**** BEGIN LOGGING AT Tue Jul 06 02:59:58 2010 **** BEGIN LOGGING AT Tue Jul 06 08:26:14 2010 Jul 06 08:26:19 ogra: i am waiting for my hw. so never try that X before Jul 06 08:26:41 cooloney, oh, sorry i tend to forget that you dont have the HW Jul 06 08:26:51 sebjan: how about you? Jul 06 08:27:00 seems the freezes happen if i dont specify mem=463M on the cmdline Jul 06 08:27:07 now it works but is very slow Jul 06 08:32:01 cooloney: do you have a beagle board? Jul 06 08:32:11 amitk: no Jul 06 08:32:20 cooloney: you will in Prague :) Jul 06 08:32:21 amitk, he'll get a panda soon Jul 06 08:32:33 i think its on its way already Jul 06 08:32:34 amitk: yeah, i will. panda and prague, hehe Jul 06 08:32:49 ogra: yeah, i hope it will arrive this week Jul 06 08:32:51 cooloney: I'm bringing you a beagle as well Jul 06 08:33:15 amitk: thanks a lot Jul 06 08:33:33 we need more XMs in the team Jul 06 08:33:45 XM support is still behind Jul 06 08:34:12 lag has one now Jul 06 08:34:16 yeah Jul 06 08:34:32 cool man Jul 06 08:34:32 i still havent found whats wrong with our u-boot yet Jul 06 08:34:46 currently the XM doesnt boot with the archive u-boot Jul 06 08:35:02 Mine does Jul 06 08:35:08 Just not initrd Jul 06 08:35:21 lag, thats equivalent to "doesnt boot" :) Jul 06 08:35:37 initrd is a requirment for ubuntu images Jul 06 08:35:47 :) Jul 06 08:36:10 mine does boot with the binary from rcn, including initrd Jul 06 08:36:29 but i cant find the difference in our sources, he seems to use the same i use Jul 06 08:36:30 ogra: I have been doing some testing with the UNE UI, is that what you mean? Jul 06 08:37:06 sebjan, well, i just wanted to know if X seems usable for you, here it freezes every time it accesses the SD Jul 06 08:38:00 i'm currently testing the panda images from http://cdimage.ubuntu.com/ubuntu-netbook/ports/daily-preinstalled/ the oem-config screens take ages to show up Jul 06 08:38:40 (note that only 20100706.1 will have the mem=463M fix in them, they are currently building) Jul 06 08:39:20 I did not have such issues the last time I tested... It was not very fast, but I suspected it was because it seems that both UNE and the regular gnome desktop UI are started (bug 596004) Jul 06 08:39:21 Launchpad bug 596004 in netbook-meta (Ubuntu) "with 2D UI (une-efl) selected, the gnome desktop toolbar is also visible and seems to be running (affects: 1) (heat: 242)" [Undecided,New] https://launchpad.net/bugs/596004 Jul 06 08:39:36 would be helpful if the panda had some indicator LED for SD access, you never know if it reads or actually is dead Jul 06 08:40:01 sebjan, yeah, that bug is on my list but shouldnt cause much slowdown Jul 06 08:40:53 * ogra triages it properly Jul 06 08:41:28 gah, now it seems to hang hard again Jul 06 08:42:01 and i was already through oem-config and in the slideshow Jul 06 08:43:39 sebjan, what kind of SD's do you use ? Jul 06 08:43:51 * ogra currently has a class 4 Jul 06 08:44:25 ogra: I also use class 4 cards (Kingston) Jul 06 08:44:39 (micro SD with adapter) Jul 06 08:45:21 hmm doesnt seem to come back to life Jul 06 08:46:18 (regular SD card for me) Jul 06 08:47:34 * ogra wishes he could at least get far enough to have a user created Jul 06 08:47:44 so i could read any logs Jul 06 08:49:44 cooloney, is the binary we have in the archive recent ? does it have all fixes and patches ? Jul 06 08:58:34 ogra: the M ti-omap4 kernel? i we got lots for fixes and patches from sebjan except the syslink module patch Jul 06 08:58:56 ogra: what's uname -a in your system Jul 06 09:00:09 sebjan: Is syslink discoverable? Jul 06 09:02:55 lag: you'll have to explain me what it means 'discoverable' :)... Jul 06 09:04:40 sebjan: i am guess, it means probable? Jul 06 09:04:51 At the moment (with the module patches applied) the kernel doesn't 'discover' the device and load the modules Jul 06 09:06:46 lag: did you add the platform device and resource in the panda board files in mach-omapX? Jul 06 09:07:34 Yes Jul 06 09:08:08 If I register the device alone I get KERNEL and UDEV uevents Jul 06 09:08:26 But I need a DRIVER uevent Jul 06 09:09:04 I've