**** BEGIN LOGGING AT Thu Jan 31 02:59:58 2013 Jan 31 03:13:18 yup, otg Id polling Jan 31 03:13:24 66k to vusb Jan 31 05:05:39 so it WAS HNP? Jan 31 05:27:20 hi mr ds2 Jan 31 05:45:42 hey mranostay Jan 31 06:28:27 Russ: no idea on the voltage, but will your work make it possible to have a nice runtime even for power switches to we can do opp_enable and opp_disable accordingly? Jan 31 06:44:57 is it me or stuff still showing up in the ELC schedule? Jan 31 06:46:30 those are not the talks you are looking for *waveshandmysteriously* Jan 31 06:47:12 The beer reception is ) Jan 31 06:47:14 :) Jan 31 06:50:13 which reminds me: tomorrow, fosdem warmup at the delirium cafe <3 Jan 31 06:51:04 * dm8tbr is _very_ happy that mozilla invited him Jan 31 06:52:53 dm8tbr: beer! Jan 31 06:58:09 indeed Jan 31 06:59:33 and friday! Jan 31 07:10:45 * dm8tbr is not able to distinguish between weekdays for at least a month now Jan 31 07:13:41 then, it's good for you :) Jan 31 07:13:56 I would like to have such a long weekend also Jan 31 07:16:23 ynezz: 'fraid it's the opposite... Jan 31 07:17:19 well once you know your company is hopeless you can enjoy your weekends more :) Jan 31 07:17:55 as in resume fixing, beach time, and side projects :) Jan 31 07:19:17 that's my project for the upcoming few months Jan 31 07:20:32 also why my ELC talk is under a fake name also.. cheap bastards wouldn't let me take time off ... :) Jan 31 07:20:48 well fake company name Jan 31 07:22:09 Fake industries Jan 31 07:22:23 ACME corp? Jan 31 07:22:34 Ranostay Industries :) Jan 31 07:22:48 i could get a LLC on that i suppose Jan 31 07:28:20 industries? not heavy industries? Jan 31 07:28:44 and you wanna be my latex salesman? Jan 31 07:29:27 Seinfeld, anyone? http://www.youtube.com/watch?v=_T35QhLx_KI Jan 31 07:31:56 coldsoup: thought that actually :) Jan 31 07:47:43 g Jan 31 07:47:44 ds2: heavy is a nice tough Jan 31 07:47:47 *touch Jan 31 07:54:14 Jan 31 07:55:33 be carefull, that will get you shot in canada Jan 31 07:55:39 that would be american Jan 31 07:55:59 nuqneH Jan 31 07:56:23 * coldsoup couldn't resist Jan 31 08:04:54 * mranostay runs a few bandpass calculations Jan 31 09:08:07 I don't like sound drivers. Jan 31 09:08:11 I don't like them at all. Jan 31 09:10:42 gm Jan 31 09:12:46 Kristina, I like them better than Video Drivers! Jan 31 09:13:06 (and anything with DT in the name) Jan 31 09:13:08 Kristina: what driver do you like? Jan 31 09:13:25 GPIO, that way I can bitbang :P Jan 31 09:13:31 +1 Jan 31 09:13:35 bitbang sound too Jan 31 09:13:41 worked on my C64 Jan 31 09:14:07 ya, worked on the ][+ too Jan 31 09:14:07 hehe Jan 31 09:14:22 av500, c64? part of any known group? Jan 31 09:14:41 ka6sox: Actually, video drivers are not that bad. Jan 31 09:14:57 Porting video drivers from Linux to Darwin is not that complicated. Jan 31 09:15:03 Kristina, says you...I've been fighting them for months Jan 31 09:15:12 ka6sox: Why? :( Jan 31 09:15:19 try from android to linux. Jan 31 09:15:55 ka6sox: Meh, you don't even need to worry about the kernelmode drivers. Jan 31 09:16:11 Just write a WSEGL shim. Jan 31 09:16:13 panto: group? as in l33t k1ds? no Jan 31 09:16:29 Or whatever that thing was called. Jan 31 09:16:59 LibHybris helps...but we were used to 2 FBs and Android only provides 1 Jan 31 09:17:01 Besides, doesn't TI already provide drivers for Linux? Jan 31 09:17:20 Kristina, "its complicated" Jan 31 09:17:48 thats the best answer I got. Jan 31 09:17:48 So what's wrong with TI's drivers? :P Jan 31 09:17:58 Do they not work? Jan 31 09:18:02 they do Jan 31 09:18:05 Kristina, lol Jan 31 09:18:05 the OMAP4 ones are interesting. Jan 31 09:18:30 Which SGX core does OMAP4 have? Jan 31 09:18:35 the issues is that they are not TI's drviers Jan 31 09:18:41 but rather imgtec ones Jan 31 09:18:49 av500, +1 Jan 31 09:19:10 av500: Not really, IMGTEC provides the core DDK while TI adds stuff onto it. Jan 31 09:19:11 540? Jan 31 09:19:41 * ka6sox can't remember because we stoped the forehead flattening exercises. Jan 31 09:19:51 Kristina: I doubt TI adds much Jan 31 09:20:02 544 in 4470 or so Jan 31 09:20:18 ya, this is 4460 we are using Jan 31 09:20:33 For example, iOS uses the Usc/Use stuff from the DDK, but the oglcompiler stuff is replaced by LLVM and the OGL implementation is replaced by GPUM. Jan 31 09:20:36 ka6sox: in what kind of product? Jan 31 09:20:54 Samsung Galaxy Nexus Jan 31 09:21:03 ah that Jan 31 09:21:15 Kristina: then I guess AAPL does for more than ti Jan 31 09:21:36 Aw, I wonder how hard it would be to get the 530 DDK to work on 544. Jan 31 09:21:51 just use the 544 DDK Jan 31 09:21:59 I don't have the 544 DDK ... Jan 31 09:22:16 av500, our issue is we don't use X11 Jan 31 09:22:26 so X11 bindings are of no use to us. Jan 31 09:22:36 And as I've gathered, obtaining it from TI is going to be a major pain. Jan 31 09:22:47 ka6sox: who is "we"? Jan 31 09:22:58 webOS Jan 31 09:23:01 ah Jan 31 09:23:10 * av500 slowly backs away Jan 31 09:23:19 he he Jan 31 09:23:26 av500: how different is 530 from 544? Jan 31 09:23:43 Kristina: I could google it :) Jan 31 09:23:57 In terms of the actual DDK crap. Jan 31 09:24:38 3.8-rc5 Jan 31 09:24:44 Kristina: no idea, never looked into that deeply Jan 31 09:24:45 will it ever end. Jan 31 09:25:14 Oh well, beagle has sgx530, so I'm fine for now. Jan 31 09:25:57 jo Jan 31 09:26:06 hey ho Jan 31 09:26:53 PandaBoard on the other hand, has sgx540. Jan 31 09:27:03 yes Jan 31 09:27:06 that the ES? Jan 31 09:27:07 4430/4460 have 540 Jan 31 09:27:11 and 4470 has 544 Jan 31 09:27:25 God damn it :( Jan 31 09:27:35 kk Jan 31 09:28:26 I hate proprietary GPUs. Jan 31 09:28:40 fact of life... Jan 31 09:29:15 at least i2s is fairly straight forward compared to the video minefield :P Jan 31 09:29:45 ka6sox so you hate all embedded gpu Jan 31 09:29:47 Kristina: are the non-proprietary ones? Jan 31 09:29:54 ups Jan 31 09:29:54 Nope :( Jan 31 09:30:16 I just hope that the DDKs are not that different. Jan 31 09:30:20 They probably aren't. Jan 31 09:30:33 woglinde, they are a necessary evil. Jan 31 09:31:20 Because I really don't feel like using binary ELF blobs on Darwin. Jan 31 09:31:48 tannenbaum would NOT like you. Jan 31 09:32:54 That requires some serious linker hacking. Jan 31 09:33:14 Well, I would probably use a separate linker instead of hacking up ELF support on top of dyld. Jan 31 09:33:24 * ka6sox hauls down the rc5 patches and sees if anything fun is in there. Jan 31 09:33:40 So have something like a shim library that invokes the ELF linker to bootstrap the blobs. Jan 31 09:33:46 okay I am off again Jan 31 09:33:58 laters woglinde Jan 31 09:34:16 fun and games with binary blobs. Jan 31 09:34:21 The great thing about Darwin is the almost-nonexistent community. Jan 31 09:34:43 There's Apple, there's the OSx86 people and there's the PureDarwin people. Jan 31 09:34:47 That's it. Jan 31 09:36:05 how many is that in total? Jan 31 09:36:11 Not that many. Jan 31 09:36:33 I think there are around 300 or so people who really know what they're doing when it comes to working with the kernel. Jan 31 09:36:37 Maybe even less. Jan 31 09:37:37 meh...nothing interesting. Jan 