**** BEGIN LOGGING AT Mon Mar 28 02:59:57 2011 Mar 28 12:08:56 hey guys can somebody explain to me how opengl works / is supposed to work when combined with framebuffer-drivers? (nvidia tegra board) Mar 28 13:34:30 GrueMaster: ping Mar 28 14:46:23 ppisati: pong Mar 28 14:49:38 GrueMaster: LP#586386 Mar 28 14:49:49 bug 586386 Mar 28 14:49:49 Launchpad bug 586386 in linux-ti-omap4 "Kernel should hand over all cmdline args to the init environment" [Medium,Fix committed] https://launchpad.net/bugs/586386 Mar 28 14:50:11 the option is there, but it doesn't spwan any getty at boot Mar 28 14:50:16 maverick, omap4 Mar 28 14:50:19 Not sure, but I think that was fixed. Mar 28 14:50:29 GrueMaster: ok, move it to fix-released then Mar 28 14:50:46 Need to double check with ogra. Mar 28 14:51:01 GrueMaster: but in my case, if i don't add the tty[OS]2 in /etc/init, i don't get any console Mar 28 14:51:02 ogra_: ^^^ Mar 28 14:51:20 GrueMaster: bug 672635 Mar 28 14:51:21 That's a different issue I believe. Mar 28 14:51:21 Launchpad bug 672635 in linux-ti-omap4 "enable CONFIG_SCSI_MULTI_LUN" [Medium,Fix committed] https://launchpad.net/bugs/672635 Mar 28 14:51:28 the same, the option is there Mar 28 14:51:37 And I tested it. Mar 28 14:51:45 k, -> fix released Mar 28 14:55:02 btw, do we the datasheet for these chips? Mar 28 14:58:01 I wouldn't know. Mar 28 19:49:13 rsalveti, is drivers/gpu/pvr moving in to kernel tree? Or did I misunderstand? Mar 28 19:49:41 robclark: not for ubuntu, we're still working as a separated module Mar 28 19:49:57 ahh, ok.. so same as before Mar 28 19:50:14 robclark: don't know if the ti android kernel maintainers are planning to push that upstream Mar 28 19:50:16 probably not Mar 28 19:50:26 it isn't really upstreamable.. Mar 28 19:50:40 yeah, so it'll probably stay as it is now Mar 28 19:50:41 but it is quite convenient to have it not as out-of-tree Mar 28 19:50:53 (the only thing I like about android kernel) Mar 28 19:51:49 Only thing? How about...er...ok I'll give you that one. :P Mar 28 20:30:39 Hi there. Sorry for the newbie question, but I have a netboot/NFS root setup and I'd like to build my own Panda kernel with the right things compiled in. Where should I start? Mar 28 20:31:00 (I need the Ubuntu kernel to get the command line MAC address setting) Mar 28 20:47:52 The easiest way to start is to use the Ubuntu kernel config and change the settings you need. Then everything else will just work. Mar 28 20:58:28 GrueMaster: thanks. So I start with kernel.org/.../ubuntu-maverick.git on the omap4 branch? Where does the .config live? Mar 28 20:59:20 Not sure. I only do image testing. rsalveti? Mar 28 20:59:55 You could also look up the ubuntu kernel build process on the wiki. Mar 28 21:00:17 https://wiki.ubuntu.com/Kernel/ Mar 28 21:02:19 michaelh1: do you need the maverick one? Mar 28 21:02:22 or nattys? Mar 28 21:02:44 rsalveti: sorry, natty. Typo. Mar 28 21:03:33 michaelh1: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=summary Mar 28 21:03:37 branch ti-omap4 Mar 28 21:03:37 rsalveti: I'd like to use the Linaro one but I need a fixed MAC address. I haven't/didn't want to open that can of worms... Mar 28 21:04:10 michaelh1: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=blob;f=debian.ti-omap4/config/config.common.ubuntu;h=f3d9425065695f7698f71d4222d5427268abe551;hb=refs/heads/ti-omap4 Mar 28 21:04:13 the config file Mar 28 21:04:31 rsalveti: OK. Copy that to .config, make oldconfig, make menuconfig Mar 28 21:04:41 michaelh1: yup Mar 28 21:04:57 rsalveti: what's the story with using the full 1 GB of RAM these days? Mar 28 21:05:22 michaelh1: still not fixed I believe, but didn't test with latest one yet Mar 28 21:05:40 OK. Can I use more than 460 MB? Mar 28 21:05:40 I remember the issue was still there at the time of rc6, rc7 Mar 28 21:05:47 768 should be ok Mar 28 21:06:07 mem=768 or some type of split mem=460, mem=256@xyz? Mar 28 21:06:20 if you're not using any gst with ducati Mar 28 21:06:33 michaelh1: the split is to leave some mem to ducati Mar 28 21:06:42 you can just use 768 directly if you're not using it Mar 28 21:06:47 Ah. These are headless so I'll drop it. Mar 28 21:08:35 Sorry again. git clone git://kernel.ubuntu.com/ubuntu/ubuntu-natty.git has completed. 