tried registering a platform driver, but because the device is unknown to the kernel the .probe function is never called Jul 06 09:15:42 lag: the .name field of the platform device in your board file should be the same of the driver platform name in syslink driver. IMHO Jul 06 09:16:47 lag: I suppose that we are missing something either in the device declaration or in the driver... this is what we need to get sorted out Jul 06 09:17:03 cooloney: It is Jul 06 09:17:18 I've moved the .probe function to the same place as the __init function Jul 06 09:17:34 The __init function is called and the driver is registered successfully Jul 06 09:17:39 But .probe is still not called Jul 06 09:18:02 .probe functions are not called if the kernel does not know about the device Jul 06 09:18:10 Hence the discoverable question Jul 06 09:18:31 sebjan: It's not even getting as far as the driver Jul 06 09:19:44 sebjan: I think we need more information in my (__init) function to tell the kernel about the existence of the hardward Jul 06 09:19:46 hardware* Jul 06 09:19:55 cooloney, well, we only had one upload yet and i cant get to a point in the image to type in uname (i cant get past user creation before the board hangs hard) Jul 06 09:20:16 I am in the process of putting the driver & device register code in the same place - see what that does Jul 06 09:20:48 cooloney, so i cant really say what uname would give, but i'm pretty sure its 2.6.34-900.1 since that was the only upload we had yet Jul 06 09:21:21 ogra: oh, please wait, i can provide a new testing kernel for you, Jul 06 09:21:24 lag, arent there any other devices you could just steal from ? Jul 06 09:21:37 cooloney, doesnt help, i'm testing images, they build from the archive Jul 06 09:21:56 cooloney, we should rather kick leann to do a new upload if the tree has so many patches Jul 06 09:21:58 ogra: ah, right Jul 06 09:22:17 ogra: ok, i think rtg will upload that Jul 06 09:22:27 lag, i could imagine rfkill to be a platform device doing something similar to what we need Jul 06 09:22:57 cooloney, ah, right, he did the last upload too, i missed that Jul 06 09:24:11 lag: send me patches of your current changes, and I'll tell you if I have an idea and try to play with it this afternoon Jul 06 09:24:49 ogra: No, I can't find any other devices that do this Jul 06 09:28:42 gah Jul 06 09:36:15 the plymouth splash looks really good on the panda :) Jul 06 11:39:35 lag, could you paste the Hardware line of a beagle XM for me ? Jul 06 11:39:42 lag, from /proc/cpuinfo Jul 06 11:42:01 Yup - 2 mins Jul 06 11:42:04 hrw, your openjdk build, how did you do the testbuild, in a clean chroot ? Jul 06 11:42:25 ogra: no - in normal BB system Jul 06 11:42:27 lag, no hurry, anytime you find the time is fine Jul 06 11:42:38 hrw, definate 2normal BB system" Jul 06 11:42:47 s/2/\" Jul 06 11:42:49 It's okay - I'm waiting for an email from linux-hotplug Jul 06 11:42:58 ogra: quite up-to-date maverick Jul 06 11:43:50 openjdk needs itself to register the ca certificates but the lucid openjdk seems to segfault here, what openjdk version did you have installed for the build ? Jul 06 11:44:31 (keytool is a java app and calls the former openjdk while installing the build deps) Jul 06 11:44:32 ogra: I did "apt-get build-dep openjdk-6", then "apt-get source openjdk-6", "cd openjdk-6-*;debuild -b -uc -us" Jul 06 11:44:42 hmm Jul 06 11:44:43 ogra: did not looked at versions etc Jul 06 11:45:22 LANG=C LC_ALL=C keytool -importcert -trustcacerts -keystore $KEYSTORE -providerClass sun.security.pkcs11.SunPKCS11 -providerArg '${java.home}/lib/security/nss.cfg' -noprompt -storepass "$storepass" -alias "$alias" -file "$cacertdir/$pem" > $log 2>&1 Jul 06 11:45:29 thats the line that segfaults Jul 06 11:45:56 and openjdk ships /usr/lib/jvm/java-6-openjdk/jre/bin/keytool Jul 06 11:46:16 (which /usr/bin/keytool links to through an alternative) Jul 06 11:46:35 keytool error: java.lang.RuntimeException: Usage error, sun.security.pkcs11.SunPKCS11 is not a legal command Jul 06 11:46:37 i dont get how it could build for you while it doesnt on the buildd Jul 06 11:46:59 