk has joined #linux-sunxi
<arokux> mnemoc: you overwrite kernel with ssh?
<mnemoc> yes
FR^2 has joined #linux-sunxi
<mnemoc> and weirdly, the CT writes my card faster than the laptop
<arokux> mnemoc: how can you tell that, it is only few MiBs?
<mnemoc> rsync'ing the modules
<arokux> mnemoc: ok
<mnemoc> which in most cases it's merely about updating the timestamp
<mnemoc> but still takes it's time
<arokux> mnemoc: what about publishing all those scripts? like I did here http://linux-sunxi.org/Possible_setups_for_hacking_on_mainline
<arokux> mnemoc: the title could be changed.
<mnemoc> I have a bunch at https://github.com/amery/git-import-help
<mnemoc> but those are for sanitizing and reconstructing trees
<arokux> mnemoc: they are not that interesting for everybody
arete74 has quit [Ping timeout: 245 seconds]
<mnemoc> I have the tendency of creating a lot of dump scripts...
<mnemoc> dumb
<mnemoc> that sprunge with built-in obsfuscation is pretty useful
<mnemoc> (or filtering)
Black_Horseman has quit [Remote host closed the connection]
<mnemoc> even if the script itself is trivial
<oliv3r> mnemoc: ohhh the gmac/emac mail for sunxi-3.4; yeah got it
JohnDoe_71Rus has quit [Read error: Connection reset by peer]
<torbenh3> arokux: do you have a git with your gmac driver handy ?
<arokux> torbenh3: sunxi-3.4 tree?
<mnemoc> stage 3.4
<torbenh3> yeah. or did it already land in sunxi-3.4 ?
<mnemoc> it's in stage/sunxi-3.4 until some issues get solved
<mnemoc> like proper mii handling
<mnemoc> or better defaults
<mnemoc> then it will be merged into sunxi-3.4
JohnDoe71rus has joined #linux-sunxi
<arokux> torbenh3: so just grab stage/sunxi-3.4
<torbenh3> got it.
<torbenh3> but there is a build error.
<arokux> ( torbenh3: and now you see how important it is to have something that works )
<arokux> torbenh3: pastebin it
<torbenh3> /media/x3/torbenh/sunxi/linux/arch/arm/mach-sun7i/pm/standby/../pm_types.h:14: error: redefinition of typedef ‘__s8’
<torbenh3> /media/x3/torbenh/sunxi/linux/include/asm-generic/int-ll64.h:19: note: previous declaration of ‘__s8’ was here
<torbenh3> sun7i_defconfig + gmac
<mnemoc> try removing the redefinition in mach-sun7i/pm/pm_types.h
<mnemoc> those errors are very toolchain specific.... my 4.6.3 doesn't even warn about it
<arokux> torbenh3: I compile with gcc-4.7.x, what is yours?
<torbenh3> -> arm-linux-gnueabi-gcc --version
<torbenh3> arm-linux-gnueabi-gcc (Debian 4.4.5-8) 4.4.5
<mnemoc> just remove the redefinitions
<torbenh3> all hell breaks loose then :)
<mnemoc> aw tends to be standby off-tree so they add some extra junk...
<mnemoc> s/be/build/
<arokux> mnemoc: maybe we should fix those redefinition to make it gcc 4.4 friendly? (patches welcome?? :p)
<mnemoc> of course we want to be 4.4 friendly
<arokux> but...?
<torbenh3> i turned off PM....
<mnemoc> so torbenh3 removes the redefines, builds, tests, submit fixes, and I apply them
<mnemoc> torbenh3: come on, just remove the extra lines
<oliv3r> never seen those at 4.6.3 either; but i recently updated my arm toolchain to 4.7 so i'm curious about new errors
<oliv3r> oh wow, that's an ol docmpiler, upgrade :)
<oliv3r> anything older then 4.6.3 doesn't deserve our time tbh :p
<mnemoc> the freaking file is even CRLF :\
<mnemoc> that's what I get by blindling pulling for-amery branches :<
<mnemoc> blindly*
binaryferret has joined #linux-sunxi
<oliv3r> lol who's branch was that?
<mnemoc> hansg
<mnemoc> torbenh3: try http://sprunge.us/IZBB
<oliv3r> bad hans!
tomee^ has joined #linux-sunxi
<torbenh3> mnemoc: i told you that all hell break loose if i do that.
<torbenh3> i use this: http://paste.debian.net/65588/
<torbenh3> but there is more brokenness in the pm..
<mnemoc> probably they miss another include
<mnemoc> we do want gcc 4.4 support, and we need YOU for that :)
arete74 has joined #linux-sunxi
<arokux> mnemoc: why do we want gcc 4.4 support?
<mnemoc> it's part of the code cleanup
<mnemoc> if it breaks 4.4 it means there is something wrong with the code
<mnemoc> proper C (gnu99) should build on any 4.x toolchain
<mnemoc> compiler weirdnesses help to disclose mistakes
<arokux> mnemoc: why it works for 4.7?
<mnemoc> works on 4.6 as well
<mnemoc> different compilers have different levels of strictness, or make different assumtions
<mnemoc> but proper code will compile on any
<arokux> mnemoc: ok.
<mnemoc> makes sense?
<tomee^> but there is also this float abi hard/softp/soft vs armhf/armel vs neon/vfpv3 weirdness
<mnemoc> that doesn't affect the C part of the kernel
<tomee^> gcc can be quite tricky on which assembly snippet it uses and how it uses it
<arokux> tomee^: take a look http://linux-sunxi.org/A20-Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<tomee^> mnemoc: oh, the kernel of course not.
<arokux> tomee^: what can you say about status of WiFi/BT?
<tomee^> mnemoc: in fact compiling kernel with any optimizations is a bad idea
<tomee^> arokux: yesterday been busy battling VDPAU
<tomee^> arokux: I can try recompiling the kernel now
<arokux> tomee^: do not rush into testing WiFi/BT, I thought you know the answer. but please update that status with some additional info.
<tomee^> I know that WiFi works.
<tomee^> ...in the cubie-supplied kernel ;)
<tomee^> no idea about BT yet
<tomee^> ok
<tomee^> should I use this branch?
<arokux> tomee^: also, you may want to ask ppl on the mailing list. lots are working on the same problems they just keep sitting quietly, but when they see something on the ML they respond.
<arokux> tomee^: Benn from Cubietech informed me they stick to https://github.com/cubieboard/linux-sunxi cubie/sunxi-3.4
<tomee^> yes but that's a fork
<arokux> tomee^: and I was asking about the status of stage/sunxi-3.4 - our kernel.
<tomee^> and you asked me to use your tree
<tomee^> that is linux-sunxi/linux-sunxi branch cubie-3.4 ?
<tomee^> erm, branch stage/sunxi-3.4 ?
<arokux> tomee^: yep
<arokux> tomee^: ethernet is there already.
<tomee^> ok
<tomee^> and cedar, which one? sunxi-cedar-mod or sun4i-cedar-mod ?
<tomee^> i've seen some differences in latest kernels
<arokux> I know nothing about cedar, mnemoc ?
<mnemoc> torbenh3: http://sprunge.us/TBPE works here
boycottg00gle has quit [Remote host closed the connection]
<mnemoc> no clue
<mnemoc> but in general we prefer unified drivers, sunxi-foo
<tomee^> ok
<tomee^> i will need to compare the cubieboard defconfing with your config parameters
<tomee^> to see if some differ
<tomee^> definitely not willing to go through menuconfig from scratch
<torbenh3> mnemoc: upps. overlooked the include linux/types.h... yeah. that one should work
<tomee^> ok, so the kernel tree is being checked out now
<arokux> mnemoc: what is this? "\ No newline at end of file"
<tomee^> meanwhile, any of you tried (wrote? :>) the vdpau driver?
<mnemoc> arokux: windows editors forget to add a new line marker at the end of the file, and vi fixes that
<arokux> tomee^: you should ask wingrime, ssvb rellla jemk etc.
<arokux> mnemoc: I see, but is this text present in C code?!
<mnemoc> no
<tomee^> are they around here sometimes? or only on mailing list/git/whatever ?
<mnemoc> `patch` deals with that
<jemk> tomee^: im here
<arokux> tomee^: yes, they are around... but you'd need to catch them.
<mnemoc> tomee^: jemk is one of the care cedarx devs
<tomee^> I see
<tomee^> anyway, I got it working, but now I am wondering if there's any way to use overlays with it
<tomee^> like subtitles
<binaryferret> If building android should I be using a different kernel branch than sunxi-3.4 for the kernel? mirror/android-3.4?
<tomee^> already implemented/planned or vlc->demux->vdpau->vram->memcpy->ram->overlays/compositing->memcpy->vram ...
<arokux> ( tomee^: you'd want to ping people so that they reply faster )
<jemk> tomee^: overlays aren't possible yet, i had tried it, but it needs a lot of memcpy which kills performance
<Turl> mnemoc: have you seen benn this week?
<jemk> tomee^: or it needs some way to interact with the xorg driver (or ignore it and take the whole screen)
<mnemoc> nope
<tomee^> jemk: neither with "pure" libcedarx?
<arokux> Turl: I had some e-mail communication with him
<mnemoc> Turl: notmart
<mnemoc> err
<mnemoc> notmart: sorry
<jemk> tomee^: i don't work on libcedarx, but it doesn't have overlays either
<tomee^> jemk: i see. so what's rendered inside the program window/fb coords cannot be in any way changed.
<tomee^> jemk: only possibility is to f*ck with memcpys which will bring the performance back to softdecoding or draw other stuff aside (e.g. subtitles below the video, given the video is not full-screen)
<tomee^> jemk: did I get that right?
<Turl> maybe you can use disp layers for that
<jemk> tomee^: yes, thats one way, the other would be to figure out some inteligent way of using disp layers, but as only two layers can be blended and the video layer already is the second one...
<jemk> for fullscreen it would be no problem at all, there video is back layer and subtitles front layer
<mnemoc> torbenh3: did that patch work for you?
<jemk> but in window mode desktop is back layer and video front and nothing left for overlay
enrico_ has joined #linux-sunxi
<torbenh3> mnemoc: i have other things todo.... testing it in the evening.
<tomee^> jemk: hmm, so in fullscreen 2 layers can be blended...
<jemk> for fully visible windows same as fullscreen, but such things can only be decided by xorg driver
<tomee^> jemk: you mean that 2 RGBA framebuffers are composited into one?
<mnemoc> torbenh3: np :) please let me know to push it
<jemk> tomee^: disp can blend two layers without copy, but only two
<tomee^> ok
<tomee^> so that would require 3 or 4 framebuffers?
<tomee^> 1 for widgets, 1 for video, another one for off-screen widget redraw = 3 ?
<tomee^> and then swapping 3->1 then compositing 1+2 ?
naobsd has joined #linux-sunxi
<tomee^> arokux: should your kernel work with the cubie bootloader? or your bootloader with the cubie kernel? or should I reinstall both at once?
<jemk> tomee^: in vdpau those are all handled by VideoSurfaces and OutputSurfaces, so that's no problem, the only problem is to bring those surfaces to display without copying them around
ganbold_ has quit [Ping timeout: 244 seconds]
<tomee^> but that's not yet implemented, right? :)
<jemk> right, because it is nearly impossible to get fully compatible vdpau. it would be easy to write it for special needs like fullscreen, but not universal
<tomee^> yeah, I know
<jemk> well, not impossible if you allow copying, but that makes watching 1080p not very funny
<tomee^> but vdpau is already somewhat more promising than cedarx... since VDPAU is at least SOMEWHAT of an abstraction layer
ganbold_ has joined #linux-sunxi
<jemk> but it's an abstraction layer that has too many possible ways of doing things that can't be represented by the limited hardware in arm socs.
<tomee^> yup
<tomee^> I cannot argue with the fact that you are right and more your knowledge is more comprehensive that mine
<tomee^> still, from my point of view, mplayer+vdpau/sunxi=no subtitles ;-)
<tomee^> while, I think, the stagefright android binaries somehow do support subtitle rendering in hardware
<jemk> possible, but android hasn't have to handle obscured windows. if the vdpau window is allways on top the same as for fullscreen works
<Turl> you could composite all desktop and subtitles by software to a layer and have the video on the 2nd one
<JohnDoe71rus> your kernel 3.4 only for Linux? suitable for android?
<jemk> Turl: i talked with ssvb some months ago about possible tricks, but they all need much work and xorg driver interaction
<jemk> like only copying the part of the video that is used for subtitles
<tomee^> Turl: couldn't that be done with acceleration? the compositing I mean.
<tomee^> I don't care about playback in a window as long as I can draw something on the video
<Turl> I'm not much into the graphics stack, but I think there's two alpha layers that get blended by sw
<Turl> if you can put the subtitles onto the desktop layer, then they could be overlayed over the video
<mnemoc> JohnDoe71rus: for android you need a different defconfig. crane in the case of a10 for example
<mnemoc> JohnDoe71rus: but the tree is androidized
<tomee^> jemk: when I played with mxplayer, I managed to get 2 sets of subtitles (one s/w, one h/w I guess), plus the top menu bar visible... this is far from Xorg complexity, but shows that the hardware can cope with that
<JohnDoe71rus> can i take defconfig from lichee 3.3 ?