31 09:38:22 So basically, most of the issues you have you have to sort out by yourself. Jan 31 09:50:23 hm, that BB OS 10 looks cute Jan 31 09:50:37 (from the app developer POV) Jan 31 09:51:29 no something crazy like java or objective-c, finally :) Jan 31 09:55:14 in my driver I have almost managed to get DT probing working, but it only probes the first of my child nodes for each spi device, do I have to iterate over them from my probe function or is my syntax wrong, pastebins coming... Jan 31 09:55:39 DT: http://pastebin.com/S7YqAcUj Jan 31 09:55:56 Driver: http://pastebin.com/dzgeP1sz Jan 31 09:56:13 Debug Output: http://pastebin.com/zYJHx86C Jan 31 09:57:47 I would have thought it would callback for each child node... Jan 31 10:04:39 Cool, so people are actually using device trees. Jan 31 10:04:41 How cool. Jan 31 10:05:26 Kristina: Attempting to ;) Jan 31 10:09:59 Kristina: do you really think there are 300 kernel trolls? Jan 31 10:10:14 Kernel trolls? :O Jan 31 10:10:38 the KT are almost certainly asleep though Jan 31 10:11:46 Are you implying that all Darwin kernel developers are "kernel trolls" (whatever that means)? Jan 31 10:14:00 Kristina: I think he's implying that the 300 people that work on the kernel wouldn't design and use devicetree just for fun :) Jan 31 10:15:09 Oh, I'm not saying that there is anything wrong with device trees. Jan 31 10:15:22 They are mandatory on Darwin anyway. Jan 31 10:15:47 I actually have the bootloader build and populate the DT based on an XML spec. Jan 31 10:15:53 im not suggesting anything Jan 31 10:15:55 Rather than doing it with that DT tool. Jan 31 10:16:22 Kristina: fyi, #beagle is a pseudoname for #middleearthdevelopers Jan 31 10:16:42 so we have have trolls, hobbits and dwarfs Jan 31 10:16:50 when Jan 31 10:17:03 sometimes some nasty goblins turn up Jan 31 10:17:22 unforutnatly we cant' finsih the story, untill Part II of the hobbit comes out Jan 31 10:18:27 dont we have the Nazgul too? Jan 31 10:19:40 "Nine he gave to Mortal Men, proud and great, and so ensnared them. Long ago they fell under the domination of the One, and they became Ringwraiths, shadows under his great Shadow, his most terrible servants." Jan 31 10:20:12 thats technically not hobbit, its LoTR Jan 31 10:20:51 I know what a "troll" is. Jan 31 10:21:11 I'm just not certain about what that would mean in a kernel development context. Jan 31 10:21:31 its a compliment Jan 31 10:21:33 re: driver issues, all fixed now - I experimented and found that you do have to iterate over the nodes from within the probe Jan 31 10:22:13 I get called insane for thinking that integrating an XML parser into a bootloader was a good idea. Jan 31 10:22:27 * av500 calls Kristina insane :) Jan 31 10:22:34 It also has a microkernel. Jan 31 10:22:57 Take that, U-boot :P Jan 31 10:24:36 ynezz: hey, objc is cool! Jan 31 10:25:51 It's funny, at Apple some developers seem to have a thing against Foundation/ObjC so they stick with CF and C. Jan 31 10:26:16 And then other developers start writing ObjC code when they join those teams. Jan 31 10:31:23 No one seems to be interested in Darwin on Beagle :( Jan 31 10:32:49 Kristina: what SW would I run on it? Jan 31 10:33:30 Well, technically, you could run iOS on it ... apart from the fact that all the drivers are missing so it wouldn't quite work. Jan 31 10:33:57 so, what would I practically do with it? Jan 31 10:34:29 I don't know, I think it's just cool. Jan 31 10:34:38 Like, I can run iOS recovery ramdisks on it. Jan 31 10:35:21 Want to try running a full iOS 2.0 image but that would require some work. Jan 31 10:35:36 As in, not fully working, but just get the UI (SpringBoard) to show up. Jan 31 10:35:42 Which is not actually that hard. Jan 31 10:35:49 Compared to getting things like Audio to work. Jan 31 10:37:12 http://pastebin.mozilla.org/2100477 Jan 31 10:37:13 :D Jan 31 10:39:30 I am terrible. Jan 31 10:40:32 Kristina: cool? It's like Perl, humans can't parse the code... Jan 31 10:41:15 ynezz: I have never had any issues. Jan 31 10:43:34 Here is some objc code I wrote a while back: http://pastebin.mozilla.org/2100478 Jan 31 10:43:43 What's wrong with it? Perfectly readable! Jan 31 10:44:34 * panto actually likes objc Jan 31 10:44:47 cause it's OO done right (at runtime where it belongs) Jan 31 10:45:23 It's A LOT nicer than C++. Jan 31 10:46:36 true that Jan 31 10:46:59 if only it caught on Desktop Linux space as well Jan 31 10:47:45 I was not talking about c++ Jan 31 10:47:54 that's insane and mess as Perl Jan 31 10:48:28 only sane usage of C++ I've seen so far, is that used in Qt :] Jan 31 10:48:30 I was actually thinking of building my own Darwin based OS. Jan 31 10:48:47 Inspired by iOS. Jan 31 10:48:55 Kristina: yes, that snippet is quite readable Jan 31 10:48:58 Kristina: have you talkes to Dr Schaller lately? Jan 31 10:49:04 talked* Jan 31 10:49:47 koen: I don't understand all those references :( Jan 31 10:50:10 Kristina, do you like movies with gladiators? :) Jan 31 10:50:22 No. Jan 31 10:50:43 do you like the movie Airplane then? Jan 31 10:51:23 Kristina: http://projects.goldelico.com/p/mySTEP/ Jan 31 10:51:25 Never see it. Jan 31 10:51:56 seen* Jan 31 10:52:04 koen: is it like exactStep? Jan 31 10:52:13 koen: Hm, I'm not a big fan of GNUstep. Jan 31 10:52:19 It's too GNU-ey. Jan 31 10:59:25 I prefer PureFoundation + Apple's CoreFoundation. Jan 31 10:59:41 who doesn't like the move airplane Jan 31 11:00:01 I don't like movies in general. Jan 31 11:03:19 yep, waste of time Jan 31 11:03:41 * ynezz prefers pubs instead Jan 31 11:27:17 lumdidum Jan 31 12:25:55 does anyone know if it is possible to add child nodes to a DT file at a later stage Jan 31 12:25:57 e.g. Jan 31 12:26:18 am33xx.dtsi has the pinctrl for the deault setup Jan 31 12:26:40 I want to add my own pinctrl but from my own dtsi file which includes the am33xx.dtse Jan 31 12:27:08 reason: I don't want to have a delta against an upstream file Jan 31 12:27:09 isn't that how the dts files in firmware/capes do it? Jan 31 12:27:25 hmmm, I haven't checked the capes Jan 31 12:29:37 koen: is overlay generic, or is it specific to !capebus Jan 31 12:30:58 hmm seems to be DT overlay specific (in the name I suppose....) Jan 31 12:31:24 however, I'm not interested in changing the devicetree live, just an overlay at compile time Jan 31 12:35:48 then look at home am335x-beaglebone-common.dtsi does it Jan 31 13:33:27 ok, my DT looks like this: http://www.hastebin.com/jowiwogujo.xml Jan 31 13:34:03 and I call devm_pinctrl_get_select_default from the probe but settings the pins fails Jan 31 13:34:18 I just have dummy pins in at the moment so don't worry about that Jan 31 13:34:49 but is there something fundamentally wrong with my DT, or am I bumbling the driver code Jan 31 13:36:24 how does it fail? what is the error return? Jan 31 13:36:41 and is that a cape fragment? Jan 31 13:36:49 or part of the base tree? Jan 31 13:37:13 thats my dtsi which includes the bone-common.dtsi and am33xx.dtsi Jan 31 13:37:50 my driver is here: http://www.hastebin.com/nokiworajo.coffee and line 85 is the pinctrl Jan 31 13:38:10 the pinmux nodes must be child nodes of the am33x_pinmux