ogra: want shell access to that beagle? Jul 06 11:47:04 no Jul 06 11:47:34 so it doesnt segfault for you apparently but just complains Jul 06 11:48:26 does apt-get install --reinstall ca-certificates-java work for you without any errors ? Jul 06 11:57:03 Setting up ca-certificates-java (20100412) ... Jul 06 11:57:03 Processing triggers for initramfs-tools ... Jul 06 11:58:39 very weird Jul 06 11:59:07 http://hrw.pastebin.com/4ZSwj2zP Jul 06 11:59:17 on the buildd /var/lib/dpkg/info/ca-certificates-java.postinst fails Jul 06 12:00:15 root@beagle:/mnt/hrw/UBUNTU# /var/lib/dpkg/info/ca-certificates-java.postinst configure Jul 06 12:00:18 creating /etc/ssl/certs/java/cacerts... Jul 06 12:00:21 done. Jul 06 12:01:04 http://hrw.pastebin.com/7KqebZ5e shows ver Jul 06 12:04:23 morning Jul 06 12:06:00 hrw, aha ... Setting up default-jre-headless (1.6-38) ... Jul 06 12:06:05 from the build log Jul 06 12:07:52 ogra@osiris:~/Devel/branches/jasper-initramfs-0.10$ apt-cache show ca-certificates-java|grep Depends Jul 06 12:07:52 Depends: ca-certificates (>= 20090814), openjdk-6-jre-headless (>= 6b16-1.6.1-2) | java6-runtime-headless Jul 06 12:08:11 ogra@osiris:~/Devel/branches/jasper-initramfs-0.10$ apt-cache search java6-runtime-headless Jul 06 12:08:11 default-jre-headless - Standard Java or Java compatible Runtime (headless) Jul 06 12:08:11 openjdk-6-jre-headless - OpenJDK Java runtime, using Hotspot JIT (headless) Jul 06 12:08:11 sun-java6-jre - Sun Java(TM) Runtime Environment (JRE) 6 (architecture independent files) Jul 06 12:08:42 so for whatever reason default-jre-headless gets installed on the buildd instead of openjdk-6-jre-headless Jul 06 12:17:41 hrw, the versions in http://hrw.pastebin.com/4ZSwj2zP cant be right Jul 06 12:17:59 hrw, 6b20~pre1-0ubuntu2 didnt build Jul 06 12:18:06 where did you get it from ? Jul 06 12:18:24 smells like you have your testbuild installed now Jul 06 12:18:52 the last one that built on the buildd was 6b18-1.8-2ubuntu2 Jul 06 12:19:45 I built and installed Jul 06 12:20:23 yeah, i need to know what version was used during installation of ca-certificates-java Jul 06 12:21:05 which indeed came in with your apt-get build-dep Jul 06 12:21:43 probably 6b18 Jul 06 12:22:10 * ogra doesnt get why it didnt segfault for you Jul 06 12:48:05 ogra: do you know if for maverick we have different kernel trees for omap? or both omap3 and omap4 should work with the ti-omap4 branch? Jul 06 12:48:56 * rsalveti looking at maverick's kernel Jul 06 12:49:09 the omap3 binary is built from the normal ubuntu tree Jul 06 12:49:14 rsalveti: I asked recently, and omap3 was broken in the omap4 branch Jul 06 12:49:25 omap4 comes from the ti-omap4 branch Jul 06 12:49:57 rsalveti, omap4 is maintained by cooloney, omap3 by lag and mpoirier Jul 06 12:50:19 oh, ok, so different branches for both Jul 06 12:50:35 or it should all work at the same branch? Jul 06 12:50:39 well, only a different branch for omap4 actually Jul 06 12:50:48 it's just that I just saw one ti-omap4 branch at the gitweb Jul 06 12:51:14 the big target is that omap4 goes to mainline at some point, then both binaries will be built from the ubuntu branch Jul 06 12:51:31 but since omap4 is over 700 patches that will take a while Jul 06 12:51:35 oh, sure :-) Jul 06 12:51:39 and is unlikely to happen for maverick Jul 06 12:51:50 got it now, at least for omap 3 we're in a better situation Jul 06 12:51:57 now the problem is just omap4 Jul 06 12:52:02 yeah, don't think so Jul 06 12:52:03 so currently TI takes the ubuntu maverick branch and rebases their patches on top of that Jul 06 12:52:11 for omap4 Jul 06 12:52:35 which makes it being a bit behind in versioning Jul 06 12:52:39 ukleinek: I don't get why a non working image would be built Jul 06 12:52:52 ukleinek: see my rational in my answer Jul 06 12:53:03 ukleinek: uboot needs to be fixed Jul 06 12:53:34 ogra: yeah, right Jul 06 12:53:49 tons of patches, maybe for 36 or 37 Jul 06 12:53:55 right Jul 06 12:55:18 note: basic panda support is in tmlind's tree for 36.. not sure if it boots, waiting for ups myself.. Jul 06 12:55:59 rcn-ee, "basic" isnt