<tomee^> Turl: say, an mplayer playing transparent video file + subtitles overlaid on top of the proper vdpau-drawn video?
<Turl> tomee^: I mean one layer with your desktop, mplayer window, a hole instead of video, and subtitles inside the hole
<Turl> and a second layer with the video
<tomee^> arokux: CC [M] drivers/net/wireless/bcmdhd/dhd_linux.o drivers/net/wireless/bcmdhd/dhd_linux.c: In function ‘dhd_os_prealloc’: drivers/net/wireless/bcmdhd/dhd_linux.c:5192:2: error: implicit declaration of function ‘wl_android_prealloc’ [-Werror=implicit-function-declaration]
<tomee^> arokux: I guess you should ifdef the DHD_OS_PREALLOC option for non-android builds... or I did something wrong
<jemk> Turl: possible, but not nice, especially as vdpau redraws all layers each frame, so the subtitles get redrawn too
<mnemoc> http://sprunge.us/ESHh .... wee... and I broke every driver now because of node duplication :p
<arokux> tomee^: cubietech guys just grab our u-boot i think.
<arokux> tomee^: i do not know what they have preinstalled
<mnemoc> i would bet they just use allwinner's SDK like olimex
<arokux> tomee^: but if using an uSD card you need to flash the bootloader anyways
<arokux> tomee^: so grab ours, it has support for CT, however it won't see 2GiB of RAM, there is patch for that.
<arokux> tomee^: oliv3r said if you are lucky the kernel will still see 2GiB of RAM
<mnemoc> mine doesn't...
<arokux> tomee^: where is that error msg comming from?
<arokux> oliv3r: how to be lucky so that the kernel sees 2GiB on CT?
<mnemoc> <6>Memory: 448MB 512MB = 960MB total
<mnemoc> <5>Memory: 830820k/830820k available, 152220k reserved, 270336K highmem
<mnemoc> the mem info comes from the ATAGs written by u-boot
<arokux> mnemoc: same for me, I think.
<mnemoc> and from the .dtb in the case of 3.10+
<tomee^> root@cubietruck:/sata/build/ffmpeg/ffmpeg-1.2.4# cat /proc/meminfo
<tomee^> MemTotal: 1869208 kB
<tomee^> MemFree: 1337728 kB
<tomee^> well, I'd prefer 2GB over 1GB ;)
<tomee^> so I will try sticking with the cubie bootloader
<tomee^> btw I wonder why they downclocked the bootloader for CT to 432MHz
<tomee^> if cubieboard2 is @480MHz
<arokux> mnemoc: tomee^ http://sprunge.us/RISD
<arokux> tomee^: how can you stick to their bootloader and use sunxi-3.4 kernel?
<tomee^> hmm, why not?
<arokux> oh, you just hit a key so you get prompt and then boot our kernel?
<tomee^> their bootloader is a fork of linux-sunxi uboot anyways I think
<mnemoc> allwinner's bootloader does not pass ATAGs
<mnemoc> but fixing u-boot-sunxi's cubietruck's dram struct would solve it based on oliv3r's comment
<tomee^> mnemoc: you mean just replacing 432 with 480 in the code?
<tomee^> or is the patch more extensive?
<arokux> tomee^: we are talking about size, not freq
<arokux> tomee^: Turl has some patches handy for increasing freq, though, you may want to take a look at those
<tomee^> arokux: 432MHz. that's the speed at which my board runs with sunxi bootloader, too. but look here:http://dl.cubieboard.org/software/a20-cubietruck/common/ct-v101_sys_config.fex
<tomee^> arokux: I simply do not know why 432MHz if the hardware can do 480... they downclocked it for security purposes in the first batch?
<arokux> tomee^: sorry, I have no idea.
<tomee^> arokux: I've heard somewhere that someone complained about overheating of the prototype
<arokux> tomee^: Turl may know.
<Turl> dunno, we should ask benn
<tomee^> arokux: and about the error you asked me about
<tomee^> arokux: it comes from the bcmdhd driver compilation
<tomee^> arokux: when I disabled buffer preallocation for it, it went through
derethor_ has quit [Read error: No route to host]
<arokux> tomee^: but first you enabled that driver (bcmdhd), right?
<arokux> here is my effort to create more docs: http://linux-sunxi.org/CedarX/Misc_Docs
<arokux> mnemoc: how do I insert line breaks? :)
<plaes> <br>
<plaes> or use <pre>
<arokux> thanks plaes !
<tomee^> arokux: of course
chkktri has quit [Ping timeout: 264 seconds]
<arokux> tomee^: ok, thanks for the input. if you find out if the blootooth works, i'd be awesome!
<tomee^> arokux: CONFIG_DHD_USE_STATIC_BUF=y broke the compilation. CONFIG_DHD_USE_STATIC_BUF=n made it work. maybe this flag should depend on android or some other build environment
<tomee^> arokux: ok I will try in a few minutes
derethor has joined #linux-sunxi
rings_IIV has quit []
<tomee^> arokux: http://linux-sunxi.org/CedarX/Misc_Docs - nice, you think my questions are worth posting to the website? :)
<arokux> tomee^: why not?
<arokux> tomee^: this irc channel is logged and is publicly available anyway
chkktri has joined #linux-sunxi
<tomee^> oh my
* tomee^ checks his make-up
<arokux> tomee^: yep, take a look at the topic.
<tomee^> read the topic? and what else, maybe read the documentation too? :P
JohnDoe71rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<arokux> tomee^: topic of this irc channel
ganbold_ has quit [Ping timeout: 245 seconds]
<tomee^> I know I know
<tomee^> that was supposed to be a joke ;)
<arokux> tomee^: ok, sorry :)
<tomee^> not a good one apparrently ;/
<arokux> so there you have it: http://linux-sunxi.org/A20-Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<arokux> mnemoc: ^ patches welcome!
<arokux> Turl: can you be so kind and update Memory frequency bullet? http://linux-sunxi.org/A20-Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<arokux> Turl: I'll then try to poke Benn
Turl changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask and wait! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
<Turl> arokux: what do you want me to write?
<arokux> Turl: correct it, add some more questions if any. or just tell it is fine.
<Turl> arokux: you quoted my opinion on the subject, "<Turl> dunno, we should ask benn" :p
<arokux> Turl: ok, so I mail Benn now.
<mnemoc> [[Ask Benn]]
rz2k has joined #linux-sunxi
ganbold_ has joined #linux-sunxi
<arokux> mnemoc: we should document more and more. please try to push that line.
wolfy has left #linux-sunxi ["Paradoxul homosexualilor este ca, desi nu reusesc sa se reproduca, sunt in fiecare an din ce in ce mai multi. (Tristan Bernard)"]
<mnemoc> we can add a bunch of those pages... [[Ask Eva]], [[Ask ...]], ...
<arokux> mnemoc: I do not mean info about asking docs only. any info on sunxi. we must have it somewhere not just buried in IRC logs.
<arokux> mnemoc: first ugly dumps will do, hopefully somebody will improve them.
<juanfont_> tomee^, jemk these overlay problems also affect xbmc when running standalone?
<arokux> hi juanfont_
<arokux> please add your findings here: http://linux-sunxi.org/CedarX/Misc_Docs
<juanfont_> hi :)
<juanfont_> it was a question
<arokux> juanfont_: you'll get the answer hopefully and so I kindly ask you to update that page with it. :) thanks!
<juanfont_> i'll update it, sure :)
<mnemoc> arokux: that's exactly why we started the wiki. to have an structured central but open source of all sunxi related information
<arokux> mnemoc: right, so it should be used even more heavily
<mnemoc> ack
paulk-collins has joined #linux-sunxi
<jemk> juanfont_: if you have xbmc running fullscreen or even without xorg it is much easier
<jemk> juanfont_: but xthe upstream bmc uses opengl for rendering, so it's totaly different there anyways
<tomee^> arokux: about irclogs, it would be nice if the wiki had an irc colorizer or allowed pasting html code. but that's not very important
<arokux> tomee^: what is interesting to colorize in IRC logs, it is not a code?
<juanfont_> jemk but, XBMC+VDPUA is not possible due to the lack of the required OpenGL extension in OpenGLES, is it?
<arokux> tomee^: there is some color here: http://linux-sunxi.org/Bootable_SD_card
<jemk> juanfont_: think so, but i didn't look too deep into xbmc
<tomee^> arokux: colorize nick names.
Black_Horseman has joined #linux-sunxi
<arokux> tomee^: I'd like those dumps to be rewritten as doc, so making dumps nicer isn't priority.
<tomee^> i see
<tomee^> i just find it much easier to read colorized logs than plain-text
<tomee^> i never know who is who without colors ;)
<arokux> tomee^: sure. please improve it.
<tomee^> I don't have admin rights on the wiki.
<tomee^> and it was just a wild suggestion. sorry.
<Turl> mnemoc: maybe we can add a wiki prefix for those (like Talk:)
juanfont_ is now known as juanfont
<mnemoc> go ahead :)
<mnemoc> this is fully community driven
<tomee^> arokux: so far the kernel didn't boot (but the sd card was a mess). I saw penguins, then nothing. will try with another one.
<Turl> I need to leave now :p but I'll do so when I come back if nobody opposes by then
<arokux> tomee^: but it booted before...?
AreaScout has joined #linux-sunxi
<tomee^> yes
<tomee^> wait a sec
Black_Horseman has quit [Quit: Αποχώρησε]
<arokux> ( tomee^: I think you should definitely use serial connection to see early boot log )
<ssvb> tomee^: 480mhz memory clock speed is too fast for cubieboard2
<ssvb> tomee^: it needs to be reduced, or we will keep getting reports about occasional stability issues from some users
<tomee^> ssvb: and cubietruck too?
<tomee^> ssvb: even though the spec says its ok?
<oliv3r> the spec lies!
<ssvb> tomee^: the spec says 400mhz for dram
<tomee^> arokux: this would require either dragging my amplituner and tv to my workstation, or dragging my workstation there, or wiring up an arduino in between, or whatever. not that simple ;)
<ssvb> tomee^: but the spec also says 1.2ghz for the cpu, which is overly optimistic :)
<tomee^> ssvb: ok, maybe I misread. or the spec was wrong.
<tomee^> ssvb: yeah. what's the safe speed? 1008? 912 ?
<tomee^> given I have a radiator and and airflow
<ssvb> tomee^: what's more important is that some real users are having real problems at 480mhz dram clock
<arokux> oliv3r: please resolve confusion! http://sprunge.us/CaCB
<arokux> oliv3r: see this: http://sprunge.us/hMNZ
<arokux> ssvb: do you have something to add to the Memory freq bullet here: http://linux-sunxi.org/A20-Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<arokux> ssvb: do you have something to add at all there? please update if so.
<tomee^> ssvb: i see. so 432 for ram is safe. and for cpu?
<oliv3r> woa lots of links pasted
<arokux> oliv3r: last one is the most up to date.
<oliv3r> dram_para needs to be set up correctly, it can be set up for 2 GiB without problem
<oliv3r> let me explain the u-boot bug paste on that dram para you pasted
<arokux> oliv3r: in fex?
<oliv3r> sunxi_dram_init() checks the memory size (dramc_init does)
<oliv3r> what you pasted, dram_cubietruck.sc
<oliv3r> now, the cubietruck has 2 GiB of ram, but we return an int
<arokux> ssvb: how are those numbers picked?
<oliv3r> as you know, int (in contrast to unsigned int!) runs from -2 GiB to +2 GiB -1
<arokux> oliv3r: I get it. how to make kernel see all 2GiB, fex parameter?
<oliv3r> so the int is overflowing :)
<ssvb> arokux: these numbers are from Tsvetan2, I have no idea how he got them
<arokux> ssvb: how did you get 432?
<oliv3r> arokux: .size is 2048 MiB
<ssvb> arokux: 432?
<oliv3r> arokux: lol no, nothing is needed for the kernel, uboot needs to pass the parameter properly
<arokux> ssvb: <tomee^> ssvb: i see. so 432 for ram is safe. and for cpu?
<oliv3r> u-boot tells the kernel it has minus 2 GiB, but the kernel is smart enough to figure ou tthat that's wrong :p
<oliv3r> (it's not, but if you cast an int into an unsigned in you get lucky and it works)
<arokux> oliv3r: this is apparently not the case. sooo... how to make kernel see 2GiB?
<oliv3r> use my u-boot
<tomee^> 432=dram clock speed
<oliv3r> kernel doesn't need anything for 2 GiB
<oliv3r> oh wait yes it does
<arokux> oliv3r: you said the kernel should still be fine with old u-boot - or I got it wrong?!
<ssvb> arokux: somebody has put 432 dram clock setup into u-boot for cubietruck by default, ask hno and/or oliv3r where they got it from :)
<oliv3r> it is fine, but you need:
<ssvb> arokux: I think benn planned to do some memory stability/overclocking tests on a large set of boards
<arokux> ssvb: I've poked at him, see ML
<tomee^> arokux: hmm, wifi wont come up
<arokux> tomee^: but now you have network, right?