node Jan 31 13:38:39 where do these pinctl nodes added? Jan 31 13:38:39 so I need to change the am33xx.dtsi to include my pinctrl, I can't overlay in a later dtsi? Jan 31 13:39:00 I didn't say that Jan 31 13:39:10 you can overload it on the board dts Jan 31 13:39:43 with the &foo { }; syntax Jan 31 13:39:53 checkout how am335x-bone.dts etc does it Jan 31 13:40:05 ok, will do thanks Jan 31 14:10:42 when I measure the voltage on the extension pins the the XM board (while in u-boot) they are at 1.8 volt. why is that? Jan 31 14:11:47 why should they not? Jan 31 14:15:06 doesn't sound safe as default Jan 31 14:16:07 av500, ours is not to ask why Jan 31 14:17:18 ok, so my pincntrl call is returning -19 which is apparently device not found, but I'm passing it the spi_device directly from the probe, is the real problem being hidden somewhere else - is -19 just a filler for a more quirky error? Jan 31 14:18:33 keesj: default should be input state Jan 31 14:18:46 so they are maybe getting pulled high Jan 31 14:18:52 so I guess it's a pull-up Jan 31 14:18:53 or they are high-z and float high Jan 31 14:26:00 quote of the day: "You can redistribute the documents to your customers, but only to potential ones" Jan 31 14:26:34 I've seen would-be customers with no potential Jan 31 14:35:06 well, let's hope, that someone who's doing git pull of the source code is potential customer :) Jan 31 14:39:11 how do you convert am335x pins numbers to the hex you find in the DT? I'm going to set the correct pins and then hope it falls into place :) Jan 31 14:40:35 ah no worries I've just managed to find a wiki page Jan 31 14:41:16 jackmitchell: use a macro Jan 31 14:41:17 bummer, you've almost missed white smoke badge Jan 31 14:42:48 it's register offset, sigh... of the TRM I go Jan 31 14:44:00 jackmitchel, it's an offset from 0x800... Jan 31 14:44:26 that'll help correlate the addresses in the TRM Jan 31 14:49:08 hi emeb Jan 31 14:59:10 hi woglinde Jan 31 14:59:35 things are slow here - my ISP's DNS servers are flaky this morning. Jan 31 15:01:05 _never_ use the ISP DNS servers Jan 31 15:05:44 yes Jan 31 15:05:49 use /etc/hosts Jan 31 15:06:00 mru: whose DNS to use then? Jan 31 15:06:16 emeb_mac: mine Jan 31 15:06:17 bind is free software Jan 31 15:06:23 also trust my CA Jan 31 15:06:30 hee Jan 31 15:07:08 btw, Google now has its own CA, so if you use 8.8.8.8, they own you :) Jan 31 15:07:58 there are these things called root servers... Jan 31 15:08:39 google already owns me Jan 31 15:09:26 I'd trust google more than some random isp Jan 31 15:11:43 "fred's dns shed" Jan 31 15:11:48 "trust no one" Jan 31 15:12:45 shyster mcshysterson's house o' dns - our motto "trust me, oh yeah..." Jan 31 15:13:11 "free candy" Jan 31 15:13:30 jo panto Jan 31 15:13:37 hi woglinde Jan 31 15:29:02 mdp: would you mind taking a very quick look to see if you notice anything out of place with my pinctrl, I did it almost exactly the same as your st77 driver, but it keeps failing Jan 31 15:29:09 driver: http://pastebin.com/5rnn0PaQ Jan 31 15:29:15 dt: http://pastebin.com/zLNDpn4e Jan 31 15:29:30 the only difference is that I have multiple child nodes Jan 31 15:35:22 jackmitchell: by inspection, it looks ok..but.. Jan 31 15:36:00 are you sure you don't have an "off-by-one" error? where the gpio references don't line up to the actual IP block in the TRM Jan 31 15:36:21 &gpio1 is actually GPIO0 Jan 31 15:36:26 something to doublecheck Jan 31 15:36:39 ah yes, that is fine I have accounted for it, the devm_ function is what keeps failing Jan 31 15:37:05 returns -19, which is apparently NO_SUCH_DEVICE Jan 31 15:37:07 ok, just checking..honestly, my brain keeps forgetting that here even though I know the problems exists Jan 31 15:37:11 interesting Jan 31 15:37:38 but I pass it the spi->dev directly from the probe, so that should be ok Jan 31 15:38:57 ah wait, I think I might know Jan 31 15:41:25 bah, no luck - I had one of the spi_pins references wrong, but it didn't help Jan 31 15:56:50 ok, if it is failing in the pinctrl part, then I reckon it must be somewhere in pinctrl_lookup_state_locked Jan 31 15:57:14 as that is the only fail state which also doesn't have an print error and bail on ENODEV Jan 31 16:03:55 ok, it's fixed I think it may have been a devicetree problem, and I fixed it with one of my earlier changes Jan 31 16:04:24 but I was making the rookie mistake of scp'ing the dtb over and then not syncing the mmc before hard rebooting the board Jan 31 16:04:34 * jackmitchell hangs head in shame Jan 31 16:06:57 * mdp recommends nfsroot Jan 31 16:08:35 jackmitchell, sync and mmc don't always do what you think it does Jan 31 16:08:37 :) Jan 31 16:08:46 sd cards are dumb Jan 31 16:10:06 You wish they where dumb Jan 31 16:10:54 they are mostly too clever Jan 31 16:11:07 mdp: yes, I've been meaning to set it up for some time now, been just getting round to it for the past few months..... Jan 31 16:11:31 jackmitchell: I think you'll find it to be a superior workflow. Jan 31 16:12:03 and power cycle to your heart's content! Jan 31 16:13:23 some people like the poor man's alternative which is sshfs, but that's dependent on the thing being booted and such Jan 31 16:13:35 you can edit stuff host side, etc..much easier to work that way Jan 31 16:14:01 it's why people doing embedded linux on boards with ethernet typically have always used nfsroot for devel Jan 31 16:14:26 I tftp the uImage in uboot and use mmcroot Jan 31 16:14:28 arm stuff has this legacy of having no ethernet or almost non-working smsc parts slapped on them ;) Jan 31 16:15:11 koen, I do not approve...but I'm willing to accept you as you are. :) Jan 31 16:16:58 can't we all get along? :) Jan 31 16:16:58 nfsroot people are the reason the mmc drivers suck Jan 31 16:17:39 nfsroot is darned convenient mmc drivers or not Jan 31 16:18:58 I guess the TI cpsw guy uses mmcroot as well Jan 31 16:19:08 that's must be why that driver sucks so much Jan 31 16:19:29 not net no nfs Jan 31 16:23:22 koen: that servo hack is to expose the pot value on a wire? Jan 31 16:25:19 yes Jan 31 16:25:52 so you can read the position before powering it on Jan 31 16:25:54 or moving it Jan 31 16:28:07 ah Jan 31 16:30:45 does anyone have a good prize question? Jan 31 16:31:18 for a fosdem contest to win some beagle hw? Jan 31 16:33:59 koen, what is the air-speed velocity of an unladen swallow? Jan 31 16:35:11 54 Jan 31 16:35:23 ahhhhhhhh Jan 31 16:35:26 koen: what is 26+26 Jan 31 16:35:51 Does OMAP3 provide a eMMC-like interface to the NAND controller? Jan 31 16:35:52 av500 to much insider Jan 31 16:36:01 Or do you have to do stuff like FTL inside the driver? Jan 31 16:38:56 Kristina: yes Jan 31 16:39:00 in SW Jan 31 16:39:16 but if you have a BB-XM, there is no NAND Jan 31 16:40:14 koen, mmcroot people are to blame for starving puppies Jan 31 16:40:22 hateses them! Jan 31 16:40:52 koen: wifi capes? Jan 31 16:41:54 koen, you are citing test methods from a group of people that consider boot and mount of rootfs from jffs2 as a complete test of hardware BCH support ;) no joke Jan 31 16:42:27 Hm, I have a crazy idea. Jan 31 16:42:57 Move ALSA to the userland and then map audio IO