**** BEGIN LOGGING AT Tue Nov 12 02:59:58 2013
Nov 12 09:03:16 <lumag> hello ant_work
Nov 12 09:03:23 <ant_work> gm
Nov 12 09:03:33 <lumag> I reflashed collie with old kexecboot.
Nov 12 09:03:46 <lumag> Now I have a problem.
Nov 12 09:03:57 <lumag> Can it read the kernel from mmc?
Nov 12 09:04:17 <ant_work> yes but only up to 2.6.31
Nov 12 09:04:54 <ant_work> after that detects, boots, crashes on boot ofc
Nov 12 09:05:12 <lumag> Yes, I used 2.6.31 from kexecboot.org
Nov 12 09:05:42 <lumag> According to debug log it detects 'vfat' file system but then it 'can't mount the device...'
Nov 12 09:05:51 <ant_work> ah.. vfat...
Nov 12 09:06:08 <ant_work> there must be a bug with fat/vfat :/
Nov 12 09:06:09 <lumag> I can reformat the card, if that is an issue
Nov 12 09:06:12 <ant_work> for all models
Nov 12 09:06:19 <lumag> ok
Nov 12 09:06:54 <ant_work> I have to debug that, yocto adds some patches for short filenames
Nov 12 09:07:17 <ant_work> fwiw the same sd card can be mounted by linux-yocto_3.10 (not on collie)
Nov 12 09:08:07 <ant_work> so the 3.10 kernels are ok
Nov 12 09:08:27 <ant_work> lumag: question about bootlogo
Nov 12 09:08:34 <lumag> yes,
Nov 12 09:08:53 <ant_work> about pushing backlight to the end of the Makefile
Nov 12 09:09:10 <ant_work> can it be that it makes the transition "smoother"?
Nov 12 09:09:24 <ant_work> I mean, after kexec the fb is b0rked for some msec
Nov 12 09:09:40 <ant_work> it was not noticeable with 2.6.3x kernels
Nov 12 09:09:57 <ant_work> maybe baclight and logo must be put in some specific place?
Nov 12 09:10:36 <ant_work> the objects I mean
Nov 12 09:11:59 <ant_work> now they are both at the top of drivers/video/Makefile
Nov 12 09:12:56 <lumag> we can try lowering them.
Nov 12 09:13:10 <ant_work> btw
Nov 12 09:13:12 <lumag> Then you would also need to remove 'if machine_is_poodle()' line
Nov 12 09:13:14 <ant_work> obj-y                             += backlight/ display/  up 3.3
Nov 12 09:13:23 <ant_work> then is just
Nov 12 09:13:24 <ant_work>  16 obj-y                             += backlight/
Nov 12 09:13:47 <lumag> They just removed unused display/ stuff
Nov 12 09:14:06 <ant_work> ok
Nov 12 09:14:23 <lumag> Hmm.
Nov 12 09:15:20 <lumag> I got oom oopses
Nov 12 09:15:46 <lumag> Maybe because I used my kernel with bundled initramfs
Nov 12 09:19:44 <ant_work> lumag: again about 3.10 in OE: can you pls add some text for the missing patche for m62332?
Nov 12 09:20:18 <ant_work> and locomolcd needs to be patched or not?
Nov 12 09:20:49 <ant_work> I've booted with only the first two patches you've sent me (the 2 on the ML)
Nov 12 09:21:50 * ant_work is reading datasheet about buffer programs on flash...
Nov 12 09:40:26 <ant_work> hm.. Status Register on strataflash seems different
Nov 12 09:42:34 <lumag> ant_work, If you move backlight/ to the end of Makefile, you will most probably need to remove the machine check in probe and always call locomolcd_on()
Nov 12 09:43:25 <lumag> What is the best compression? lzma or lz4?
Nov 12 09:45:09 <ant_work> today? dunno..
Nov 12 09:45:26 <ant_work> what is trendy in kernel today? xz or lz4?
Nov 12 09:46:02 <ant_work> note there are limits for devices with 32MB ram or they won't boot
Nov 12 09:46:35 <ant_work> lumag: I've spent much time to pack all in that 2.6.31
Nov 12 09:46:47 <ant_work> iirc it fits just for 2-3 kb...