<ssvb> arokux: good, but it would have been better if you did some search for the already known information first :)
<tomee^> arokux: gmac.
<arokux> ssvb: sorry, not enough time, and i'm still noob.
<tomee^> arokux: yes.
<arokux> tomee^: so do this: dmesg | curl -F 'sprunge=<-' http://sprunge.us
<arokux> tomee^: and post any other errors you've got while setting up wlan adapter.
Soru has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
<ssvb> arokux: benn apparently has access to ~100 boards for running tests - http://irclog.whitequark.org/linux-sunxi/2013-11-08#5494435;
HeHoPMaJIeH has quit [Remote host closed the connection]
<oliv3r> arokux: something like Memory split (3G/1G user/kernel split)
<oliv3r> but i don't remember exactly
<oliv3r> you need the 'high memory' stuff basically
<arokux> oliv3r: have you tested it?
<oliv3r> of course
<arokux> ssvb: I've updated http://linux-sunxi.org/A20-Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<oliv3r> but i forgot the exact kernel option; i'm looking it up for you
<arokux> oliv3r: not for me, for everybody :p
<oliv3r> not everboydy has a board :p
<oliv3r> only 'us' :)
<oliv3r> for now
<arokux> oliv3r: no.. tomee^ and some others have the board already!
<oliv3r> arokux: enable CONFIG_HIGHMEM and see if it works
<arokux> oliv3r: so either we quickly fix sunxi-3.4 or loose them forever.
<oliv3r> it should be CONFIG_HIGHMEM if memory serves me right
<oliv3r> and that's not a sunxi problem!
<oliv3r> it's a general 'what if i have 2 gb or more memory on a 32bit architecture'
<arokux> oliv3r: what is not sunxi problem?
<oliv3r> but it should be documented; i think i've set the proper .config settings in the 3.10 kernel
<arokux> oliv3r: so if CONFIG_HIGHMEM it should be added to sun7i_defconfig?
<oliv3r> test it to be sure; but then yes
<oliv3r> sun4i too
<tomee^> arokux: http://sprunge.us/gXEK <- this is what it looks like on a kernel I downloaded from cubieboard.org
<oliv3r> arokux: is hopefully busy building a highmem enabled kernel :)
<arokux> oliv3r: no, tonight.
<oliv3r> tomee's log looks good though: [ 0.000000] Memory: 448MB 1536MB = 1984MB total
<oliv3r> sounds about right no?
<arokux> oliv3r: that is cubietech kernel.
<oliv3r> Memory: 1868996k/1868996k available, 162620k reserved, 1318912K highmem
<tomee^> no.
<oliv3r> ohh
<oliv3r> ok nvm then
jemk has quit [Quit: Verlassend]
<oliv3r> then we absolutly need to enable highmem in default buidls
<tomee^> this is your kernel
<tomee^> this is cubie
<arokux> tomee^: thanks, I've added it to TODO: http://linux-sunxi.org/User:Arokux#TODO
<oliv3r> hmm iVaP also sees 2 GiB of ram
<tomee^> yes
<arokux> mnemoc: how your CT's dmesg looks like?
<oliv3r> so what's the problem exactly?
<arokux> tomee^: you also wanted to test bluetooth
<tomee^> arokux: I will retry the wifi/bt on another usd card/linux environment but I suspect the same outcome.
<arokux> oliv3r: mnemoc said his kernel wont see 2GiB, so thought I...
<arokux> tomee^: hm.. they are on the same chip, so you can be right.
<arokux> oliv3r: why cubietech isn't submitting patches?
Soru___ has joined #linux-sunxi
<arokux> :((
<oliv3r> arokux: *sigh* who knows :( sometimes they do
<arokux> tomee^: the cubietech kernel you test comes from: https://github.com/cubieboard/linux-sunxi/commits/cubie/sunxi-3.4
<arokux> tomee^: ?
Soru has quit [Ping timeout: 248 seconds]
<tomee^> as you see by the compilation date in dmesg
<tomee^> it must have been pulled out of their prebuilt linux image
<tomee^> so... don't know
nove has joined #linux-sunxi
<arokux> tomee^: can you post their config?
mkutsevol has left #linux-sunxi [#linux-sunxi]
mkutsevol has joined #linux-sunxi
<tomee^> ha, something's strange. when I explicitly specified the same fw file the module won't load. but on the other hand, maybe it just won't load twice (ie. the hardware can not be re-initialized)
<arokux> tomee^: which kernel? module for wifi?
<tomee^> arokux: http://sprunge.us/hHFH
<tomee^> arokux: yes
<tomee^> arokux: (this is their config)
<tomee^> CONFIG_BCMDHD_FW_PATH="/lib/firmware/ap6210/fw_bcmxxxx.bin" CONFIG_BCMDHD_NVRAM_PATH="/lib/firmware/ap6210/nvram_apxxxx.txt"
<tomee^> this is funny
<gzamboni> did anyone already manage to make spidev work ?
<tomee^> yet it works WITHOUT specying those parameters on cmdline
<arokux> juanfont: please paste your conversation on overlays to hat Misc_Docs pages, thanks.
<arokux> tomee^: their kernel is not perfect, they hacked on it till it work then just shipped. that is why you'd like to use ours, our priorities are different.
<tomee^> root@cubietruck:/sys/module# for file in /lib/firmware/ap6210/fw_bcmxxxx.bin /lib/firmware/ap6210/nvram_apxxxx.txt; do test -f $file && echo "$file exists" || echo "no $file"; done
<tomee^> no /lib/firmware/ap6210/fw_bcmxxxx.bin
<tomee^> /lib/firmware/ap6210/nvram_apxxxx.txt exists
<tomee^> arokux: yeah, I can see all around that there are hacks and poor documentation on the cubietech side
<gzamboni> i have the /dev/spidev0.0 but when i try to test it using the compiled spidev_test.c example i just get: can't send spi message: Invalid argument Aborted
<tomee^> arokux: and I would prefer your kernel too ;-)
<arokux> tomee^: first we just need to go to their kernel and cherry pick stuff. it shouldn't be hard. then we fix it - takes more time.
<tomee^> yes...
<tomee^> i spent dozens of hours on the board already
<tomee^> it's not that their stuff even works properly ;/
<arokux> tomee^: hw is nice though :)
<arokux> tomee^: what was your original goal, btw?
<tomee^> and while a10 and a20-cubieboard2 are not "top sellers", 90 bucks for a cubietruck is really a competitive price
<tomee^> for a board with dual-core GPU, dual-core CPU, 2GB RAM, 8GB NAND, SD, BT, WiFi, SATA, GPIO...
<tomee^> gigabit ethernet, two tv-out ports and most importantly to me, TOSLINK
<tomee^> arokux: xbmc. not there yet but I see light at the end of the tunnel.
<Montjoie> hello, I have perhaps hit a bug in sunxi-3.4, does someone with CRYPTO_USER_API want to test my poc http://pastebin.com/33MKW7R9?
<hramrach> there are problems @ 480MHz even on A10, more on a20. So they downclocked for the new board and pray that they get fewer issues I guess
hipboi has joined #linux-sunxi
<arokux> Montjoie: This paste has been removed!
<hramrach> but there is no extensive stability testing with *any* clock speed I suspect. They make a handful of prototypes, test them, do test batch, and we are currently testing that :p
<arokux> hramrach: how do they pick that numbers?
<hramrach> from thin air
<Montjoie> arokux, here http://pastebin.com/n2kmXcD3
<hramrach> you know, the 480MHz on A10 worked like 90%
<hramrach> that's decent but not good enough
<Montjoie> my problem is that after 164864 - 1024 bytes, sendmsg block forever
<arokux> Montjoie: look how nice it could have been: http://sprunge.us/cahH?c
<tomee^> hramrach: yeah. and how do the ram problems manifest themselves? so far I've only had trouble with make -j2 once - the gmac stopped responding, but upon reloading it was OK
<hramrach> and they don;t ahve the budget to make a batch of thousands boards (or at least hundreds) and extensively test all of them
<hramrach> tomee^: I have no idea. I don;t have problems. You can try searching the ML for some problems that were blamed on ram
<arokux> Montjoie: please extend the last bullet here with instructions etc, I will test tonight: http://linux-sunxi.org/User:Arokux#TODO
<arokux> Montjoie: please specify which boards/kernels should I use.
<hramrach> but there are reportedly 'stability issues'
<Montjoie> ok arokux
<tomee^> hramrach: maybe I will face them too... since its the 1st batch...
<hramrach> tomee^: sometimes problems with -j2 are also makefile problems
<hramrach> and they are not reliably reproducible either
<hramrach> if it's it never happens with -j1 don't blame ram
<tomee^> hramrach: so far, I know for sure that it ran ok for 24hr+ seeding torrents (legal stuff of course), and a couple of hours playing video
<tomee^> hramrach: I suspect the driver is flawed and simply shut down on race conditions
<hramrach> you can try multiple -j1 builds in parallel for testing. Or run a cedar using video player while building for extra memory stress testing
FDCX has quit [Remote host closed the connection]
<tomee^> hramrach: and with make -j2 + ethernet + ddrt + wifi2.4 + openwrt + wifi5ghz + ssh session there is a big chance of retransmissions
<hramrach> also if gmac shuts down it might just be bug in the driver
FDCX has joined #linux-sunxi
<ssvb> tomee^: gcc compilation of itself, 7-zip (p7zip) benchmark and memtester can be used to spot obvious memory stability problems
<hramrach> I like booted the board and sshd in but that's pretty much everything I did
<Montjoie> arokux, I do not have the rights to edit this page
<hramrach> Montjoie: everyone does
<hramrach> just register and don't post links
<Montjoie> I have just registred
<ssvb> tomee^: also the simultaneous use of mali has been observed to make it break at a lower dram clock frequency than with just cpu alone
<hramrach> arokux: on arm you need higmem for 1G
<tomee^> ssvb: yeah, a lot of memset & memcpy in video
<hramrach> was it dropped from sun4i?
<tomee^> ssvb: zeroing a large parge of nonbuffered ram could even lead to capacitor/voltage regulator problems
<tomee^> ssvb: this is stuff you face even in "enterprise" servers if you use desktop memory chips
<hramrach> Montjoie: ok, that page has links already so to edit it you might need to ask mnemoc to make an exception for you
<hramrach> because the spam filter pretty much flags every link as spam
<hramrach> tomee^: that's why there are server memry sticks with less power draw :p
<tomee^> hramrach: ... and buffering and ECC-R.
<ssvb> tomee^: the dram chips themselves are rated as DDR3-1333, the problematic part is likely inside of the A20 SoC
<tomee^> ssvb: why can't it be related to the voltage/power circuitry?
<tomee^> ssvb: I mean, the same stuff works in thousands of samsung smartphones, right?
<tomee^> ssvb: but they probably spent some time designing the power supply
<ssvb> tomee^: which same stuff?
<tomee^> Mali GPU ?
<hramrach> even if it's power circuitry it's the way it is on the board
<oliv3r> arokux: re: your mail to benn; is there still confusion or has it been resolved/
<hramrach> but the amount of ram on the board is ridiculously small and so is the frequency
<arokux> oliv3r: ppl are saying the correct freq it determined experimentally.....
<ssvb> tomee^: how is mali related other than providing additional stress on the memory controller and maybe having some problematic memory access pattern which triggers the issue more easily?
<tomee^> hramrach: if you "forget" a capacitor somewhere... or better yet, power the device from a impulse 0,5A "universal usb charger" for $5...
<oliv3r> arokux: yeah, the chip is somewhat buggy sometimes, depends on the quality of your chip, so the 'safest' freq. is 380 :p i think standard is 432, but some can clock upto 480. I think tom even had samples that would run at 520 MHz but I don't think it was fully stable :)
<ssvb> tomee^: btw, cubieboard and cubieboard2 are using identical PCBs, but cubieboard2 has more problems with dram overclocking
<tomee^> ssvb: I admit, it might be not. I just don't know if it's allwinner's flaw really... or the PCB
<arokux> oliv3r: please reply like this to my request to benn, then we make a wiki page out of the ML thread
Soru___ has quit [Read error: Connection reset by peer]
<tomee^> ssvb: identical PCBs but the hardware on it is different
<ssvb> oliv3r: I kinda successfully overclocked dram in my cubieborad2 to 552MHz, but this was only possible with 552MHz mbus :)
<tomee^> so this would suggest a problem with the PCB
<ssvb> tomee^: which part is different other than SoC?