want we need for maverick :) Jul 06 12:56:12 also the tree we use is blaze+panda capable Jul 06 12:56:23 rcn-ee: yeah, just saw the commits Jul 06 12:56:38 yeah, but not everyone use the default maverick kernel. ;) Jul 06 12:57:34 * rsalveti is looking forward to put my hands at a panda board Jul 06 12:57:43 rcn-ee, well ... Jul 06 12:57:59 rsalveti, no fun yet ... multiple hangs etc Jul 06 12:58:08 700 patches, I believe maverick will be much more than the basic Jul 06 12:58:17 definately Jul 06 12:58:20 yay, sounds like fun.. ;) just like my custom imx515 board... Jul 06 12:58:25 rsalveti: at any pandaboard or at own pandaboard? Jul 06 12:58:44 ogra: hangs are always fun, at some way haha :-) Jul 06 12:58:52 heh Jul 06 13:00:00 hrw: the omap4 pandaboard :-) Jul 06 13:00:41 rsalveti: by 'any' I meant 'not own' Jul 06 13:01:10 ogra: That's not true Jul 06 13:01:19 I am working on omap3 & omap4 Jul 06 13:01:40 * ogra hangs his head in shame Jul 06 13:02:05 rsalveti, correction: lag woprks on *all important kernels* :) Jul 06 13:02:48 Now you're talking Jul 06 13:03:39 rcn-ee: what custom imx515 board? Jul 06 13:03:50 rcn-ee: want to send me patches to add board support? ;) Jul 06 13:12:17 amitk: someone offered efikamx few days ago... Jul 06 13:12:29 hrw: the board or code? Jul 06 13:13:25 boards Jul 06 13:14:07 #linaro:02:10 < zumbi> if someone fancies an efikamx board to test out linaro, let me know Jul 06 13:14:37 Saturday 3rd July Jul 06 13:15:19 hrw: oh, sorry :-) do you know when pandaboard will be public? like beagleboard Jul 06 13:15:52 rsalveti: nope Jul 06 13:16:03 rsalveti: should be Q4 Jul 06 13:16:47 hrw: Do you know if the efiamx comes with documentation to enable in mainline? Jul 06 13:17:04 Q4? wow Jul 06 13:18:51 amitk: it doesnt Jul 06 13:20:14 armin76: pity. I don't want to spend time on it then Jul 06 13:26:24 amitk: no idea Jul 06 13:55:37 npitre: I'm not sure anymore about ZRELADDR, IIRC something was broken with ZBOOT_ROM images for U-Boot Jul 06 13:58:45 ogra: Sorry for the delay - I had to make and re-flash a new card Jul 06 13:58:49 ogra: Hardware : OMAP3 Beagle Board Jul 06 13:58:56 perfect Jul 06 13:58:59 same as C4 Jul 06 13:59:12 * ogra is happy he doesnt need to special case for XM Jul 06 13:59:18 :) Jul 06 13:59:25 I'm pleased you're pleased Jul 06 13:59:31 :) Jul 06 13:59:55 perisa Jul 06 13:59:59 Doh Jul 06 14:00:24 Actually: lol & ogra Jul 06 14:00:30 I need a word ;) Jul 06 14:00:40 * ogra pulls up the dictionary Jul 06 14:00:46 You've wasted a working day! Jul 06 14:00:55 a long one or rather a short one ? Jul 06 14:00:55 Leading me up the garden path Jul 06 14:01:07 REF: /lib/udev/rules.d/80-drivers.rules, DRIVER!="?*", ENV{MODALIAS}=="?*", RUN+="/sbin/modprobe -b $env{MODALIAS}" Jul 06 14:01:16 Do you know what that means? Jul 06 14:01:36 lool* Jul 06 14:02:02 lool, ogra: Do you subscribe to linux-hotswap? Jul 06 14:02:03 if DRIVER is set run modprobe ? Jul 06 14:02:10 NOOOOOOOOOOOOOOOo Jul 06 14:02:12 lag, nope, i'm no kernel guy Jul 06 14:02:24 It doesn't mean that Jul 06 14:02:32 It's what you told me it meant Jul 06 14:02:34 so what does it mean ? Jul 06 14:02:36 Arhgggggggggggggggggggggggggg Jul 06 14:02:41 :) Jul 06 14:02:56 educate me :) Jul 06 14:04:04 ugh, it means if DRIVER is *not* set, right ? Jul 06 14:04:12 It means: This rule is skipped only if DRIVER is set Jul 06 14:04:21 Hang on Jul 06 14:05:52 lag: Oh right, DRIVER!=, I missed this bit Jul 06 14:05:57 right, i see that now, if DRIVER is not set but MODALIAS exists in the environment, call modprobe Jul 06 14:06:00 Follow this: http://www.spinics.net/lists/hotplug/msg03946.html Jul 06 14:06:03 lag: I read DRIVER= Jul 06 14:06:07 * ogra too Jul 06 14:06:07 Arrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr Jul 06 14:06:10 :( Jul 06 14:06:11 * lag crys Jul 06 14:06:16 * lag cries* Jul 06 14:06:17 lag: But the MODALIAS bit is ok, right? Jul 06 14:06:37 Follow that link I posted Jul 06 14:06:50 This has cost me a day :( Jul 06 14:07:20 lag: But the upside is that we're all three educated on what DRIVER is used for, and you have a path forward! Jul 