/srv/irclogs.ubuntu.com/2012/02/03/#ubuntu-arm.txt

=== trelane is now known as Guest63137
krosswindzI am having issues when trying to compile Ubuntu kernel in ARM cross chroot04:58
krosswindzThe error I am getting is http://pastebin.pandaboard.org/index.php/view/4345800604:58
krosswindzI noticed a bug report for precise https://bugs.launchpad.net/ubuntu/+source/tree/+bug/90476304:59
ubot2`Launchpad bug 904763 in tree "Unable to build tree for armel" [Undecided,New]04:59
krosswindzI was wondering on how do I change the CFLAGS for the same05:00
person987having trouble with initial boot of ubuntu on a pandaboard...06:49
person987during the initial "resizing the filesystem" stage, it says "Errors were found while checking the disk drive for /."06:50
infinityperson987: If the resize is failing, the original write to the SD was bad.07:14
infinityperson987: Which could be either (A) because something went wrong, or (B) the card is dead/dying.07:15
infinityperson987: My bet's usually on B, but you can try rewriting and see if it likes you the second time.07:15
infinitykrosswindz: That bug looks bogus to me (or, rather, their "fix" for it).07:22
infinitykrosswindz: I suspect dropping optimisation would be enough to solve your problem.07:23
infinitykrosswindz: But, honestly, if you have to do such nasty things to cross-compile, you might want to rethink things and compile natively. :P07:24
person987I have tried writing the image to this SD card twice or 3 times now.  Maybe you're right and the card is bad, I'll get another one tomorrow.07:24
person987Has anyone had problems with hard-locks using ubuntu 11.10 on a pandaboard?07:25
infinityUnder seriously heavy load (ie: running as a buildd), we manage to lock them once in a while.07:26
infinityBut that's fairly rare, compared to the pain they're being put through.07:26
person987I have another SD card which works but I have been getting regular hard-locks.  Probably once every half hour.  I was trying to build a new image on the new SD card to see if maybe I've gotten a bad driver or something like that (pretty new to Ubuntu/Panda)07:28
person987I'm often running some OpenGL code which renders Kinect data when I'm using the pandaboard07:29
infinityWell, I'd suggest installing to a real hard drive at some point to rule out bad media.07:30
infinityBut if you're doing intense OpenGL stuff, it may well be that.07:30
person987My typical usage is to write some code, run for a bit, write some more code, maybe a couple web searches, etc.  It will lock up during any of the above, including just typing.07:30
infinityLocking up during typing, on the other hand, definitely shouldn't be happening.07:31
infinityWe have machines with uptimes in the weeks and months around here.07:31
person987I suspect the graphics driver/opengl stuff.  I tried to install the drivers to get opengl hardware acceleration and it didn't seem to fully work.  For example, I'd get one boot where the desktop was really high res, then the next would be back to 1024x768, the 3D performance is really bad too; only 20fps for a trivial rotating quad; so I don't think I've got the video stuff working correctly.07:33
person987Good to know that you have not had lockups, I have a friend who said his pandaboard has never locked.  I tried my SD card on it and it eventually locked on me :-)07:34
infinityYeah.  May well be your software/driver setup.07:34
infinityIf you're having troubles with the TI binary drivers, I'd recommend poking ndec.07:34
infinityI don't touch them.07:34
person987what is ndec?07:34
* infinity points at ndec.07:35
infinityHe's a who, not a what.07:35
person987hehe07:35
infinityTI developer, packages the binary stuff.07:35
XorAa who can become a what after application of numerous alcoholic drinks :-D07:37
person987oh great!  well you have been a huge help.  I will get another card and re-install everything.07:37
janimojcrigby, rsalveti the 3.1 kernel for mx5 has been running all night. I have not tested it heavily but I'd say it is good to  upload to precise08:10