memory into the usermode. Jan 31 16:43:46 why use alsa at all then Jan 31 16:43:55 av500, +1 Jan 31 16:44:11 av500: what else is there to use? Jan 31 16:44:19 it's obviously possible to drive anything from userspace with mmap Jan 31 16:44:20 pulseaudio? :P Jan 31 16:44:38 provided you win the fight with the kernel for control of the clocks Jan 31 16:44:56 and power where applicable Jan 31 16:45:11 Oh, right, I forgot that audio is very time sensitive. Jan 31 16:45:26 that's not what I'm talking about Jan 31 16:45:48 Well, I meant in terms of latency. Jan 31 16:45:54 oh, and you'll have a hard time dealing with interrupts in userspace Jan 31 16:46:01 panto: http://dominion.thruhere.net/koen/angstrom/beaglebone/Angstrom-Cloud9-IDE-GNOME-eglibc-ipk-v2012.12-beaglebone-2013.01.31.img.xz Jan 31 16:46:07 You have to respond to interrupts in a very short timeframe otherwise you get lag. Jan 31 16:46:16 that's with 3.8.0-rc5 Jan 31 16:46:24 koen, nice Jan 31 16:46:28 Or something, I have no idea how this audio stuff works. Jan 31 16:48:20 Kristina: if you dont want to deal with alsa complexity, there is also an OSS driver for the TWL4030 Jan 31 16:48:23 http://gitorious.org/archos/archos-gpl-gen6-kernel/blobs/master/sound/oss/omap2-audio-twl4030.c Jan 31 16:48:46 that and of course you need the McBSP stuff Jan 31 16:48:59 Hm, is there a datasheet that explains it from a programmer's perspective? (like, base addresses of peripherals, memory layout, register addresses, what registers do, etc) Jan 31 16:49:17 the omap3 TRM has all that Jan 31 16:49:52 It's not NDA'd, is it? Jan 31 16:50:02 no Jan 31 16:50:14 I dont know if the TWL stuff is public though Jan 31 16:50:17 av500, McBSP with fries? Jan 31 16:50:46 Kristina: MsBSP is kinda like a standard TI "part" you will even find no-os drivers for it Jan 31 16:50:51 Mc* Jan 31 16:51:22 mcdriver Jan 31 16:51:57 StarterWare is the answer Jan 31 16:52:03 Hmmm, that OSS driver looks pretty straightforward. Jan 31 16:53:14 mdp: is the gpio number fortan too, i.e. GPIO0_5 would be &gpio1 6 Jan 31 16:54:02 jackmitchell: no, only partial fortan numbering is implement in the DT Jan 31 16:54:18 ok, so it would be &gpio1 5? Jan 31 16:54:50 correct Jan 31 16:56:18 Hehe, people seem to be struggling with device trees. Jan 31 16:56:31 I prefer my shitty XML based thing. Jan 31 16:56:37 there's no struggle Jan 31 16:56:41 DT won Jan 31 16:57:47 mdp: I know people that are refusing to register their board files, they are hoarding them for the day that evil DT comes after them Jan 31 16:58:03 LOL Jan 31 16:58:08 ok, there are warts with DT, but XML? :) Jan 31 16:58:19 XML makes my eyes bleed Jan 31 16:58:31 XML and DT is orthogonal Jan 31 16:58:38 one could write DT files in XML Jan 31 16:58:45 to have the best of both worlds Jan 31 16:59:03 hi koen! is there some overview about status of the branches in your repo https://github.com/koenkooi/linux.git ? Jan 31 16:59:05 XML is like the perl of data layouts...write only Jan 31 16:59:22 mdp: all the XML parser authors disagree Jan 31 16:59:47 not human readable, thanks for confirming ;) Jan 31 16:59:48 * av500 sees your XML and raises you a strstr() Jan 31 17:00:33 panto's extensions for dt overlays start making DT unreadable too Jan 31 17:00:39 it was only readable in simple cases Jan 31 17:00:46 I resent that Jan 31 17:01:00 I clarified after I used your name Jan 31 17:01:02 no blame there Jan 31 17:01:13 mdp, the dma of a sglist that loops around the end of a kfifo never triggers the dma_callback, have you noticed? Jan 31 17:01:24 there's absolutely no syntax changes besides a simple /plugin/ declaration at the start of a cape fragment Jan 31 17:02:01 and __overlay__ Jan 31 17:02:03 blame DT maintainers for the archaic syntax Jan 31 17:02:08 yes, I am Jan 31 17:02:22 mdp, it's not a syntax change Jan 31 17:02:24 do I still have to hardcode hex values? Jan 31 17:02:28 it's a just a name Jan 31 17:02:28 you work to make things useful, exposes how confusing it will become Jan 31 17:02:45 panto, people look at the minimal boot dtses and think it's trivial Jan 31 17:02:57 I agree there Jan 31 17:02:59 wait until they see a fully loaded pinmux node with alternates etc. Jan 31 17:03:03 * mdp boggles Jan 31 17:03:25 but as long as there's no-one to push for a saner syntax (or a friggin macro facility) there will be no change Jan 31 17:03:45 * av500 heard macro Jan 31 17:03:49 * av500 proposes m4 Jan 31 17:04:07 panto, I don't know if you recall this, but Mentor was chatting with FSL back when we were all friends...and mentioning how their multi-core DT management was an absolute nightmare Jan 31 17:04:33 of course it was Jan 31 17:04:37 panto, they had already found that their customers needed a high-level tool to create DT data Jan 31 17:04:51 that's the wrong way to go Jan 31 17:04:53 which got Mentor mgmt horny cause they like tools..they are tools Jan 31 17:05:06 if you need a tool to spit out a config file you got problems... Jan 31 17:05:40 mhaberler: yes, all are: don't use Jan 31 17:05:48 I'm sure if they did spit out a tool it's going to be obsolete already Jan 31 17:06:08 mhaberler: pick a branch from https://github.com/beagleboard/kernel/branches, that is the latest and greatest Jan 31 17:06:24 it's actually a similar problem to what the OMAP maintainers cited when they first started learning about DT..."um, it requires a separate DT for each core in the system, and they all need to cooperate!" Jan 31 17:06:32 I need a 3.2.21 or 3.5.7 base for a xenomai patch Jan 31 17:07:00 panto, right, I told them any tool would have to be a FOSS effort...they will fail at a commercial tool Jan 31 17:07:26 ExpressDTmaker Jan 31 17:07:43 DTcomposerStudio Jan 31 17:07:55 ExpressDTMaker Pro Extreme - iPhone Edition Jan 31 17:07:58 www.dtovernightdelivered Jan 31 17:08:03 www.dtovernightdelivered.com Jan 31 17:08:13 DTCloudBuilderEXTREEEEMEEEEE Jan 31 17:08:15 VisualDTStudio 2019 Jan 31 17:08:15 "we tree in your sleep" Jan 31 17:08:23 that's the release year if they start coding *now* Jan 31 17:08:29 VisualStuDTio" Jan 31 17:08:42 mmm...studlyCaps Jan 31 17:08:46 DTforDummiesTool Jan 31 17:08:53 openDT Jan 31 17:08:56 freeDT Jan 31 17:08:57 netDT Jan 31 17:09:03 gnuDT Jan 31 17:09:03 OpenFirmware Jan 31 17:09:04 DTToolForTools Jan 31 17:09:25 of course, the FOSS version would be "yadtt" Jan 31 17:09:37 eDT Jan 31 17:09:50 50% better than DT! Jan 31 17:09:52 iDioT Jan 31 17:10:08 mdp: eDT is the "embedded" DT fork Jan 31 17:10:17 and of course "systemdT" Jan 31 17:10:17 how can we work POS into the name? ;) Jan 31 17:10:47 I'm confident that systemd will be loading dt overlays for you very soon Jan 31 17:11:52 mdp: please add that on the deeteepedia wiki Jan 31 17:12:26 no, deeteepedia must never be updated with current info Jan 31 17:12:42 arf, our tree got put on deeteeleaks Jan 31 17:13:04 "deetee wants to be free!" Jan 31 17:13:14 deetee phone home Jan 31 17:13:35 not even Friday...we're doing great, folks Jan 31 17:13:51 friday is a state of mind Jan 31 17:14:50 "Get me deetee for my boothole!” Jan 31 17:14:52 amen brutha, pass the flask Jan 31 17:15:03 koen: is this your private experimental