Soru has joined #linux-sunxi
<ssvb> oliv3r, Turl: higher mbus frequency seems to mean better stability for dram when running at the higher clock speed, at least that's what I have observed with my board
<tomee^> ssvb: I didn't analyze thoroughly
<tomee^> ssvb: but cb1 came with 0,5GB ram at first I think
<tomee^> ssvb: plus the A20 SoC comes with "dual-core" Mali
<tomee^> which would mean "need more juice at peak"
<tomee^> of course I may be mistaken
<tomee^> but I have played around with AVR chips, I²C buses and the like
<tomee^> and it is really funny how the hardware can behave if you power it improperly
<oliv3r> ssvb: what about higher ddr voltage, 1.3V instead of 1.25V that should increase it all too
<tomee^> or have signal interference/current leaks through pulldown resistors/whatever
<oliv3r> obviously allwinner prefer the lower voltage for better battery life
<ssvb> tomee^: Tsvetan2 sees to say that ~20% of A20 chips have drastically worse dram overclocking capabilities
<buZz> tomee^: yeah try 230V AC for once ;)
<tomee^> buZz: what, in your opinion, is the voltage of a 5V "phone charger" ?
<ssvb> tomee^: and replacing A20 chips on the PCB resolved the issues for the problematic boards
<tomee^> buZz: especially plotted against current(A) draw?
<buZz> tomee^: usually? like 4.6V
<tomee^> buZz: I'd say 3,5 to 8v
<buZz> :)
<buZz> yeah especially the cheap ones
<arokux> tomee^: olimex boards are flexible in that respect.
<tomee^> buZz: and the current/voltage is not a straight line. it's a mess. it is power regulator's job to flatted it out.
Soru_ has joined #linux-sunxi
<Dapsaille> ssvb => do you know the minimum pixel clock on an a13 board ?
<ssvb> tomee^: I apparently got a "good" A20 chip, because it can run dram up to 552mhz, but the other people are not so lucky
<Montjoie> arokux, I have updated my file with instructions http://sprunge.us/bTdP
<tomee^> ssvb: which olimex boards?
<ssvb> tomee^: cubieboard2
Soru__ has joined #linux-sunxi
Soru has quit [Ping timeout: 244 seconds]
<arokux> tomee^: boards require voltage in the 6 to 16V
<tomee^> arokux: yeah
<tomee^> arokux: because they have a decent voltage regulator with some BIG FAT DIODES
<arokux> Montjoie: ok. will sun4i also do? sun7i - I'll test on cubietruck.
<tomee^> arokux: "olinuxino" and header layout suggests electronics - and those boards are usually designed that way
<tomee^> arokux: (olinuxino sounds like arduino, which also can be fed with 5 do 30V)
<Montjoie> arokux, I have only a cubieboard2, no other board
Soru_ has quit [Read error: Connection reset by peer]
<ssvb> tomee^: I don't have olimex sunxi boards myself, but any conclusions pointing to the A20 chip being the culprit should apply to cubieboard2/cubietruck too
<arokux> Montjoie: do you want me to test sun4i or you know it won't work?
<Montjoie> you can test
<tomee^> ssvb: yes.
<arokux> Montjoie: ok!
<Montjoie> I want to be sure that it is not myself the problem:)
<Montjoie> this bug is blocking all my progress on the security system
Soru___ has joined #linux-sunxi
<tomee^> ssvb: I just think it may be ONLY cubieboard design flaw. or a little error, say, the spec from allwinner said "+/- 10%" whereas it runs only on "+/- 5%", cubie implemented it as written in the spec (or even "oh, let's make it 15%") and hence the trouble
<arokux> Montjoie: I want to see it working, that is why i help!
<Montjoie> On x86_64 3.10.17 the bug is not here
Soru__ has quit [Ping timeout: 244 seconds]
<tomee^> ssvb: also, just count the capacitors and diodes surrounding ICs on the olinuxino, and compare that to the cubie PCB... and then look at the monstrous electrolytic capacitor that olinuxino has...
<tomee^> ssvb: if I have time, I will try overclocking with a) a phone charger b) a decent phone charger (LTE router charger) c) a professional transformer-based stabilized 5V PSU
Soru__ has joined #linux-sunxi
<Dapsaille> phone chargers are generally crappy
<Dapsaille> imho
Soru___ has quit [Ping timeout: 240 seconds]
<tomee^> Dapsaille: frankly speaking, everything that weighs less than 0,5kg and costs less than $30 can't be considered a dependable power supply
<Dapsaille> :)
<tomee^> and people often don't realize that
<tomee^> for example, plug a Xenon HID to a non-stabilized PSU and see what happens
<tomee^> and they don't sell stabilized PSUs, because they are big and costly
<Dapsaille> i've just replaced one of them in a diy thermal/humidity controller .. he just burned ....
<tomee^> yeah
<mnemoc> arokux: I broke all drivers here, so my dmesg does look nice :p
<tomee^> drawing too much current from a cheap impulse PSU can cause anything from voltage drops to the PSU becoming a pass-through of AC voltage ;)
<binaryferret> Anyone had any luck getting ektf2k touch screen driver to work with sunxi kernel? It keeps throwing a wobbler about 'Unknown symbol register_early_suspend' which I've read was taken out in an earlier linux kernel version.
<binaryferret> I grabbed the driver from another source tho as I've been trying a few things. Just wondering if anyone can help.
<Dapsaille> for christmas i hope to have a lab regulated power supply of 0 - 30 V DC / 0 - 5 A 150 W
<Dapsaille> instead of having 243636631234 chargers lying around :)
<mnemoc> arokux: http://sprunge.us/YbYJ
<tomee^> Dapsaille: great idea ;)
<arokux> mnemoc: strange this is tomee^ dmesg: http://sprunge.us/IVaP and he has 2GiB
<mnemoc> using which u-boot?
<tomee^> cubie
<oliv3r> mnemoc you not seeing 2 G in linux?
<oliv3r> mnemoc: have you enabled hihgmem? :)
<mnemoc> defconfig + gmac only
<mnemoc> 1m
<oliv3r> double check CONFIG_HIGHMEM
<arokux> mnemoc: ^
<mnemoc> grep HIGHMEM build_sun7i/.config
<oliv3r> hmm, strange
Soru_ has joined #linux-sunxi
<arokux> tomee^: so tell exactly how do you boot our kernel with cubie-u-boot, do you hit a key in u-boot and get theirs prompt?
Soru_ has quit [Read error: Connection reset by peer]
Soru____ has joined #linux-sunxi
Soru__ has quit [Ping timeout: 240 seconds]
<tomee^> arokux: no. there is an ext2 partition that holds uImage, uEnv.txt and script.bin
<tomee^> arokux: this particular partition also has a boot.scr and other stuff (came from a cubieboard2 image)
<arokux> tomee^: so you overwritten that stuff?
<speakman> Any idea which is the lower voltage limit of a "high" signal on GPIO input pins?
<tomee^> arokux: I overwrote the bootloader (8-2048s)
Soru_ has joined #linux-sunxi
<tomee^> arokux: and I replaced uEnv, uImage and script.bin -> this way it booted with the cubie kernel. I didn't care about the rest of crap on the partition (hdmi_drv etc)
<tomee^> arokux: by putting your uImage on that partition, your kernel also booted
<rz2k> [19:04:21] <speakman> Any idea which is the lower voltage limit of a "high" signal on GPIO input pins? - VDD_IO of the board
<rz2k> lower limit is around 0.3 or something
<speakman> rz2k: Hm. So that's the lower limit? I'm on a OLinuXino A20
Soru____ has quit [Ping timeout: 252 seconds]
<speakman> I've got 2v2 right now and it's interpreting is properly. I just wonder how close to the limit I really am.
<nove> disappointing, the h264 encoder only get to encode 1920x1080 at around 20fps
<arokux> nove: whos limitation is this?
apo_ has joined #linux-sunxi
<tomee^> disappointing?
<tomee^> on a core i7 I *encode* 1,5Mbit/s 700x400 xvid streams at ~100fps
<nove> arokux, no idea, maybe memory pressure, 1280x720 get 46fps
apo has quit [Ping timeout: 248 seconds]
<nove> this if i get the memory clock correct at 320Mhz, A13
<arokux> nove: what software is used for encoding, some blobs or community oss?
<nove> tomee^, it should be able to decode and encode at fullHD in real time, that is what hardware is for
<arokux> nove: so 20fps is done by open source software developed by community?
<nove> arokux, github.com/patrickhwood/h264encoder
<nove> arokux, no the blob binary
ZetaNeta has quit [Ping timeout: 252 seconds]
<tomee^> it's supposed to *encode* at full-hd 1080p? hmm, ok, maybe, didn't know.
<tomee^> but it also depends "what is full hd"
<tomee^> what bitrate... how much vector optimizations, keyframes, ...
<rz2k> speakman: iirc we never had full hardware manual for the voltage part
<arokux> nove: hm.. so the promised performance cannot be even achieved with the blob? how do you know the hardware can really do more?
<rz2k> this is directly interacts with vdd_io and depends on the actual implementation of the analog part
Soru___ has joined #linux-sunxi
<arokux> oliv3r: so give us 2GiB with sunxi-3.4!
<nove> arokux, i don't know, i am only showing my disappointing with the first result
Soru_ has quit [Ping timeout: 245 seconds]
<arokux> nove: and the promise just comes from the data sheet?
<oliv3r> it shoul dwork allready! :p
<arokux> oliv3r: your last message was: <oliv3r> hmm, strange
<tomee^> ... from a pdf brochure ;p
rz2k has quit [Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/]
<nove> arokux, no promise, only my wish
soul has joined #linux-sunxi
<tomee^> "up to full-hd performance at up to 25fps in up to real-time *" * - depending on software, hardware configuration and current planet constellation
soul is now known as Guest74573
Guest74573 has quit [Read error: Connection reset by peer]
<arokux> nove: "nove> tomee^, it should be able to decode and encode at fullHD in real time, that is what hardware is for"
<arokux> nove: how do you know what is the hw for?
<oliv3r> arokux: i don't know why it won't for mnemoc it works for everybody else :)
Soru_ has joined #linux-sunxi
Soru___ has quit [Ping timeout: 248 seconds]
<arokux> oliv3r: defined everybody else?
<tomee^> arokux: funny, I thought hardware was only for running software. else can be used as a paperweight or a stool or coffee table.
<oliv3r> arokux: tomee^ mee, probably you :)
dalee has joined #linux-sunxi
<arokux> oliv3r: tomee^ uses u-boot from cubiethech! me has 1GiB as well as mnemoc! :)
<arokux> oliv3r: so... give us 2GiB in sunxi-3.4 please! :)
<nove> arokux, the hardware is for what we want, if the hw doesn't do what we want is not good hardware
<oliv3r> are you using my u-boot?
<tomee^> nove: agreed.
<dalee> have some1 problems with UVC capture from WebCam? I did compile kernel with UVC and preview working but capture gets green image on saving
egbert_ is now known as egbert
ZetaNeta has joined #linux-sunxi
ZetaNeta has quit [Changing host]
ZetaNeta has joined #linux-sunxi
<dalee> this is result from capture
<tomee^> but then comes the definiton of full hd. blu ray is full hd, but uses an MPEG1/2 codec which your kitchen oven could deal with.
<tomee^> then there's h264 which is really, really sophisticated and makes that 30GB of movie fit into 3GB...
<arokux> oliv3r: no. i'm using sunxi-u-boot, but you said it should be fine too and that kernel is smart enough to figure out the memory size by itself. is it not true?
Soru__ has joined #linux-sunxi
<oliv3r> it doesn't figure anything out, but the overflown int cast back to an int makes it so that it shouldn't be an issue
<oliv3r> the kernel NEEDS u-boot to tell it its correct size; for that to work, the .size and .density parameters need to be correct
Soru_ has quit [Ping timeout: 272 seconds]
rellla3 has joined #linux-sunxi
rellla3 has quit [Client Quit]
Soru_ has joined #linux-sunxi
<WarheadsSE> arokux: and others. Status of CB2 integration to linux-sunxi from the prior 3.3 pile of $%^&
<arokux> oliv3r: so sunxi-3.4 doesn't have a chance to see 2GiB with current u-boot. YES/NO?
<arokux> WarheadsSE: what? :)
<WarheadsSE> how is the CB2 (linux sun7i) for the 3.4 tree?
Soru__ has quit [Ping timeout: 272 seconds]
<arokux> WarheadsSE: no idea, I just recently got truck.
<WarheadsSE> yeah, i know :P
<WarheadsSE> Whats the SoC on that again?
<arokux> A20
<arokux> sun7i
<arokux> same as on CB2
<arokux> WarheadsSE: just tell here what isn't working and it will get fixed (soon)
<WarheadsSE> K
<tomee^> hmm
<WarheadsSE> Well, we're still packaging the 3.3 CB2 kernel
<WarheadsSE> we'd like to move to the proper linux-sunxi/sun7i 3.4+
<tomee^> arokux: so you say that cubie has a repo of the bootloader, but the crucial part comes from allwinner and is closed-source?
hno has quit [Excess Flood]
wingrime has joined #linux-sunxi
hno has joined #linux-sunxi
jemk has joined #linux-sunxi
Soru__ has joined #linux-sunxi
<arokux> WarheadsSE: so you want somebody to submit you an updated .config?
Soru_ has quit [Ping timeout: 245 seconds]
atiti has quit [Ping timeout: 245 seconds]
<arokux> tomee^: I never said that. I have no idea what bootloader they use, mnemoc had some ideas.