jcrigbyjanimo, ok will do08:11
XavBperson987: we (TI) have made a major update few hours ago, let us know if you are still seeing bad perf. Note that you will boot with TI kernel with DVFS on-demand by default. Then we are having perf impact with metacity compositing on and with unity-launcher running. More improvements shall come.08:14
XavBNote: I am working with ndec... ;)08:15
=== chrisccoulson_ is now known as chrisccoulson
burliHi, is it possible to run Ubuntu on this tablet? http://www.zenithink.com/Eproducts_C71.php10:03
burliIt has an Amlogic 8726-M and 512MB RAM10:04
Ziomuschiohi there11:10
Ziomuschiomay I ask some info about a problem with echi-omap when installing ubuntu-omap-extras in pandaboard?11:10
=== Quintasan_ is now known as Quintasan
=== dduffey_afk is now known as dduffey
rsalvetijanimo: great, thanks14:03
krosswindzinfinity: It takes forever to compile on the board, the reason I setup an ARM cross chroot on my x86 laptop14:55
krosswindzinfinity: The fix might appear bogus, I will try to disable optimization and see if it builds14:55
krosswindzinfinity: there definitely is some issue as I am running into exactly the same problem14:55
krosswindzI am seeing oops when I try to either halt or reboot the board using this kernel.15:12
krosswindzThe trace is available here http://pastebin.pandaboard.org/index.php/view/9445053815:12
person987XavB: thanks, I'm sorry but I don't understand most of your comment (DVFS on-demand? metacity? unit-launcher?)  My background is graphics programming but I'm a complete noob to Ubuntu.  I'd love to take a look at your latest update.  It would be great to solve my "hard-lock" problems as well as get some improved 3d performance.  If I can help you in any way with logs or details on my system15:16
person987config I would love to.  (you might have to give me detailed instructions tho! :-)  How exactly do I get your latest update?15:16
XavBperson987: ok, no pb. So first step would be to upgrade to last sw and let us know if it improves the situation.15:18
XavBthen there are several settings you can do to improve graphics performance15:19
XavB1- modify DVFS that will decide about CPU ussage strategy, it is by default set to ondemand into file /etc/init.d/ondemand15:20
XavBIf you comment the setting or force performance value, you might gain few fps.15:20
XavBThen by default you will be using Unity2D so metacity. Metacity is using compositing so graphics performance are impacted, you can install gconf-editor, then go into apps->metacity and uncheck box compositing manager. You will gain few more fps.15:22
person987I can try this right now, first step was to "upgrade to last sw", how?  (sorry I'm new!)15:24
XavBLast point is about unity panels, they are having an impact on graphics, you can kill them to gain few more fps: "sudo killall unity-panel-service unity-2d-panel unity-2d-launcher" 2 times.15:24
XavBperson987: You have installed TI PPA already right?15:24
XavBor "TI OMAP Addons"15:25
person987In the "Ubuntu Software Center", I had a package named "ubuntu-omap4-extras".  About 1/2 hr ago I removed that and rebooted to see if the hard-locks would go away.  Is that what you're referring to?15:27
person987in software sources, I have ppa.launchpad.net/tiomap-dev...15:28
XavBperson987: yes15:29
OlivierN1person987: fine, then re-install this package, and run an upgrade15:29
XavBSo perform an "apt-get update" then install manaually "ubuntu-omap4-extras"; it will be fine.15:29
person987Ok, installing, it gave a warning about future updates not including some other linux kernel stuff.15:34
XavBperson987: normal15:41
XavBYou will have a TI kernel15:42
person987Ok great.  I'm not too concerned about perf, my requirements are not too high.  The only reason I mentioned it was that I wasn't sure if I had things installed correctly.  So this is helping greatly.  I will probably not kill the panels but I'll try out disabling compositing. Almost done installing.15:46
XavBperson987: of course you'll need to reboot at the end of install16:02
