=== mcasadevall is now known as NCommander | ||
=== sonicmctails is now known as mcasadvall | ||
=== ogra__ is now known as ogra_ | ||
shaw | mcasadevall: hi! | 20:51 |
---|---|---|
mcasadevall | hi shaw | 20:51 |
mcasadevall | suihkulokki, ping | 20:52 |
shaw | mcasadevall: I'd like to join the arm porting team so I can do some ARM builds in my PPA. | 20:56 |
mcasadevall | shaw, ARM enabled PPAs are unfortunately restricted to Canonical staff at this time | 20:56 |
mcasadevall | (Xen isn't available for ARM) | 20:56 |
shaw | mcasadevall: oh. dang! I fall asleep every time I do local builds. ;) | 20:57 |
mcasadevall | shaw, :-) | 20:57 |
Martyn | mcasadevall : Um, Xen _is_ available for ARM | 21:36 |
Martyn | mcasadevall : I'm working with it this very second even. | 21:36 |
mcasadevall | Martyn, rephrase | 21:36 |
mcasadevall | Martyn, its not available for what we're using in the data center :-) | 21:36 |
mcasadevall | Martyn, and if you got a link, I'd be greatly interested. | 21:37 |
Martyn | http://wiki.xensource.com/xenwiki/XenARM | 21:37 |
Martyn | enjoy | 21:37 |
Martyn | I'm currently working on getting it limping along on a cortex-A9 ( FastModel sim ) | 21:38 |
mcasadevall | Martyn, we have iMX.51 boards, not 21 | 21:38 |
* mcasadevall however has a 21 board ... | 21:38 | |
Martyn | it works in Babbage. | 21:38 |
mcasadevall | Martyn, You have a Cortex-A9? | 21:38 |
mcasadevall | *drool* | 21:38 |
Martyn | Yep. | 21:38 |
Martyn | Quad core | 21:38 |
mcasadevall | .... | 21:38 |
* mcasadevall stares | 21:38 | |
mcasadevall | I'll trade you my SMP ia64 for it :-)? | 21:38 |
Martyn | runs slow (125Mhz, not too speedy) but it works in FastModels nicely | 21:38 |
Martyn | we also have the even SLOWER (but real silicon) PDX .. runs at all of 70Mhz and has two cores | 21:39 |
mcasadevall | Martyn, actually, if you could do me a favor, could you dump /proc/cpuinfo and shoot me an email with it | 21:39 |
mcasadevall | I'd like an example of it when on SMP systems for when I get around and write a landscape stub for ARM | 21:39 |
Martyn | mcasadevall : let me check if that violates the NDA | 21:39 |
Martyn | (I don't think it does, because you can synthesize that output from the arm kernel tree) | 21:39 |
Martyn | email? | 21:43 |
mcasadevall | Martyn, michael.casadevall@canonical.com | 21:44 |
mcasadevall | Martyn, I've not been able to find an SMP ARM system to dump its cpuinfo, so this is really handy :-) | 21:44 |
Martyn | sending the damned keymap takes for bloody ever at 125Mhz | 21:46 |
Martyn | There's seriously no reason this part of the boot should take so long | 21:46 |
mcasadevall | Martyn, hardware issue? | 21:46 |
Martyn | No, don't think so | 21:47 |
Martyn | it just takes a while, then kicks over to the event manager and drivers as it should | 21:47 |
Martyn | but keymap takes >2min | 21:47 |
mcasadevall | Martyn, sounds like my IA64 while its still in EFI | 21:48 |
mcasadevall | Martyn, how difficult is it to get Xen going? (and how much kernel customization is needed) | 21:49 |
mcasadevall | Since a .21 kernel won't boot on a Babbage 1 | 21:50 |
Martyn | There you go | 21:51 |
Martyn | Plenty | 21:51 |
Martyn | I have to get Xen working on the multicore, which is a real PITA | 21:51 |
Martyn | because getting the hypervisor to abstract all the cores is a pain | 21:51 |
mcasadevall | Martyn, *groan* | 21:51 |
mcasadevall | Martyn, so much for easily recycling the existing build hardware in the DC | 21:51 |
Martyn | Why? | 21:51 |
kblin | evening folks | 21:52 |
Martyn | You can build mx31/mx51 on the old hardware .. all it takes is a x-compiler | 21:52 |
mcasadevall | Martyn, we're using non-FSL hardware in the DC | 21:52 |
Martyn | what's the hardware in use? | 21:52 |
kblin | you people build all your packages native, right? | 21:52 |
mcasadevall | Marvell | 21:52 |
mcasadevall | kbingham, yeah | 21:52 |
mcasadevall | er, kblin | 21:52 |
mcasadevall | Martyn, I forget the SoC off the top of my head | 21:52 |
Martyn | Not the shivaplug thing? | 21:52 |
