IRC log of #cubox of Wed 21 Nov 2012. All times are in CET < Back to index

00:52 ptl anyone has a kernel with device-mapper (thus LVM) support?
00:52 ptl for ubuntu 12.10
00:52 ptl or... clear directions on how to compile a recent kernel with all the needed drivers for cubox to work properly?
01:15 ptl no?
01:18 shesselba rmk_: kirkwood spdif registers on DT, now comes the testing
01:51 shesselb 01:51 * shesselba wonders what clk-gates should also be enabled for i2s1..
02:01 shesselba rmk_: it is not the clk gates but something different that locks up the cubox when using aplay .. I guess something is wrong with my DT stuff, will investigate later
02:15 shesselba hmm, dco doesn't lock but all clocks are enabled..
05:50 ralix morning
05:50 dotarray hello ralix!
05:51 dotarray had a chance to play with your new cubox yet?
05:51 ralix Good Morning dotarray! I installed Archlinux on my new Cubox :)
05:52 dotarray fabulous!
05:53 dotarray i'm just writing about archlinux for a new blog post, actually!
05:53 ralix oh yes, i am so happy!
05:53 dotarray did you use the CuBox installer, or go through all of the individual steps?
05:54 ralix I use the installer, was very easy
05:55 dotarray great news :D
05:56 ralix I have a problem at the moment with the correct location for the advancedsettings.xml. I put the xbmc home to "/var/xbmc" because I problme with "/var/empty". I have installed xbmc-cubox-git.
05:57 dotarra 05:57 * dotarray wonders if anyone is around to help
05:58 ralix But I hope in the next few hours ;-)
05:58 dotarray fingers crossed!
05:58 ralix :) thx!
05:59 dotarray i wish it was something i was more familiar with :)
05:59 dotarray soon, maybe :D
06:01 ralix No problem, have already received very good help here. I can practice my English with some nice conversations with you.
06:02 dotarray i don't think i ever caught up with you after you went to the football the other weekend. did your team win?
06:04 ralix Unfortunately lost the last game, but I have hope for the next weekend. But this is here at home :) No hour journey to nowhere location.
06:06 dotarray oh good :)
06:07 ralix oh yes :)
06:08 dotarray which team is it, by the way?
06:12 ralix Babelsberg 03 (14.) vs. Münster (06.)
06:19 ralix I love coffe at morning :)
06:27 dotarray i love tea all the time! :)
06:27 dotarray sorry, distracted by this archlinux stuff :)
06:32 ralix I just hope rabeeh, N30N or ... it can help me today :)
06:33 dotarray i'm crossing my fingers for you!
06:35 ralix It's so cold here. I can not wait 2 weeks I'm in Jakarta. Then it's finally warm again.
06:35 ralix thx thx thx !
06:37 dotarray ah, yes, jakarta would definitely be warm :)
06:38 ralix oh yes :)
06:38 ralix 3 weeks
06:40 ralix Now quick shower and then to work. I'm back in 1 hour.
06:40 dotarray have fun!
06:41 ralix thx ;-)
08:06 ralix morning again
09:04 ralix dotarray, A tip for your Arch Linux article. If you want to reduce the CPU load at idle you have to make three adjustments in the advancedsettings.xml.
09:04 ralix vi .xbmc/userdata/advancedsettings.xml
09:04 ralix
09:04 ralix
09:04 ralix
09:04 ralix 3
09:04 ralix 0
09:04 ralix
09:04 ralix
09:04 ralix false
09:04 ralix
09:04 ralix
09:04 ralix I just have to find the right location for the advancedsettings.xml;-)
09:05 jnettlet ralix, is the XBMC UI using GLESv2 for rendering?