06 14:07:38 I guess there's always an upside Jul 06 14:07:39 and your MODALIAS is wrong i'D say Jul 06 14:07:53 Why is it wrong? Jul 06 14:07:56 "platform:" Jul 06 14:08:02 What should it be? Jul 06 14:08:07 the module name Jul 06 14:08:16 Uh no Jul 06 14:08:28 thats what modprobe -b calls in the rule above Jul 06 14:08:34 The real MODALIAS is: platform:syslink Jul 06 14:08:44 I ****'ed out the syslink Jul 06 14:08:44 and the module name is syslink Jul 06 14:09:04 that's ok, you just need to modalias the .ko now Jul 06 14:09:14 /sbin/modprobe -b platform:syslink will error out, no ? Jul 06 14:09:46 ogra: Well not if some module has an alias for platform:syslink ? Jul 06 14:10:02 in sysfs ? Jul 06 14:10:19 no, in the .ko Jul 06 14:10:19 i.e. where would modprobe get that from ? Jul 06 14:10:20 ogra: modinfo e1000e|grep alias Jul 06 14:10:22 ah Jul 06 14:10:38 oh, right Jul 06 14:11:01 grep MODALIAS /var/log/udev Jul 06 14:11:40 yeah yeah :) Jul 06 14:11:53 * ogra admits again that he's no kernel guy Jul 06 14:12:47 * lag goes to work Jul 06 14:12:58 * lag takes a big breath then --^ Jul 06 14:15:35 * ogra goes to eat some icecream Jul 06 14:19:11 lool: The driver is called syslink_ipc.ko Jul 06 14:19:22 module == driver Jul 06 14:19:48 MODALIAS=platform:syslink_ipc Jul 06 14:20:03 Surely there should be an association? Jul 06 14:20:18 KERNEL[946692859.898376] add /devices/platform/syslink_ipc (platform) Jul 06 14:20:19 lag: That doesn't work? Jul 06 14:20:25 No Jul 06 14:20:38 KERNEL[946692859.898376] add /devices/platform/syslink_ipc (platform) Jul 06 14:20:38 UDEV_LOG=3 Jul 06 14:20:38 ACTION=add Jul 06 14:20:38 DEVPATH=/devices/platform/syslink_ipc Jul 06 14:20:38 SUBSYSTEM=platform Jul 06 14:20:40 MODALIAS=platform:syslink_ipc Jul 06 14:20:42 SEQNUM=735 Jul 06 14:20:44 That is my uevent Jul 06 14:20:51 But udev isn't loading the module Jul 06 14:20:52 lag: But platform:syslink_ipc != syslink_ipc Jul 06 14:21:04 Correct Jul 06 14:21:06 lag: Try modprobe platform:syslink_ipc for yourself Jul 06 14:21:13 And modinfo syslink_ipc Jul 06 14:21:22 ubuntu@panda:~$ modinfo syslink_ipc Jul 06 14:21:22 filename: /lib/modules/2.6.34-901-omap4/kernel/drivers/dsp/syslink/multicore_ipc/syslink_ipc.ko Jul 06 14:21:22 license: GPL Jul 06 14:21:22 author: Texas Instruments Jul 06 14:21:22 srcversion: 7329950ABDD680C3701426C Jul 06 14:21:23 depends: syslink_proc,omap_notify,syslink_proc4430,syslink_platform,notify_ducatidriver,syslink_ipu_pm Jul 06 14:21:26 vermagic: 2.6.34-901-omap4 SMP preempt mod_unload ARMv7 Jul 06 14:21:28 parm: ipc_name:Device name, default = syslink_ipc (charp) Jul 06 14:21:30 parm: ipc_major:Major device number, default = 0 (auto) (int) Jul 06 14:21:32 parm: ipc_minor:Minor device number, default = 0 (auto) (int) Jul 06 14:21:42 You probably want an actual alias then Jul 06 14:21:57 I think you need to define that in your kernel module Jul 06 14:22:10 So an alias | and entry in udev rules? Jul 06 14:22:19 You shouldn't need an entry in udev Jul 06 14:22:28 OR Jul 06 14:22:33 AIUI, udev runs modprobe platform:syslink_ipc but that fails to find anything Jul 06 14:22:49 lag: Right, but you probably dont want an udev rule, just an alias Jul 06 14:23:14 I don't know how to do that Jul 06 14:23:16 * lag looks Jul 06 14:23:36 Or you change the kernel to emit another event which triggers the load, perhaps another rule would match; I think the modalias is the easiest path Jul 06 14:24:00 MODULE_ALIAS("platform:" DRV_NAME); Jul 06 14:24:03 Bingo! Jul 06 14:25:33 Eh right, was grepping for something similar in drivers/, but I should have looked in arch/arm Jul 06 14:25:39 arch/arm/mach-omap2/mailbox.c has another instance Jul 06 14:26:59 That was in /drivers Jul 06 14:27:17 lag: linux/mod_devicetable.h #defines PLATFORM_MODULE_PREFIX, not sure using it is worth though Jul 06 14:27:22 I mean it would have been quicker to grep arch/arm Jul 06 14:28:00 #define PLATFORM_MODULE_PREFIX "platform:" Jul 06 14:28:01 Lol Jul 06 14:28:14 ukleinek: Not widely used it seems (drivers/base/platform.c) Jul 06 14:28:23 MODULE_ALIAS(PLATFORM_MODULE_PREFIX, DRV_NAME);# Jul 