<WarheadsSE> the whole tree needs moved from 3.3 to 3.4
<WarheadsSE> I'd rather know it was sound to move, and everything was functioning
<tomee^> arokux: definitely uboot ;) but /me no idea as well.
shineworld has joined #linux-sunxi
notmart has quit [Read error: Operation timed out]
BluesBoy has quit [Quit: HydraIRC -> http://www.hydrairc.com <- s0 d4Mn l33t |t'z 5c4rY!]
<arokux> WarheadsSE: I see. it is hard to tell everything works I think. is it an option to do something like -testing and push it to your users so we know how good it is?
<WarheadsSE> well, then I would not push it to my users.. I would make it individually available
<arokux> WarheadsSE: "individually available" -- plz, explain
<libv> do we have a tool to make noise on the serial line for android, or does anyone remember such a thing?
<libv> it should be easy to create, but still, would be better to re-use something than to create it from scratch
<wingrime> jemk: ping
<arokux> wingrime: meet new page and give it some love: http://linux-sunxi.org/CedarX/Misc_Docs
<wingrime> arokux: already done
<jemk> wingrime: ask, not ping
<arokux> wingrime: where? http://linux-sunxi.org/index.php?title=CedarX/Misc_Docs&action=history
<WarheadsSE> arokux: ala a tarball
<wingrime> jemk: I tryed render 4k video with more VBV buff
<wingrime> jemk: but result still same
<arokux> WarheadsSE: for those interested to test?
<wingrime> jemk: but, it looks little faster than before
<arokux> wingrime: have you seen? nove> disappointing, the h264 encoder only get to encode 1920x1080 at around 20fps
<WarheadsSE> arokux: yes
<wingrime> jemk: but 1MB not realistic
<arokux> WarheadsSE: ok.
<WarheadsSE> arokux: also, keep me abreast of cedar stuff && packaging possiblities.
<wingrime> arokux: PLL can be tuned
<arokux> wingrime: have you achieved more fps?
shineworld has left #linux-sunxi ["Leaving"]
<arokux> WarheadsSE: I've added it too my TODO http://linux-sunxi.org/User:Arokux#TODO
<wingrime> arokux: I have no FPS stats
<wingrime> jemk: we need Statistics from cedar
<nove> wingrime, i think when they say supports 2160p, then mean 3d 1920x1080, not 4K
<jemk> wingrime: start the cycle counter at decode start and read it at end, then divide by ve freq and you have frametime, 1/frametime = fps
tomboy64 has joined #linux-sunxi
rellla has quit [Quit: Nettalk6 - www.ntalk.de]
tomboy64 has quit [Client Quit]
<wingrime> jemk: can you add this with enabling using env vars
<arokux> wingrime: can you please share the status of cedar? :) what are you working on?
<wingrime> arokux: RE blob
<wingrime> look at traces
<wingrime> testing videos
<arokux> wingrime: how many % is done?
<arokux> :)
<hramrach> hello
<hramrach> wingrime: you were working on disp for u-boot?
<wingrime> hramrach: only tryed with hdmi init
<wingrime> hramrach: but still something missing
<hramrach> hdmi is a good start
<hramrach> well, if it worked
<hramrach> you have a repo somewhere?
<wingrime> hramrach: no, thats not worked
<wingrime> hramrach: I have testing app for uboot
<wingrime> hramrach: I still don't know what it need too
<hramrach> hmm
Soru__ has quit [Read error: Connection reset by peer]
tomboy64 has joined #linux-sunxi
Soru has joined #linux-sunxi
<hramrach> what does not work?
<nove> wingrime, registers were updated in viewer
rellla2 has quit [Ping timeout: 248 seconds]
<arokux> hramrach: what is your ultimate goal with sunxi world if I may ask? :)
rellla has joined #linux-sunxi
<hramrach> arokux: just play with some moderately cool hardware. It's not particularly useful for anything but hacking as it turns out.
<arokux> hramrach: it will become soon, won't it? :)
<arokux> hramrach: hardware is powerful.
<hramrach> no. so long as it cannot play video it's not useful as desktop and I have a router already.
<wingrime> nove: nice
<hramrach> I mean, really play random videos which somebody encoded with random options they like to tick in their encoding app
<wingrime> hramrach: no signal on my screen
<wingrime> hramrach: but PLL/gate/ are configured
<hramrach> hmm
<hramrach> so there is something unspecific broken
<hramrach> do you have source for the app yo used for testing?
<arokux> hramrach: I meant in general.
tomboy64 has quit [Quit: Uhh ... gotta go.]
<hramrach> arokux: in general I wanted to try to use it as desktop and see if it's too limited or not
<arokux> hramrach: have you seen images posted by focus ?
<hramrach> no
<hramrach> or what kind of images?
<arokux> his Ubuntu images
pfdm has joined #linux-sunxi
<hramrach> I have Debian image on my A10. It works as much as you can expect of an A10
<hramrach> meaning that Gnome is sluggish - this is expected and tunable
<hramrach> and video decoding is buggy as hell
<pfdm> arokux: I've seen your todo list, I have a pkgbuild for arch kernel 3.4.61 with cedar patches, if you want to update the official repo.
<hramrach> the gnome performance is not a showstopper. You can use less demanding desktop and wiht Lima doing proper buffer management you can get doublebuffering and probably better performance
<hramrach> but there is no known way to decode video on an a10/a20
<hramrach> 10bit media is common and cedar either cannot decode that or it is not known how to feed it such media
<arokux> pfdm: what are those patches for?
arete74_ has joined #linux-sunxi
<arokux> oliv3r: so sunxi-3.4 doesn't have a chance to see 2GiB with current u-boot. YES/NO?
arete74 has quit [Read error: Connection reset by peer]
<pfdm> arokux: no, I mean based on the pat's branch.Oh I see, didn't see the updated one.
<arokux> pfdm: why those patches are not submitted to sunxi-3.4?
<arokux> pfdm: do you have a repo for you PKGBUILD?
<arokux> tomee^: have you seen this? http://docs.cubieboard.org/tutorials/cb3/start
tomboy64 has joined #linux-sunxi
<pfdm> arokux: there's a misunderstanding , I think pat's branch pathces are merged to sunxi3.4. But I was using it from mid october , so i have a pkgbuild for it. Now the patches have been merged into sunxi3.4 so it's ok.
tzafrir has joined #linux-sunxi
<hramrach> arokux: also on a10 the low memory bandwidth that prevents 1080p scanout might be showstopper for some. I don't use that large resolution so I don't mind that much
<arokux> pfdm: are you sure they are merged already? just want to be sure.
<arokux> hramrach: is it better with A20?
<pfdm> arokux: No I don't have a repo, but it's nothing special. Let me verify. What is not merged yet, is montjoie patch for axp209 temperature sensors.
<hramrach> unknown. there is possibility to tune mbus which seems to yield better memcpy performance but how it affects scanout and what speeds are still stable is untested
<Dapsaille> regarding resolutions of a13 i've found this .. does it mean that a13 can output 1920x1080 ? http://service.awbase.com:8000/faq/index.php/A13_LCD%E6%A8%A1%E7%BB%84%E5%88%97%E8%A1%A8
<Dapsaille> may output .. sorry :)
<hramrach> and since I don't have a 1080p screen I won't test this I guess
<arokux> hramrach: i see.
<hramrach> it can output 1080i without issue. with 1080p you may have issues when the system is not idle. might be better with 50Hz over 60Hz ans 16bpp over 32bpp. ymmv
<Dapsaille> ty
<arokux> pfdm: montjoie patch isn't a bug fix, so it is not that critical.
Soru__ has joined #linux-sunxi
<pfdm> arokux: ok. From what i see in the log, all the cedar patches have been merged. So looks ok from me.
<Dapsaille> do you know if the A20 had a real VGA output signals ?
<hramrach> Dapsaille: as on a10 VGA is wired to TVout
<arokux> pfdm: perfect. how can I contact you so that you can test the up-to-date kernel package?
<jemk> torbenh3, arokux: any progress with gmac at u-boot? i don't get it to send anything...
<arokux> pfdm: as I understand you own cb2?
<arokux> jemk: what u-boot are you using?
<Dapsaille> does it mean that we can get vga signals from tvout ? i don't understand :x
<hramrach> Dapsaille: if you configure the registers correctly you should have VGA signal
Soru has quit [Ping timeout: 272 seconds]
<Dapsaille> ok, i'm trying to get a 15.6khz signal from a13 vga out but no luck .. black screen
<jemk> arokux: linux-sunxi/u-boot-sunxi with designware at the moment, but sun6i_gmac was the same
tomboy64 has quit [Quit: Uhh ... gotta go.]
<pfdm> arokux: yes I'm on cb2.
<Dapsaille> so i wonder if it will be easy with an A10 or A20
<Dapsaille> easier . sorry
tomboy64 has joined #linux-sunxi
<jemk> arokux: mii info works on both, phy receives all ok and even sends to mac, but mac doesn't talk to phy
<arokux> jemk: I am hacking around sun6i_gmac, no success so far, I was able to see something.. but its not working yet.
<hramrach> I have no idea about a13 or VGA. never got it working. presumably the parts that are bot on a10 and a13 are mostly the same
<arokux> jemk: oh.. you know more then me then. do you have your code handy at github?
<hramrach> note that at least on HDMI not every clock frequency is supported and if your screen is picky it won't sync
<arokux> jemk: this is my branch, very ugly: https://github.com/arokux/u-boot-sunxi/tree/sunxi-gmac
<Dapsaille> hramrach => i use olimex a13 board with onboard hack lcd> vga .. but i'm affraid that this hack is only for standard vga signals (34khz) .. so 15.6 don't work ...
ZetaNeta has quit [Ping timeout: 252 seconds]
<Dapsaille> and i cannot get help from olimex chan :x
notmart has joined #linux-sunxi
notmart has joined #linux-sunxi
notmart has quit [Changing host]
<hramrach> Dapsaille: try the mainlinglist. also try normal VGA screen and modes first if you did not
<jemk> arokux: http://sprunge.us/TUXa
<Dapsaille> hramrach => VGA works perfectly for me :) i will try ml , thanks for advices
<arokux> jemk: yesterday I've added this and it got better, but still not working.
rz2k has joined #linux-sunxi
<jemk> arokux: i verified that tx/rx clock is running correctly by measurement today, so clocks are okay
Soru__ has quit [Read error: Operation timed out]
<arokux> jemk: how can you measure this?
<jemk> arokux: also received data is send between phy and gmac, but gmac doesn't react on it
Soru__ has joined #linux-sunxi
<jemk> arokux: the clocks are at R129 and R132, so i probed there and had a nice 25MHz clock signal as expected for 100Mbps
<pfdm> arokux: for the cedarx , I am trying to make it work with gles output without copy. I haven't had much time lately, but it's possible to use a texture from UMP memory, has to see the perf now.
Soru____ has joined #linux-sunxi
<arokux> jemk: I need to take a look at home if you got this right: #define GMAC_AHB_BIT0x00020000
Gerwin_J has quit [Quit: Gerwin_J]
<jemk> arokux: it is right, otherwise the gmac regs wouldn't even be accessible
Soru__ has quit [Ping timeout: 244 seconds]
<arokux> jemk: u're right. 17th bit, I have the same.
Soru____ has quit [Read error: Connection reset by peer]
<arokux> pfdm: I see, sorry I know very little about cedar&gpu
<arokux> jemk: luckily there is stage/sunxi-3.4 where GMAC works.
Soru_ has joined #linux-sunxi
<arokux> jemk: will gmac send something to phy?
<jemk> arokux: no, only phy to gmac, but the tx clock is generated properly by gmac
ZetaNeta has joined #linux-sunxi
<jemk> arokux: so i believe there is some problem doing dma right
<arokux> jemk: ok, I cannot tell you something new. you know much more than me, sorry.
Gerwin_J has joined #linux-sunxi
* tomee^ been afk
<jemk> arokux: i only verified the electrical signals, i don't know more about the software side
<arokux> i've PMed you
<mnemoc> o_O
<mnemoc> arokux: mail subject and date please :p
<arokux> mnemoc: sorry, that is everything I've seen :p
<arokux> mnemoc: look for sensors
popolon has quit [Quit: Quitte]
pfdm has quit [Ping timeout: 250 seconds]
Soru__ has joined #linux-sunxi
Soru_ has quit [Ping timeout: 245 seconds]
wolfy has joined #linux-sunxi
Soru_ has joined #linux-sunxi
arokux2 has joined #linux-sunxi
Soru__ has quit [Ping timeout: 248 seconds]
<arokux2> mnemoc: have you found it?
eebrah has joined #linux-sunxi
Soru_ has quit [Ping timeout: 245 seconds]
<oliv3r> arokux2: I don't remember. But since i have a patch, I didn't look into it really, there's a current working u-boot patch that works ;)
Soru_ has joined #linux-sunxi
<arokux2> oliv3r: but it is not yet in sunxi-u-boot, so I've thought kernel can solve this problem, but it seems not to be the case.
<arokux2> oliv3r: if you are satisfied with you patch, can you push it to our u-boot-sunxi, please?
<arokux2> oliv3r: if testing is needed let me know.