09:06 ralix i hope so, i use the latest archlinux snapshot und xbmc-cubox-git from portage
09:14 ralix 12:59:02 T:3013521408 NOTICE: Setup SDL
09:14 ralix 12:59:02 T:3013521408 NOTICE: Checking resolution 16
09:14 ralix 12:59:10 T:3013521408 NOTICE: GL_VENDOR = Marvell Technology Group Ltd
09:14 ralix 12:59:10 T:3013521408 NOTICE: GL_RENDERER = GC600 core
09:14 ralix 12:59:10 T:3013521408 NOTICE: GL_VERSION = OpenGL ES 2.0
09:14 ralix 12:59:10 T:3013521408 NOTICE: GL_SHADING_LANGUAGE_VERSION = OpenGL ES GLSL ES 1.00
09:14 ralix 12:59:10 T:3013521408 NOTICE: GL_EXTENSIONS = GL_OES_depth24 GL_OES_stencil1 GL_OES_stencil4 GL_OES_fragment_precision_high GL_OES_compressed_ETC1_RGB8_texture GL_OES_compressed_paletted_texture GL_OES_vertex_half_float GL_OES_depth_texture GL_EXT_texture_format_BGRA8888
09:14 ralix GL_OES_EGL_image GL_OES_packed_depth_stencil GL_OES_rgb8_rgba8 GL_VIV_shader_binary
09:14 ralix this is from xbmc.log
10:04 ralix jnettlet, was that what you wanted to know?
10:05 jnettlet ralix, yeah, thanks
10:05 ralix ok
10:06 ralix I hope it comes soon one can help me with my problem-;-)
10:25 Punkley_Chillin look at me using the installer to build the ubuntu 12.10 image I built :)
10:30 jnettlet Punkley_Chillin, did you get the graphics driver compiled and working?
10:30 Punkley_Chillin just trying that now actually
10:31 Punkley_Chillin compiled yes
10:31 Punkley_Chillin rebooted and the box didn't come back
10:31 Punkley_Chillin so im doing a clean install
10:31 jnettlet okay...the graphics driver shouldn't be effecting that. Maybe no display but everything else should work.
10:31 N30N ralix: advancedsettings.xml gose in the userdata folder, so for you that'll be /var/xbmc/.xbmc/userdata/
10:31 Punkley_Chillin apt-get install ubuntu-desktop
10:32 N30N ralix: In my testing algorithmdirtyregions didn't seem to help improve performance.
10:33 Punkley_Chillin i do have some kind of a kernel issue as booting up I still don't have a fb
10:34 N30N ralix: make sure you enable the vmeta-clk@667 service.
10:34 ralix So it is in the right place. Unfortunately the RssFeed still runs.
10:34 ralix [root@alarmcubox userdata]# pwd
10:34 ralix /var/xbmc/.xbmc/userdata
10:34 ralix [root@alarmcubox userdata]# ls -la advancedsettings.xml
10:34 ralix -rw-r--r-- 1 xbmc xbmc 234 Nov 20 07:08 advancedsettings.xml
10:34 ralix [root@alarmcubox userdata]#
10:35 ralix Ok but the RssFeed helps;-)
10:36 N30N I think there's a gui option in the skin setting to turn off RSS.
10:43 ralix I did that on my x86 with GeeXboX / ubuntu way and it worked.
10:44 ralix I look at times where you have to write in the otherwise might / should.
10:46 ralix [email protected] - Set vmeta clock to 667
10:46 ralix Loaded: loaded (/usr/lib/systemd/system/[email protected]; enabled)
10:46 ralix Active: active (exited) since Tue, 2012-11-20 12:58:31 CST; 14h ago
10:46 ralix Process: 135 ExecStart=/bin/sh -c echo %I000000 > /sys/devices/platform/dove_clocks_sysfs.0/vmeta (code=exited, status=0/SUCCESS)
10:46 ralix CGroup: name=systemd:/system/[email protected]/667
10:46 ralix :)
10:52 ralix I now have time in the guisettings.xml: 16485 xbmc 20 0 225m 37m 23m S 4.6 5.4 0:13.15 xbmc.bin
10:52 ralix The load is has become smaller, I hope the RssFeed off. I'm not at home, unfortunately.
11:53 rmk 11:53 * rmk_ wonders if Rabeeh saw my question about vmeta power/clk control
11:53 rmk_ Rabeeh: what's the story behind your 'hack' comments in vmeta.c ?
