**** BEGIN LOGGING AT Sun May 13 02:59:59 2007 May 13 08:11:33 03blaster8 * r833 10kernel/ (58 files in 11 dirs): Archive old versions, add 2.6.22 devel patchset May 13 08:15:37 03blaster8 * r834 10kernel/ (72 files in 4 dirs): Archive 2.6.21, development concentrated on 2.6.22 May 13 08:25:04 03blaster8 * r835 10kernel/trunk/patches/2.6.22/ (3 files): Reorder numbering in series May 13 12:25:54 03blaster8 * r836 10kernel/trunk/patches/2.6.22/ (7 files): Add i2c-gpio driver support patches May 13 12:26:12 blaster8: hi there May 13 12:26:16 hey May 13 12:26:21 quick question May 13 12:26:31 k May 13 12:26:49 I'm considering starting on the fsg-3 soon (as you may have noticed from the nslu2-linux-dev list) May 13 12:26:55 but buttons/leds aren't working May 13 12:26:57 cool May 13 12:26:59 ok May 13 12:27:08 I''l look in hte board support code now May 13 12:27:21 I remember that there was some trick to figure out the GPIOs for the buttons from redboot May 13 12:27:33 (by reading memory somewhere) May 13 12:27:34 that I can't help you with May 13 12:27:40 I arrived after that all turned up May 13 12:27:44 aha May 13 12:27:47 let me think for a minute May 13 12:27:59 you can poke random memory addresses from linux May 13 12:28:19 find the GPIO addresses, mmap and poke from there May 13 12:28:40 nasty but it'll work May 13 12:29:04 I believe the GPIO addrs are accessible from within redboot May 13 12:29:11 apparently it has an LED at 15 May 13 12:29:14 I tried using redboot to figure out the buttons on the DS101 May 13 12:29:29 But that was *ages* ago ;) May 13 12:29:34 how many lights does it actually have May 13 12:30:19 I think mine has 5 May 13 12:30:27 but mine has wifi, so it might have an extra one May 13 12:30:28 eek May 13 12:30:30 lots then May 13 12:30:46 You'd have to dig through a redboot manual, sorry May 13 12:31:12 np May 13 12:31:17 I now roughly how to access them 'bare-metal' through linux but I don't know redboot May 13 12:31:25 lots of work to do in the kernel repo btw May 13 12:31:32 it's probably easier to try to figure it out from the old kernel sources May 13 12:31:52 old header files are your best bet May 13 12:31:59 yeah May 13 12:32:02 from Freecom's release May 13 12:34:17 I'm just sick of freecoms firmware May 13 12:34:29 that bad? May 13 12:34:34 eek May 13 12:34:42 the wireless falls out constantly May 13 12:34:55 wireless is so dodgy May 13 12:35:01 so I want to replace the wifi-card with something I know is more stable May 13 12:35:09 only works reliably in my experience on OS X and FreeBSD May 13 12:35:16 yeah May 13 12:35:27 I'm hoping that the new stack will get things a bit more stable May 13 12:35:45 hopefully yeah, but it's been a long time coming May 13 12:35:59 good things come to those who wait :P May 13 12:36:06 I'm building libertas / zd1201 / zd1211 usb support in 2.6.22 May 13 12:36:12 lots of nice in-kernel usb wifi drivers May 13 12:36:19 Marvell libertas? May 13 12:36:23 indeed May 13 12:36:49 and NO_HZ / HRTIMERS May 13 12:36:50 6>mwl0: Marvell Libertas MB/CB35P Wireless LAN adapter: mem=0x48000000, irq=24 May 13 12:36:59 That's the minipci-card in the fsg-3 May 13 12:37:08 mm, it's a USB driver AFAIK May 13 12:37:17 ah, ok May 13 12:37:24 zd1211 or zd1211rw? May 13 12:37:28 not that the mini-pci driver won't be that different May 13 12:37:33 zd1211rw May 13 12:37:35 ok May 13 12:38:04 though I may use the mac80211 zd1211 if it gets Master mode support May 13 12:38:38 yeah May 13 12:38:52 we're *really* close to getting full support upstream May 13 12:38:59 ooh, cool May 13 12:39:01 we can do it by 2.6.23 May 13 12:39:06 what's missing now? May 13 12:39:13 ixp_npe driver May 13 12:39:33 ah May 13 12:39:44 well, Krzysztof's driver is clean May 13 12:39:45 Which of the two free drivers would be best? May 13 12:39:49 ah May 13 12:39:53 but not LE May 13 12:39:57 upstream wise, Krzysztof's one May 13 13:14:06 * NAiL is very annoyed by the fsg-3 wireless... It fails silently, and the only way to recover is to pull the power May 13 13:31:19 "Marvell Libertas 8388 802.11a/b/g cards" but only USB is support in-kernel atm May 13 14:18:59 03blaster8 * r837 10kernel/trunk/ (5 files in 3 dirs): Remove squashfs from 2.6.22, update defconfig May 13 14:44:12 03blaster8 * r838 10kernel/trunk/patches/2.6.22/series: Update series with rtc probe information May 13 16:51:28 03bzhou * r6033 10optware/trunk/make/py-yaml.mk: py-yaml: 3.04 -> 3.05; added py25 subpackage May 13 16:58:58 Hi all. May 13 17:05:32 03bzhou * r6034 