**** BEGIN LOGGING AT Wed Aug 15 02:59:59 2012 Aug 15 13:10:41 ppisati, i see bug 1018907 with your 3.5 kernel as well, could we somehow not provide omapdrm if there is no backend attached to it ? Aug 15 13:10:44 Launchpad bug 1018907 in plymouth "plymouth in quantal on arm does only boot with black screen" [High,Triaged] https://launchpad.net/bugs/1018907 Aug 15 13:14:32 oh, apart from the fact that omapfb seems to be gone ... drat Aug 15 14:39:28 ppisati, so is there any reason to not upload the new kernel already ? Aug 15 14:42:34 ogra_: i'm fixing audio ATM Aug 15 14:42:59 well, even without audio it wuld be helpful to the PVR driver guys to have it Aug 15 14:49:29 ogra_: my target is end of week Aug 15 14:49:34 ok Aug 15 14:49:37 ogra_: if i can get audio working, good Aug 15 14:49:49 yeah, just upload what you have by then Aug 15 14:49:55 it already looks awesome Aug 15 14:50:00 ogra_: besides, i won't be around from 27th Aug to 8 Sept Aug 15 14:50:09 ogra_: vacation Aug 15 14:50:29 hmm, so we should do a test session in friday then Aug 15 14:50:54 oh, wait, 27th is still a week out Aug 15 14:52:03 yep Aug 15 14:52:13 if i have it in the archive by friday, we have one week for testing Aug 15 14:52:26 great Aug 15 15:53:02 robclark, do you have any idea about bug 1018907 (and why the plymouth kvm handler doesnt work with omapdrm) Aug 15 15:53:02 Launchpad bug 1018907 in plymouth "plymouth in quantal on arm does only boot with black screen" [High,Triaged] https://launchpad.net/bugs/1018907 Aug 15 15:56:12 ogra_: mee too, but on amd64 with cryptsetup which mangles grub fb options (?!) stuff =) Aug 15 15:56:36 xnox, ouch Aug 15 15:56:58 though i doubt it is the same issue ... mine seems to be on the kernel driver level Aug 15 15:57:10 (or the missing peieces of libdrm) Aug 15 15:57:12 ogra_: e.g. the new crypt stuff I did for desktop fails to reboot in a VM =( Aug 15 16:21:13 ogra_, if you have the setup to reproduce that issue, maybe you can somehow grab kernel traces w/ drm.debug=7? Aug 15 16:21:45 robclark, will do, no prob, is it supposed to work ootb with the kvm driver of plymouth ? Aug 15 16:22:29 I'd expect so, unless plymouth is trying to special case every drm driver Aug 15 16:22:43 if it uses libkms like the drm test apps do, it should work fine Aug 15 16:23:03 no, i think it tries the opposite, but if a driver has special needs it seems to ship a HW specific libdrm-$hw Aug 15 16:23:17 if it is trying to special case every drm driver, and only knows about desktop drm drivers, then plymouth might need some changes Aug 15 16:23:32 it shouldn't even need libdrm_$hw Aug 15 16:23:57 libkms will fall back to using the "dumb" (ie. not tiled) scanout buffer allocation Aug 15 16:23:58 well, it needs it for radeon and nouveau apparently ... thats what made me wonder Aug 15 16:24:06 (if there is no libkm_$hw backend) Aug 15 16:24:20 right Aug 15 16:24:43 it is possible that plymouth pre-dates the "dumb" buffer allocation ioctl.. that is why I was asking what plymouth is doing :-P Aug 15 16:24:58 ah, well, let me produce that log Aug 15 16:25:00 if you point me at de codez, I can go have a look Aug 15 16:45:24 ogra_: we'll need to fix it anyway, we need to use omapdrm Aug 15 16:45:37 also, integrating xserver-xorg-video-omap will not solve anything Aug 15 16:46:02 that helps later, and it's needed by pvr, but it's not used by default Aug 15 16:46:14 it needs a xorg file just to load the driver, which is something we don't want Aug 15 16:46:26 I sent an email yesterday to the xorg-dev ml about this issue Aug 15 16:49:46 * robclark suspect issue should be easy to fix.. I just need to look at the plymouth code and then I should be able to send a patch Aug 15 16:50:57 robclark: https://code.launchpad.net/~plymouth-dev/plymouth/trunk :-) Aug 15 16:51:07 thx Aug 15 16:57:04 robbiew: http://bazaar.launchpad.net/~plymouth-dev/plymouth/trunk/view/head:/src/plugins/renderers/drm/ply-renderer-libkms-driver.c#L288 where it fails Aug 15 16:58:06 robclark: ^ Aug 15 16:58:06 ;) Aug 15 16:58:10 rsalveti, hmm, that *should* work Aug 15 16:58:14 sorry ;-) Aug 15 16:58:22 maybe a libkms backend missing? Aug 15 16:58:39 does it need a backend for omap? Aug 15 16:58:49 no, the 'dumb' backend should work fine Aug 15 16:59:13 (and I think they should