**** BEGIN LOGGING AT Sun Jul 06 02:59:59 2014
**** BEGIN LOGGING AT Sun Jul 06 04:03:50 2014
Jul 06 22:14:02 <ant__> lumag, fyy, removed mem=64M from collie then I get strange OOM..
Jul 06 22:14:05 <ant__> see,,
Jul 06 22:14:16 <lumag> hello ant__
Jul 06 22:14:26 <lumag> long time, no time for anything.
Jul 06 22:14:35 <ant__> I know the times...
Jul 06 22:14:48 <ant__> Memory: 10496K/16384K available (4053K kernel code, 173K rwdata, 1060K rodata, 230K init, 146K bss, 5888K reserved)
Jul 06 22:14:52 <ant__> instead of
Jul 06 22:15:03 <ant__> Memory: 64MB = 64MB total
Jul 06 22:15:03 <ant__> Memory: 55500k/55500k available, 10036k reserved, 0K highmem
Jul 06 22:15:03 <ant__> Virtual kernel memory layout:
Jul 06 22:15:15 <ant__> why does it default to 16??
Jul 06 22:15:41 <ant__> again because of the strange dev models prolly
Jul 06 22:16:01 <lumag> Yes. If I remember correctly, this is passed through atags (or through pre-atags config).
Jul 06 22:16:11 <ant__> then the 5000 owners must set 32M
Jul 06 22:16:30 <ant__> I wanted to make it more compatible...imagine...
Jul 06 22:17:05 <ant__> do you think we should revert and go back to 64M default?
Jul 06 22:17:23 <lumag> The collie firmware 3.xx (i think so) have enabled 'ram disk' - split of the memory to 16M/the rest, with the ram disk used for read-write storage.
Jul 06 22:18:01 <ant__> I see
Jul 06 22:18:13 <ant__> maybe the bootloader sets 16
Jul 06 22:18:19 <lumag> That would require patching of either bootloader (don't do so) unless you want to replace it with uboot or of kernel startup files. See arch/arm/kernel/atags-compat.c
Jul 06 22:18:28 <lumag> Yes, it does if I remember correctly.
Jul 06 22:18:48 <ant__> ok, let's go back to 64M
Jul 06 22:18:56 <ant__> 5500 was far more popular
Jul 06 22:19:02 <lumag> One can use fixup to really fix memory map during kernel boottime.
Jul 06 22:19:08 <lumag> See arch/arm/mach-pxa/tosa.c
Jul 06 22:19:16 <ant__> only for 5000 is possible?
Jul 06 22:19:45 <lumag> That would require detecting model (by reading pins of one of chips) and then setting memory map to either 32 or 64 megs.
Jul 06 22:20:03 <ant__> yes, in proper english is that ;)
Jul 06 22:20:25 <ant__> worth doing that?
Jul 06 22:21:12 <ant__> I'd keep this in low priority for our to-do
Jul 06 22:21:29 <ant__> fwiw 3.16 boots
Jul 06 22:21:33 <ant__> on collie
Jul 06 22:21:33 <lumag> :)
Jul 06 22:21:35 <lumag> great.
Jul 06 22:21:49 <ant__> and anarsoul fixed opie calibration, almost
Jul 06 22:22:00 <lumag> Didn't have time for zaurii since long ago.
Jul 06 22:22:05 <ant__> I wanted to test but was getting OOM ...
Jul 06 22:22:11 <lumag> I should refresh and resend at least locomo patches.
Jul 06 22:22:20 <ant__> I can send you the opie-image 3.14 if it works
Jul 06 22:22:32 <ant__> yes, please check the clock stuff
Jul 06 22:22:36 <lumag> :))
Jul 06 22:22:49 <ant__> testing in 5 mins
Jul 06 22:23:06 <lumag> No need, probably. I had a custom console image to help in kernel development.
Jul 06 22:23:36 <ant__> I'll upload on github some 'good' images anyway
Jul 06 22:23:42 <ant__> for me as well ;)
Jul 06 22:23:49 <lumag> That is a usual problem for me - either I can work with userspace, or I hack kernel with console userspace
Jul 06 22:23:52 <lumag> Good.
Jul 06 22:24:01 <ant__> now I understand
Jul 06 22:24:12 <ant__> I normally come to prompt and reboot
Jul 06 22:24:18 <lumag> Do we have an organization on GitHub?
Jul 06 22:24:18 <ant__> thats' my userland
Jul 06 22:24:27 <ant__> hmm, no, good idea
Jul 06 22:25:00 <ant__> lumag, http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto_3.14.bbappend
Jul 06 22:25:04 <lumag> It is free if all you need is public repos & webpages.
