**** BEGIN LOGGING AT Tue Dec 04 02:59:59 2012 Dec 04 10:47:11 hi bluelightning Dec 04 10:48:15 hi ant_work Dec 04 10:48:22 now that koen has committed pls pull the meta-handheld kernel patches Dec 04 10:48:47 ant_work: ok, will do at lunchtime Dec 04 10:49:09 ah, iirc there is another one of dromede floating Dec 04 10:52:01 ah, yes, 39165 Dec 04 10:52:04 ok, thx Dec 04 13:23:21 hi channel Dec 04 13:23:34 android on Cxx0 models, what do you guys think about that? Dec 04 13:24:03 i was recently contacted by a guy who did this: Dec 04 13:24:07 http://www.xda-developers.com/android/htc-universal-gets-rom-with-cyanogenmod-7/ Dec 04 13:24:11 slow? :) Dec 04 13:24:26 he wants to port ZGrom to the Universal Dec 04 13:24:54 this will happen, Cxx0 models and the Universal are very similar Dec 04 13:24:56 shouldn't be too tricky I would have though Dec 04 13:24:58 thought Dec 04 13:25:40 that guy said that donut (android 1.6) runs fairly weel on his Universal Dec 04 13:25:45 so at least that is doable Dec 04 13:25:59 the biggest problem is the lack of ram Dec 04 13:26:48 so my inital approach would be to start with 2.3 and just butcher it as much as i can to reduce size and memory footprint Dec 04 13:27:27 from my point of view, all we need is the ability to run a certain portion of the android app ecosystem Dec 04 13:27:54 so basically, we need just a barebone built with dalvik, nothing else Dec 04 13:28:01 *build Dec 04 13:35:19 sounds like an interesting experiment :) Dec 04 13:35:29 ant_work: you just sent two patches right? Dec 04 13:35:39 ant_work: plus the one from dromede_ Dec 04 13:36:11 hi Dec 04 13:36:17 bluelightning: right Dec 04 13:36:23 ok, just checking Dec 04 13:36:32 l-k_3.4 following this night/tomorrow Dec 04 13:37:17 I'm extracting the 'kexecboot' specific fragments Dec 04 13:37:35 dromede_: there was Omegamoon playing with Android Dec 04 13:37:50 http://www.omegamoon.com/blog/ Dec 04 13:41:54 ant_work: ping Dec 04 13:42:19 I am trying to use the DTB file with KEXEC/KDUMP on ARM cortex-A15 Dec 04 13:42:50 hi, yes, I've seen on the ML Dec 04 13:43:10 omegamoon - cortez, did have donut running on his spitz Dec 04 13:43:18 but that was a standard build Dec 04 13:44:13 ant_work: thanks Dec 04 13:44:45 nareshbhat: I've read you cannot append cmdline if you're using dtb :/ Dec 04 13:45:31 Ah..ok..But I have tried the same with --cmdline option too Dec 04 13:47:10 we use --command-line in kexecboot Dec 04 13:47:47 honestly I have never tried with devicetree/dtb kernels Dec 04 13:47:51 agree, my typo mistake..I will confirm once again Dec 04 13:48:20 ok, I can see patches pushed couple of months back Dec 04 13:48:29 under kexec-tools Dec 04 13:49:49 for me it looks like very much bleeding edge..not sure anybody tested it :) Dec 04 13:50:44 I've tested master recently on armv5 Dec 04 13:51:34 but nowadays it looks like --atags is needed Dec 04 13:51:42 nareshbhat: check this old log Dec 04 13:51:44 http://de.pastebin.ca/raw/1714957 Dec 04 13:51:50 at the end Dec 04 13:52:27 try to pass console there Dec 04 13:53:35 not ethat kexecboot appends here the commandline specified in boot.cfg Dec 04 13:54:03 i.e. --command-line=root=/dev/mmcblk0p1 rootfstype=ext2 rootwait mtdparts=sharpsl-nand:7168k(smf),99328k(root),-(home) Dec 04 13:54:07 ^^ this is kexecboot Dec 04 13:54:23 console=ttyS0,115200n8 console=tty1 noinitrd debug Dec 04 13:54:29 ^^ this comes from boot.cfg Dec 04 13:54:53 ok thanks I will try with the provided combination Dec 04 13:57:04 nareshbhat: another thing is, recent kernels allow to 'merge' commandline passed by booltloader while once it was only possible to use it or override with static cmdline Dec 04 13:57:09 this for arm Dec 04 16:52:22 Jay7: I see a lot of commits touching fb.c :/ https://github.com/zeldin/kexecboot/network Dec 04 16:52:49 and afaik there are more patches in your e-mail Dec 04 16:54:03 well, there are just 5 forks :) Dec 04 16:54:31 is hackaton on 8th confirmed? I'll be around for testing Dec 04 18:38:36 heh.. fb.c is very popular place to patch Dec 04 18:42:06 well, let's plan kexecboot hackaton on 8th Dec Dec 04 23:19:06 argh Dec 04 23:19:07 configure: error: unrecognized option: `---enable-bpp=16' Dec 04 23:21:36 dromede: +EXTRA_OECONF_poodle = "--enable-zaurus --enable-delay=2 --enable-evdev-rate=1000,250 ---enable-bpp=16" Dec 04 23:22:00 bluelightning: can