12:03 jnettlet rmk_, are on vacation or something?
12:10 rmk_ jnettlet: more grammar required to understand that sentence :) who is the subject of that question?
12:12 jnettlet rmk_, sorry my wireless keyboard is low on juice...I need to go and get more batteries
12:12 jnettlet if I type to fast I lose words
12:12 jnettlet are you on vacation?
12:12 rmk_ no :)
12:12 jnettlet ah okay...just busy got it
12:13 rmk_ why do you ask?
12:13 jnettlet you have just gotten very into this recently. figured it was your vacation project
12:14 rmk_ umm... "recently"... not really, it's been something I've been playing with since July.
12:14 rmk_ which is why I'm currently at 190 odd commits for the kernel alone
12:17 jnettlet excellent.
12:17 rmk_ I started hacking on it because what was supplied on the initial install was diabolical; tried on two TVs and one HDMI monitor and none of them would recognise the output from the cubox if they were powered on when the cubox booted.
12:17 rmk_ what I found in once case was I had to boot the cubox with HDMI disconnected, let it fully boot, and then plug the monitor in.
12:18 rmk_ I decided that this was utter crap, and that's when I decided I'd convert both the kernel and X server to DRM.
12:18 rmk_ because DRM can handle hotplugged monitors correctly
12:20 rmk_ and it rather snowballed from there
12:25 jnettlet I am surprised you are still using the bmm driver
12:26 rmk_ well, vmeta requires contiguous chunks of memory, and nothing else in my system hands out contiguous chunks of memory anymore
12:26 rmk_ all my X pixmaps are SHM-backed
12:26 rmk_ except for the scanout buffers
12:27 jnettlet how does galcore like that? it wants its memory mapped as uncached
12:27 rmk_ err, no it doesn't
12:28 rmk_ galcore gets on just fine; it has its own MMU, and if you deal with it sensibly it doesn't matter if its cached or not
12:29 jnettlet I was having all sorts of corruption if the memory allocated it for drawing wasn't coherent
12:30 rmk_ that's because you need to track the ownership of pixmaps
12:30 rmk_ and keep them owned as much as possible by the GPU (iow, do as much on the GPU as possible)
12:31 rmk_ its also very very probable that you're hitting bugs in the galcore kernel layer; it's ability to wait for operations to complete is... broken.
12:31 jnettlet yeah I started with that but found that performance was too slow. I would be curious how fast a default gtkperf runs under your setup.
12:32 jnettlet rmk_, yes I know that very well
12:33 rmk_ I'd run it now, but vlc is paused on the cubox since last night atm :) (and that means vlc is repeatedly sending the same video frame via Xv) so it won't give correct results
15:42 rmk_ hmm, I think I need to push the sdhci changes into mainline too
17:53 rmk_ Andrew Lunn has just confirmed to me that Marvell have confirmed my "no race free way to clear interrupts" issue
17:57 jnettlet rmk_, I missed that episode. Which interrupts can't be cleared?
17:57 rmk_ umm. wrong way round. they can be cleared, even ones you didn't know were pending.
17:58 rmk_ http://lists.arm.linux.org.uk/lurker/message/20121118.162944.a6dd4b4b.en.html
17:59 rmk_ that patch is heading for the stable kernel trees too now
21:08 Punkley_Chillin quietly screens to to world, so close
21:18 shesselba rmk_: the reason why my DT enabled i2s fails is, because I also registered kirkwood-dma on the same pdev (as in tegra30_i2s)
21:18 shesselba and that overwrites the dai private_data with pcm private_data
21:18 shesselba So all bugs introduced by me.. I
21:19 rmk_ ok, so nothing to worry about as far as those patches go
21:19 shesselba nope
21:20 shesselba OTOH this makes me think how to register kirkwood-dma on DT. It is not a device as i2s is..
21:20 rmk_ that's the big silly problem of asoc :(
21:21 rmk_ asoc's structure forces a non-hardware on everything, which is wrong.