person987XavB: That brings up another thing, it hasn't been shutting down cleanly.  I always eventually have to pull power because it just sits on the ubuntu screen with the dots that cycle.  Is that normal?16:12
XavBperson987: I can observe wrong reboot with canonical kernel, new one shall "reboot fine"16:14
XavBafter reboot "uname -a" shall return: Linux ubuntu-desktop 3.1.0-1282-omap4 #10 SMP PREEMPT Tue Jan 24 15:52:14 CET 2012 armv7l armv7l armv7l GNU/Linux16:14
XavBperson987: is your reboot ok? UI up etc...16:15
person987XavB: actually the update was still going.  It failed: Not all updates can be installed - run a partial upgrade to install as many updates as possible.16:51
person987XavB: I'm going to have to get to work but I'll build a new image this weekend and get back to this point.16:52
XavBperson987: if you do it in 2 steps (2 dist-upgrades) it shall work, no need to reboot between both16:52
loolHey folks17:09
loolThe web indices for e.g. http://uec-images.ubuntu.com/precise/20120203/ which I think are generated from cdimage code say "For ARMv5t processors and above"; it's because for "armel" images we say "For ARMv5t processors and above." -- which was true in jaunty; since we don't really have any official ARM images for anything older than lucid which is ARMv7t2, I propose that we change it to ARMv7; is that ok?  would you rather have a different wording?  a17:11
loolinfinity, ogra_: ^17:11
* ogra_ has no clue about uec images, nor why there would even be arm ones17:12
ogra_better ask #ubuntu-server ?17:13
ogra_we didnt enable that17:13
loolok17:14
morphisheyho19:17
morphisI am wondering wether it's possible to build currently for armhf with uploading packages to my ppa19:18
morphisanyone knows if it's possible?19:18
infinitymorphis: We don't currently offer ARM PPAs to the public for security reasons (we don't have virtualised ARM builders)19:19
infinitymorphis: That's in the works, but not done yet.19:19
infinitymorphis: If you want armel and armhf builds, your only option right now is to build locally.19:19
morphisinfinity: ok, thats good to know as it's very hard to find details about building for armhf and ubuntu today19:20
morphisinfinity: is there a definitive way how to build packages and images?19:21
morphisI saw there is support for pbuilder to build packages with qemu19:21
infinitymorphis: If you don't have native hardware, qemu is certainly the easiest way to go.19:23
morphisinfinity: ok19:23
morphisinfinity: can you tell me also which hardware is recommended for building? I read about debian is using some i.MX53 boards but whats with the panda boards?19:24
morphiswhats cannoncial using for the armhf bootstrapping?19:24
infinitymorphis: We support both the Panda and mx53.  Ubuntu's buildds are Pandas, Debian's are mx53.19:25
morphisok19:25
morphisinfinity: you work for cannonical?19:25
infinitymorphis: If you can get your hands on a Pandaboard ES, they're well supported in Precise, and probably the fastest generally-available hardware right now.19:25
infinitymorphis: I do, yes.19:25
infinitymorphis: I was the person who did the armhf bootstrap.19:25
morphisinfinity: ah :D19:26
morphisinfinity: I already though about the Pandaboard ES19:26
morphisI have done a lot with OpenEmbedded in the past were everything is cross compiled19:27
infinityYeah, we don't treat any of our ports as "embedded" targets.  armel/armhf are still complete Ubuntu builds and self-hosting.19:27
infinitySo, it's no different than x86 in that regard.19:27
morphisthats somethine I really like19:27
morphiscross compilation is sometimes very hard19:27
infinityHonestly, most "embedded" work these days isn't. :P19:27
infinityIt's general-purpose OSes.19:27
morphisas you have to map all the different build systems against the cross compilation (automake/cmake/...)19:28
infinityYeah, I know.  I used to work in scratchbox a lot with Maemo.  It was vile.19:28
morphis:)19:28
infinityAnd there was no good reason for it, once maemo was targetting devices as fast as the N900.19:28