06 14:28:25 :) Jul 06 14:28:32 lag: no , please Jul 06 14:28:45 ? Jul 06 14:28:48 lool: yeah, I know, just stumbled over it some time ago Jul 06 14:28:54 MODULE_ALIAS(PLATFORM_MODULE_PREFIX DRV_NAME); Jul 06 14:29:33 Oh, I see Jul 06 14:29:40 I'm not going to use it anyway Jul 06 14:29:48 I was just messing :) Jul 06 14:30:04 My first attempt is the most common Jul 06 16:55:34 I'm trying to install ubunt-arm on the beagleboard via netinstall. Unfortunately, my usb-nic-adapter (it is a moschip type) is not recognized as a network device, although it worls perfectly on a x86 linux system. Where can I find the kernel sources/patches and kernel config in order tu build my own uImage file? Jul 06 16:56:48 try maverick first? Jul 06 16:59:08 robschi: do you know the module responsible for it at x86 already? Jul 06 16:59:44 there is a fix for it but the netinst image hasnt been regenerated yet Jul 06 17:00:13 either do a server install or use maverick Jul 06 17:00:37 or wait for lucid .1 Jul 06 17:02:41 rsalveti: lsmod replies with things like Jul 06 17:02:49 mii 4032 3 mcs7830,usbnet,r8169 Jul 06 17:02:50 ukleinek: I never used ZBOOT_ROM with U-Boot Jul 06 17:03:17 ukleinek: I think U-Boot is getting too much in the way for that feature Jul 06 17:04:05 server install does not work, becaus the installer insists on using a cd-rom drive Jul 06 17:04:12 robschi: I had filed a bug on this in Lucid, but The number (and status) currently escapes me. Jul 06 17:04:14 * npitre thinks that U-Boot is often too much in the way, period. Jul 06 17:04:51 robschi, server install was tested by several people, it should just work Jul 06 17:06:23 robschi, did you follow https://wiki.ubuntu.com/ARM/BeagleServerInstall ? Jul 06 17:06:48 ogra_cmpc: yes, i did Jul 06 17:07:29 thats strange, the released image was heavily tested Jul 06 17:07:58 lag: I have my syslink modules auto-loading finally :). The implementation is not clean yet, but the mechanism is functional at least. I'll send you my changes so that we can discuss them together. Jul 06 17:08:29 I have also finished :) Jul 06 17:08:36 * ogra_cmpc thinks lag found the right solution too today :) Jul 06 17:08:40 ogra_cmpc: i'm installing over serial port, but this should not influence the behaviour of the installer, or does it? Jul 06 17:08:44 snap .... Jul 06 17:09:02 lag: cool :) Jul 06 17:09:08 What did you do? Jul 06 17:09:09 robschi, as long as the boot.scr stays intact Jul 06 17:10:53 (and you dont drop the preseed values from the kernel cmdline) Jul 06 17:11:14 ogra_cmpc: i will check this again. Jul 06 17:11:42 lag: I'll send you a patch shortly, I am doing some quick reformtating and squashes. But is basically is what we talked about and what you dealed about here this afternoon ("platform:") Jul 06 17:12:13 Oh, you thought you'd sneak in there with my info - cheeky! Jul 06 17:13:48 lag: I have worked on this in dotted line since last week, and sure would have spent some time on this platform thing :) Jul 06 17:15:04 lag: but I am sure the patch needs re-disgn to put things in proper files at least Jul 06 17:15:14 So which patch should we use? I can submit mine either today or tomorrow morning if you like? Jul 06 17:15:22 My code is only a few lines Jul 06 17:15:42 I think it's 10 lines of code Jul 06 17:18:11 lag: I am ok to use yours. Please wait until I send you mine so that we can cross-check what we did is correct. I'send send you my patch in a few minutes. Jul 06 17:18:42 No problem sebjan Jul 06 17:26:34 lag: I just sent it to you Jul 06 17:26:49 I received it Jul 06 17:27:39 sebjan: I think platform_device_alloc() is deprecated Jul 06 17:29:57 lag: ok, I did not check that (the device creation part is a rought copy/paste from some syslink code) Jul 06 17:34:19 sebjan: http://kernel.ubuntu.com/git?p=lag/ubuntu-maverick.git;a=commitdiff;h=b8bec3a6239f0d9d23250eb0265bd56f77198384;hp=05cd276ae6bfe20745244f2285fab2b3d40ca1e3 Jul 06 17:36:30 ogra_cmpc: yes, the bootargs in boot.scr have not been correct, because I bootet the system manually without boot.scr. Now it installs fine. Thank you for your help. Jul 