Soru_ has quit [Read error: Connection reset by peer]
_massi_ has quit [Quit: Leaving]
Soru has joined #linux-sunxi
<arokux2> wingrime: how much of blob was RE already?
<oliv3r> arokux2: no, kernel can't solve it without a working u-boot
<arokux2> oliv3r: ok (you were saying the opposite that is why I was hoping...)
<oliv3r> arokux2: i am satisfied and I think it works, if you are using it and concider it stable, i'd love to; but it's quite an invasive patch to u-boot core, that goes beyond sunxi-u-boot and requires atleast hno's blessing
<arokux2> oliv3r: you are even scared to push it to u-boot-sunxi?
<oliv3r> nah; :p
<oliv3r> if hno would say 'ok lets push it and see what happens' i'd be happy to do that too
<oliv3r> I was hoping, months ago when starting it, that the patch would have landed upstream allready, and we could have only pushed our board specific patches
enrico_ has quit [Quit: Bye]
<arokux2> oliv3r: you should resubmit now as plain text
<oliv3r> arokux2: yeah as git send-email :)
<oliv3r> i will do that right now
<arokux2> oliv3r: I wonder, no ARM board had 2GiB before?!
<oliv3r> yeah u-boot claims 'it works fine'
<oliv3r> but i think it's because the overflow issue at 2 GiB isn't really big, and some cast magic actually fixes it below the line; it's still a bug
Soru has quit [Ping timeout: 248 seconds]
<arokux2> oliv3r: u-boot as u-boot maintainers?
<oliv3r> but you saw the mails; they where arguing about random shit without even looking at the stuff
<oliv3r> u-boot maintainers say 'we have other boards, it works fine'
<oliv3r> but i think if you skip the memory detection check, you don't hit the bug
<arokux2> oliv3r: yes, hopefully they will be more focused if they see code.
<oliv3r> which is what they are arguing about
Soru_ has joined #linux-sunxi
<oliv3r> and now for the secret to be revealed, mailing list time
<arokux2> oliv3r: +1
<arokux2> oliv3r: have you tried to skip that detection on CT?
hipboi has quit [Quit: Leaving]
<oliv3r> memory detection should be used; and butchering u-boot over it will be a lot of useless work
<oliv3r> btw, i can confirm current u-boot will not ever work with more then 1 GiB right now, as it's limited to not ever check more then 1 GiB :)
Soru_ has quit [Ping timeout: 272 seconds]
Soru has joined #linux-sunxi
<arokux2> oliv3r: good argument for you cover letter.
<oliv3r> well that's limited to sunxi for now ;)
<oliv3r> *barf*
Soru_ has joined #linux-sunxi
<arokux2> oliv3r: oh, could you improve that sunxi limitation only? ppl are really pissed off if we tell them kernel sees only 1GiB, even if the second GiB is not really used.
<oliv3r> i'll prepare my patches, resend them to u-boot ML and then push it ot ours
<oliv3r> we can always remove the patch(es)
Soru_ has quit [Read error: Connection reset by peer]
Soru has quit [Ping timeout: 272 seconds]
Soru has joined #linux-sunxi
<torbenh3> jemk: ping
<torbenh3> jemk: you said you were having troubles with gmac transmitting stuff in u-boot ?
<jemk> torbenh3: yes, did you get it work?
<torbenh3> jemk: i am seeing the same thing in mainline, i think
<torbenh3> jemk: i see packets being reveived fine.
<jemk> torbenh3: with the driver from cubieboard ported to mainline?
<torbenh3> jemk: yes.
<torbenh3> jemk: but transmission of packets doesnt work.
<torbenh3> looks ok, from the dma pov.
<jemk> you really receive them in software, or only the led blinks?
<torbenh3> driver debug features, print out the contents of rx packets.
<torbenh3> i see them.
<jemk> ok, thats better then u-boot, there i don't see anything at the software side
<torbenh3> hmm... ok. then we are seeing seomthing different i guess.
Soru_ has joined #linux-sunxi
<jemk> torbenh3: do you correctly set up the gmac module clock (0x01c20164 = 0x00000006), as this is responsible for the transmit clock, but NOT for the rx clock
Soru has quit [Ping timeout: 245 seconds]
Soru_ has quit [Read error: Connection reset by peer]
Soru has joined #linux-sunxi
<arokux2> jemk: this driver is working in sunxi-3.4, it should be correct then?
<arokux2> techn_: ping
wolfy has quit [Ping timeout: 252 seconds]
<techn_> arokux2: pong
<jemk> arokux2: it should, but it doesn't work so something must be different
<jemk> afk, bbl
<arokux2> techn_: you've just replied to the e-mail, that guy had another patch. I'm asking myself it resolve hackberry issue too.
<hno> <hno> If it looks sane and works for two people then I am happy.
<hno> <hno> oliv3r^
<hno> <hno> but changes to u-boot core should really be sent to u-boot mailinglist for discussion.
* hno really should get a vpn of some sort to his IRC bouncer to avoid disconnects.
<oliv3r> hno: ok, rgr
<techn_> arokux2: might solve.. I was thinking if it need to be fex file variated
<oliv3r> hno: i did send the patch and even discussed it, but they kinda ignored the content of the patch, had some small talk; and then just gave up
<oliv3r> hno: so would have been happy if you would have weighted in on u-boot ML; i'll resend the patch again though.
<oliv3r> hno: p.s. HansG did review and test it
soul has joined #linux-sunxi
soul is now known as Guest87208
Soru has quit [Ping timeout: 245 seconds]
Guest87208 has quit [Read error: Connection reset by peer]
<oliv3r> arokux2: can you build and test a u-boot if I give you a repo now?
Soru has joined #linux-sunxi
<arokux2> oliv3r: yes
Gerwin_J has quit [Quit: Gerwin_J]
<oliv3r> arokux2: ok cool let me fix this compile error and then i push --force it to my repo; if you test it also (i will too) i will push it to sunxi :)
<arokux2> oliv3r: I'll test it right away.
<oliv3r> ok i'm getting mmc compile errors
<oliv3r> did something get pushed that breaks things?
<oliv3r> oh wait, let me switch back to an older compiler
<oliv3r> i think i'm hitting some gcc 4.7.3 u-boot bugs
<oliv3r> no that's not it
<oliv3r> arokux2: what do you get if you compile current head?
* arokux2 compiles current head
<oliv3r> seen this?
<mnemoc> arokux2: [linux-sunxi] [PATCH][RFC] Add standalone driver for the A20 Soc TP embedded temperature sensor
<arokux2> oliv3r: for which board should I compile.. (if it matters)
<mnemoc> arokux2: oct. 7th..... but ... was it accepted?
<oliv3r> Cubietruck
wingrime has quit [Ping timeout: 245 seconds]
pfdm has joined #linux-sunxi
<oliv3r> mnemoc: i think the author was in talk with the hansg, who had some issues/pointers, i think in the end we agreed for 3.4 it didn't matter much
<arokux2> Montjoie is here
<mnemoc> so I just apply it?
<mnemoc> or wait for a v2?
<oliv3r> mnemoc: i don't remember the conclusion
<arokux2> oliv3r: everything compiles find for me.
<mnemoc> :\
FR^2 has quit [Quit: Connection reset by peer]
<arokux2> oliv3r: fine*
<oliv3r> arokux2: your not usign the BSP are you
<arokux2> oliv3r: no!
<arokux2> oliv3r: make distclean
<mnemoc> oliv3r: will you push your fixes/branches to revive the bsp?
<oliv3r> arokux2: the BSP uses O(output)=build/${BOARDNAME} so you don't ever hae to dot hat
<oliv3r> mnemoc: i think i pushed them to a WIP branch
<oliv3r> still get the error on the timer
<arokux2> oliv3r: where is your u-boot branch, I can compile it.
<oliv3r> ok let me push it
<oliv3r> branch wip/2GiB
<arokux2> oliv3r: do not forget to ping me
<oliv3r> arokux2: ping :)
<oliv3r> hno: btw, the next u-boot merge with upstream will be interesting, there where quite some cleanup patches (whitespaces/tabs) on boards.cfg for example
<arokux2> oliv3r: compiles fine
Soru_ has joined #linux-sunxi
<arokux2> oliv3r: http://sprunge.us/dUcH
<oliv3r> awesome, so it may be a bsp thing :(
Gerwin_J has joined #linux-sunxi
<oliv3r> arokux2: can you boot it :)
rings_IIV has joined #linux-sunxi
Soru has quit [Ping timeout: 248 seconds]
<oliv3r> i still rather think my compiler is somehow broken :(
<arokux2> oliv3r: go to u-boot-sunxi and build it directly!
<oliv3r> i could :)
<oliv3r> but i have little hope
<montjoie[home]> mnemoc dont appy it, I think I dont have resend the version with ioremap
<oliv3r> kernel compiles fine
<oliv3r> arokux2: ohhhh wait, you might be right, distclean may have been the issue; I did a make build_all to test the patchset
<mnemoc> montjoie[home]: ok. I'll wait for your v2
<oliv3r> Image Name: U-Boot 2013.10-rc2-08404-g5298fe
<oliv3r> arokux2: can you boot that uboot?
<arokux2> oliv3r: http://sprunge.us/PEiS
<oliv3r> strange
<oliv3r> no clue what's causing that
<oliv3r> or where the fault is in that
<hramrach> maybe *that* is compiler ;-)
<arokux2> oliv3r: DRAM: 1 GiB
<arokux2> oliv3r: please make it be 2 :)
<oliv3r> you sure your compiling the right hting?
<oliv3r> this is the full 2 GB patch thing
<arokux2> oliv3r: this "DRAM: 1 GiB" is our u-boot
<arokux2> oliv3r: yours gives crash
<oliv3r> that is my u-boot
<oliv3r> i just compiled it and flashed it
<mnemoc> anyone has the a10meminfo from a running 2GB CT?
<oliv3r> mnemoc: was merged a while ago
<arokux2> mnemoc: a10meminfo where do I get it?
<mnemoc> but why does it insist in telling 1GB?
<arokux2> oliv3r: have you managed to compile it?
<oliv3r> make distclean fixed it
<hramrach> oliv3r: for nits - it causes problem for 2G or more. you only get 2G on arm but that's a problem already
<oliv3r> because I did make all_boards a while ago
<oliv3r> hramrach: que?
<hramrach> the commit text
<oliv3r> what's wrong there?
<hramrach> more than 2G
<arokux2> oliv3r: U-Boot 2013.10-rc2-08404-g5298fe2 <--- what is this hash?
<hramrach> it should be 2G or more
<oliv3r> that's the git hash from git that was used during the compile
<oliv3r> hramrach: let me read commit log to ompiare :) i wrote that 2 months ago
<arokux2> oliv3r: hashes here are different: https://github.com/oliv3r/u-boot-sunxi/commits/wip/2GiB
<arokux2> mnemoc: https://github.com/maxnet/a10-meminfo <-- want me to run it?
<mnemoc> arokux2: on a 2GB CT booting the stock OS
<arokux2> where do I get correct MAC address for Cubietruck?
<mnemoc> but oliv3r said it's already apploed on u-boot
<oliv3r> arokux2: from wherver you want :)
<mnemoc> so I don't know
<hramrach> arokux2: /dev/urandom
<arokux2> mnemoc: :$ how can i do something in the stock os..?
<oliv3r> mnemoc: it's uploaded to u-boot, it's uploaded to sunxi-boards
<mnemoc> :\
<mnemoc> hno!
<hramrach> arokux2: what problem are you having with the stock os?
<oliv3r> mnemoc: what's wrong?
<arokux2> hramrach: I do not know how can I get a nice shell
<hramrach> I did get a shell on serial
<arokux2> oliv3r: give me the hash of you wip branch!
<mnemoc> oliv3r: CPU: Allwinner A20 (SUN7I)
<mnemoc> Board: Cubietruck
<mnemoc> DRAM: 1 GiB
<mnemoc> I2C: ready
<hramrach> or with adb in the case of andriod
<arokux2> hramrach: without prompt?
<arokux2> mnemoc: +1
<oliv3r> ok now i have 3 people i have to fix things for :p
<mnemoc> U-Boot 2013.10-rc2-08400-g8a4621c (Nov 04 2013 - 22:45:46) Allwinner Technology
<arokux2> mnemoc: what is that hash at the end?
<mnemoc> i can live with 1GB... but something is clearly wrong
<mnemoc> arokux2: the last commit on the tree I think
<arokux2> mnemoc: why is it prepended with g???
<hramrach> git
<arokux2> hramrach: thanks!
<oliv3r> flash that to an SD card (bs=1024, seek=8) and see if that changes things
<mnemoc> *g*
<arokux2> mnemoc: dd if=u-boot-sunxi-with-spl.bin of=/dev/sde bs=1024 seek=8 && sync && eject /dev/sde
<arokux2> mnemoc: /dev/sde WATCH OUT!
<mnemoc> :)
<mnemoc> ok, i'll try oliv3r's spl
* arokux2 tries it too
<mnemoc> i used the one of the nightlies made out of the head of the sunxi branch
<arokux2> oliv3r: success: http://sprunge.us/PPEa
<oliv3r> mnemoc: sunxi branch only does 1 G
<mnemoc> fix it ;-)
<arokux2> hramrach: where do they get MAC addresses for the stickers they put on your RJ-45?
