**** BEGIN LOGGING AT Sat Oct 31 02:59:57 2009 Oct 31 03:35:53 * mwester-laptop kicks OpenPlugLogger Oct 31 04:03:44 mwester-laptop: why you kicking? Oct 31 04:42:43 no more logging lol.. Oct 31 09:33:23 eFfeM: Hey Oct 31 09:34:59 Hmm, the OpenRD kernel seems to have a slight misconfiguration. Oct 31 09:37:31 eFfeM: We dont have enough buildds to build another ARM flavour right now, and we don't have the time to QA it, nor do we have a technical solution to mix and match various armel builds Oct 31 09:38:20 What's this, Ubuntu? Oct 31 09:38:27 eFfeM: I think it's best to use jaunty or debian for armv5 hardware until v7 hardware is more widely available (and it will be); the cool stuff out here is happening on v7; latest iphone, n900 etc. Oct 31 09:38:35 Erant: Yes Oct 31 09:39:25 Ah, right. MVDove should be out soon-ish. Oct 31 09:43:48 lool, hi, didn't expect you here Oct 31 09:45:39 I understand the time issue and as no-one is sponsoring this ... Oct 31 09:46:02 guess plug users should use debian (or openembedded :-) ) Oct 31 09:48:18 lool, guess with an arm9 (which is an armv4T iirc) I will be completely out of luck Oct 31 09:48:29 * eFfeM is looking at a hawkboard (www.hawkboard.org) Oct 31 09:48:35 EFfeM: ARM926EJ-S? ARMv5 Oct 31 09:49:08 omap l138 (ti) Oct 31 09:49:28 ARM926EJ-S Core Oct 31 09:49:51 erant and this was indeed about ubuntu Oct 31 09:50:22 ARM926EJ-S is one of the most prominent ARM9s. And ARMv5te Oct 31 09:51:01 what was the misconfig you had in the openrd kernel? (and which kernel, if it is the OE one I can fix, if it is the marvell one clone the git and build yourself) Oct 31 09:51:16 Yeah, it's sad. ARMv5 and v6/v7 are fairly different, especially as far as caches go. Oct 31 09:51:23 thtanks for the cpu info Oct 31 09:51:52 It's the Marvell one, it was something about IPv4 Oct 31 09:51:54 And cookies Oct 31 09:52:04 found the doc: http://www.arm.com/products/CPUs/ARM926EJ-S.html Oct 31 09:52:34 for marvell try to bother Dhaval (maybe just put it on the mailing list) Oct 31 09:52:41 CONFIG_SYN_COOKIES=y <-- That was originally not set, made Ubuntu barf. Oct 31 09:52:45 guess your chances are best if you include a fix Oct 31 09:52:50 ah ok Oct 31 09:53:06 let me check what the oe kernel has Oct 31 09:53:37 Makes it go: * Setting kernel variables (/etc/sysctl.d/10-network-security.conf)... error: "net.ipv4.tcp_syncookies" is an unknown key Oct 31 09:53:53 Turn it on and: * Setting kernel variables (/etc/sysctl.d/10-network-security.conf)... [ OK ] Oct 31 09:54:12 OE has it set :-) Oct 31 09:55:14 Nice. Does OE have an OpenRD branch? Oct 31 09:57:47 Erant: OE has support for sheevaplug, openrd base and openrd client machines and offers the angstrom distro; I have it running on my sheeva and tried it on openrd, the kernel on my openrd is also an OE kernel Oct 31 09:58:33 note that this is still work in progress client does not have audio or video, basically it is the same as in the git of eInfoChips Oct 31 09:58:43 and base is not tested, I do not have base hw Oct 31 09:58:50 Erant: yes, I have base here Oct 31 09:59:04 I have a base, so I might get around to installing it. Oct 31 09:59:12 $ cat /proc/version Oct 31 09:59:12 Linux version 2.6.32-rc4-g45722c3-dirty (frans@linux-suse) (gcc version 4.2.1) #13 PREEMPT Fri Oct 23 17:19:44 CEST 2009 Oct 31 09:59:36 Currently very happy to just have everything booting off SATA. (NFS booting