06 17:37:12 sebjan: I have to go - speak tomorrow Jul 06 17:37:34 lag: ok, me to, talk to you tomorrow Jul 06 17:38:13 Have a good night all Jul 06 17:38:16 * lag out Jul 06 17:39:07 anyone knows if maverick gcc supports -mfloat-abi=hard? Jul 06 17:40:44 hrw@home:~/xdeb$ arm-linux-gnueabi-gcc hello.c -mfloat-abi=hard Jul 06 17:40:44 /usr/lib/gcc/arm-linux-gnueabi/4.5.1/../../../../arm-linux-gnueabi/bin/ld: error: /tmp/cckGhdno.o uses VFP register arguments, a.out does not Jul 06 17:40:48 /usr/lib/gcc/arm-linux-gnueabi/4.5.1/../../../../arm-linux-gnueabi/bin/ld: failed to merge target specific data of file /tmp/cckGhdno.o Jul 06 17:41:13 markos_: you need to rebuild all software with it iirc Jul 06 17:42:31 hrw@home:~/xdeb$ arm-linux-gnueabi-gcc -mfloat-abi=hard test.c -o test.o -c Jul 06 17:42:51 hrw: I know, the question is if it supports it, previously it didn't Jul 06 17:42:59 the error msg you gave is a good sign :) Jul 06 17:43:11 previously it just said "not supported" Jul 06 17:43:17 previously = karmic Jul 06 17:43:26 markos_: well, iirc you can't make gcc not support it Jul 06 17:43:34 http://pastebin.com/TXTBgRij is assembly of generated source Jul 06 17:43:41 it was included on gcc-4.5, so... Jul 06 17:43:43 armin76: with an old gcc version you can :) Jul 06 17:44:10 http://pastebin.com/2S6ztE87 is source used Jul 06 17:44:28 markos_: codesourcery != gcc :) Jul 06 17:45:31 ssvb: I know, I was just curious if I could use maverick's gcc instead of codesourcery if it supported hardfp Jul 06 17:46:16 hrw: what flags did you use (for comparison) Jul 06 17:46:33 markos_: hrw@home:~/xdeb$ arm-linux-gnueabi-gcc -mfloat-abi=hard test.c -o test.o -c -s^C Jul 06 17:46:53 arm-linux-gnueabi-objdump -D test.o Jul 06 17:47:09 markos_: The gcc-4.4 we prepare for maverick armel has CodeSourcery patches (gcc-linaro) Jul 06 17:47:10 ok, just wanted to check optimizations settings Jul 06 17:47:31 lool: great to know Jul 06 17:47:50 basically each FSF gcc release has some of codesourcery patches ;) Jul 06 17:49:12 lool: does gcc-4.4 show good performance? Jul 06 17:50:03 ssvb: here I get up to 35% faster than softfp in pure fp benchmarks -using codesourcery Jul 06 17:50:38 markos_: hardfp contra softfp? Jul 06 17:51:56 in my limited benchmarks, gcc-4.4 has major performance regressions on arm platform for some tests when compared to gcc 4.3 and 4.5 Jul 06 17:52:32 hrw: yes Jul 06 17:52:57 markos_: on A9 it is less difference afaik but who has A9... Jul 06 17:53:34 hrw: me :D Jul 06 17:53:45 hrw: http://www.powerdeveloper.org/forums/viewtopic.php?p=13602#13602 Jul 06 17:53:56 hrw: i don't have an a9 (yet) to test hardfp Jul 06 17:54:07 perhaps the difference is smaller, I don't know Jul 06 17:54:25 markos_: yes, I know :) I also have seen the speedup on some artificial benchmarks that I did myself Jul 06 17:55:51 is ubuntu going to use -mfloat-abi=hard ? Jul 06 17:56:29 ssvb: there were some discussions about making such rebuild for tests Jul 06 17:56:31 irc Jul 06 17:56:33 iirc Jul 06 17:57:12 good Jul 06 17:57:21 ubuntu itself will likely not change in maverick, in maverick+1 we might take as default what linaro develops atm Jul 06 17:57:41 since they do the toochain changes during maverick Jul 06 17:57:53 and test them etc Jul 06 18:04:44 ssvb: know of any way to detect the current abi on a system? Jul 06 18:04:52 or in a file for that matter Jul 06 18:06:31 markos_: what do you mean? Jul 06 18:06:41 markos_: via gcc predefined macros? Jul 06 18:07:45 markos_: you can use 'readelf -a' to get abi of some binary Jul 06 18:09:07 ssvb: that's what I wanted yes Jul 06 18:09:22 binaries using hard vfp will have "Tag_ABI_VFP_args: VFP registers" Jul 06 18:10:36 yep Jul 06 18:10:56 ssvb: I would be interested in discussing your performance regressions Jul 06 18:11:07 ssvb: cool thanks Jul 06 18:11:31 ssvb: I'm animating the toolchain WG in Linaro, I'd love if we could check your performance regressions against our Linaro toolchains Jul 06 18:12:02 ssvb: I know 4.5 is much better than 4.4, and I believe Linaro 4.4 / CodeSourcery 4.4 has much of 4.5 improvements