all get built in to one .so.. /me just checking) Aug 15 16:59:35 and we're using the latest libdrm code Aug 15 16:59:42 2.4.38 Aug 15 16:59:51 rsalveti, do you see the same issue if you try to run modetest from libdrm? Aug 15 16:59:58 it should be working in exact same way Aug 15 17:00:00 ogra_: when was the last time you reproduced the bug? Aug 15 17:00:24 yeah, looks like there is nothing special to enable when building libdrm to have the 'dumb' backend Aug 15 17:02:00 rsalveti, fyi, this is what modetest does, which works: http://hastebin.com/wucabeqike.avrasm Aug 15 17:02:28 ahh, hah, I see the issue Aug 15 17:02:44 remove these two lines: Aug 15 17:02:45 ply_array_add_uint32_element (attributes, KMS_PITCH); Aug 15 17:02:45 ply_array_add_uint32_element (attributes, (uint32_t) *row_stride); Aug 15 17:03:01 then use: kms_bo_get_prop(bo, KMS_PITCH, stride); Aug 15 17:03:19 the dumb backend will fail out if you try to pass KMS_PITCH into the constructor like that Aug 15 17:03:21 hm, ok Aug 15 17:03:34 row_stride is passed as an argument Aug 15 17:03:37 probably nouveau, radeon, etc, don't check for that.. they have their own custom libkms backends Aug 15 17:03:43 right, it shouldn't be like that Aug 15 17:04:56 actually, looking at latest libdrm, I sort of think it should be failing on desktop too Aug 15 17:06:41 for both radeon and nvidia they are using their own calls Aug 15 17:06:46 like radeon_bo_open and such Aug 15 17:07:07 now it'd be interesting to check on a generic use case as well on the desktop Aug 15 17:07:27 brb, will test it in a few Aug 15 17:08:02 ahh Aug 15 17:08:17 plymouth should be able to use libkms for everything Aug 15 17:08:30 and not have to special case for nouveau, radeon, etc (fwiw) Aug 15 17:40:13 interesting, wonder why we have special cases for them Aug 15 17:40:23 could be because libkms wasn't working properly at that time? Aug 15 17:43:36 could anyone point me to some sample code that shows how to do fullscreen GLES2 on the beagleboard without running X ? Aug 15 17:44:52 I went to google and typed in "beagleboard GLES2 sample" and a bunch of my own web sites popped up as the top matches lol Aug 15 17:53:26 rsalveti, I'd guess maybe they pre-date libkms? Not sure Aug 15 17:57:05 just fyi, we are still taking feedback on the PandaBoard-NC - http://groups.google.com/group/pandaboard/browse_thread/thread/74b5c6cc761d2e3c# Aug 15 18:32:56 rsalveti, this afternoon Aug 15 18:33:44 ogra_: just tested here with the latest image and it seems that it worked just fine Aug 15 18:33:59 http://cdimage.ubuntu.com/daily-live/current/quantal-desktop-armhf+omap4.img Aug 15 18:34:13 let me check the logs Aug 15 18:34:57 you have a splash after install ? Aug 15 18:34:59 maybe because of the latest libdrm package Aug 15 18:35:07 hm, not yet after Aug 15 18:35:10 but at the installer Aug 15 18:35:13 during the first boot Aug 15 18:35:29 unless the set of packages used by the installer is not the same Aug 15 18:35:50 you want to edit the cmdline in preEnv.txt btw, fki doesnt handle changing it yet Aug 15 18:36:09 (after install before firest boot) Aug 15 18:36:24 yeah, noticed that Aug 15 18:36:46 i didnt have any splash after instaklll and nothing with the 3.5 kernel from paolo either Aug 15 18:36:56 [ply-renderer.c] ply_renderer_open_plugin:could not query rendering device for plugin /lib/plymouth/renderers/drm.so^M Aug 15 18:36:56 [./plugin.c] close_device:closing device^M Aug 15 18:36:56 [./plugin.c] unload_driver:unloading driver^M Aug 15 18:37:08 got another error here at the installer's splash Aug 15 18:37:09 weird Aug 15 18:38:24 so that's why it worked Aug 15 18:38:33 it gave up on drm, and went over framebuffer Aug 15 18:38:52 but that's still not the desired behavior Aug 15 18:39:03 let me finish the installer and will check what happens after that Aug 15 18:40:03 looks like i have 2.4.38-0ubuntu1 of all the libdrm bits here Aug 15 20:19:16 Hello, I have a performance problem running Ubuntu 12.04 on my Pandaboard. When idle the load average is 0.5 or even 1.3 ish and when browsing or apt-get it goes to 2.0 or even 3. Aug 15 20:20:21 as reference, what load average (command top), should I expect to have after a fresh intall and leave it idle for 15min? Aug 15 20:20:39 I came to this step: http://omappedia.org/wiki/Ubuntu_PPA