was no fun) Oct 31 09:59:51 Erant: oe/base support is at 2.6.31, I still need to push the PCIe support patch to OE Oct 31 10:00:01 Very pleased to see uBoot being able to pull my kernel off my SATA/EXT2 as well. Oct 31 10:00:12 likewise: cool Oct 31 10:00:12 likewise: What was wrong with PCIe? Oct 31 10:00:53 Linux version 2.6.32-rc4-g45722c3-dirty (erant@Erants-MacBook-Pro.local) (gcc version 4.4.1 (GCC) ) #2 PREEMPT Sat Oct 31 10:31:55 CET 2009 Oct 31 10:01:09 likewise does that impact defconfig? hrw complained about ext4 missing and so I added it and some other things to defconfig; about to push that one, but if you feel other things should be set let me know (or add it afterwards) Oct 31 10:01:24 Is there anyone else stupid enough to do this shit on MacOSX btw? I didn't find anything. Oct 31 10:01:42 Might have to document some stuff, like building your own toolchain, finding mkimage and shit. Oct 31 10:01:44 Erant: thinking of it, I think this kernel is not the oe kernel (i'm juggling with too many systems atm) Oct 31 10:02:16 booting the sheeva :-) Oct 31 10:02:16 Heh Oct 31 10:03:36 How many docs has Marvell released to the public btw? Oct 31 10:05:04 ok this is what the sheeva is running; fully angstrom: Linux version 2.6.32-rc4 (frans@linux-suse) (gcc version 4.3.3 (GCC) ) #1 PREEMPT Sat Oct 24 21:38:26 CEST 2009 Oct 31 10:05:33 Erant: not too much, what is on the cd and what is on the wiki Oct 31 10:05:38 Aight. Close to what I'm running. Oct 31 10:05:56 eFfeM: Alright, so nothing like the Functional Spec? Oct 31 10:08:33 I have some of these under armwrenching NDA, so I was just wondering if anything had changed. Guess not :P Oct 31 10:11:18 Erant: I know the average Marvell NDA, you have my sympathies :-) Oct 31 10:11:36 Yeah, I'm down an arm, a leg and a soul. Oct 31 10:11:57 actually I think sheeva is something moving a lot a more open direction Oct 31 10:12:04 it used to be much worse Oct 31 10:23:32 So how are people doing development for these things, Marvell HAL? Oct 31 10:40:16 not ? Oct 31 10:42:48 Right. Oct 31 10:44:03 i've never played with marvell hal Oct 31 10:44:36 Well, I don't see any need to do deep down development just yet. I'll be playing some with PCIe though, and that might require me to dive slightly deeper. Oct 31 10:46:23 basically i have two approaches: if I need it have the info for work because we are aiming at the product with it then the chip proivder should provide the needed info. if not most likely we will select a different vendor; and if it is info i need privately and it is not there, I move to a different (read more open) platform. Oct 31 10:46:49 btw for sheeva I have no problems, it runs all I want to run on it and all seems pretty standard Oct 31 10:47:46 openrd is somewhat different and I am actually slightly irritated on the slow support when it comes to adding drivers for some of the hardware (in combination with the lack of doc) Oct 31 10:48:49 What drivers are lacking right now? Oct 31 10:49:16 at least audio and video Oct 31 10:49:50 Ahright. Yeah, I was wanting to try 'n get an HD4350 radeon chipset working on this thing. Oct 31 10:49:56 and there does not seem to be any doc of the Z11 chip either Oct 31 10:50:30 4350 would be nice, do you have one ? guess there is a linux driver for that one Oct 31 10:50:38 I need to work out if I need to change the powersupply though. GlobalScale gave me a 24W