'git branch' only lists master. How do I select the ti-omap4 branch? Mar 28 21:11:04 michaelh1: git remote -a should list the remote branches Mar 28 21:12:03 then just give git branch ti-omap4 remotes/origin/ti-omap4 Mar 28 21:12:15 and with --track if you still want to use pull on it Mar 28 21:13:28 Ah, success. Thanks. For some reason git 1.7.1 (Maverick) doesn't have git remote -a. Mar 28 21:14:37 rsalveti: is this written up somewhere? Otherwise I'll blog it so I remember it... Mar 28 21:15:10 michaelh1: don't know if it's posted somewhere already, just blog it :-) Mar 28 21:15:17 rsalveti: will do Mar 28 21:45:56 Hmm. All of my PandaBoards end up with the same MAC address of 72:b9:94:3b:51:04 Mar 28 21:49:44 michaelh1: you can change that by a module parameter or by changing using normal debian tools Mar 28 21:50:20 rsalveti: the old .macaddr module parameter has gone away. I'm doing NFS root with no uInitrd so it has to be done in the kernel. Mar 28 21:51:03 I think the parameter is still working with our kernel Mar 28 21:51:27 rsalveti: the code for smsc95xx.macaddr=xxx is gone. There seems to be something re: USBNET_PLATDATA_FLAG__USE_MAC but I'm still tracking it down... Mar 28 21:52:07 rsalveti: the function smsc95xx_is_macaddr_param() is gone as is the module parameter macaddr in drivers/net/usb/smsc95xx.c Mar 28 21:53:04 michaelh1: yeah, it's applied but then reverted: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=commitdiff;h=4c35e56b6bd6a8f372e24c4ed21dc4bce6e7c444;hp=497525b16697ed40b8e4c2bcc2b634472f6a663d Mar 28 21:53:55 Huh, not the most useful comment. Once you've got userspace you can probably fix this. Mar 28 21:55:33 michaelh1: hm, http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=commitdiff;h=6114ac31d9805956f6115d4eceba78edd03a8792;hp=321610a00ebdfacc6df4c7537626ab03ec0278b5 Mar 28 21:55:48 it should be "unique" Mar 28 21:57:01 at least with our latest kernel Mar 28 21:57:49 The trick is to see if it is persistent & unique. Mar 28 21:58:25 OK. Same uImage, different boards. Here's board 1: [ 6.402801] smsc95xx 1-1.1:1.0: eth0: register 'smsc95xx' at usb-ehci-omap.0-1.1, smsc95xx USB 2.0 Ethernet, 72:b9:94:3b:51:04 Mar 28 21:58:36 * michaelh1 swaps boards Mar 28 22:00:29 [ 7.566955] smsc95xx 1-1.1:1.0: eth0: register 'smsc95xx' at usb-ehci-omap.0-1.1, smsc95xx USB 2.0 Ethernet, 72:b9:94:3b:51:04 Mar 28 22:00:45 so it's not working :-) Mar 28 22:02:02 Well, not to the depth required. Not sure how it is getting this data, but if it is using something from the cpu that is burned in at wafer fab time, it could be unique to each wafer. Mar 28 22:02:37 Not as rare in a multiplatform deployment, but still enough for normal usage. Mar 28 22:03:07 http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=commitdiff;h=321610a00ebdfacc6df4c7537626ab03ec0278b5;hp=ed89bb009173a8a4bf0567311e616a6100b1353f Mar 28 22:04:03 Yip. I see in struct platform_async_platform_data panda_async_pdata_map that it refers to usb1/1-1/1-1.1 and this seems to be usb1/1-1/1-1.0 Mar 28 22:08:09 Hacking it and setting .device_path = "usb1/1-1/1-1.0" gives unique MAC addresses... Mar 28 22:08:19 Heh. I'd like to see them use 00-00-36 as the starting mac address. Mar 28 22:09:31 Hmm? It's currently in the 80:xxx series which are private... Mar 28 22:09:55 (sorry, was looking at the ieee mac oui database. That was for Atari. ) Mar 28 22:10:37 Heh. Disregard the above - I've now broken it as the MAC addresses are now random. It seems the 1.1 was right, but the addresses are not unique. Mar 28 22:15:26 michaelh1: makes sense, now why die_id is the same for all your pandas I'm not sure Mar 28 22:16:04 I just printed out the die ID and they're identical: 1b95c d00f0451 1140 0 (four hex words) Mar 28 22:22:48 tap_base is fc002000. Is that correct? Mar 28 22:38:53 Huh, id.c seems to be wrong. According to the TRM, the DIE ID is at 208, 20C, 210, and not 218, 21c, 220. It's actually reading the product ID which is why it's consistent. Mar 28 22:41:58 michaelh1: cool Mar 28 22:43:16 Right, I have a hack-around. I can't find the die ID in the OMAP3 manual to see if it's always been wrong. Mar 28 22:47:41 Huh. The die IDs are a6caea0 0 105e0023 300f0 and a6caea0 0 105e0023 300f0 so only id_2 is unique... Mar 28 22:56:33 Hmm. mem=768M causes a halt at early boot: http://pastebin.ubuntu.com/586652/ Mar 28 22:58:07 michaelh1: weird, let me check the options Mar 28 22:58:26 I have: "console=tty0 console=ttyO2,115200n8 root=/dev/nfs earlyprintk fixrtc nocompcache vram=0 mem=768M ip=dhcp" Mar 28 22:58:56 I also have