<hramrach> arokux2: do you get sicker on your board?
<hramrach> sticker
<oliv3r> arokux2: you buy them from IEEE or whatever
<mnemoc> arokux2: MAC addresses are sold
<arokux2> hramrach: on Mele A1000 - yes, there was a sticker.
<oliv3r> arokux2: cubietech SHOULD have bought a MAC (per board) but haven't
<mnemoc> oliv3r: thanks! 2GB
<hramrach> if you have ROM to burn them to
<oliv3r> ok arokux2 can you repull my tree, and build that?
<hramrach> which most A10/A20 boards don't
<arokux2> oliv3r: !! your git hub wasn't up to date!
<oliv3r> olimex has i2c rom, but doesn't buy mac either; these are 'dev boards' and don't get mac's. then again Mele doesn't buy a mac either
<oliv3r> arokux2: you pushed to fast :p
<oliv3r> pulled
<hramrach> no change here
<mnemoc> arokux2: just set a MAC on userspace
<oliv3r> if mnemoc and arokux2 can compile and use this u-boot; i will push
<oliv3r> mnemoc: rather set it via boot env of u-boot
<oliv3r> a little cleaner imo :)
<mnemoc> true
<arokux2> oliv3r: Mele did put sticker
focus2 has joined #linux-sunxi
<oliv3r> arokux2: so maybe they did buy a valid mac; good!
<mnemoc> arokux2: mele bought addresses, and write them on the u-boot of each device
<mnemoc> u-boot env*
<oliv3r> arokux2: basically, companies buy a mac 'block' like 10 million or whatever mac addresses, they buy the 'prefix', in USB words, the vendor ID; they can then use as many mac addresses as they want from that pool
<arokux2> mnemoc: do you already see 2GiB in kernel?
<oliv3r> arokux2: btw, if you boot android from flash, just do 'ls' fromt he console, it will work :)
<mnemoc> arokux2: yes, with oliv3r's .bin
<oliv3r> one hugely annoying feature of having a battery backed cubietruck, is you can't just yank the power cord to reset/power it off :)
<oliv3r> ok fixing commit messages, then pushing to sunxi u-boot
<arokux2> oliv3r: there is RESET button!!!
<mnemoc> <6>Memory: 448MB 1536MB = 1984MB total
<mnemoc> <5>Memory: 1869132k/1869132k available, 162484k reserved, 1318912K highmem
<mnemoc> ^---- looks sexy
<arokux2> mnemoc: can we do something like RSS? and tell subscribers: 2GiB now visible in U-Boot and kernel?
<hramrach> just edit the CT status I guess
<mnemoc> sunxi G+ group?
<arokux2> mnemoc: nooo please, noooo
<mnemoc> linus and greg use a G+ group as official Linux blog
<oliv3r> arokux2: power down ;) but yeah i learned to love the reset button
<arokux2> mnemoc: ok, why should we?
<arokux2> oliv3r: I cannot get prompt
rellla3 has joined #linux-sunxi
<mnemoc> arokux2: you wanted to post news ;-)
<oliv3r> arokux2: no u-boot prompt?
<arokux2> oliv3r: ^
<mnemoc> arokux2: if the ML is not enough for announcement, G+ groups provide feeds ;-)
<arokux2> mnemoc: btw, our wiki provides RSS too.
<arokux2> oliv3r: hm.. but everything was fine with your *.bin!
<arokux2> oliv3r: what do you get as git describe?
<oliv3r> mnemoc: how do i git send-email patches 3-5 from head (e.g. not head, patch head-1 or head-2)
<oliv3r> v2011.09-sun4i-8404-gaea66a1
<oliv3r> weird line
<arokux2> oliv3r: git format-patch --thread --numbered --cover-letter -o outgoing -s 497a5f88d1a4fda8042b585879e36ccd5a1156e4^
<arokux2> oliv3r: how can you get this??????????
<oliv3r> i did just rebase to update all commit messages
<oliv3r> arokux2: does that include HEAD? i think it does, i only wanna send a few patches, but not the top few; i thought the syntax was 'from hash'-'to hash' but never gotten that to work
<mnemoc> just make a hash range
<oliv3r> mnemoc: ho
<oliv3r> hramrach: so works for old stuff too
<arokux2> oliv3r: then my line is wrong - it includes from hash and up
<oliv3r> yeah i tried to do hash-ranges before, i never gottten it to work
<mnemoc> 497a5f88d1a4fda8042b585879e36ccd5a1156e4..HEAD^ ?
<oliv3r> ok lets try
<hramrach> I use something like HEAD~3 to get 3 patches
<hramrach> but why HEAD^ ?
<arokux2> hramrach: what compiler are u using, plz run --version
<hramrach> excluding last patch?
<oliv3r> git format-patch 43cafbd^..da7a3b6
<oliv3r> works, angel
<montjoie[home]> If someone know what exactly means the last value of sunxi_dma_config, I will love it (what means data block size...)
<arokux2> hramrach: disregard
<hramrach> gcc (Debian 4.7.2-5) 4.7.2 (emdebian)
<arokux2> oliv3r: got prompt, now do not know why.
<oliv3r> so, you guys play with what i have in my git hub now (forced push 30 seconds ago, should be identical, just some commit messages changed) and i'll push it after bathroom break
<oliv3r> hramrach said it's all okay
<arokux2> oliv3r: I confirm everything works, so please push it.
<arokux2> montjoie[home]: now I'll test you hackery
ZetaNeta has quit [Ping timeout: 240 seconds]
rellla has quit [Ping timeout: 246 seconds]
FR^2 has joined #linux-sunxi
<oliv3r> not gonna push da7a3b6 Update all references to get_ram_size to unsigned long
<oliv3r> since that's about all other boards
<oliv3r> pushed
<oliv3r> mnemoc: can you pull and test so atleast someone tried HEAD without the patch i ommited :)
<oliv3r> arokux2: and no comment about the big secret?
<arokux2> oliv3r: :)
<arokux2> oliv3r: it depends on the outcome, will we get new devs? :)
<oliv3r> arokux2: ideally, yes
<arokux2> oliv3r: it is just awesome you will go to FOSDEM!!!!
rellla3 is now known as rellla
<arokux2> oliv3r: please ping me to update status or do it yourself: http://linux-sunxi.org/Cubietruck#Status_of_the_community_kernel_.28sunxi-3.4.29_.2F_U-Boot
<mnemoc> oliv3r: forcing an update of the nightlies
pirea has joined #linux-sunxi
<mnemoc> 8a4621c..7c867fd sunxi -> sunxi
<arokux2> mnemoc: can we somehow have `git describe` instead of "+" here "Linux alarm 3.4.67+ ...." ?
<mnemoc> that's the standard way of linux versioning
<mnemoc> we don't touch it...
<arokux2> mnemoc: I do not like it :(
<arokux2> mnemoc: we never know what ppl run
<mnemoc> argue that on lkml :p
<oliv3r> mnemoc: Warning: apc_store(): Potential cache slam averted for key 'sunxi-mw:user:id:454' in /srv/http/sunxi/linux-sunxi.org/wiki/includes/objectcache/APCBagOStuff.php on line 59
<oliv3r> arokux2: can't update, it's allready there
<oliv3r> * 2GiB RAM works with u-boot-sunxi git
<mnemoc> arokux2: i think there is a CONFIG_ to get the hash appended
<mnemoc> arokux2: we can add it to the defconfigs
<arokux2> mnemoc: we should do it.
<oliv3r> i agree, it's a realyl good idea
<oliv3r> u-boot built from git does it too
<hramrach> do we have battery status form axp?
* hramrach had a battery pack somewhere
<oliv3r> I use an old iPhone 3gS battery with a connector soldered on
popolon has joined #linux-sunxi
<hramrach> I have a battery from a mobile DVD player
<hramrach> 2x 1300 mAh
<arokux2> scripts/setlocalversion
<mnemoc> arokux2: find the CONFIG_ and submit the patch
<arokux2> mnemoc: in progress
<oliv3r> arokux2: sounds promessing
<oliv3r> i wouldn't be supprised if you use that script to build a local version
<arokux2> oliv3r: don't get distracted from git send-mail, it should happen today!
<oliv3r> lol
<oliv3r> busy atm :)
<oliv3r> i did what I could
<oliv3r> we got 2 GiB local u-boot; we're good :)
<arokux2> [root@alarm ~]# uname -a
<arokux2> Linux alarm 3.4.67-00006-gd9e27bc #28 SMP PREEMPT Wed Nov 1[root@alarm ~]# uname -a
<arokux2> Linux alarm 3.4.67-00006-gd9e27bc #28 SMP PREEMPT Wed Nov 13 21:02:41 CET 2013 armv7l GNU/Linux3 21:02:41 CET 2013 armv7l GNU/Linux
<arokux2> mnemoc: ^
<arokux2> oliv3r: ?? you said you'll post patches to ML tonight?
<oliv3r> arokux2: i assume you call setlocalversion after/during the make
<oliv3r> arokux2: i will try :p
<oliv3r> arokux2: other things to do
<arokux2> oliv3r: grep CONFIG_LOCALVERSION_AUTO .config
<arokux2> that is all I do
<oliv3r> awesome
<arokux2> oliv3r: you've promised, so please, do it :p
<arokux2> mnemoc: how can I get board name from user space?
<tomee^> do you even think it exists somewhere in the SoC? I doubt it
<arokux2> tomee^: no, but U-boot can pass it somehow.
<tomee^> arokux2: wouldn't mount /dev/mmcblko0p1 /foo && bin2fex /foo/script.bin|grep board be your best shot?
fredy has quit [Excess Flood]
<arokux2> tomee^: uhg, yes, but I'd better write board name myself :)
<tomee^> I guess U-boot is configured for a particular board but its macro-based, unsure if its stored anywhere...
<arokux2> tomee^: oliv3r just pushed 2GiB-fix to our u-boot, so you can liberate yourself from cubietechs u-boot.
<tomee^> strings and binwalk are your friends
<tomee^> arokux2: ok, will give it a shot definitely.
<arokux2> Montjoie: montjoie[home]: I tested your request, please visit http://linux-sunxi.org/User:Arokux
Fusing has joined #linux-sunxi
<arokux2> tomee^: ok, just grab u-boot-sunxi
<tomee^> arokux2: will do, but unsure if today
<tomee^> right now I am messing around with VLC and XBMC
fredy has joined #linux-sunxi
<tomee^> too bad mplayer only support vdpau with X11. and I am too stupid to produce a working ffmpeg string that would test the vdpau capability ;-)
rellla2 has joined #linux-sunxi
rellla has quit [Ping timeout: 260 seconds]
<arokux2> sun4i won't boot anymore with latest stage/sunxi-3.4 http://sprunge.us/bjZe
<arokux2> oh.. I think I've used wrong kernel.......
popolon has quit [Quit: Quitte]
<arokux2> tomee^: have you had a chance to test bluetooth?
<oliv3r> arokux2: i promised no such thing!
<oliv3r> arokux2: the BSP can get the board name form userspace :)
<tomee^> arokux2: will try if xbmc compilation succeeds ;-)
eebrah has quit [Ping timeout: 248 seconds]
<montjoie[home]> thanks arokux , Now I am using cryptodev for testing/benching
<arokux2> <oliv3r> arokux2: yeah as git send-email :)
<arokux2> <oliv3r> i will do that right now
<arokux2> <oliv3r> i'll prepare my patches, resend them to u-boot ML and then push it ot ours
<arokux2> montjoie[home]: will test on sun4i in a second
<jemk> sun7i# ping
<jemk> Using mii0 device
<jemk> host is alive
<jemk> :)
<arokux2> jemk: !!!!
<arokux2> jemk: so what did you do? :)
<arokux2> jemk: push it, i'll test it right away
<Dapsaille> may be out of subject but .. i Rockchip make an opensource SOC, is there is any chance to see it supported by sunxi ?
<arokux2> Dapsaille: haven't you seen/been to #linux-rockchip ?
<Dapsaille> no :)
<jemk> arokux2: 1) disable dcache ^^ 2) the gmac version seems to be different from the designware.c one, so some more bits need to be set
<Dapsaille> thanks
<arokux2> Dapsaille: take a look at Radxa board
<jemk> arokux2: i'll clean it up and push it, but needs some time, the experimenting made code horrible
<oliv3r> arokux2: no promise there :p but a lil busy atm
<arokux2> oliv3r: well.. you said it, not me :(
<arokux2> jemk: I can imagine!
<arokux2> jemk: but please do it asap, fiddling with uSD makes me crazy. :)
popolon has joined #linux-sunxi
rz2k has quit []
<jemk> but, is disabling dcache a good idea? if not i have to figure out what needs to be flushed
<arokux2> jemk: what is dcache anyway? you cash something already compiled?
<arokux2> montjoie[home]: what is the state of the cyphering stuff?