10optware/trunk/make/vim.mk: vim: 7.0 -> 7.1 May 13 17:57:26 * rwhitby returns from week-long business trip May 13 17:58:04 wb May 13 18:06:29 thx **** ENDING LOGGING AT Sun May 13 18:43:39 2007 **** BEGIN LOGGING AT Sun May 13 18:43:59 2007 May 13 18:53:46 NAiL: any news on the ram? :) May 13 21:21:54 Someone should add http://www.pjsip.org/ to optware May 13 22:04:54 pluesch0r: yes, I've removed it, and gotten a bit further May 13 22:05:02 right now, I'm having problems with gphoto2 May 13 22:05:08 which is required by sane May 13 22:06:22 NAiL: ow. May 13 22:07:14 Probably solvable, I'm testing on a different arch/distribution May 13 22:10:15 i really appreciate your help. :) May 13 22:10:30 np, I want scanner-stuff working too May 13 22:10:43 got a scanner that I want to attach to one of my nas-devices again at a later point May 13 22:10:54 only problem is that it only works when running in little-endian mode May 13 22:11:00 hm May 13 22:11:41 quite a few of the backends appear to have endianness-issues May 13 22:11:50 which result in screwy images May 13 22:12:09 that would explain the problem that i've got with my scanner. May 13 22:12:26 have you tried running the slug in LE? May 13 22:12:37 ie, slugos/le or "proper" debian? May 13 22:12:41 and also the patch-line that patches exactly the backend that i need to use in my current setup. May 13 22:13:00 for the current openslug saned, that is. May 13 22:13:10 NAiL: i installed debian. that's where the screwed images occur. May 13 22:13:20 how did you install debian? May 13 22:13:21 everything's fine in openslug, though. May 13 22:13:29 oh, it is? May 13 22:13:32 yeah. May 13 22:13:36 hang on... May 13 22:13:44 so it's the debian saned that is screwy? May 13 22:14:01 i've got a canon canoscan lide 50 over here .. works like a charm with openslug. doesn't work with debian (screwed images). May 13 22:14:09 aha May 13 22:14:13 and i've got a canon canoscan lide 25, too. that one isn't recognised by openslug. May 13 22:14:16 well, in that case I can't help you May 13 22:14:21 it's recognised in debian, but the image is also screwed. May 13 22:14:24 waitwait, i'm not yet finished. May 13 22:14:47 and *that's* why i figured that i need a newer version of saned on openslug. so that the lide25 works too. :) May 13 22:14:54 aaah May 13 22:14:56 ok May 13 22:15:15 I can't help you with the debian screwy stuff, but hopefully we can get your lide25 working May 13 22:15:22 so .. i don't need my debian issue to be fixed .. for reasons beyong my knowledge, the debian slug is very very slow. May 13 22:15:28 yeah. May 13 22:15:38 at least, i bought it 'cause the sane homepage said it was 100% supported. May 13 22:15:40 the debian slug is slow because of two things May 13 22:15:57 low mem and low cpu? May 13 22:15:57 :) May 13 22:16:32 It's running in LE mode (network speed -30%), and debian insists on using hard-float which has to be handled with exceptions in the kernel, slowing down everything that does floatingpoint math a *lot* May 13 22:17:07 oh. May 13 22:17:41 the floatingpoint-bit should be fixed (more or less) once debian migrates to arm eabi, which is the new binary stuff May 13 22:17:46 sigh. religion. a terrible topic. May 13 22:17:49 but that'll probably be a long while to wait May 13 22:19:17 i'd rather set my hopes on you. :) May 13 22:19:21 hehe May 13 22:19:57 this is actually the first time in my whole open source history that someone stepped in and did something for me. May 13 22:20:08 aah May 13 22:20:15 not kidding. :) May 13 22:20:23 i know, i know, it's not 'for me' etc. May 13 22:20:38 anyways. after midnight, no native speaker. May 13 22:20:42 well, there's a lot of helpful people around. You just gotta find the persons that have a vested interest in fixing it ;) May 13 22:20:46 thanks, man. May 13 22:20:57 I'm no native speaker either, and it's 00:21 ;) May 13 22:21:06 same timezone. scary. :) May 13 22:21:11 haha May 13 22:21:15 there are a few europeans in here May 13 22:21:16 nighty! May 13 22:21:18 oddly enough ;) May 13 22:21:19 nite May 13 22:21:40 \o May 13 22:22:32 NAiL, got some urls re: foonas ? May 13 22:23:36 xkr47: nope, got any questions? May 13 22:24:19 you said it was NAS-centric, does it mean it aims on services related to distributing files in different ways on a LAN ? May 13 22:25:02 nas-centric as in built for nas-devices mostly ;) May 13 22:25:21 more generic than slugos ? May 13 22:25:48 well, slugos is pretty generic May 13 22:26:00 foonas has a modular webinterface May 13 22:26:08 