Nov 12 09:47:18 <lumag> Mine initramfs is 4.5M uncompressed
Nov 12 09:47:27 * lumag is using core-image-minimal-initramfs
Nov 12 09:47:45 <ant_work> kernel will be put in diet today...
Nov 12 09:52:55 <ant_work> ok ,I think I see the problem: SR.6 Suspended block erase should be resumed first
Nov 12 09:53:12 <ant_work> seems this dual page needs an extra check (or two...)
Nov 12 09:53:31 <ant_work> happily there are the flowcharts :)
Nov 12 09:53:47 <lumag> ok, kernel has grown too much.
Nov 12 09:54:11 <lumag> Stripped kernel is 5.2 Mb
Nov 12 09:55:00 <lumag> That is with rootfs (i think)
Nov 12 09:55:42 <lumag> http://pastebin.com/8GyfWNSi
Nov 12 09:55:50 <ant_work> yes, kernel is 3,x MB
Nov 12 09:57:27 <ant_work> ok, it fails to decompress the cpio apparently
Nov 12 10:00:32 <ant_work> lumag: so we take the 5 patches for linux-yocto-next?
Nov 12 10:00:44 <ant_work> plus some MTD fixes from 3.10
Nov 12 10:01:30 <lumag> I think so
Nov 12 10:02:15 <ant_work> ok, I'll sync in meta-handheld so we'll rebuild linux-kexecboot
Nov 12 10:02:58 <ant_work> ah, note that with the old 2.6.31 kernel when in kexecboot collie does not poweroff. Restarts
Nov 12 10:03:08 <ant_work> at least recharges
Nov 12 10:03:11 <ant_work> ;)
Nov 12 10:12:26 <ant_work> wow, you caught RMK attention
Nov 12 10:12:31 <ant_work> great
Nov 12 10:12:46 <ant_work> (for me it took 1-2 months last time...)
Nov 12 10:13:02 <ant_work> betetr that YOU send all upstream ;)
Nov 12 13:37:17 <lumag_> ant_work, collie has 64Mb of RAM, right?
Nov 12 13:38:00 <lumag_> I re-checked my oops log
Nov 12 13:38:02 <lumag_> Memory: 5792K/16384K available
Nov 12 13:38:09 <lumag_> This is really funny
Nov 12 13:45:15 <lumag_> Ah, I see that you pass 'mem=64M' as a bootarg
Nov 12 13:47:02 <ant_work> yes, the old mtd-ram use
Nov 12 13:47:38 <ant_work> fwiw I got OOM on poodle 32M with core-image-sato
Nov 12 13:53:14 <ant_work> lumag_: if you build with OE be sure to have  http://cgit.openembedded.org/openembedded-core/commit/?h=master-next&id=81002f41a0ddf5eb0c5b5cb4865c3a863c306f8e
Nov 12 13:55:36 <ant_work> omg, now we have patch abundance ;)
Nov 12 13:56:01 <ant_work> I somehow feel sa1100 will shine in 3.13
Nov 12 13:59:49 <lumag_> 3.13 in oe, aim for most of the fixes in 3.14 in mainline
Nov 12 14:00:24 <lumag_> Hmm.
Nov 12 14:00:46 <lumag_> I build oe-core in a distroless mode, but never had issues.
Nov 12 14:00:51 <lumag_> Anyway I pulled today.
Nov 12 14:01:07 <lumag_> regarding poodle: does it have only 32Mb of RAM?
Nov 12 14:01:30 <lumag_> Ah, atag on collie is wrong because it reserved memory for mtdram?
Nov 12 14:03:48 <ant_work> distroless = wrong formfactor = empty /etc/defaults/rotation = unrotated fbconsole / wrong psplash etc
Nov 12 14:04:07 <lumag_> When using GPE on tosa I always enabled small swap partition.