<jemk> arokux2: data cache of cpu, so gmac didn't see new dma descriptor because it did not get written to ram but only to cache
paulk-collins has quit [Quit: Ex-Chat]
<arokux2> jemk: ugh, you are smart.
<jemk> arokux2: had the same problem with cedar, but needed some time till i got this idea again
<arokux2> jemk: i see, it is very cool you get it sorted.
<arokux2> jemk: do you think torbenh3 can profit somehow from your findings?
<jemk> arokux2: i don't think so, cache should be handled correct in linux and the extra bits needed i got from the sunxi-3.4 driver
<arokux2> jemk: I see. another thing: but how did sun6i_gmac work in u-boot? hm.. but we actually do not know if it worked.
<arokux2> montjoie[home]: Montjoie: see Test outcome, Mele A1000 (sun4i) - same shit.
<jemk> arokux2: at least the bits are correct there too, and maybe they also disabled cache, would be worth a look
<arokux2> jemk: here is a link handy for you: http://git.hands.com/?p=u-boot.git;a=tree;h=refs/heads/allwinner-sunxi-a31;hb=refs/heads/allwinner-sunxi-a31
<arokux2> montjoie[home]: Montjoie: sun4i_defconfig doen't have CONFIG_CRYPTO_USER_API_SKCIPHER should it?
<tomee^> arokux2: ...unless you have a quick howto for testing a bt chip somewhere
<tomee^> arokux2: bluez-utils always frightened me and never did what I wanted ;]
<arokux2> tomee^: no, sorry. ok.. let it be, not that important.
<tomee^> but I guess that if wifi works then bt will as well
<tomee^> bt is more or less a crippled subset of 802.11b
<tomee^> bt2 is another story though
<arokux2> tomee^: jemk just added GMAC support to U-Boot ;)
<arokux2> tomee^: not yet public, but will be very soon.
Gerwin_J has quit [Quit: Gerwin_J]
<tomee^> what does that mean?
<tomee^> that the bootloader can initialize the correct PHY pins?
<tomee^> or that I can now do PXE/network boot? :-)
<arokux2> tomee^: the bootloader can load kernel over network
<oliv3r> jemk: did you use designware driver?
<arokux2> oliv3r: yes
<oliv3r> awesoem
<oliv3r> very very cool jemk
<jemk> oliv3r: yes, with one bit changed and dcache disabled
<arokux2> oliv3r: plz, submit patches, otherwise you'll never do it :(((
<tomee^> arokux2: neat ;-)
<arokux2> jemk: have you looked at aw's u-boot, did they disable cache?
<oliv3r> jemk: well we do need dcache on; why did you didsable it?
<jemk> arokux2: i'm not sure, they enable it only #ifdef CONFIG_SPL
<arokux2> oliv3r: why we need dcache in u-boot?
<oliv3r> the kernel assumes its on
<jemk> oliv3r: do we really need dcache in u-boot. if yes we need a way for dma coherent memory which i havent seen in u-boot
Quarx has quit [Read error: Connection reset by peer]
<mripard> oliv3r: no.
<oliv3r> mripard: oh! that's good then
<oliv3r> mripard: and hi!
<oliv3r> jemk: i wonder why you had to disable dcache though
<mripard> hi :)
<jemk> oliv3r: because gmac needs dma
<arokux2> hm.. why was it then enabled by default in u-boot
<arokux2> if the kernel doesn't want it to be enabled...
<hno> mnemoc, oh, there is highmem in 2GB boards..
pirea has quit [Quit: Leaving]
<arokux2> hno: what?
<oliv3r> jemk: ahh, how does stm handle it? do they enable dma?
<oliv3r> arokux2: hno is backreading
<arokux2> oliv3r: I thought he is saying you have missed something.
<hno> yes.
<hno> backreading.
<hno> but it's fairly important. highmem stresses kernel quite a bit more.
<oliv3r> mripard: also your a sneaky lurker, but good that you do :)
<hno> jemk, there is cache operations for dealing with DMA.
<jemk> oliv3r: good question
notmart has quit [Quit: notmart terminated!]
<hno> you need to flush affected cache ranges on DMA operations.
<mripard> oliv3r: I don't like to speak when I have nothing to say :)
<oliv3r> mripard: i speak too much; when I shouldn't say :p
Fusing has quit [Quit: Nettalk6 - www.ntalk.de]
<arokux2> oliv3r: and you have promised to resubmit patches to the u-boot ML tonight
Black_Horseman has joined #linux-sunxi
naobsd has quit [Quit: Page closed]
<jemk> hno: ok, i only searched for dma coherent alloc and such things, but flush might do.
<oliv3r> jemk: can't spy on stm driver more? :)
<oliv3r> they should be in the same situation
<jemk> oliv3r: i use the stm driver, the cache things are done by cpu/board functions
<jemk> oliv3r: but a grep for dcache_enable in spear related directories gave nothing, so i guess they don't enable it
<arokux2> jemk: git grep dcache_enable | grep spear
<arokux2> board/spear/common/spr_misc.c:dcache_enable();
<oliv3r> ah ok
<oliv3r> jemk: can i see your patch?
<oliv3r> mripard: are you going ot be @ fosdem2014?
<mripard> I'm seriously considering it
<jemk> arokux2: oups
<oliv3r> mripard: Data cache must be off. does that mean dcache will have to be turned off after u-boot is done to load the kernel?
<oliv3r> mripard: linux-sunxi
<oliv3r> mripard: @fosdem :D
<mripard> but I think I already have a training planned around that time, so I don't know if there's a collision or not
<oliv3r> weekend training?
<mripard> if not turned off, at least flushed
<mripard> no, but the training starts first time, on monday morning, so it might take some time to fly there
<mripard> this is what I have to check :)
<oliv3r> a1 & 2 feb
maz_ has joined #linux-sunxi
<mripard> yep, I know that
n01 has joined #linux-sunxi
<arokux2> my gf's birthday is on 1st feb :$
<oliv3r> 10th :p
* andoma missing fosdem this year :\
<andoma> or .. next year or whatever
<jemk> so, am i allowed to disable cache or do i have to rework designware driver and find all places to add flush
<arokux2> jemk: push with cache disabled, then there will be comments
<oliv3r> jemk: depends on what stm wants, as both boards will have to work with the one driver
fredy has quit [Excess Flood]
<arokux2> mm.. so maybe it is a good idea to leave designware untouched first?
<jemk> oliv3r: there is no flush, so they don't NEED it, and i think the only way to not need flush is to disable cache
fredy has joined #linux-sunxi
<oliv3r> jemk: why do we enable it
wolfy has joined #linux-sunxi
<jemk> oliv3r: don't ask me...
<oliv3r> jemk: :)
<arokux2> jemk: git show 0283ea70
<arokux2> oliv3r: http://sprunge.us/TLTJ
<libv> hrm. that's quick: http://opentablets.org/ "The time has come for OpenTablets.org to shutdown.."
<arokux2> oliv3r: you are from the phone, right? :p
<arokux2> libv: what was there?
<jemk> arokux2: hm
<arokux2> jemk: so the reason was: "No "WARNING: Caches not enabled" any more"
<arokux2> jemk: :D
<libv> arokux2: everything to do with the spark^W vivaldi kde community tablet
<libv> i have that hw, i used it to bring up mali400 support up a year or two ago
<libv> not knowing then that this was going to be the vivaldi tablet, i just needed something cheap with mali400
<oliv3r> arokux2: my guess is; we don't even need it enabled; so don't
<libv> i was looking for some board pictures, as i am thinking about using my freshly written serial_noise "tool" to find serial on a device
<libv> guess i need to use some telechips hw
<libv> as one of the early lima targets is mostly disassembled as it is
<arokux2> libv: haven't you heard about another kde tablet?
<arokux2> libv: aren't you on the arm-netbook list?
<libv> arokux2: i have constantly _heard_ about it
<arokux2> :)
<arokux2> now it is for real.
<libv> and i know that lkcl is hard at work
<arokux2> libv: and you know Aaron Seigo is behind it?
<libv> but this makeplaylive thing does cast a pretty bad light on it all
<libv> yes.
<arokux2> libv: will you go to FOSDEM?
<Turl> yay fosdem :)
<libv> arokux2: nah
<libv> arokux2: FOSDEM is stupid
<Turl> libv: but there's gonna be a sunxi talk this year
<arokux2> Turl: GMAC works in u-boot thanks thanks to jemk
<oliv3r> libv: buspirate has that too i think
<Turl> arokux2: :D what was it?
<libv> arokux2: there's this great big bastard called libv who organizes a devroom there
<arokux2> Turl: http://sprunge.us/TLTJ came into way and additional bits needed to be set.
<libv> oliv3r: 120locs, built statically for android, 50 of which is actual code :)
rellla2 has quit [Quit: Nettalk6 - www.ntalk.de]
<libv> anyway, about this vivaldi story, it's great that we might finally get some hardware, but one would expect these two websites to still be there
<libv> one is alive, but clearly not active, the other is dead
<libv> that's not very confidence inspiring
<libv> that they would no longer get as much hype as they would've gotten two years earlier, that's clear.
<libv> but this is definitely not helping, and something that is or would've been easy to fix
<Turl> mripard: mike sent his pull to linus.. without our patches :/
hramrach has quit [Ping timeout: 240 seconds]
maz_ has quit [Ping timeout: 244 seconds]
<mripard> Turl: seriously? ...
hramrach has joined #linux-sunxi
<arokux2> jemk: still there? I have nice contact in #u-boot and asked about caches, why they are needed etc.
<arokux2> jemk: so the dcache is solely needed "to significantly speed up the boot process"
<jemk> arokux2: half here, preparing push
<arokux2> jemk: "in general, you need to sprinkle flush/invalidate_dcache_range() whereever you use DMA in your driver"
<oliv3r> well you HAVE to turn it off before loading the kernel, the kernel assumes its off; so the only speed up is u-boot loading itself
<arokux2> oliv3r: yes. exactly.
<jemk> arokux2: yes, but that needs aligned memory and the designware driver doesn't care about aligned memory, so much work
<oliv3r> so disable it :D
<oliv3r> revert the patch
<arokux2> patch cannot be reverted, it was overwritten.
<arokux2> i.e. part of the original patch was deleted.
<arokux2> i.e. reformatted.
shineworld has joined #linux-sunxi
shineworld has left #linux-sunxi [#linux-sunxi]
<oliv3r> arokux2: yeah but i ment ..
<oliv3r> but yeah
<oliv3r> arokux2: libv always goes to fosdem :p
<arokux2> I wonder what the speed up with dcache is anyways.
<arokux2> oliv3r: I didn't know. I always ask if I don't know something. :)
<arokux2> oliv3r: and I do not promise things without doing them :p
wolfy has left #linux-sunxi [#linux-sunxi]
<oliv3r> lol
<arokux2> oliv3r: what is this for? sun4i_crane_defconfig
[7] has quit [Disconnected by services]
TheSeven has joined #linux-sunxi
<oliv3r> android it hink
<arokux2> mnemoc: ^
tomboy64 has quit [Ping timeout: 240 seconds]
tomboy64 has joined #linux-sunxi
n01 has quit [Ping timeout: 245 seconds]
<jemk> arokux2: sorry, i'll finish and push it tomorrow morning, need some sleep
<arokux2> jemk: ok, np! good night.
Soru_ has quit [Ping timeout: 272 seconds]
nove has quit [Quit: nove]
jemk has quit [Quit: night all]
AreaScout has quit []
<oliv3r> arokux2: got the mail ready to go out, just gotta write the cover letter; it's open in my editor I will think about a good cover letter and send it off tomorrow
<oliv3r> nna ll
<arokux2> oliv3r: +1
iamfrankenstein has quit [Quit: Nettalk6 - www.ntalk.de]
<Dapsaille> well .. i've made some progress regarding A13 Vga to 15khz crt ... it seems that i lack csync and i cannot find anything in docs regarding csync in sunxi
focus2 has quit [Quit: Leaving]
<Dapsaille> and of course my arcade cab monitors need csync :)
<Dapsaille> time to get some rest for now ... see ya tomorrow, code safe o/
iamfrankenstein has joined #linux-sunxi
* tomee^ go 2 sleep
tomee^ has quit [Quit: I Quit.]
Black_Horseman has quit [Quit: Αποχώρησε]
Black_Horseman has joined #linux-sunxi
iamfrankenstein has quit [Read error: Connection reset by peer]
iamfrankenstein has joined #linux-sunxi
pfdm has quit [Ping timeout: 250 seconds]
<arokux2> Dapsaille: night.
FR^2 has quit [Quit: und weg...]
dalee has quit [Quit: Page closed]
<Turl> ssvb: were you using my mbus patches? maybe it's the higher voltage and not the mbus
<arokux2> Turl: did you get your CT?
<ssvb> Turl: no, these were the old results from a long time ago
<Turl> arokux2: nope
<Turl> ssvb: ah ok
<arokux2> Turl: can you track it?
<arokux2> ssvb: please update this thread with new results, if any, I'll create a doc based on that.
<Turl> arokux2: yes I can, it's on customs
<arokux2> Turl: is Patrick Wood working for Cubietech?