space ? I'd rather avoid wasting time, as some others already did Jan 31 17:15:24 mhaberler, wasting time is what we do best here Jan 31 17:15:43 mhaberler: you are new to IRC? :) Jan 31 17:15:49 yes Jan 31 17:16:07 welcome, mhaberler Jan 31 17:16:14 hi! Jan 31 17:16:16 ah give it a month you'll be like us Jan 31 17:16:31 Hi all! Know anybody how to enable the BeagleBoard to wake up by the GPIO line? Jan 31 17:16:52 sounds like a school project Jan 31 17:16:56 I'm not keen on the kernel branch bingo - so I'd rather ask Jan 31 17:17:12 right, interrupts can suck it till tomorrow. Have a good day everyone Jan 31 17:17:22 +1 Jan 31 17:17:35 mhaberler: are you wanting to use xenomai on beaglebone? wasn't clear on your platform Jan 31 17:17:59 yes, actually I have one, although on an arago-3.2 base Jan 31 17:18:44 djlewis: my daughter does beagleboard gpio wakeup in her 3rd grade class..definitely a school project. Jan 31 17:18:58 :) Jan 31 17:19:15 Jan 31 17:19:47 mhaberler: I understand how xenomai's hugely invasive patches land on some magic kernel versions..soo... Jan 31 17:20:09 Bravo for your daughter :O Jan 31 17:20:29 mdp: meaning what? Jan 31 17:20:44 you have one good choice. you choose the 3.2-ish arago vendor kernel to go with their 3.2 patch Jan 31 17:20:59 mdp: similar wakeup request recently Jan 31 17:21:36 dapm: class project is a personal fusion reactor, but that requires the full team. Jan 31 17:21:58 well this is in fact what I did - it lacks on some cape suppport though: http://git.mah.priv.at/gitweb/linuxcnc-kernel.git/shortlog/refs/heads/beaglebone-3.2.21-xenomai-2.6.2 Jan 31 17:22:55 mhaberler: your option for proper cape support is to port xenomai 3.5.7 to the 3.8 branch. I can assure you that will suck Jan 31 17:23:01 but if you need it, you need it Jan 31 17:23:22 that's the price of out-of-tree things like xenomai though Jan 31 17:23:48 mdp: what about using the https://github.com/beagleboard/kernel/tree/beagleboard-3.5 branch? as a base? makes sense? Jan 31 17:24:33 that's beagleboard only Jan 31 17:24:55 until you get to the 3.7 branch...all previous versions only work on the named board Jan 31 17:25:13 beaglebone-3.6 is the first one for bone Jan 31 17:25:20 ah Jan 31 17:25:23 since 3.2, that is Jan 31 17:27:54 so it's either beagleboard/beaglebone-3.2 + 3.2.21, or massage xenomai into beagleboard/3.8 ? Jan 31 17:28:12 yes Jan 31 17:28:34 ok, makes sense - thanks a lot Jan 31 17:29:03 do the koen branches have any bearing on this? Jan 31 17:29:24 make sure you check closely in the xenomai community, you may find somebody porting it to the current kernel..I dunno how they pick versions..haven't followed that in some time Jan 31 17:30:49 I do - today somebody came up with a 3.2.21 Xenomai patched kernel based on some koen linux-ti33x-psp-3.2.21-r13g+gitr720e07b4c1f687b61b147b31c698cb6816d72f01 branch, and I wasnt sure this is a good idea Jan 31 17:32:32 given that some of that stuff in the base branch doesnt even compile - copy & paste errors - I have some doubts Jan 31 17:32:46 it's suboptimal if you need features from 3.8, definitely ;) Jan 31 17:33:30 somebody else does, I wont do the grunt work; however, bets are rather off by starting on the wrong leg Jan 31 17:34:52 mru: your solution to the font quiz? :) Jan 31 17:40:04 off to the salt mines Jan 31 17:44:08 mranostay: you're not going to guess? Jan 31 17:49:03 mdp: are there any code base differences between the beagleboard/* branches and the Angström distro? sorry if this sounds confused - I am Jan 31 17:49:34 mhaberler: nfc how those map Jan 31 17:49:47 ah. Jan 31 17:49:54 I feel less alone ;) Jan 31 17:49:58 I use an angstrom rootfs and nothing else Jan 31 17:50:46 most of my work involves u-boot/kernel devel so I'm using my own trees and just need a sane userspace Jan 31 17:51:20 I had tried the angström distro, and port a package relying on lots of debian/ubuntu packages, and the Angström package stream was fairly useless for the case Jan 31 17:57:35 koen: ping Jan 31 17:57:45 koen: i need a beagleboardtoys.com logo asap Jan 31 17:58:06 prpplague: ask hieu@circuitco.com, that's the fastest way Jan 31 17:58:21 prpplague: I have received no news on wifi boards either Jan 31 17:58:23 koen: thanks i'll email now Jan 31 17:59:47 mdp: any idea how forward porting Xenomai 3.5.7 to https://github.com/beagleboard/kernel/tree/beaglebone-3.6 might pan out? Jan 31 18:00:12 koen: cc'd you Jan 31 18:00:31 mhaberler: http://26-26-54.hardwarebug.org/30 :) Jan 31 18:00:49 well maybe somebody tried Jan 31 18:01:10 or has experience with the differences between 3.5.7 and 3.6 Jan 31 18:04:45 prpplague: got it, thanks Jan 31 18:07:42 koen: see /msg Jan 31 18:09:21 mhaberler git diff helps Jan 31 18:09:38 right Jan 31 18:10:12 well I might just apply the 3.5.7 ipipe patch to v3.6 and see what the fallout is Jan 31 18:10:57 what is a ipipe? Jan 31 18:11:13 kind of a virtualsation patch Jan 31 18:11:32 runs under linux - linux becomes a glorified IRQ handler Jan 31 18:11:53 Xenomai is based on it; RTAI too Jan 31 18:24:23 mdp, have you tried gpevt with a sglist of more than one element? Jan 31 18:24:57 jsabeaudry: no, it's not supposed to be a robust case, could be problems with just increasing it Jan 31 18:25:05 it's only to verify the dma event fires Jan 31 18:25:22 but SG lists of greater size work in the mmc drivers Jan 31 18:28:07 mdp, ok I'll take a look again at the mmc drivers cause here, all my dma with 1 sg all work fine but no callback with dma that has 2 sg entries Jan 31 18:29:19 I once made the mistake of trying to explain ipipe to karim Jan 31 18:29:33 thankfully he said "I designed it" early on Jan 31 18:33:17 beaglebone-black Jan 31 18:33:20 ? Jan 31 18:33:59 somebody dropped one in a bucket of paint ...then the rumor started that there's a different beaglebone Jan 31 18:34:01 all lies Jan 31 18:34:02 lies Jan 31 18:37:19 is there some kind of HR violation going on here? Jan 31 18:38:59 Russ: eh? :) Jan 31 18:39:09 heh Jan 31 18:39:15 It's too weak of a joke to even explain Jan 31 18:39:15 jeffro makes a good point Jan 31 18:39:27 "exact steps":i t's the technical version of being Rickrolled Jan 31 18:40:19 Russ: kinda like 'that is what she said' :) Jan 31 18:40:40 ever try explaining that one to a foreigner :) Jan 31 18:41:21 stuff like that rarely crosses borders Jan 31 18:41:40 I was in leeds once and the busses all said "white power" in big block letters Jan 31 18:42:22 didn't catch at first that they were ads for leeds united Jan 31 18:45:16 russ *g* Jan 31 18:46:38 hehe Jan 31 18:51:01 Russ: even there it would seem not to be PC :) Jan 31 18:55:12 hello. I need to contact with some software technical supporter for the beagle bone. Jan 31 18:55:31 Nobody has answer me to my question in the discuss forum and I need the answer asap Jan 31 18:56:17 https://github.com/CircuitCo/BeagleBone-Black-RevA4/blob/master/BeagleBone_Black_SRM_A4.pdf Jan 31 18:56:24 javier we all ned answers asap Jan 31 18:56:35 but we do not get them all asap Jan 31 18:56:41 Yes, but how many of you will buy 10.000 units per year? Jan 31 18:57:03 I guess nobody here is your ti sales person