Nov 12 14:04:24 <lumag_> I think that even 128Mb or 256Mb helps a lot
Nov 12 14:05:14 <ant_work> I have not spent much time in userland but surely swap helps
Nov 12 14:05:43 <ant_work> I'm still at 'calibrate the damn touchscren' in X ;) now we have xinput-calibrator
Nov 12 14:06:15 <ant_work> as first step it would be nice to have Opie working on collie/poodle
Nov 12 14:06:20 <ant_work> with 3.12
Nov 12 14:06:39 <ant_work> poodle is almost ok, apart some shutdown/power issue
Nov 12 14:07:17 <ant_work> clamshell are ok (tested -sato on c7x0 last week)
Nov 12 14:07:32 <ant_work> tosa?
Nov 12 14:19:22 <lumag_> Checked few weeks ago.
Nov 12 14:19:34 <lumag_> It required some patches, esp. in backlight part.
Nov 12 14:19:43 <lumag_> and ac97
Nov 12 14:20:02 <ant_work> ah, yes, sound is missing on all models I think
Nov 12 14:20:12 <lumag_> It looks like I overloaded linusw with questions
Nov 12 14:20:22 <ant_work> is justthe module-autoload stuff missing
Nov 12 14:22:30 <ant_work> I think he works from home so is not always on kb
Nov 12 14:23:43 <ant_work> ah, btw, all these patches mean that I better keep my fingers off the locomo-spi ;)
Nov 12 14:24:00 <ant_work> we have to try again with linusw
Nov 12 14:49:02 <lumag_> ant_work, I mostly refactored locomo stuff to a saner state.
Nov 12 14:49:12 <lumag_> I just need to test that on real hw.
Nov 12 14:49:54 <lumag_> After that one can bring forward locomo-spi.
Nov 12 14:51:17 <ant_work> great
Nov 12 14:55:08 <ant_work> lumag_: so for 3.10 in Oe should I also add the m62332 fix?
Nov 12 15:14:18 <lumag_> I think so
Nov 12 15:16:17 <lumag_> The problem is that it was never present in 2.6/3.x and nobody noticed
Nov 12 15:23:28 <ant_work> what for subject? do you send a patch for meta-hh?
Nov 12 15:29:25 <lumag_> Too many patches...
Nov 12 15:29:43 <lumag_> Subject. It incorrectly toggles SCL line instead of SDA.
Nov 12 15:29:54 <lumag_> Or vise verse. I don't remember ATM.
Nov 12 15:33:01 <ant_work> ok thx
Nov 12 16:14:25 <ant_work> lumag_: oe-core has all the fixes now btw
Nov 12 16:28:47 <lumag_> ant_work, great
Nov 12 16:30:41 <ant_work> this evening I'll send the missing ones for meta-handheld - only 3.10 atm
Nov 12 16:56:52 <ant_work> Jay7: I'm reading things like : is it possible that kernel memcpy is much faster than the one from glibc?
Nov 12 16:57:10 <ant_work> we had issues with klibc's memcpy too slow? Can't remember
Nov 12 16:57:54 <ant_work> wow, we are on-topic ;)
Nov 12 16:59:18 <ant_work> "I just ported Nicolas Pitre's ARM memcpy from kernel and magically I reduced time spent by application on copying memory by about 30%"
Nov 12 17:00:27 <ant_work> hm
Nov 12 17:00:28 <ant_work> https://github.com/kexecboot/kexecboot/blob/master/fb.c#L289
Nov 12 17:00:40 <ant_work> * NOTE: klibc uses 8bit transfers that breaks image on tosa
Nov 12 17:00:40 <ant_work> * So we will use own memcpy.
Nov 12 17:00:50 <ant_work> lumag_: ^^ ?
Nov 12 17:01:37 <lumag_> on tosa you should access framebuffer using u16 _only_.
Nov 12 17:02:11 <ant_work> I see
Nov 12 17:03:20 <ant_work> not that we draw much on screen...
Nov 12 17:04:49 <ant_work> bbl
Nov 12 19:16:00 <Jay7> lumag_: u16 only? i.e. no u32?
Nov 12 19:16:28 <lumag_> According to docs, yes
Nov 12 19:18:03 <Jay7> hm.. then we should build kexecboot for tosa with special configure options
Nov 12 19:18:20 <lumag_> Hmm.
Nov 12 19:18:25 <lumag_> Maybe I'm mistaken.