mcasadevall | Martyn, I think it was the dev boards that became the shivaplug | 21:53 |
kblin | mcasadevall: ok, so I'm doing the "correct thing" atm :) | 21:53 |
Martyn | Gah | 21:53 |
mcasadevall | Martyn, yeah | 21:53 |
Martyn | The 'kirkwood' | 21:53 |
Martyn | piece of crap SoC | 21:53 |
mcasadevall | No | 21:53 |
mcasadevall | Not kirkwood | 21:53 |
mcasadevall | Hold on | 21:53 |
mcasadevall | I got cpuinfo somewhere | 21:53 |
mcasadevall | Processor: Feroceon rev 0 (v5l) | 21:54 |
mcasadevall | Hardware: Marvell DB-78x00-BP Development Board | 21:54 |
Martyn | Feroceon .. hmmm. | 21:54 |
mcasadevall | Martyn, wow, the SMP cpuinfo is different then anything else I've ever seen. | 21:54 |
mcasadevall | BogoMIPS: 999.42 - they are sexy :-) | 21:54 |
Martyn | bogomips are totally bogus :) | 21:54 |
mcasadevall | True | 21:54 |
Martyn | It''s just a timing loop | 21:54 |
mcasadevall | I think they're either 1Ghz or 1.2Ghz boards | 21:55 |
Martyn | the 124 bmips of the simulated Cortex-A9 doesn't really reflect anything either | 21:55 |
mcasadevall | They're faster than the Babbage, but limited to ARMv5TE+VFP | 21:55 |
Martyn | Yep | 21:55 |
Martyn | they are based on Armv5's | 21:55 |
mcasadevall | Babbage is 600 or 800Mhz if memory serves | 21:55 |
suihkulokki | mv78x00 is similar to kirkwood, up's it with dual-issue core, 64bit 800Mhz ddr2 memory channel and vfp | 22:00 |
mcasadevall | suihkulokki, we've seen some oddness with timing issues on that SoC :-/ | 22:05 |
mcasadevall | https://bugs.edge.launchpad.net/ubuntu/+source/libipc-sharelite-perl/+bug/299847/ | 22:06 |
ubot4 | Launchpad bug 299847 in libipc-sharelite-perl "Shared memory operations on very fast ARM hardware suffer from non-atomic operations and race conditions." [Low,Confirmed] | 22:06 |
Martyn | re | 22:11 |
Martyn | 600 | 22:11 |
Martyn | [ 15.758593] Brought up 4 CPUs | 22:13 |
Martyn | [ 15.758654] SMP: Total of 4 processors activated (498.07 BogoMIPS). | 22:13 |
Martyn | Mmmmmm | 22:13 |
lool | Is there a generic uboot MMC command for init/listing parts etc.? On EVM it's mmcinit / fatls, but on Marvell Uboot that doesn't seem to exist | 22:29 |
=== dyfet_ is now known as dyfet | ||
lool | Hmm it doesn't see my USB mass storage | 22:36 |
* Martyn muh-hu-has | 22:38 | |
mcasadevall | lool, help should have the command list | 22:38 |
lool | It does | 22:38 |
mcasadevall | lool, do we have source yet? | 22:38 |
lool | No | 22:38 |
lool | I'm looking at the sheeva marvell tree | 22:38 |
lool | But it doesn't really relate | 22:38 |
mcasadevall | oh | 22:38 |
lool | On my uboot, "ide" actually manages sata (no sata commands) | 22:39 |
Martyn | source for which? u-boot? | 22:47 |
lool | Yup | 22:47 |
lool | Hmm I think the command is mtdparts | 22:47 |
lool | It's hard to tell whether my SD card is in the slot though; I think the slot is made for thiner cards, sigh | 22:47 |
mcasadevall | lool, probably is an MMC slot | 22:52 |
mcasadevall | Vs. an actual SD slot | 22:52 |
lool | Yeah, my phone has the same thing | 22:54 |
lool | Except it's for mini SDs, so I don't have any single MMC I can try, sigh | 22:55 |
mcasadevall | lool, I haven't see an actual MMC in years :-/ | 23:05 |
mcasadevall | If we need them, that's going to be a massive PITA to locate | 23:05 |
lool | Same here | 23:08 |
lool | Nearest store doesn't have them, usual online website neither | 23:08 |
lool | I found a bunch at some amazon partners | 23:09 |
mcasadevall | lool, ugh, I *really* hope we don't need MMCs | 23:09 |
lool | We don't "need" them, but the board uses that | 23:09 |
mcasadevall | Oh right | 23:09 |
mcasadevall | this isn't the babbage and its fun SD slot :-) | 23:09 |
lool | Until it sees USB devices, we're going to need tftp, or sata, or mmc for installs | 23:10 |
mcasadevall | woo :-/ | 23:13 |
lool | Well ext2fs always returns blank, doesn't even fail, so I give up there too | 23:24 |
lool | Only tftp seems to work | 23:24 |
lool | Blah | 23:24 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!