Jan 31 18:57:03 so Jan 31 18:57:04 You will just have to wait until it is ready! Announce and release date istargeted for end of April at Design West.Stay tuned! Jan 31 18:57:04 Gerald Jan 31 18:57:07 heh Jan 31 18:57:24 I do not need any sales person. We already have the offer Jan 31 18:57:25 * mranostay releases the trolls Jan 31 18:57:41 release the kraken! Jan 31 18:57:59 TI is not answer me to my questions by the special channel for Design Partners Jan 31 18:58:10 hobbit puts on the ring, goes invisible, so sales seeking goblin can't see me any more Jan 31 18:58:35 well this isn't a TI channel, technically :) Jan 31 18:58:59 You offered the boards to my customer! Jan 31 18:59:04 #beagle is #middlearth Jan 31 18:59:18 mranostay: you sellign boards, you bad troll Jan 31 18:59:30 <-- not-TI :) Jan 31 18:59:47 boards! boards! buy your boards! Jan 31 18:59:58 mranostay the beagle troll has boards.. Jan 31 19:00:04 I do not need help with the AM335x it is so clear. I need help with the beagle bone. My customer will not use it as a testing board but as the board for contrilling his product also for production Jan 31 19:00:23 christ already what is your question? Jan 31 19:00:56 mranostay: do you get paid to do this IRC thing? Jan 31 19:01:17 indirectly i suppose :) Jan 31 19:01:31 You get sexual favors? Jan 31 19:01:31 gerald was commenting on the beaglebone-black, it seems it might be true Jan 31 19:02:02 javier_: ask what you probme is Jan 31 19:02:17 topic topic! Jan 31 19:02:18 if someone knows they mnight tell you what th answer is Jan 31 19:03:00 my problem is with the wake up Jan 31 19:03:42 and? Jan 31 19:04:29 there's 184 people waiting for a question to be asked. Jan 31 19:04:38 We need to wake the board up from suspend with an interrupt in the GPIO0 port, but using Angstrom it is not working. And we must not modify the kernel. Is it possible? Jan 31 19:04:57 no Jan 31 19:05:15 no, why? Jan 31 19:05:29 because you have a constraint of not modifying the kernel Jan 31 19:05:32 so, no Jan 31 19:06:26 why if I define an input in GPIO0 and I configure an interrupt the micro does not wake up? I have never seen any processor which do not respond to an external interrupt in a GPIO port available for that Jan 31 19:06:46 thankyou for playing javier_... our next contestent comes from Sycamore, IL... and is wanting to know if he can use a beagle in his tractor Jan 31 19:07:26 javier_: I'm sorry, you didn't listen. I said it's not possible with your constraint Jan 31 19:07:32 you have to be willing to change the kernel Jan 31 19:08:10 your binary kernel software does not support that...end of story. next! Jan 31 19:08:25 mrpackethead_: hehe Jan 31 19:09:06 javier_: maybe go and visit the good folks at #$35_buys_you_pi_which_can_do_anything_you_want Jan 31 19:09:12 double points on the big board today Jan 31 19:09:20 hazard pay Jan 31 19:09:42 javier_ you need to get familiar with angstroem build process and where to get the kernel sources Jan 31 19:09:52 okay mdp makes not anybody happy Jan 31 19:10:31 woglinde: can't do that, "we must not modify the kernel" Jan 31 19:10:39 the trolls have driven him away Jan 31 19:11:13 ah maybee thats why the payed support did not answer him Jan 31 19:11:17 question 2 Jan 31 19:12:21 ok, so there is not any possibility to wake a beagle bone up without changing the kernel... not by touchscreen, not by GPIO0, not by UART? Jan 31 19:12:48 wakeup from what? Jan 31 19:13:17 sleep of death Jan 31 19:13:37 wake up from any event Jan 31 19:13:52 an alarm input, an button pushed, etc. Jan 31 19:14:00 why do you need to have it alseep Jan 31 19:14:33 javier2: when do you have to have your homework handed in by? Jan 31 19:14:47 I supose that you have a celular. And int it, there is a button to sleep your phone and when you press it again, it wakes up, right? For that Jan 31 19:15:14 For the middle of next week Jan 31 19:15:34 oh, so this is a homework question.. Jan 31 19:15:45 and your "customer" is who exactly Jan 31 19:15:48 * panto pushes the big red button Jan 31 19:16:02 not, really. I am the CTO of a company and I am helping my engineers to solve the problem Jan 31 19:16:10 lol Jan 31 19:16:33 mrpackethead_: do you work for beagleboard.org? Jan 31 19:17:18 wheeeeeeeeeeeeeeeeeeee Jan 31 19:17:39 i'm just working on an answer for that question Jan 31 19:17:57 my customer is an electronics manufacturer Jan 31 19:18:10 . . . . . . . . . . . ...... no ... Jan 31 19:18:20 javier2: hire some smarter engineers Jan 31 19:18:37 I sense a repeat of the 54 scenario Jan 31 19:18:46 i work for "troll recruitment services" Jan 31 19:18:49 that'd be the 53! ;) Jan 31 19:18:50 maybe in Budapest the engineers are smarter than in Germany right? Jan 31 19:19:12 would you like to hire a troll Jan 31 19:19:22 i have several fine trolls i can sell you Jan 31 19:19:37 over in the corner is mranostay.. hes one of the best trolls out Jan 31 19:19:52 I suppose you have time to lose... me not. bye Jan 31 19:20:02 i personally perfer hobbits Jan 31 19:20:36 *blushs* Jan 31 19:22:21 the moral of the story: don't ask TI for help Jan 31 19:22:39 thats a bit harsh.. i've found them to be very helpful sometimes. Jan 31 19:22:54 koen, agreed Jan 31 19:24:26 mrpackethead_: TI doesn't manufacture or distribute the beagle* boards Jan 31 19:24:27 there is of coruse Russ's patch.. https://groups.google.com/forum/?fromgroups=#!topic/beagleboard/GXqO9rRsyCg Jan 31 19:24:46 koen: they do sell it through the Ti.store.. Jan 31 19:24:54 is that enough guilt Jan 31 19:25:27 TI is just like amazon...they'll sell anything Jan 31 19:25:36 below cost Jan 31 19:25:40 SCNR Jan 31 19:25:48 make it up in volume Jan 31 19:25:49 just make it up in volume Jan 31 19:25:52 jinx! Jan 31 19:28:48 don't see what the big deal is here - the beagle family are just usages of the TI ARM chips Jan 31 19:34:51 dinner time Jan 31 19:40:05 ds2, all I can say is that #beagle has a disappointed customer so the big board will reflect everybody's poor support this month Jan 31 19:40:15 * mdp is deeply disappointed in the team today. Jan 31 19:41:39 * djlewis sniffles and walks away... Jan 31 19:41:54 we can't run from this problem, folks Jan 31 19:42:02 but you can walk away from it, djlewis ;) Jan 31 19:42:14 ah I know there the problem is Jan 31 19:42:28 his TI sales person did not till that beagles are toys Jan 31 19:42:34 no support for toys Jan 31 19:42:42 just tying to hide my dissapointment :( Jan 31 19:43:32 sadly, I couldn't mention that starterware has examples for this stuff, but since his constraint was to use Angstrom's unmodified binary kernel.... Jan 31 19:43:48 what a strange constraint Jan 31 19:44:13 Russ, we're just the engineers. give us problem constraints and we tell you what can be done. ;) Jan 31 19:44:15 shouldn't he be using the AM33x EVM? Jan 31 19:44:33 ds2, no, I think somebody told him what he should be using...RPi Jan 31 19:44:45 nobody is going to make money of a sale to that guy Jan 31 19:44:46 ds2 no he said the customer said he wants the beagletoys Jan 31 19:44:57 for production Jan 31 19:45:12 how cute...little toy beagles Jan 31 19:45:16 ahhhh Jan 31 19:45:43 