21:21 shesselba yeah, I put spdif-tranceiver within the "simple-bus" node.. just to have it parsed by someone
21:22 rmk_ these things like dma drivers and i2s interface drivers should all be libraries which the 'soc' stuff (like kirkwood-spdif) uses
21:22 rmk_ but that introduces one issue: there's some i2s interfaces which are entirely separate from their dma backends...
21:23 rmk_ but if you start out as having all these mini-drivers as libraries in the first place, you can stitch them together whatever way you need
21:23 shesselba yeah, I ll find another way to store dma private_data somewhere
21:25 rmk_ I'll see about chucking out the sdhci changes in the next day or so; it won't be DT because Rabeeh's kernel + v3.6 isn't DT for dove.
21:27 rmk_ hmm, except the driver has been DT-ized since v3.6
21:31 shesselba sdhci is on DT
21:35 rmk_ just converting it over to also support DT
21:36 rmk_ if you have a DT based setup on the cubox, can you test this patch?
21:36 shesselba yes
21:36 shesselba but there is already a of_device_id table for sdhci-dove
21:37 rmk_ yes, just missing the gpio cd stuff
21:37 shesselba great! I have the rev1 with gpio cd bug ;)
21:37 shesselba sdio0 cd routing bug
21:40 rmk_ now, this will currently support both DT and the legacy method...
21:40 rmk_ http://www.home.arm.linux.org.uk/~rmk/cubox/sdhci.diff
21:40 rmk_ I haven't run that myself yet, so...
21:40 rmk_ but it does build
21:41 rmk_ and getting rid of the legacy stuff is fairly trivial
21:41 rmk_ and before it gets submitted I'll move that damned gpio.h include :)
21:42 rmk_ like that... there, includes sorted alphabetically
21:43 rmk_ is this just for fixing that rev1 bug then?
21:45 shesselba yes, it allows to overwrite present bit with the value of the gpio that also has sdio1 cd..
21:51 shesselba hmm, I guess polarity should be also configurable..
21:51 rmk 21:51 * rmk_ has no idea what revision my cubox is
21:53 rmk_ note that many of these sdhci platform support modules are broken
21:54 rmk_ they're all making the same old mistake: they publish the device to the world and then do the setup
21:54 shesselba yeah, that was my mistake you already fixed ;)
21:54 rmk_ we covered that in the very first few kernel summits
21:55 rmk_ and here we are in 2012 and almost all the sdhci stuff make that mistake because sdhci bringup didn't propagate my two-stage initialization stuff
21:56 rmk 21:56 * rmk_ didn't create the two stage mmc_alloc_host...mmc_add_host just for kicks 8/
21:57 rmk_ sdhci_pltfm_register really needs to die.
22:01 shesselba hmm, if DT gpio is wrong, i.e. returns <0, sdhci_dove_readl panics
22:02 rmk_ why is your gpio_is_valid() returning true for gpio numbers < 0 ?
22:02 shesselba DT gpio-mvebu
22:02 shesselba maybe
22:03 rmk_ static inline bool gpio_is_valid(int number)
22:03 rmk_ {
22:03 rmk_ return number >= 0 && number < ARCH_NR_GPIOS;
22:03 rmk_ }
22:03 rmk_ is the asm-generic definition of that
22:04 rmk_ which, afaics, should be the one in use for dove
22:06 shesselba it also fails if gpio is correct (12) and valid
22:06 rmk_ send oops :)
22:07 shesselba http://pastebin.com/DXVNpe1t
22:08 rmk_ argh. fsck. you've hit what I was talking about above.
22:08 rmk_ sdhci-pltfm needs to be fixed for this to work
22:09 shesselba hmm, host and host->ioaddr seem to be correct at least
22:09 rmk_ oh, it _can_ be fixed already
22:09 rmk_ expect a new patch shortly
22:17 rmk_ new patch there; I've redone the probing entirely
22:18 shesselba same uri?