morphisbut OpenEmbedded adds a very easy to use abstraction layout on top of it19:28
* infinity rnus off to lunch.19:30
morphisinfinity: Bon app├ętit!19:30
GrueMasterogra_:  Can you think of any omap/omap4 specific boot parameters that should be kept during netboot install?  I have "console=", "mem=", and "fixrtc".19:40
infinityvram=32M mem=456M@0x80000000 mem=512M@0xA0000000 fixrtc19:51
infinity^-- Should do it.19:51
GrueMasterI'm trying to avoid hardcoding as much as possible.  If they weren't on the netboot boot.scr, they won't get passed through, although I would also like to get the preseed "d-i debian-installer/add-kernel-opts" section.19:53
GrueMasterNot sure how to get that info in f-k-i.19:54
infinityI suppose f-k-i could grow preseed support (at least, for that one option)19:55
GrueMasterYea, I'm looking into it now.19:56
GrueMasterI have other changes I will be pushing soon (expect a merge request before EOD).19:57
infinityGrueMaster: Looks like user-params from debian-installer-utils may be what's needed.20:04
GrueMastergrmbl.  Netboot is having some serious issues this morning.  It is failing to install packages and sometimes losing connection to my mirror.20:04
GrueMasterOk, I'll look at that.20:04
infinityOh, maybe not.20:05
infinityThat's the inverse.20:05
infinityOr, rather, what it does doesn't seem to match the README...20:07
infinityAhh, no.  I misread.20:07
infinityThe README says it "doesn't include preseeded values", what it really means is it literally doesn't include pressed/command/syntax=foo stuff on the commandline it spits out, which is what you want.20:08
infinityBut it does include stuff that's been preseeded *from* add-kernel-opts.20:09
infinitySo, yeah.20:09
infinityuser-params is probably what we want to run in f-k-i to transfer the installer commandline to the final system.20:09
infinityLooks like grub-installer uses it, so that's comforting.20:11
GrueMasterI just ran it and it is missing a few params.  It only spews fixtrc, console=, and mem= lines.  Missing vram= and (in one case) smsc95xx.macaddr=20:19
GrueMasterOh, nevermind.  It pulls that from the preseed (duh).20:23
GrueMaster(I didn't look at the preseed of my currently running system).20:23
=== panda is now known as Guest88129
=== dduffey is now known as dduffey_afk
=== shaola is now known as apk
=== apk is now known as shaola
niklasfihi, obviously my beagleboard is ooming with the syslog repeatedly statign "eth0 kevent 2 may have been dropped", when doing a lot of io. I have had this issue since i got my beagle board back in 2010 and noticed the same behaviour on debian. did someone else notice this?22:03
krosswindzI have noticed that on the pandaboard that I got 2 weeks back22:04
niklasfikrosswindz: ohh good. then you saved me from making a bad purchase22:04
niklasfii was hoping this would be resolved due to more throughput22:05
krosswindzniklasfi: not sure what the issue is google isnt helpful on it22:05
niklasfikrosswindz: well not really, but as you may have noticed there is a workaround22:05
krosswindzniklasfi: yes I noticed the work around but no real solution22:06
niklasfikrosswindz: what value did you have to pump min_free_kbytes up to?22:06
infinityWell, I see it very very infrequently on the Panda, and it has no adverse effects.22:06
niklasfiinfinity: it totally locks up my machine to the point that i have to press the reset button22:06
krosswindzniklasfi: I havent played with that yet22:06
krosswindzniklasfi: I am still trying to get my kernel compiled for it22:07
infinityBut, really, heavy I/O on a system where almost everything is on the same USB bus (such as the Beagle and Panda) will occasionally hitch up.22:07
infinityniklasfi: I never have to reset, though I suppose if I was impatient, I might.22:07
infinityniklasfi: (As in, the machine might seem unresponsive for a bit, but it always comes back)22:07
niklasfiinfinity: is waiting for half an hour impatient?22:07
krosswindzniklasfi: I havent had the panda lock up yet22:08
infinityniklasfi: No, probably not.  Unless you're doing half an hour of long I/O. :P22:08
