=== XorA is now known as XorA|gone === rsalveti_ is now known as rsalveti [18:51] Are you familiar with bugs like this on omap: [ 2361.163635] BUG: soft lockup - CPU#0 stuck for 61s! [dpkg:1737] [19:03] nosse1: I've seen that error reported for all sorts of architectures. Unfortunately, I don't know the cause. I'm fairly certain it's not OMAP-specific though. [19:04] persia, thanks. I'd just wanted to check if it was something familiar on omap. [19:04] what's the error? [19:04] sorry, just popped in, so I don't have the backscroll [19:05] [ 2361.163635] BUG: soft lockup - CPU#0 stuck for 61s! [dpkg:1737] [19:05] nosse1: which kernel are you urins? [19:05] BUG: soft lockup - CPU#0 stuck for 61s! [19:05] *using [19:05] nosse1: the TI one? [19:05] I'm running the TI-one yes, the "official" lucid ti-omap won't work for AM3517 [19:06] It could seem to me that the lockup occurs on heavy disk activity. And since I'm running NFS root, it could perhaps be related to that? [19:09] could be, but hard to say without more information from the kernel [19:09] Yeah, I know. Except the kernel doesn't tell any more [19:10] Another thing: I have X up and running (yeah!). [19:10] Martyn: did i ask you if you had sound running on the tegra? [19:11] What apt target do you guys use when testing? "ubuntu-netbook" has unknown packages which doesn't seem to be available yet [19:11] Do you have a all-in-one target which pulls down the gnome desktop? [19:12] That'd be ubuntu-desktop, but ubuntu-netbook is significantly lighter-weight in terms of requirements. [19:12] Which bits of ubuntu-netbook aren't installing? [19:12] persia, hold on [19:15] persia, cancel that. It seems all packages are available now [19:16] There's sometimes architecture skew with the mix of arch:all and arch:any binaries (because the arch:all binaries are built on i386), but it doesn't typically last that long, unless something broke (which means it needs fixing). [19:16] If you get a persistent issue with metapackage installation, it may well be a bug needing fixing. [19:16] (but I'm glad to hear this isn't the case right now) [19:17] That said, according to https://launchpad.net/builders there are 19 packages that are out-of-date on armel (waiting to build). [19:19] Are there any commercial developers here in this forum? (perhaps not on sundays) [19:19] I'm uncertain if we should register our products on the ARM machine registry or not [19:20] and I'd like to know what other mfgs are doing [19:21] armin76 : Since I'm very server-oriented, I haven't even tried to get sound working on the tegra [19:21] armin76 : I've even disabled video, to get the memory back [19:21] nosse1: Do you need a special MachineID for your hardware? If so, you'll want one. If not, there's little point. [19:22] nosse1 : I'm a 'comercial developer' [19:22] Martyn: nod :) [19:23] persia, I'm not sure i need one. You use machineid whenever the kernel isn't aware of the HW its running on, right? If that is true, I guess we don't need one [19:24] I don't really understand when it's needed, but I think it's part of kernel compilation: if you can run a multi-board kernel, I think you don't need it. [19:40] Alas, it seems NFS root is the culprit of the kernel crashes [19:41] * nosse1 wish that the AM3517 had a SATA disk connector... [19:49] Anything designed to have disks should have such a controller, but MTD rules when it comes to local data. [20:16] persia, do you mainly use USB or SD MTDs? [20:16] I've never used an MTD that wasn't on-board. [20:17] I'm fairly certain that neither USB nor SD allows one to actually use the MTD drivers. === Meizirkki_ is now known as Meizirkki === hrw|gone is now known as hrw [20:29] persia, you need a fairly large mtd if you want to do development on it. E.g. for building packages (since Ubuntu builds natively). Do you know how it's done in the build farm? [20:29] morning [20:30] (I'm somewhat bothered here having to reboot my target whenever I do a larger apt install) [20:31] nosse1: I'll say that an MTD is inappropriate for doing development: for that you want spinning disks. [20:31] yes [20:31] I believe the build farm uses SATA-over-USB but I may be mistaken. [20:31] connecting 1.5TB disk over usb is easy [20:31] That's probably overkill :) [20:31] Ah. [20:32] persia: I have such one connected to sheevaplug [20:32] I'd rather use an esata-bearing sheeva. =รพ [20:32] or rather had as I have some problems with sheeva so disk is connected to desktop [20:32] Because we need to make one ourselves unless we somehow can build packages natively on our large build farm [20:32] DanaG: I bought sheevaplug year ago when it was new stuff [20:33] nosse1: connect sata-over-usb and pray that TI ehci will work [20:34] So it's about praying for one or the other. One=NFS, Other=ehci.... NFS crashed once more *sigh* [20:34] ehci is probably more stable than NFS in most ways. [20:34] nosse1: 16GB sd is other option [20:34] I remember that NFS was discussed for the LP build farm and rejected, but I don't remember why. [20:35] hrw: If you like buying new SD cards :) (but 4-8G is enough for most packages) [20:35] nosse1: quite cheap, easy to buy and omap3 can drive 3 mmc/sd slots [20:35] I have here device with omap3 and 3 sd slots. mmc0=rootfs, mmc1=userfs, mmc2=sdio/wifi/bt [20:35] How's wear levelling handled on SD? Is it the SD firmware, or is it up to the host SW? [20:36] SD itself [20:36] ...and from what I've heard it spans from excellent to horrible [20:37] yes [20:37] Reliable brands: SanDisk, Kingston ? [20:38] * nosse1 needs to shop SD cards tomorrow [20:38] sandisk [20:38] avoid kingston rather [20:39] hmm, where can you get a cheap sdio wifi? [20:40] I wanna' try one in my host computer's sdhc controller. [20:40] DanaG: no idea [20:40] DanaG: here I have marvel8686 in two devices - both on sdio [20:40] and 3rd one has it on spi bus [20:40] Something weird with the Ricoh reader in Windows: it assumes the whole disk is one partition (even if it's not true)... and then can't comprehend what partition type. [20:41] common [20:41] I end up having to use "rw-everything" (awesome tool, by the way) to essentially "setpci" as I would have to on Linux... and then it reattaches to the SDHC controller instead of the Ricoh controller. [20:41] brand new cards sometimes do not have partition table [20:41] http://intr.overt.org/blog/?p=59 [20:42] The card I was testing with, was my beagle (two partitions). [20:43] Once I did that, it then showed partitions properly. [20:43] I'm curious to try some SDIO card with the thing. [20:43] DanaG: nice [20:45] I did the writing value on the mmc controller function, specifically. [20:56] but - is each sd controller also sdio capable? [20:59] hrw, After skimming the RM from TI, it seems so [21:00] nosse1: I know that i.mx31, omap3, pxa2xx are sdio capable. no idea about others === kevin- is now known as styol === hrw is now known as hrw|gone