**** BEGIN LOGGING AT Tue Jan 21 02:59:59 2014 Jan 21 08:58:28 lumag: finally to have stable ubifs I did one step back Jan 21 08:58:31 http://tinyurl.com/oclv4rl Jan 21 09:01:54 lumag: instead of removing features I've also tried to add the fixes described in "Adapting the Linux Kernel for Micron® P30, P33, and J3 Flash Memory" Jan 21 09:02:26 but the only stable setup is like in the patch Jan 21 09:03:59 ant_work, so you finally set to single partition :( Jan 21 09:04:15 I was convinced by you that we would benefit from two partitions. Jan 21 09:04:21 Let's live with one however. Jan 21 09:04:25 disabling SuspendCmd it works even with 4 Jan 21 09:04:47 but I think there is some Errata Jan 21 09:05:15 I can't properly see if they correspond to the Intel Series Jan 21 09:05:43 The Strataflash had bugs with suspend/resume up to revision "with a B in the 9 place" Jan 21 09:05:54 our flash: LH28F640BF Jan 21 09:06:11 ^ Jan 21 09:06:24 so we have a B... Jan 21 09:06:49 If I could find a table with Year and Revisions... Jan 21 09:06:58 anyway, it is fast Jan 21 09:07:05 much faster than CF ! Jan 21 09:07:48 about benefits, there are docs explaining in detail Jan 21 09:08:20 I'll upload Jan 21 09:08:42 lumag: btw, I discovered cfi_udelay != udelay Jan 21 09:09:04 anyway, delays won't help Jan 21 09:20:42 lumag: console-image-minimal.ubifs is about 55% of flash space (no modules) Jan 21 09:21:41 jffs2 image is also 6,7-7MB sized but there is more space Jan 21 09:22:10 ..and I've noticed there are 0 kb for root. We need to fix this in OE Jan 21 09:24:26 lumag: I'll have to clean much comments before sending the patch upstream... would you remove the comments about the +3 offset for the nr of partitions? Is pointless now... Jan 21 09:26:24 I would leave those parts commented (remember, c comments, not c++, they bite me few times). Jan 21 09:26:37 Or even enabled. Jan 21 09:26:52 Because next time somebody won't lose time for those. Jan 21 09:30:29 when reading nr=2 the code loops twice, like as for 2 chips Jan 21 09:31:27 but I'll retest. With 0 -> 1 all is stable and rather quick Jan 21 09:31:43 I'll fix the comments Jan 21 09:33:28 lumag: if you have time, google for the little Jan 21 09:33:29 Adapting the Linux Kernel for Micron® P30, P33, and J3 Flash Memory Jan 21 09:33:55 they add 0xFF instead of delay before suspend and the udelay(500) like for the cmdset_0002 Jan 21 09:34:07 for the latency after resume Jan 21 09:34:25 but yet..no help..those seem 'modern' chips ;) Jan 21 09:34:54 maybe Sharp really used an old bad batch for Collie Jan 21 09:34:58 ;) Jan 21 09:36:06 morning ant_work, lumag, all Jan 21 09:36:21 hello Jan 21 09:37:20 lumag: using it, ubifs on collie seems faster than on poodle Jan 21 09:37:26 much snappier Jan 21 09:37:52 just a first impression, no benchs Jan 21 09:39:50 bluelightning: I think koen is waiting for dracut in meta-initramfs Jan 21 09:39:59 tbh I didn't build it yet Jan 21 09:40:14 ant_work: I'm happy to merge it if you are... Jan 21 09:40:39 heh, was the hype couple of years ago Jan 21 09:41:03 at least we should do a build-test before Jan 21 09:41:05 ... Jan 21 09:57:42 ah, ok, 3.13 is out Jan 21 09:58:07 we'll get rid of -dev and 3.10 kernels soon **** ENDING LOGGING AT Wed Jan 22 02:59:59 2014