niklasfiis there a way to disable syslog? i think writing thousands of syslog messages does not help the situation22:08
infinityniklasfi: But I was referring to the behaviour being mostly harmless on Pandas.  I don't have a Beagle.22:08
infinityWhat's writing thousands of syslog messages?22:08
infinitysyslog culls duplicates.22:09
niklasfiinfinity: when the system freezes, it does. and no it does not filter out the duplicates22:09
infinityWeird.22:10
infinityAnd if it's flushing buffers and writing logs, I'd hardly call that "frozen".22:10
infinityJust "really busy".22:10
niklasfiinfinity: if you connect via serial port you just see thousands of messages, and if you use ssh, it does not respond22:11
niklasfiok. between 11:57 and 12:00 the system froze and generated 10000 lines of syslog. then again at 12:04 the system froze for another five minutes leaving another 10000 lines of syslog22:14
krosswindzwow that is a lot22:14
krosswindzwhat i/o load were you generating around then?22:15
niklasfijust for today i have 69418 lines of syslog22:15
niklasfiumm. i don't really know actually most of today was downloading stuff from the internetz so i guess maybe 800kB/s22:15
niklasfiover a long period of time22:15
niklasfithe system is capable of doing up to 4MB/s for short intervals, of maybe 1 or two minutes before locking up22:16
mythosis canonical paid for to develop this SDK "http://software-dl.ti.com/dsps/dsps_public_sw/ezsdk/latest/index_FDS.html" for TI?22:16
infinitymythos: If they are, I don't know about it (and I'm "them").22:17
mythoshmm... thx infinity. i only ask, because my boss said, that citrix said, that ti is paying canonical for this SDK...22:18
niklasfiis there a board available that is more stable maybe even with (e-)sata support?22:19
infinitymythos: That's a rather roundabout list of saids. ;)22:19
mythosinfinity, yeah, that's why i asked =)22:19
infinityniklasfi: The i.MX53 QuickStart has SATA.22:19
GrueMasterniklasfi: The Freescale Quickstart has SATA and we support it.22:20
infinityFor some value of "support".22:20
infinityBut we have images, and they work.22:20
niklasfiinfinity: can i boot from sata?22:21
mythosso you are canonical's "arm-team", infinity? ^^"22:22
niklasfisadly it does not have gigabit ethernet :(22:22
niklasfiis that possible though with current arm hardware?22:25
infinitymythos: I'm one of them. :P22:25
infinityniklasfi: It's not that it's not "possible", there are two primary reasons why it's not often done.22:25
niklasfiinfinity: what are they?22:26
infinityniklasfi: 1) Most of these cheap dev boards like to hang all their devices off USB.  USB can't do GigE (well, it could, but not at full-speed, so why bother?)22:26
infinityniklasfi: 2) GigE is wildly more expensive than 100baseT, and they're trying to keep the dev boards affordable.22:26
niklasfiam I the only one who thinks these boards make perfect home servers?22:26
infinityniklasfi: If you want something designed to be a server instead of a cell phone dev platform, you might want, say, a Trimslice.22:27
infinityniklasfi: They're a tiny bit (not much) more expensive, but come with GigE and SATA options.  And actual cases.22:28
mythosso, if i have a dev board with 2 gbit ports, i have a very expensive one? o.o22:28
infinitymythos: From the POV of cost-cutting, yes. :P22:28
infinitymythos: (Which board is that?)22:28
mythosinfinity, http://software-dl.ti.com/dsps/dsps_public_sw/ezsdk/latest/exports/DM814x_AM387x_EVM_Quick_start_guide.pdf22:29
mythosthere are some images in the pdf too22:29
mythosrather nice. i boot it via sdcard and mount the rootfs via nfs22:30
infinitymythos: Shiny.22:32
* infinity is trying to resist buying a Trimslice.22:34
mythosre22:34
mythoswhat was the last thing, i wrote here?22:34
infinity15:30 < mythos> rather nice. i boot it via sdcard and mount the rootfs via nfs22:35
mythosoh, ok... thx22:35
niklasfiinfinity: are they that tempting?22:36
infinityniklasfi: I like toys.22:36
niklasfiinfinity: but i don't know. i really can't trust these guys if they make their webpage that crappy22:37