22:18 rmk_ yep
22:19 rmk 22:19 * rmk_ changes his nick to his old linuxnet nick, as my present nick appears to be taken here :(
22:22 _rmk 22:22 * _rmk_ chuckles at the IET's latest publication... this month on climate change.
22:23 shesselba _rmk_: it boots without oops, you want me to pull the card? it's my rootfs ;)
22:23 _rmk_ whatever makes you feel happy that you've tested it sufficiently :)
22:25 _rmk_ (if you haven't heard, people are being flooded out of their homes for the third time since August... again, another year of uk floods)
22:29 shesselba http://pastebin.com/LjUJGwS2
22:29 shesselba boot without usd
22:29 shesselba oops
22:30 shesselba could also be related with gpio-mvebu
22:31 _rmk_ usd?
22:32 shesselba micro (u) sd
22:32 shesselba but I remeber that gpio-mvebu is now irq controller itself
22:33 _rmk 22:33 * _rmk_ wonders...
22:33 _rmk_ so, the tasklet is only initialized in sdhci_add_host, not in sdhci_alloc_host
22:34 _rmk_ that makes it unsafe to register the irq before sdhci_add_host :(
22:34 shesselba but why does it work with usd plugged in?
22:34 _rmk_ it probably doesn't trigger an irq immediately upon request in that circumstance
22:36 shesselba if I pull out the usd after sdhci, it detects removal and insertion correctly
22:37 shesselba yeah, but without usd the oops is back
22:38 _rmk_ whoa, tegra's sdhci cleanup doesn't look good
22:40 _rmk_ maybe third time lucky
22:41 _rmk_ new patch at same url
22:44 shesselba testing in a second, about kirkwood-dma: ok to put prdata into kirkwood-i2s priv data?
22:44 shesselba that is sitting in snd_soc_platform_get_drvdata() when I register dma within i2s
22:44 shesselba to avoid DT node for dma
22:45 _rmk_ for now, I think so; I think the longer term solution would be to integrate the two together properly
22:48 shesselba sdhci without usd still panics
22:49 _rmk_ in the same way?
22:49 _rmk_ argh, sorry
22:49 shesselba yes
22:50 _rmk_ guess who copied the patch file to the web server _before_ creating the replacement patch
22:50 _rmk_ the right one's there now. sorry about that.
22:50 shesselba \me wasn't ;)
22:50 shesselb 22:50 * shesselba of course
22:55 shesselba guess what :)
22:55 shesselba without usd it does boot and wait for rootfs.. I plug the usd in and it gets recognized and mounted :)
22:56 _rmk_ yay, success.
22:57 shesselba and dco locks again (or I should better say, kirkwood-dma doesn't f*cks up priv->io and kirkwood-i2s now looks up the correct register)
22:58 shesselba aplay /usr/share/sounds/alsa/Front_Center.wav runs through.. now I need to carry my cubox to my audio receiver
23:05 shesselba 2ch pcm works
23:08 shesselba ac3 passthough too
23:14 _rmk_ yay!
23:15 shesselba looking for my dts sample..
23:17 shesselba mplayer -ao alsa:device=iec958 -novideo -ac hwdts -aid 0 dts-starts-later-than-ac3.m2ts
23:17 shesselba Forced audio codec: hwdts
23:17 shesselba Opening audio decoder: [hwac3] AC3/DTS pass-through S/PDIF
23:17 shesselba hwac3: switched to DTS, 1536000 bps, 48000 Hz
23:17 shesselba AUDIO: 48000 Hz, 2 ch, ac3le, 1536.0 kbit/100.00% (ratio: 192000->192000)
23:17 shesselba Selected audio codec: [hwdts] afm: hwac3 (DTS through S/PDIF)
23:18 shesselb 23:18 * shesselba listens to DTS over SPDIF on cubox ;)
23:27 shesselb 23:27 * shesselba didn't even know his audio receiver can play back DTS-HD HR 7.1
23:27 shesselba _rmk_: great work on kirkwood-i2s :)
23:54 shesselba wondering if 5% cpu of mplayer playing truehd sample can only be passthrough.. unfortunately my receiver doesn't tell me the audio format