we need some kraken toys Jan 31 19:45:57 mdp koen has tentacles Jan 31 19:46:04 uhhhhhhh Jan 31 19:46:18 ahh, tentacles! Jan 31 19:46:18 i read that wrong first... Jan 31 19:46:48 yeah I am poor so my english is even poorer Jan 31 19:58:04 mdp, what's your board you want? Jan 31 19:58:13 not here Jan 31 19:58:34 I accidentally erase a reply, need to get the energy to retype ;) Jan 31 19:58:47 * bradfa accidentally erases hard disk Jan 31 19:59:03 bradfa: that is your story to the FBI? Jan 31 19:59:24 * bradfa has bright light shining on him while handcuffed Jan 31 19:59:30 you want my exact steps to board, I PM them Jan 31 19:59:37 mdp, I do Jan 31 19:59:49 did I not PM you? Jan 31 19:59:56 * bradfa fails at irc all the times Jan 31 20:00:01 If I see nothing on the gpmc bus during my gpmc dma, could it explain why it does not work with sg_len > 1 (dma transfer fails for the first param set and thus the transfert does not link to the next) ? Jan 31 20:00:13 gezzz who let the hardware dude in here? Jan 31 20:00:15 to javier's qeustion, why can't the tps wake the system up? Jan 31 20:00:23 push big red PWR_BUT Jan 31 20:00:44 voila, wake up from SLEEP or OFF (granted, not from RAM staying alive without some help) Jan 31 20:01:56 Do I need more than request_mem_region and ioremap_nocache in a DT scenario? Jan 31 20:05:43 jsabeaudry: yep, the param set link won't be followed if that transfer doesn't complete Jan 31 20:06:15 and only the last one in the set of links has the interrupt flag set such that you get a callback Jan 31 20:06:24 mdp, ok good, that explains why i dont get the interrupt Jan 31 20:06:52 now, how can I troubleshoot why the transfer fails, all the dmaengine stuff succeeds Jan 31 20:07:56 I noticed that there is no call to set the mask for the dma with request_slave_dma_channel, is the mask supposed to be in the DT? Jan 31 20:09:01 jsabeaudry: the OF DMA helpers handle all that behind the curtains Jan 31 20:09:42 jsabeaudry: there's a woefully complex translation of DT properties into a controller channel etc. to get the proper channel Jan 31 20:15:04 mdp, so if my scope shows nothing on the gpmc during dma, but every other gpmc acces in my driver works, any hunch what could be the problem? Jan 31 20:16:05 jsabeaudry, what board are you using? Jan 31 20:16:18 ka6sox-away, bone A3 Jan 31 20:16:37 what kernel? Jan 31 20:16:45 3.8rc5+ Jan 31 20:17:13 then we are working on the same thing. Jan 31 20:18:07 ka6sox-away, you do gpmc dma ? Jan 31 20:18:24 not yet! Jan 31 20:19:56 used to work fine on 3.2, I guess I missed something when switching to DT/dmaengine Jan 31 20:22:52 av500, http://uk.reuters.com/article/2013/01/31/uk-serbia-pm-prank-idUKBRE90U14V20130131?feedType=RSS&feedName=worldNews Jan 31 20:26:01 <_av500_> yeah Jan 31 20:27:45 mdp, Can you see a problem with me using <&edma 40> ? Jan 31 20:31:55 I tried <&edma 30> also with the same results Jan 31 20:32:53 * koen tries to remember what panto set about net-dma claiming all the channels Jan 31 20:33:17 oh yeah Jan 31 20:33:48 but I assume his request_channel is succeeding if he gets to the transfer part Jan 31 20:34:04 <_av500_> you can pry GPT12 out of my cold, dead hands Jan 31 20:34:36 jsabeaudry: exactly what event are you trying to use? Jan 31 20:35:24 30 is UART2 and 40 is DCAN0 Jan 31 20:35:55 mdp, hmm, I just want to do dma and have a callback Jan 31 20:36:10 aha Jan 31 20:36:31 well, let me apologize first..I was under the assumption that you were doing event driven dma Jan 31 20:37:10 the slave_sg transfer type is only designed to work with that Jan 31 20:38:02 all the supported drivers for peripherals right now generated dma events that drive the actual transfer itself Jan 31 20:38:11 So I have to go back to the private edma api ? Jan 31 20:38:54 your fpga doesn't provide a dma request when the fifo is ready to drain? Jan 31 20:39:23 everything I've ever seen externally had an interface like that to plug into the soc external bus connection Jan 31 20:39:52 mdp, yes but some of the fifos I want to drain every 200ms even when they are not half-full Jan 31 20:40:43 Back to private edma api, right? Jan 31 20:40:47 I see, I recall this one stereovision processing asic that had a timer for that case Jan 31 20:41:02 but yeah, you'll have to use the private edma api now Jan 31 20:41:27 I'll think about that one a bit Jan 31 20:42:18 Alright I'll leave that work in a branch and move back to private edma Jan 31 20:42:24 thanks for all the help Jan 31 21:16:19 gn Jan 31 21:56:44 * bradfa needs a new compiler :( -> /usr/lib/gcc/arm-linux-gnueabi/4.4.5/include/stdarg.h:40: internal compiler error: Segmentation fault Jan 31 21:59:10 mdp: you're on lwn Jan 31 22:07:22 https://lwn.net/Articles/535153/ Jan 31 22:07:35 He's (got code that's) been in space AND lwn Jan 31 22:07:37 He's a superstar! Jan 31 22:12:00 Hmmm Jan 31 22:14:23 theres bit banging, but is there bit-listening? Jan 31 22:15:43 any idea why this won't work? http://pastebin.ca/2309436 Jan 31 22:15:56 simple edma + sg stuff Jan 31 22:25:57 LOL Jan 31 22:26:10 bbl Jan 31 22:26:29 jsabeaudry: laughing at lwn comment not you Jan 31 22:29:54 If I recall correctly, there was some other proprietary library used by TI for OMAP FB stuff, wasn't there? Jan 31 22:37:13 Hmmmm 21AH batteries Jan 31 22:41:36 Hm, would an ICE debugger let me control execution before bootrom switches out of monitor mode? **** ENDING LOGGING AT Thu Jan 31 22:46:41 2013 **** BEGIN LOGGING AT Thu Jan 31 22:48:23 2013 Jan 31 23:15:15 @%#%@#$@#$%#@$@#$$@##!@#!@;2Q$!#@# ELC Jan 31 23:15:43 ah yes ds2 ? Jan 31 23:15:44 they are in a collusion to kill people... ELC - SCALE - MWC all within 2 weeks Jan 31 23:16:18 and that is ignoring ABS before ELC Jan 31 23:16:54 Why doesn't that ^%$# link? http://pastebin.ca/2309468 Jan 31 23:17:07 some people are going to the linaro conf in hong kong 1st week of march... Jan 31 23:17:19 mru: so it is 3 weeks straigh Jan 31 23:17:33 mru: do you have the pleasure of a barcelona vacation after ELC? Jan 31 23:17:41 I'm going home for a few days between Jan 31 23:18:16 mru: are they all visa waivers for you? Jan 31 23:18:50 yes Jan 31 23:29:01 later . . . Jan 31 23:33:22 u could be conferenced out and do no work Jan 31 23:38:36 that isn't an option when one may be working it Jan 31 23:38:50 ds2, see you at ELC and SCALE :) Jan 31 23:39:09 Russ: I am not going to SCALE Jan 31 23:39:09 stalker Jan 31 23:39:17 I got to do stuff for MWC Jan 31 23:39:52 bummer, I think I'll be the only one at scale besides ka6sox Jan 31 23:40:27 I'd goto scale if they could move it Jan 31 23:47:09 Is it me or do all the examples in the TRM have TCINTEN even on params that link to other params Jan 31 23:47:36 Same goes for the edma example code Jan 31 23:50:47 well I have to fly back to LA after ELC anyway Jan 31 23:51:48 Russ: when are you leaving ELC? Jan 31 23:52:25 my flight isn't till 9:15p on friday Jan 31 23:54:40 Russ: doing bart to the airport? Jan 31 23:55:33 yup Jan 31 23:57:07 Russ: did you look at the $1 mega bus option? Jan 31 23:57:27 I can't even parse that sentence Jan 31 23:57:40 Russ: did i read you had done soeme work on wake/sleep