powersupply. Oct 31 10:50:59 Not yet, but RadeonHD and Radeon drivers both support it (And work on ARM. Debian has a package for it even) Oct 31 10:51:36 cool Oct 31 10:51:54 So I'll probably spend today bootstrapping a debian system. Oct 31 10:52:20 Another interesting thing, as, like I said, I'm probably the only idiot stupid enough to try this on a Mac. Oct 31 10:53:35 So far I've only been unable to build u-boot, but I don't really need that anyway. Current u-boot does all I want. Oct 31 10:54:22 i haven't tried u-boot either Oct 31 10:54:53 and probably i'm the only idiot stupid enough to try this on open-suse 11.2 prerelease Oct 31 10:54:55 :-) Oct 31 10:54:57 :D Oct 31 10:55:31 Well, at least you could use the dev environment that was on the DVD? I've had to rebuild from scratch :/ Gives me the latest GCC though, which I always try to get. Oct 31 11:12:30 yeah saw you gusesd gcc 4.4., some people state it generates larger code, haven't compared myself Oct 31 11:12:57 I'll start caring about larger codesize when I'm working on a platform with 4kb RAM. Oct 31 11:13:17 (btw some of the things I do block the chat window, hence the slow response Oct 31 11:13:34 more codesize often also means more instructions so slower code Oct 31 11:14:18 Not always true. Loop unrolling for example generates faster code, generally. (It gets complex when you factor in the cache misses) Oct 31 11:15:52 true **** ENDING LOGGING AT Sat Oct 31 11:39:06 2009 **** BEGIN LOGGING AT Sat Oct 31 11:39:30 2009 Oct 31 12:54:27 Hmm, that was fairly trivial. I could debootstrap a Debian squeeze install from the Ubuntu jaunty install I had. Oct 31 12:55:03 Erant: onto the NAND JFFS2 filesystem? Oct 31 12:55:16 twb: I'm not using the NAND at all. Oct 31 12:55:26 Well. Only for u-boot Oct 31 12:55:28 I had to install onto an external USB key, which is sucky. Oct 31 12:55:46 I installed onto a 250GB SATA drive. :P Oct 31 12:55:53 Bleh, moving parts Oct 31 12:56:26 I'd have used an SSD if I had one lying around, or wern't twice as much as the OpenRD itself. :P Oct 31 13:06:47 is there a package which delivers QtOpenGL (i must say I do not fully understand the variants in the qt recipes) Oct 31 13:08:18 libqt4-opengl - Qt 4 OpenGL module? Oct 31 13:08:35 oops wrong channel, this was for #oe Oct 31 13:09:03 :) Oct 31 13:15:46 anyone ever attempted to run netbsd on a sheeva ? Oct 31 13:16:41 Doesn't netbsd run on everything? Oct 31 13:32:02 ansx: Afair, a fair few Sheeva/Feroceon based routers run NetBSD Oct 31 13:34:45 yeah but i can't find any info on google about how to run the kernel from uboot Oct 31 13:36:50 ansx: What happens when you just wrap the kernel in a uImage container? Oct 31 14:35:46 I didn't tried yet Oct 31 14:36:30 I don't know bsd enough to be able to debug it in case it got some issues with the sheeva Oct 31 14:40:10 ansx: the same way you boot freebsd ? Oct 31 14:40:22 with the kernel on a tftpd Oct 31 14:41:02 by defaut uboot doesn't support ufs, so you're not going to boot from flash/sd/usb Oct 31 14:41:44 you might be able to put the kernel on a ext3 partition Oct 31 14:43:54 yeah I was thinking about it Oct 31 14:44:10 or maybe flash the kernel to the nand with the uimage thing Oct 31 14:44:48 you're not going to be able to put / on flash in any case Oct 31 14:45:35 freebsd has no 'jffs2' driver, i don't think any bsd has one Oct 31 14:46:06 freebsd doesn't have