infinityAnd it would replace my mx53 as my archive mirror.22:37
infinityYour trust is based on the quality of people's web design? :)22:37
infinityIt's actually not that bad for a small manufacturer.22:37
infinityI've seen much worse sites. :P22:37
niklasfiinfinity: yes, it is very much in deed. if people can't set up a proper site using a markup language, how can i trust them with turing complete languages, or even hardware22:38
infinityHardware nerds don't tend to care much about pretty web sites, in my experience.22:40
infinitySo, really, what you're looking for is "did they hire a good web monkey", to which the answer appears to be "no".22:40
infinity(Hint, Canonical employs web developers to work on the website, and they're not the same people who work on Ubuntu, so judging one by the other would, again, be silly, right?)22:40
niklasfiinfinity: i guess the problem is that i am no hardware nerd. I sometimes would very much like to understand what you guys do, but i haven't even gotten to learning all of the boot options i need to manually start my ubuntu from grub22:41
mythosuboot is/was new to me too22:42
niklasfiinfinity: yes. but it gives me information about how much they care about their user base. if they are alright with making them suffer by showing them crappy web pages, they are probably alright with making them suffer by giving them bad support22:42
mythosbut with some time, you can learn everything22:43
infinityniklasfi: Most of the support for the Trimslice is via their wiki.  Then again, that's not much worse than the situation with most dev boards. :)22:49
infinity(To be clear, I have no affiliation with them, and I don't even own one, but they look neat and, like I said, I like toys)22:50
GrueMasterinfinity: On this f-k-i bootparm issue, I'm trying to figure how best to ensure some kernel parameters are passed from cmdline or enabled via preseed.  I need to make sure none are duplicated.  Suggestions?22:57
GrueMasterI really don't like scripting in shell.  Especially busybox.22:58
infinityGrueMaster: Well, I get the impression that user-params smooshes together both command line and preseed into one.22:58
infinityGrueMaster: I didn't look closely, but I assume it removes dupes?  Dunno.  If not, it should.22:59
GrueMasterNope.  Just preseed from what I can tell.22:59
infinityGrueMaster: No, not just preseed.  It starts by parsing cmdline.22:59
infinityGrueMaster: Most of the script is about parsing and filtering cmdline, then it adds the preseed bits at the end.23:01
infinityGrueMaster: Although, I think it only parses and collects options after the "--" (ie: user-added options)23:02
GrueMasterPossible.  I'm looking at it now.23:02
infinityGrueMaster: Which makes sense, since f-k-i or grub-installer, etc, should be what's responsible for making sure platform-specific options are in place.23:02
GrueMasterSo the mem= lines should be on by default for omap4.  makes sense.  I can do that.23:03
infinityGrueMaster: Anyhow.  Avoiding duplicates is easy enough, we can just loop through our options in f-k-i.23:04
GrueMasterWhat I don't see though is things like module parameters, or console parameters.  Not that either are critical for first boot afaict.23:04
infinityGrueMaster: Well, if you were to add module params manually, they'd land after the -- ... Or they should.23:05
infinitySo, user-params would pick them up.23:05
GrueMasterOk.  The kernel shouldn't care I guess.23:06
=== rsalveti` is now known as rsalveti
GrueMasterd-i sucks in so many ways.  http://paste.ubuntu.com/828261/23:57
infinityGrueMaster: I don't see anything sucking there.23:58
infinityGrueMaster: Everything's listed the same number of times it's provided.  We can easily make them unique later.23:59
GrueMasterOnly that user-params is spewing both /proc/cmdline and preseed.23:59
infinityGrueMaster: It's supposed to.23:59
infinityGrueMaster: Like I said, everything after the -- on cmdline, plus preseed.  That's intentional.23:59

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!