Jul 06 18:12:19 lool: http://hardwarebug.org/2009/08/05/arm-compiler-shoot-out/ Jul 06 18:12:28 lool, "animating" ? Jul 06 18:12:34 like http://www.youtube.com/watch?hl=de&v=QAue4hnH8-A ? Jul 06 18:12:54 I generally don't trust random blogs, but it matches the results of my own benchmarks Jul 06 18:13:14 in some cases gcc 4.4 produces code which is up to 30% slower Jul 06 18:13:39 ssvb: so I guess a way to detect is a system is softfp/hardfp would simply grep Tag_ABI_VFP_args: VFP registers in /lib/libc.so.6 Jul 06 18:15:30 lool: and what is very easy to verify, gcc 4.4 always generates bigger binaries (maybe it's one of reasons why it is so bad for some applications) Jul 06 18:17:33 lool: both gcc 4.3 and 4.5 generate code which is either a lot faster than 4.4 or show about the same performance Jul 06 18:17:48 lool: and 4.5 is faster than 4.3 Jul 06 18:22:06 ssvb: thanks for sharing tihs link Jul 06 18:22:58 ssvb: Ok, well I will bring these up at our next meeting, I think it's good feedback Jul 06 18:23:17 ssvb: we carry the CodeSourcery patchset in Linaro GCC, so it's rather good for us it seems Jul 06 18:23:24 But the RVCT tests are interesting too Jul 06 18:32:00 lool: ffmpeg may be not a very good example because it's coding style is not so typical Jul 06 18:32:05 but my own benchmarks included pixman library (its regression tests) and a simple C++ roguelike game :) Jul 06 18:32:42 I have not tried thumb2 though, maybe it would show some totally different results Jul 06 18:33:18 lool: what do you generally use to evaluate toolchain performance? Jul 06 18:33:59 have a nice rest of day Jul 06 18:41:58 ssvb: hello world :D Jul 06 18:42:06 ssvb: In my experience, it's hard to get definitive informationon who uses what benchmarks Jul 06 18:42:10 ssvb: Kind of secret sauce Jul 06 18:42:31 ssvb: So in Linaro we aim at building tools to measure more things, such as size of binaries Jul 06 18:42:53 ssvb: thumb2 is really important, the size difference might be just enough to fit in cache, and make a huge difference Jul 06 18:43:13 lool: but you lose performance with thumb2 Jul 06 18:43:51 markos_: A very small amount, and it's decreasing with new SoCs Jul 06 18:44:08 markos_: And it's only lose if you don't consider cache/memory pressure Jul 06 18:44:21 plus, storage loading times Jul 06 18:44:27 I see Jul 06 18:44:28 it all adds up, it's not much but ... Jul 06 18:45:00 * ssvb wonders if gcc can (or does) generate arm code for "hot" functions Jul 06 18:45:10 it would make a lot of sense Jul 06 18:46:36 I think you can annotate fhes,e or use compiler flags for source files, but I'm not sure you can do that at runtime with gcc yet, while I think it is possible with LLVM, but I have no experience with it Jul 06 18:46:57 npitre: apropos of u-boot, we at pengutronix try to make it better with barebox. At least I like it better Jul 06 18:47:18 I guess mixing arm and thumb2 intelligently may provide some really good results Jul 06 18:47:49 ssvb: BTW I'm not sure you can build ffmpeg with thumb2 Jul 06 18:47:56 but it's not expected to make a big difference anyway Jul 06 18:48:12 ssvb: i think we switched to ARM for ffmpeg due to so much handcoded assembly that it would be hard to fix Jul 06 18:48:51 lool: http://hardwarebug.org/2009/03/25/thumbs-up/ :) Jul 06 18:53:09 eh Jul 06 18:56:59 fwiw, it depends on if it is inline asm or just separate complete functions implemented in arm asm.. Jul 06 18:57:13 it is no problem to call arm functions from thumb code Jul 06 18:57:59 I think it's just a lot of places with assembly which need to be reviewed Jul 06 18:59:12 ahh, gotcha Jul 06 18:59:30 I think some selected applications and libraries just need good performance, so they may be better compiled for arm just for this reason Jul 06 19:00:35 btw, on a related topic.. I recommend -O3 when you are building libfaac.. Jul 06 19:00:50 seemed to help a lot for me.. like 2x+ Jul 06 19:01:17 (the faac code uses a lot of floating point.. not sure if that makes much difference in how effective the compiler optimizations are..) **** ENDING LOGGING AT Wed Jul 07 02:59:56 2010