a driver for the SD either Oct 31 15:46:32 unrealMac: What, like JFFS2 is the only filesystem that works on NAND? Oct 31 15:55:48 well what about ubifs? Is that supported by freebsd? Oct 31 16:11:03 it's software. everything is supported if you want it bad enough Oct 31 16:27:14 WARNING: at arch/arm/kernel/process.c:171 cpu_idle+0x78/0xc0() <-- wtf is this? Oct 31 16:30:55 a bug Oct 31 16:32:01 Ya think. Oct 31 17:40:24 Erant: what i meant, was that there's no nand fs for freebsd Oct 31 18:48:54 eFfeM: Well it will take a rebuild of Ubuntu for it to work on your platform, yes Oct 31 18:49:07 eFfeM: Debian is probably the easiest to get working on ARMv4 in this case Oct 31 18:53:23 /dev/sda3 on /var type ext4 (rw) (wewt) Oct 31 18:53:52 I love it when everything 'just works'. Minus a kernel recompile, but mjeh. Oct 31 18:54:21 in a sheeva, is the nand faster so slower than sdhc ? Oct 31 18:56:26 s/so/or Oct 31 18:56:45 faster Oct 31 18:57:24 anything is faster than most sdhcs Oct 31 18:57:29 =) Oct 31 18:57:47 OpenRD:/# dd if=/dev/mtdblock1 of=/dev/null bs=2k Oct 31 18:57:47 2048+0 records in Oct 31 18:57:47 2048+0 records out Oct 31 18:57:47 4194304 bytes (4.2 MB) copied, 1.01887 s, 4.1 MB/s Oct 31 18:58:07 try a larger bs ? Oct 31 18:59:34 OpenRD:/# dd if=/dev/mtd2ro of=/dev/null bs=128k Oct 31 18:59:34 ^C280+0 records in Oct 31 18:59:34 279+0 records out Oct 31 18:59:34 36569088 bytes (37 MB) copied, 7.12247 s, 5.1 MB/s Oct 31 19:00:09 that's no so good Oct 31 19:00:43 this on Linux OpenRD 2.6.32-rc4-g45722c3-dirty #3 PREEMPT Sat Oct 31 19:32:01 CET 2009 armv5tel GNU/Linux Oct 31 19:01:24 As a comparison: Oct 31 19:01:25 OpenRD:/# dd if=/dev/sda1 of=/dev/null bs=128k Oct 31 19:01:25 ^C4251+0 records in Oct 31 19:01:26 4250+0 records out Oct 31 19:01:26 557056000 bytes (557 MB) copied, 7.07576 s, 78.7 MB/s Oct 31 19:01:50 (Just to show that it's not something funky with my kernel) Oct 31 19:01:51 that's not a sdhc ;) Oct 31 19:02:10 Obv. not. That's a regular SATA disk. :P Oct 31 19:08:33 so yeah, I confirm the nand performance is pretty bad Oct 31 19:09:13 strange. I hadn't checked before, but read on the forums someone saying the performance was really good Oct 31 19:09:40 I'm wondering where to install my stuff Oct 31 19:10:04 the slower is ok for kernel and storage, but not for applications Oct 31 19:13:00 ansx: Have you considered installing a SATA disk? Oct 31 19:13:31 Erant: on a sheevaplug? Oct 31 19:13:47 armin76: OpenRD Oct 31 19:14:07 I'd like to use sdhc + nand only for main OS Oct 31 19:14:10 did he mention he has an openrd? Oct 31 19:14:19 and have a sata hdd as storage (which won't be mandatory then) Oct 31 19:14:40 armin76: Nope. Oct 31 19:15:02 and I got a sheeva :p Oct 31 19:15:11 then no sata :) Oct 31 19:15:29 with an usb adapter it will be ok Oct 31 19:15:29 There is, just not connected to an actual SATA connector. :P Oct 31 19:19:38 ansx: depends on the sdhc you have (aka the class) Oct 31 19:19:45 class 2 = 2MB/s Oct 31 19:19:57 class 4 = 4MB/s and so on Oct 31 20:01:18 ok Oct 31 21:13:56 Erant: were those nand measurements using jffs or ubifs ? Oct 31 21:17:12 tinker-f595: Ehm. You did see what I used as a commandline, right? Oct 31 21:47:04 did not pay much attention to other than the speed mesaurement Oct 31 21:47:58 It's raw device speed Nov 01 01:10:47 what happened to me voice? **** ENDING LOGGING AT Sun Nov 01 02:59:56 2009