| === ScriptRipper_ is now known as ScriptRipper | ||
| === WyrM is now known as WM | ||
| === WM is now known as drwyrm | ||
| === drwyrm is now known as WyrM | ||
| === ApOgEE- is now known as ApOgEE | ||
| ogra | cbrake, looks like your host system is missing proper timezone configuration (the script tries to determine your TZ from /etc/timezone of the host) | 10:48 |
|---|---|---|
| ogra | cbrake, as a quick fix run: sudo dpkg-reconfigure tzdata ... before you run the script | 10:50 |
| cbrake | ogra: yup, that worked -- thanks | 14:15 |
| cbrake | is there a standard way for making modifications to an embedded Ubuntu rootfs? | 17:44 |
| cbrake | I need to add some files for serial console, and set up the default networking | 17:45 |
| ogra | untar it, make your changes, tar it again ;) | 17:45 |
| cbrake | obviously, I can hack these files into the tarball, but wondering how support for various macnhines is done | 17:45 |
| ogra | as long as its config changes that will work just fine | 17:45 |
| persia | cbrake, As much as possible, we attempt to make the system able to handle all machines. | 17:45 |
| persia | So, for instance, the kernel and udev should be populating the devices you need, and you should be able to pass kernel command-line options for the serial console. | 17:46 |
| persia | Networking is a bit trickier: mac-based DHCP is a handy way to go. | 17:46 |
| cbrake | persia: I can get kernel messages out the serial port with kernel command line, but it does not start a getty on the serial port unless I do https://help.ubuntu.com/community/SerialConsoleHowto | 17:47 |
| ogra | yeah | 17:48 |
| persia | Hrm. | 17:48 |
| ogra | the rootfs build script only rolls a rootfs and adds a set of defaults ubiquity would add otherwise | 17:48 |
| persia | I wonder if there's a sane model for having that enableable on boot. Might be handy for server folk as well. | 17:48 |
| ogra | there isnt without creating the file in event.d | 17:49 |
| ogra | i could add a switch to the script, but that still leaves the serial options | 17:49 |
| ogra | i dont want to crowd the script with to many options | 17:49 |
| ogra | i.e. serial speed etc that should be set in the ttyS0 file | 17:49 |
| persia | Oh, right. | 17:49 |
| persia | I was sorta-kinda-hoping there would be some way to feed it without hardcoding, for general deployment. | 17:50 |
| ogra | but i surely can add a switch that at least creates it with a bunch of defaults | 17:50 |
| cbrake | I'm coming from the Openembedded world where the concept of machine support is very prominant, so trying to get re-aligned ... | 17:50 |
| ogra | yeah, ubuntu is rather focusing on "as broad as possible" support | 17:51 |
| persia | cbrake, Yeah, it's a different way of looking at things. We tend to focus on having a (mostly) general solution, rather than lots of machine-specific versions. | 17:51 |
| ogra | which indeed leaves off a lot of optimization | 17:51 |
| ogra | but gets nearly anyone the option to run it | 17:52 |
| ogra | (as long as your arch matches the minimal reqs ... i.e. ARMv5 in jaunty, ARMv6+ in karmic etc) | 17:52 |
| cbrake | interesting, so ARMv5 is only for jaunty | 17:53 |
| ogra | yeah, karmioc will do v6 ... likely vfp and NEON | 17:53 |
| cbrake | my default installs are running at 380MB -- is there any way to trim that down? | 17:53 |
| ogra | well, its not OE :) | 17:54 |
| cbrake | nod :-) | 17:54 |
| ogra | you can cut down by ripping out packages indeed | 17:55 |
| cbrake | so is that typically done by apt-get remove after the system is booted? | 17:55 |
| cbrake | or is there a way to remove packages when generating the rootfs? | 17:55 |
| ogra | no | 17:55 |
| ogra | after booting | 17:55 |
| cbrake | ok, thanks for all the help | 17:56 |
| ogra | unless you build a qemu image first, trim that, loop mount it and tar it up | 17:56 |
| ogra | the prob is that you cant just chroot into the rootfs ... you need a vm | 17:56 |
| cbrake | yes, qemu is pretty useful for that vs the target machnine | 17:56 |
| ogra | but the script has an option to build a qemu only image | 17:56 |
| persia | Well, unless you're building the rootfs on an armel system :) | 17:57 |
| ogra | which you then can boot in qemu and trim | 17:57 |
| ogra | oh, indeed | 17:57 |
| ogra | if you have powerful arm HW thats indeed the quicker option | 17:57 |
| persia | well, you might have to go through the rootfs game to bootstrap, but once there... | 17:58 |
| ogra | indeed ... you wouldnt use the script at all then | 17:59 |
| persia | Why not? Wouldn't the script speed the bootstrap? | 18:02 |
| ogra | it would still run qemu | 18:04 |
| persia | Well, yeah, but you have to do the initial boot somehow, unless debootstrap just works on the BSP. | 18:05 |
| ogra | oh, right | 18:06 |
| Martyn | re | 18:36 |
| Martyn | ogra : have you ever had this error pop up? (gcc 4.2) LD init/built-in.o | 18:37 |
| Martyn | LD .tmp_vmlinux1 | 18:37 |
| Martyn | `.cpuexit.text' referenced in section `.ARM.exidx.cpuexit.text' of arch/arm/mach-realview/built-in.o: defined in discarded section `.cpuexit.text' of arch/arm/mach-realview/built-in.o | 18:37 |
| ogra | nope | 18:37 |
| persia | Martyn, You mentioned yesterday that you found a power issue with idle_pm vs. kernel_idle. I just wanted to poke you to file a bug about that (as I hear the kernel *should* do the right thing with kernel_idle) | 18:37 |
| Martyn | persia : As soon as we're done with the tests (Fri) I'll do so. | 18:39 |
| persia | That soon? Excellent! Thanks. | 18:39 |
| Martyn | persia : More things have been discovered since yesterday, that may be errata in the i.mx51 and RealView_Cortex_A9 | 18:39 |
| persia | I always like it when it's a hardware problem :) | 18:39 |
| Martyn | it's a very high priority now | 18:39 |
| Martyn | well, it's more than a hardware problem. If its a fundamental design issue in the SoC, it's something my company will have to carefully watch for as we create ours. | 18:40 |
| Martyn | power domain issues are so.. fucking.. finicky | 18:40 |
| Martyn | ogra : Thoughts on booting SMP -- right now, I'm having a whale of a time getting the Cortex-A9MP kernel to go SMP, because for some reason ARM expects the bootloader (u-boot in this case) to hold the non-main cores in WFI (I use WFINE) | 18:41 |
| Martyn | ogra : however, the other cores never get the software interrupt they need to get going . and even THEN .. the WFI resumes right where it left off .. in the bootloader. | 18:42 |
| Martyn | I think it's the kernel itself that should hold the other processors back in WFI .. right? | 18:42 |
| Martyn | so that when it restarts the PC's of the other cores are pointed at kernel code .. not bootloader code. | 18:42 |
| ogra | hmm, amitk might be able to tell | 18:43 |
| Martyn | *nod8 | 18:44 |
| Martyn | My boss made nosies at ARM about getting us a babbage2 | 18:44 |
| Martyn | ogra : Question though .. in order to unify the Ubuntu/ARM universe, are we standardising around RedBoot? | 18:45 |
| Martyn | or was RedBoot just convenient for the Babbage? | 18:45 |
| Martyn | OR .. are we shifting to UEFI? | 18:45 |
| === cbrake is now known as cbrake_away | ||
| === Nicke_ is now known as Nicke | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!