so it's customizeable May 13 22:26:46 something between unslung and slugos ? May 13 22:26:56 kinda May 13 22:27:06 It's what I always wanted slugos to be ;) May 13 22:27:10 :D May 13 22:30:33 * xkr47 tries to get a grip of whether he wants to use SlugOS/BE or /LE May 13 22:31:00 BE is fast but has portability issues, LE is slower but less endianness-issues May 13 22:31:07 from what I understand it seems that you can get the optware packages that are also available for unslung through this procedure: http://www.nslu2-linux.org/wiki/Optware/Slugosbe May 13 22:31:23 LE is slower? is that a cpu feature ? May 13 22:31:56 or "well you end up using generic arm code which is not optimized for the cpu" ? May 13 22:32:09 LE is slower because the cpu has to byteswap network traffic (which is BE) May 13 22:32:52 but without the byteswapping part, it's just as fast ? May 13 22:33:11 I *believe* so. May 13 22:33:20 * xkr47 wants to believe too ;) May 13 22:33:39 and afaik, there are no specific optimisations for the ixp4xx May 13 22:33:47 it's an armv5t May 13 22:34:22 how about the dsp stuff it's supposed to include ? May 13 22:34:31 dsp? May 13 22:35:35 "The IXP420 has some DSP features like ... May 13 22:35:41 http://www.nslu2-linux.org/wiki/Info/CPUOverview May 13 22:35:51 ah intel-only :) May 13 22:35:52 sorry May 13 22:36:38 not really applicable on the slug I think May 13 22:36:53 is the processor in either LE or BE mode or can you swap mode easily ?-) May 13 22:37:09 You don't want to switch while it's running ;) May 13 22:37:17 that's done right before the kernel runs May 13 22:37:20 too many cycles ? May 13 22:37:44 no, the kernel is either LE or BE -compiled, so switching at that point is impossible May 13 22:38:01 some hard-coded instructions for network stuff would be cool :) May 13 22:38:14 wellwell May 13 22:38:22 I guess it's only a matter of a few percents May 13 22:38:39 the network speed difference is not just a few ;) May 13 22:38:48 oh? May 13 22:38:54 more like 30% May 13 22:39:00 oh my May 13 22:39:23 yes May 13 22:39:24 I got some 1.6 MB/s when scp:ing files from the flash to another machine May 13 22:40:07 (and 7.00M/s when I tried: dd if=/dev/sda1 of=/dev/null) May 13 22:42:15 * xkr47 wonders what the thumb cpu instruction set is May 13 22:44:06 thumb is a 16(?)-bit reduced instruction-set May 13 22:44:25 as opposed to the regular 32-bit May 13 22:44:27 smaller binaries than arm ? May 13 22:44:45 yes May 13 22:45:00 possibly also speed benefits, not entirely sure about that bit though May 13 22:45:01 how about speed ? May 13 22:45:03 right May 13 22:46:14 are there any stats on how many people use the various variants of firmware ? May 13 22:49:56 www.nslu2-linux.org/presentation.pdf May 13 22:50:10 oh yeah now I remember :D May 13 22:50:20 read that twice already :) May 13 22:50:36 great stuff May 13 22:50:49 NAiL: there are people who have done BE/LE speed measurements on nslu2, and found negligible difference due to other slowness in the system May 13 22:50:57 really? May 13 22:51:07 interesting May 13 22:51:17 well, I remember blaster8 quoting you on that the other day May 13 22:51:20 actually May 13 22:51:33 trouble is I can't actually find the person who did it now May 13 22:51:41 neither can i find the person who said 30% May 13 22:52:09 so there are *no* reliable figures as far as I am concerned, and someone should document a performance measurement and comparison on the wiki May 13 22:52:40 Well, I benchmarked samba and got approx. 30% speed difference May 13 22:52:57 let's pick up some "profiling linux kernels on arm system" document and .. ;) May 13 22:53:21 NAiL, maybe samba does nasty/slow swapping in userspace May 13 22:53:29 in addition to the kernel May 13 22:53:36 byteswapping* May 13 22:56:40 good night, sleep now! May 13 23:10:08 * xkr47 is addicted :P May 13 23:13:20 what is the difference between the optware packages available from say the slugosbe and unslung feeds ? May 13 23:20:42 xkr47: kernel version, libc version, packaging, lots of stuff. May 13 23:20:47 gcc version May 13 23:21:03 do they install packages into different places ? May 13 23:21:28 (thx for the info) May 13 23:21:48 yes May 13 23:27:36 * xkr47 re-attempts sleep May 13 23:28:02 I'd like to go flash my slug with slugos/be but then I won't wake up tomorrow :P May 13 23:28:12 (one thing leads to another :) May 13 23:28:27 --> May 13 23:57:16 03bzhou * r6035 10optware/trunk/make/py-moin.mk: py-moin: 1.5.7 -> 1.5.8 **** ENDING LOGGING AT Mon May 14 02:59:57 2007