mem=768M@0x80000000 but not using the latest kernel, still 2.6.38-1206-omap4 Mar 28 22:59:02 and here it's working fine, let me upgrade it Mar 28 23:01:48 Does it interact with u-boot any? Mar 28 23:03:08 mem=512M works, mem=520M doesn't, u-boot claims 512 MB of RAM Mar 28 23:03:41 hm, maybe your x-loader? Mar 28 23:03:53 I'll replace MLO and u-boot and see what happens... Mar 28 23:03:58 I believe we're using latest linaro's u-boot Mar 28 23:06:42 Ah, that fixes it. Mar 28 23:06:46 * michaelh1 tries 768M... Mar 28 23:08:02 michaelh1: hm, what x-loader and u-boot were you using before? Mar 28 23:08:13 2010.11? Mar 28 23:08:21 u-boot now claims 1 GB of memory Mar 28 23:09:00 rsalveti: are you still around? Mar 28 23:09:08 TheUni: hey, yes Mar 28 23:09:14 rsalveti: great Mar 28 23:09:28 rsalveti: one of our devs has done some mangling of the packages, i'd like to ask you a few questions if you don't mind Mar 28 23:09:44 TheUni: sure Mar 28 23:10:17 rsalveti: basically, we have lots of packages. we would much prefer to condense them down into something like xbmc and xbmc-data... Mar 28 23:10:25 but i'm not sure what policy dictates Mar 28 23:10:43 TheUni: what are the packages you currently have? Mar 28 23:10:49 and what are they for? Mar 28 23:10:51 for example, we have some event-clients. each client is a separate binary. for ex, a bin for a ps3 remote control. Mar 28 23:11:20 in the past, we've split these out into one package per binary. so xbmc-eventclients-ps3, xbmc-eventclients-java, etc Mar 28 23:11:53 we would much prefer to just install all of these along with xbmc. but not if it would be denied by ubuntu policy Mar 28 23:12:51 TheUni: well, actually there's nothing wrong for having one major package for it, the only issue is that to use one of them you'll bring all dependencies Mar 28 23:13:16 rsalveti: well in this case, you can't use the clients without xbmc anyway. so it's kinda moot. Mar 28 23:13:35 I mean, for eventclients-java for example, java is probably necessary Mar 28 23:13:49 if java is already a dependency for xbmc, then makes no much sense Mar 28 23:13:56 oh, i see Mar 28 23:14:08 yea, understood. Mar 28 23:14:35 but it was my understanding (please correct me if i'm wrong) that in general, the debian policy is to split packages for each functional binary. Mar 28 23:14:37 package split makes more sense in the dependency level than just functionality speaking Mar 28 23:14:52 yes, in that example, it would make sense. Mar 28 23:15:20 in the best case yes, but if you're creating tons of packages with just small binaries, you could just create one major one Mar 28 23:15:37 if the dependencies are quite all the same Mar 28 23:15:41 ok Mar 28 23:16:13 yep, makes perfect sense. I'll have a look at the binaries and see. we'll construct packages based on that and no extras and see how things look Mar 28 23:16:22 *have a look at the deps Mar 28 23:16:38 cool Mar 28 23:17:21 TheUni: do you have any other updates regarding the gles support on xbmc? Mar 28 23:17:40 rsalveti: gles is fully supported. still just a matter of getting it packaged Mar 28 23:17:43 I remember topfs2 had something working, but unfortunately still didn't check Mar 28 23:17:52 TheUni: oh, that's nice, good to know Mar 28 23:18:00 TheUni: already at trunk? Mar 28 23:18:09 rsalveti: yep. it's used on the ipad port. Mar 28 23:19:07 awesome, natty is finally working with gles and 2.6.38, should be able to try it once I get some free time Mar 28 23:19:51 great Mar 28 23:43:14 dang, the HDMI KVM switch I got doesn't seem to forward EDID info Mar 28 23:50:21 Amaranth: I have the same problem with my 5 port HDMI switch. Have to make sure the switch is on the system I need before waking up the system. Mar 28 23:50:23 rsalveti: the code in id.c in linux-linaro-2.6.28 is correct both re: OMAP4 die ID and how to set a MAC address. Mar 28 23:50:48 ah, right Mar 28 23:56:55 I hope the autoswitch thing on this doesn't try to outsmart me Mar 28 23:58:42 dang, it does Mar 29 00:12:17 at least moving my rootfs to a usb stick worked correctly, my read speeds are now 11.9MB/s average according to palimpest instead of 1MB/s or whatever the SD card gives Mar 29 00:51:31 hrm, E-EDID checksum failed Mar 29 01:22:49 easily solution for that, I put other devices on the HDMI KVM so the panda can get its own direct connection Mar 29 02:52:31 rsalveti, ping **** ENDING LOGGING AT Tue Mar 29 02:59:57 2011