stuff on teh beagle? Jan 31 23:57:42 sounds like I'm in a game show Jan 31 23:57:49 mrpackethead_, yes Jan 31 23:57:59 Russ: there is a company, MegaBus, that offers SF - LA rides at $1 Jan 31 23:58:09 oh, hell no Jan 31 23:58:21 I'm not taking a 10 hour bus ride Jan 31 23:58:24 no like bus? :D Jan 31 23:58:33 a self proclaimed CTO came this morning askign about sleep/wake stuff. Jan 31 23:58:36 they probally drive fast enough that it'll be only 8hr Jan 31 23:58:41 the trolls ate him up Jan 31 23:58:58 Why is the interrupt enabled in the channel1 of the linked transfer in http://processors.wiki.ti.com/index.php/EDMA_sample_test_application ? Jan 31 23:58:59 * Russ checks the gg Jan 31 23:59:52 nothing jumps out Feb 01 00:00:39 besides you confirming my theory about OTG funny business? ;) Feb 01 00:00:52 Its the ID pin Feb 01 00:01:05 still funny business Feb 01 00:01:18 I measure resistance from ID to VUSB, I get 66k Feb 01 00:01:18 the real question is - did you find a way to stop it? Feb 01 00:01:47 no way to stop it, I could mask that interrupt source Feb 01 00:02:11 every time it happens, the interrupt handler has to do some i2c transfers to find out the source, so its a bit of an annoyance Feb 01 00:02:16 isn't that MUSB behind it? Feb 01 00:02:22 yes, musb Feb 01 00:02:51 I vaguely recall some setting to tweak that stuff Feb 01 00:03:07 I could change the timer, but for OTG to work, it has to probe ID Feb 01 00:03:28 but for your implementation, it doesn't require OTG, or am I wrong? Feb 01 00:03:39 It's generic code Feb 01 00:03:49 rewrite it Feb 01 00:03:54 I'm trying to catch wakeup sources for RTC-only sleep Feb 01 00:03:57 Russ: 1$ how is that possible? Feb 01 00:04:04 no one will notice... it is only MUSB... :D Feb 01 00:04:14 so that if during the suspend process, a wakeup even occurs, I don't sleep Feb 01 00:04:18 you can even have a chat with Felipe about it too ;) Feb 01 00:04:23 mranostay, you'd have to pay me to ride the bus Feb 01 00:04:30 ah that bad Feb 01 00:04:54 reminds me of bdale's rant on LWN wrt preinstalled windows Feb 01 00:05:07 it doesn't save a company money to leave out windows, it costs them money Feb 01 00:05:19 all the companies that pay them money to preload software Feb 01 00:06:37 ds2, lets see...5V, 66k, 90uA sink Feb 01 00:09:11 Russ: but it is the wake up Feb 01 00:09:22 Russ: btw, how low of a power consumption ahve you gotten it? Feb 01 00:09:26 plugging in a usb cable will wake from RTC only sleep Feb 01 00:09:42 heh, beaglebone can do rtc only sleep, but it isn't designed for it Feb 01 00:09:53 you need a separate power rail for ddr vdd and vvdq Feb 01 00:09:57 er, vddq Feb 01 00:10:19 so it only goes down to 100mW Feb 01 00:10:34 I can hibernate though Feb 01 00:10:37 doesn't the memory controller support self refresh? Feb 01 00:10:47 yes, I'm putting it in self refresh Feb 01 00:11:01 but the am335x ddr io rail is still powered, along with the ddr's io rail Feb 01 00:11:10 Oh Feb 01 00:11:20 that's still 30 or so mA measured Feb 01 00:11:33 100mW@5V Feb 01 00:11:45 ok... so Feb 01 00:11:58 I am looking at running the bone at ~3V Feb 01 00:12:03 seems to work ok Feb 01 00:12:52 so you are suggesting that 20mA or so is from the regulator idling? Feb 01 00:13:07 I haven't measured individual rails Feb 01 00:13:29 evmsk has better control over rails, but the pmic doesn't really support rtc only Feb 01 00:14:02 trying to understand if the burn is on the PMIC side or something else Feb 01 00:14:11 20mA seems to be quite a lot for it to be burning Feb 01 00:14:32 well I could configure it to turn off the ddr rail Feb 01 00:15:08 Can I power the BeagleBone with a 5VDC power source supplied to pins 3 or 6 on Expansion Header P9? Feb 01 00:15:13 the obvious stuff like LEDs are off, right? Feb 01 00:15:28 bpetersen: what does the SRM say about that? Feb 01 00:15:34 ds2, yes Feb 01 00:15:39 here, I'll test w/o ddr2 Feb 01 00:15:42 er ddr Feb 01 00:15:59 I think it says I can. I've never done it before and I don't want to blow it up Feb 01 00:16:22 I just wanted to double check Feb 01 00:16:40 do you have reasons to double the beagle gods? :D Feb 01 00:16:47 lol Feb 01 00:17:14 whenever I doubt the beagle gods, I reference the schematic, and I am comforted Feb 01 00:17:41 powering it by those pins are relatively uninteresting Feb 01 00:18:02 there are more interesting paths with power path management Feb 01 00:19:47 well that's not good, power consumption went up Feb 01 00:19:52 I need to do some probing... Feb 01 00:20:27 hahahahahahah Feb 01 00:20:44 maybe you need to run cpuburn to bring the power down to sub 1mW @ 5V ;) Feb 01 00:20:52 My friend wants to use a ribbon cable for the expansion header Feb 01 00:21:06 thats why I needed to know if the pins will work Feb 01 00:21:07 Can anyone confirm how many interrupts will be triggered by 2 linked slots that both have TCINTEN set? Feb 01 00:24:49 * Russ plays find that component Feb 01 00:28:10 that's a bad sign, 3V3A is at 1.35V Feb 01 00:30:12 whoa Feb 01 00:30:37 are you drawing several amps? Feb 01 00:30:50 no Feb 01 00:31:07 I think because of the weird configuration I have it in, some esd diodes are "leaking" Feb 01 00:31:45 or maybe there the parasitic diodes in the PFETs are doing wierd things? Feb 01 00:31:56 something like that Feb 01 00:32:19 I have another 3.3V ldo with 0.5V on it's output, it's enable pin is low though Feb 01 00:33:31 not all LDOs block Feb 01 00:34:08 I think I need to move a couple zero ohm jumper Feb 01 00:34:09 s Feb 01 00:36:33 uncape is in motion Feb 01 00:36:42 of course they are right between the ethernet jack and the dc jack Feb 01 00:38:25 like right where the battery expansion header is? ;) Feb 01 00:39:07 um... Feb 01 00:39:09 ya, sure Feb 01 01:47:50 more on the beaglebone black Feb 01 01:48:39 the whut ? Feb 01 01:49:10 mrpackethead: howsit , what a beaglebone black? Feb 01 01:49:59 https://github.com/CircuitCo/BeagleBone-Black-RevA4 Feb 01 01:50:34 Gerald is answering questons in teh mailing list. Feb 01 01:52:23 the document says "Not for Discolousre" Feb 01 01:52:24 opps. Feb 01 01:53:01 ohhh Feb 01 01:53:05 onboard HDMI Feb 01 01:53:31 onboard managed NAND Feb 01 01:53:37 beauty Feb 01 01:53:42 512MB RAM Feb 01 01:53:46 AM3358A Feb 01 01:54:01 yeah I see it is 4Gb Feb 01 01:54:06 and DDR3 Feb 01 01:54:34 and I see Mr HDMI! :D Feb 01 01:55:03 is it real hdmi or dvi in a smaller form factor again? Feb 01 01:55:05 "managed nand"? Feb 01 01:56:09 its way too much sensible information for it to be a hoax Feb 01 01:58:33 http://circuitco.com/support/index.php?title=BeagleBoneBlack Feb 01 02:01:09 and it will have SGX support in v5? Feb 01 02:03:27 it is all a figment of your imagination Feb 01 02:03:51 heh Feb 01 02:04:05 quite possibly Feb 01 02:07:51 and there is no way that I would have one Feb 01 02:08:31 :~ Feb 01 02:09:05 I suppose u-boot commits probably tipped people off Feb 01 02:09:41 some docs for the R5 are available already Feb 01 02:09:53 err v5 :P Feb 01 02:10:03 urrg A5 Feb 01 02:10:16 hehehe Feb 01 02:11:44 https://www.youtube.com/watch?v=tvaD-29pQBY&t=1m9s **** ENDING LOGGING AT Fri Feb 01 02:59:58 2013