the repositories were correct, but I didnt have the "Install OMAP4 addons" (is that outdated?), so i tried to install it via commandline. evertthing was installed, and when I rebooted it, the display to HDMI/tv was black and flickering... Aug 15 20:20:53 I tried taht because I thought ubuntu wasnt hardware accelerated, so slow performance in the desktop.. but now I fear its something hardware related, as even a fedora console has a high load average when idle..... Aug 15 20:21:05 Finlandia: If it's actually completely idle, then you should expect close to zero. But "when I run stuff, the CPU gets used" isn't a performance problem.. Aug 15 20:22:18 mmm, when opening firefox and watch a youtube vid, it skips alots of frame, maybe 1 frame every 2s Aug 15 20:22:43 its not usable, and i dare to say atm the rPi is even faster.. Aug 15 20:23:07 is a fresh install of Ubuntu 12.04 hardware accelerated? Aug 15 20:23:30 or do i need to install the OMAP4 addons somehow...? Aug 15 20:24:27 how can I find out what the cause of the slow performance is? Aug 15 20:25:20 Finlandia: load avg usually means i/o is slow Aug 15 20:25:43 Finlandia: and the i/o on a panda is rather on the slow side, especially with a cheap sd card Aug 15 20:26:12 s/load avg/high load avg/ Aug 15 20:26:22 so a class 10 card is useless? Aug 15 20:27:01 my "extreme pro" sandisk card has 95MB/s printed on it - it can do 20MB/s on a good day Aug 15 20:27:17 problem is the bus is 8bits 20MHz or so (IIRC) Aug 15 20:27:26 so with a slow card, it just gets worse Aug 15 20:28:10 so what do you suggest? dont use the SD card and run ubuntu from USB stick? Aug 15 20:28:24 that's probably even slower Aug 15 20:28:35 USB 2.0 isn't very fast Aug 15 20:29:24 ok, so what can I do to make it go faster Aug 15 20:29:26 seems USB 2.0 should be able to do 35MB/s - that's mostly theory - most USB sticks can't do that Aug 15 20:29:32 get a fast SD card Aug 15 20:29:37 the one I have works fine Aug 15 20:29:45 I seen yt videos with a good performance ubuntu. Aug 15 20:29:51 it's still slow compared to most other systems, but it works Aug 15 20:30:20 I got a class 10 card, and writing the image on SD card goes quick enough. Can't remember the actual MB/s anymore. Aug 15 20:31:32 but even a 15min idle (fresh install and doing aboslutling nothing except command top), uses the SD card continous (read/write) and results in a slow performance. that can't be right.... Aug 15 20:31:55 class 10 is 10MB/s Aug 15 20:32:11 meaning it can probably deliver half of that in practice Aug 15 20:32:23 IIRC I used a class 10 when I first setup my panda Aug 15 20:33:21 Finlandia: apt-get install sysstat - enable it in /etc/default/sysstat, start it Aug 15 20:33:35 that'll make "sar" give you info about how the system load is over time Aug 15 20:34:15 system load, cpu load, i/o load, whatnot Aug 15 20:36:18 ok, i'll try that, atm puting a new image on it, writing to SD card with 13MB/s... Aug 15 20:37:09 probably faster than most USB sticks can do Aug 15 20:37:57 also installing xbmc is a good test to see if the board performance is good? Aug 15 20:38:13 I can remember that I couldnt even start xbmc few weeks ago... Aug 15 20:38:28 let alone run a 720p or even a 1080p movie... Aug 15 20:39:08 even browsing with precise on a pandabiard in 1080p is a horror Aug 15 20:39:35 it's a small thing, not good for what a PC can do Aug 15 20:40:26 yeah ok, but compared to a rPi, I would expect atleast hte same and tbh a better performance :P Aug 15 20:41:42 pandaboard has a better CPU/GPU than the rPi, right? Aug 15 20:43:52 not sure - the panda has a better cpu, but the rPi may have a better GPU Aug 15 20:44:45 * RoyK bought some crap PSUs on ebay "rated" 2A, giving 4V on measuring between TP1 and TP2 on the rPi Aug 15 20:44:53 so, haven't tested it too much yet Aug 15 20:45:53 ok, cheers for the help :) Aug 15 20:59:34 ogra_: also, mem=456M@0x80000000 mem=512M@0xA0000000 is not needed anymore Aug 15 21:26:46 after todays update on omap4+armhf there is no desktop. bug filed. Aug 15 21:37:44 bug #1037306 Aug 15 21:37:46 Launchpad bug 1037306 in xorg "after todays update there is no desktop on omap4+armhf on pandaboard ES" [Undecided,New] https://launchpad.net/bugs/1037306 **** ENDING LOGGING AT Thu Aug 16 02:59:58 2012