Nov 12 19:18:28 <Jay7> some time ago I've switched to u32 transfers by default
Nov 12 19:18:40 <Jay7> it's easier to test imho
Nov 12 19:21:07 <lumag_> Can't find it in docs atm
Nov 12 19:21:13 <Jay7> np
Nov 12 19:21:23 <Jay7> let's leave it for later
Nov 12 19:21:29 <Jay7> after backlight etc :)
Nov 12 19:36:05 <lumag_> For now ant dragged me into sa1100 cleanup.
Nov 12 19:36:33 <lumag_> I hope to find time for tosa next week.
Nov 12 19:43:00 <Jay7> seems it's time to find my collie and tosa :)
Nov 12 19:53:17 <lumag_> Jay7, :)
Nov 12 19:53:27 <lumag_> I also have some kind of corgi.
Nov 12 22:06:53 <ant_home> lumag_: still problems with MTD. I hope to come to an end
Nov 12 22:06:56 <ant_home> root@collie:~# flashcp -v /initrd.bin /dev/mtd2
Nov 12 22:06:56 <ant_home> Erasing blocks: 112/112 (100%)
Nov 12 22:06:56 <ant_home> Writing data: 14336k/0k (100%)%)
Nov 12 22:06:56 <ant_home> Verifying data: 10500k/0k (14336%)File does not seem to match flash data. First mismatch at 0x00a3e800-0x00a41000
Nov 12 22:07:32 <lumag_> ant_home, Are you sure it is not just worn out?
Nov 12 22:07:53 <lumag_> do a cmp --verbose. How many errors do you see?
Nov 12 22:08:10 <ant_home> one of you 3 guys should test... ;)
Nov 12 22:08:53 <lumag_> Hmm. Ping me next night. I will have a device with me.
Nov 12 22:08:59 <ant_home> with all the patches sent to oe-dev
Nov 12 22:09:00 <ant_home> Intel/Sharp Extended Query Table at 0x0039
Nov 12 22:09:01 <ant_home> cfi_cmdset_0001: Suspend erase on write disabled.
Nov 12 22:09:01 <ant_home> Using buffer write method
Nov 12 22:09:01 <ant_home> Using auto-unlock on power-up/resume
Nov 12 22:09:25 <ant_home> I think I have an idea
Nov 12 22:09:59 <ant_home> anyway, let sleep for one night ;)
Nov 12 22:11:26 <ant_home> I guess there is some silly issue here..112 vs 113
Nov 12 22:11:28 <ant_home> Eraseblock size:                131072 bytes, 128.0 KiB
Nov 12 22:11:28 <ant_home> Amount of eraseblocks:          113 (14811136 bytes, 14.1 MiB)
Nov 12 22:12:01 <ant_home> vs. Erasing blocks: 112/112 (100%)
Nov 12 22:12:45 <ant_home> need to re-read & compare the flow charts of sharp AND Intel strataflash
Nov 12 22:13:02 <ant_home> seems our flash has an extended Status Register
Nov 12 22:13:22 <ant_home> and the bit is about buffer disponibility for writing...
Nov 12 22:13:43 <ant_home> ...not overnight, need much coffee
Nov 12 22:24:51 <ant_home> lumag_: linux-yocto-tiny-kexecboot_3.10 built
Nov 12 22:25:51 <lumag_> ant_home, do you have a tree somewhere?
Nov 12 22:30:37 <ant_home> well, yes on github but not updated.  I have just sent the patches to ML/patchwork
Nov 12 22:30:55 <ant_home> this kernel seems not booting..hmmm
Nov 12 22:44:55 <ant_home> will test more once we get mmc
Nov 12 23:02:46 <ant_home> lumag_: https://github.com/andrea-adami/meta-handheld
Nov 12 23:03:12 <lumag_> ok, I will try to test tomorrow
Nov 12 23:04:43 <ant_home> kernel still too big...tomorrow diet ;)
Nov 12 23:04:54 <ant_home> now a bit of sleep
Nov 12 23:04:55 <ant_home> gn
**** ENDING LOGGING AT Wed Nov 13 02:59:58 2013