Jul 06 22:25:08 <ant__> ^actual status
Jul 06 22:25:09 <lumag> (and wiki/issues).
Jul 06 22:25:39 <lumag> So we have custom non-Yocto defconfigs?
Jul 06 22:25:50 <ant__> old-style for you and linusw
Jul 06 22:25:57 <ant__> back on popular request ;)
Jul 06 22:26:24 <ant__> I have added a dubious patch and reverted one upstream commit
Jul 06 22:26:27 <ant__> no good :/
Jul 06 22:28:03 <ant__> lumag, plus there is one pending patch reducing kernel size
Jul 06 22:28:42 <ant__> http://patches.openembedded.org/patch/74149/
Jul 06 22:29:16 <lumag> Heh :)
Jul 06 22:29:48 <ant__> bluelightning is on holiday, still pending
Jul 06 22:29:51 <lumag> I'd suggest to also blacklist ipv6 module in userspace by default. Otherwise it will be loaded pretty much soon.
Jul 06 22:30:04 <ant__> I wanted to ask about ipv6
Jul 06 22:30:33 <ant__> I don't have modern routers so don't use yet
Jul 06 22:35:48 <ant__> fwiw, 3.16-kexecboot is too big for lzma. I have used xz and size is better
Jul 06 22:36:17 <ant__> collie needs *lot* of time to load kernel in kexecboot
Jul 06 22:36:28 <ant__> is because lzma maybe?
Jul 06 22:36:47 <ant__> or maybe some DMA/cache issue?
Jul 06 22:40:02 <ant__> lumag, in case you build those days... a recent commit in oe-core makes now absolute symlink for /boot/zImage
Jul 06 22:40:12 <ant__> kexecboot is broken by this
Jul 06 22:40:27 <ant__> Cannot open `/mnt/boot/zImage': No such file or directory
Jul 06 22:40:28 <ant__> Nothing has been loaded!
Jul 06 22:40:36 <ant__> I've sent a msg to oe-core ML
Jul 06 22:40:37 <lumag> ant__, ok, I'll remember that.
Jul 06 22:40:53 <ant__> lrwxrwxrwx 1 root root      28 lug  5 12:26 zImage -> zImage-3.14.5-yocto-standard
Jul 06 22:40:56 <ant__> ^good
Jul 06 22:41:03 <ant__> lrwxrwxrwx 1 root root      34 lug  5 12:26 zImage -> /boot/zImage-3.14.5-yocto-standard
Jul 06 22:41:10 <ant__> bad, today, sigh ^
Jul 06 22:42:33 <lumag> I don't know about IPv6 - I have it on my router and on my PC, but I don't remember using it from zaurus. Most websites work w/o IPv6.
Jul 06 22:43:58 <ant__> these were the yocto defaults
Jul 06 22:44:05 <ant__> for standard/base
Jul 06 22:44:38 <ant__> actually zaurus could have some little future for network analysys/cracking
Jul 06 22:44:55 <ant__> one security-scan distro maybe
Jul 06 22:45:22 <lumag> Regarding compression. There is a tradeoff. small kernels load faster, but take more time to decompress.
Jul 06 22:45:56 <lumag> ant__, yes. Several years ago I had plans to use one of them to connect 802.15.4 dongles.
Jul 06 22:47:52 <lumag> If blacklisted via /etc/modprobe.d/*.conf, ipv6 still can be loaded, but it will save few kibibytes for those who don't require.
Jul 06 22:47:59 <lumag> don't require it.
Jul 06 22:48:09 <ant__> well, anarsoul got it work on ipaq.opie cal still fails here on collie, I can't exit it
Jul 06 22:48:50 <ant__> we have to verify the modules infrastructure
Jul 06 22:49:16 <ant__> let's rebuild all from scratch
Jul 06 23:06:03 <ant__> lumag, finally Russell pulled the two mtd patches for collie in 3.16-rc3. Now I'm bothering the linux-mtd guys for the CFI patch
Jul 06 23:06:19 <ant__> seems taking longer than expected
Jul 06 23:06:32 <lumag> That is the usual problem.
Jul 06 23:06:34 <ant__> my bad I sent the patches too near merge-windows
Jul 06 23:07:16 <ant__> I've resent selling it as fix
Jul 06 23:07:26 <ant__> for 3.16-rc*
Jul 06 23:21:06 <ant__> ok  https://github.com/andrea-adami/meta-handheld
Jul 06 23:21:18 <ant__> I'll push here while Paul is away
Jul 06 23:21:23 <ant__> late now...
Jul 06 23:21:26 <ant__> good night
**** ENDING LOGGING AT Mon Jul 07 02:59:59 2014