you pls amend that line ? Dec 04 23:22:31 oh dear Dec 04 23:22:34 one moment Dec 04 23:23:04 I'm sorry I've missed it... Dec 04 23:23:16 I hit immediately, though ;) Dec 04 23:23:37 sorry about that Dec 04 23:23:47 np, thanks for the patch Dec 04 23:25:00 fixed Dec 04 23:25:02 np Dec 04 23:26:47 bluelightning: I'm playing now with linux-yocto-tiny-kexecboot_3.4 (as base for 3.8) Dec 04 23:26:55 now, a doubt: Dec 04 23:27:34 the -tiny kernel defaults for 'allnoconfig' so the options unspecified are not selected by default Dec 04 23:27:43 all fine until here Dec 04 23:28:09 but the standard yocto-tiny policy defines too much things for our space Dec 04 23:28:24 so we have to unset some options, not many in reality Dec 04 23:28:50 (this is valid for any arch) Dec 04 23:29:39 my doubt is: should I go further and extract the specific bits necessary for kexecboot (to boot from any fs) ? Dec 04 23:29:59 split that in generic and zaurus specific? Dec 04 23:30:40 you see, if one plans to add a new machine could reuse some frags Dec 04 23:31:26 but in treality it is much simpler to put a defconfig and ignore the yocto config checks Dec 04 23:31:51 I'd really need a second machine for testing/sharing Dec 04 23:34:20 i.e. if some ipaq fellow wants to try linux-yocto he could start from the less-grained 3.2 and then split further as done in 3.4 Dec 04 23:36:34 it seems reasonable to do the split... obviously that's more work though Dec 04 23:38:13 * bluelightning actually looks at the 3.4 setup Dec 04 23:38:21 this should pay on the long run Dec 04 23:39:22 hmm so there's quite a lot of stuff in zaurus-non-hardware.cfg Dec 04 23:40:35 lot of net drivers Dec 04 23:41:06 if you build, you'll find in /meta all the details/sources Dec 04 23:42:12 note I've split hw/non-hw according to the warnings of kernel config tools Dec 04 23:42:59 the little mismatches are bugs I've sent to Bruce Dec 04 23:46:00 * ant_home notes MTD_COMPLEX_MAPPINGS=y nd MTD_PHYSMAP=y are common to all models. patch following... Dec 04 23:46:47 this was a lot of work... well done :) Dec 04 23:47:11 btw do you think anything has changed on the collie front? Dec 04 23:47:34 yes but in worse...I got compile errors with 3.4 Dec 04 23:48:21 hmm Dec 04 23:48:33 the good news are we can simply copy and create 3.6 and it works Dec 04 23:48:53 3.8 is awaited and will be almost vanilla Dec 04 23:49:07 patches already in linux-net Dec 04 23:49:12 *next Dec 04 23:56:11 bluelightning: note that it should be possible to opt-out of tracing and profiling features... Dec 04 23:57:35 it ought to be possible...iirc there was an issue Dec 04 23:59:52 it is possible in any case including an undefine.cfg like I'm doing for the kexecboot version Dec 05 00:00:03 as last one Dec 05 00:00:51 see http://paste.debian.net/214425/ Dec 05 00:04:54 indeed, I was about to mention that earlier Dec 05 00:07:00 what's strange, linux-yocto.inc sets Dec 05 00:07:01 KERNEL_FEATURES_append = " features/netfilter" Dec 05 00:07:42 the rest (kprobe, uprobe, ...) is added by yocto-policy defaults Dec 05 00:07:55 net drivers as well Dec 05 00:09:28 it is because it defaults to KCONFIG_MODE = "--alldefconfig" Dec 05 00:09:33 probably Dec 05 00:19:01 wow, with above unsets kernel can pass the sizecheck :) Dec 05 00:19:33 I'l llaunch a collie build overnight...fingers crossed Dec 05 00:19:41 gn Dec 05 00:19:52 seems to me the config ought to be additive rather than subtractive... I think we might have talked about that before Dec 05 00:19:53 night Dec 05 00:20:15 yes, the linux-yocto recipe is only additive Dec 05 00:20:36 but for the kexecboot we are obliged Dec 05 00:21:32 results: http://paste.debian.net/214421 Dec 05 00:22:16 hmm... "required" Dec 05 00:23:12 http://paste.debian.net/214434/ Dec 05 00:23:14 ^^ Dec 05 00:24:11 CONFIG_NETWORK_FILESYSTEMS ??? Dec 05 00:24:12 http://paste.debian.net/214435/ Dec 05 00:24:19 ^^ vs 'missing' Dec 05 00:24:40 it's allright, some are bogus (x86 stuff :/) Dec 05 00:25:09 finally Dec 05 00:25:20 http://paste.debian.net/214436/ Dec 05 00:25:38 ^^ that's what does matter Dec 05 00:27:42 this is the full story Dec 05 00:27:44 http://paste.debian.net/214437 Dec 05 00:29:16 * ant_home notes cmdline needs to be fixed too Dec 05 00:31:04 note the weird options at #206 (yocto.cfg) Dec 05 00:31:38 and #216 FWIW **** ENDING LOGGING AT Wed Dec 05 02:59:59 2012