**** BEGIN LOGGING AT Mon Feb 28 00:40:52 2005 Feb 28 00:41:02 test logging Feb 28 00:41:17 works Feb 28 00:41:17 test Feb 28 00:41:35 seems so Feb 28 00:42:09 and dammit who took my username Feb 28 00:52:44 ByronT-zzZZ-Away, are you still up! Feb 28 00:52:57 or did you wake up and have to come out here. Feb 28 01:04:20 hmm - my bluetooth adapter doesn't work when connected to my 7 port hub Feb 28 01:05:14 but it does work when directly connected to the slug Feb 28 01:06:15 also, I plugged two disks into prodslug, and rebooted, and the wrong disk came up as sda1, so [g2]'s emprical data from the other day was just luck. Feb 28 01:06:46 so they dont' come up in the same order each time? Feb 28 01:07:03 nop Feb 28 01:07:18 hmm I was wondering about that. Feb 28 01:07:23 how does it decide? Feb 28 01:07:25 and they are both ehci devices? Feb 28 01:07:42 Can we force it via uuid or somehting? Feb 28 01:07:49 or label Feb 28 01:16:44 well, the bb world failed Feb 28 01:16:52 restarting to see if it was transient Feb 28 01:18:27 in a screen session? Feb 28 01:18:39 ah - it's dyoung's fault Feb 28 01:19:02 it was? Feb 28 01:19:11 umask 002 please Feb 28 01:19:11 mt-daapd broke? Feb 28 01:19:18 oh. ;-) Feb 28 01:52:44 ctorrent failed Feb 28 01:52:51 looks like world is not going to be a goer Feb 28 01:52:58 time for an openslug-packages bb Feb 28 01:54:11 :( Feb 28 01:54:33 jacques, its been running about 94% nice Feb 28 01:54:40 but 0% idle. Feb 28 01:55:49 is there a util that we can use to see the temps? Feb 28 03:19:15 I'm proposing that we put openssh in openslug instead of dropbear. Thoughts? Feb 28 03:21:53 I was thinking the same thing, but i thought i was being selfish. Feb 28 03:22:06 Any idea why [g2] used dropbear? Feb 28 03:22:22 I presume because that's what was in task-bootstrap Feb 28 03:24:09 I'm openslugged with 1.5 btw. Serial-less. Feb 28 03:24:30 now, how to ipkg install openssh when logged in via dropbear ... Feb 28 03:24:51 I could do it by serial, but that's cheating Feb 28 03:30:10 hmm - I need ipkg-utils on the slug so I can create a local file feed and just scp things up to it Feb 28 03:30:43 hmm - that might need python ... Feb 28 03:44:58 ok, I just installed openssh, and the postinst referred to /opt/bin/... ! Feb 28 03:45:07 something from unslung has got into openslug! Feb 28 03:46:04 weird. Feb 28 03:46:12 very Feb 28 03:46:19 how is that even possible? Feb 28 03:46:25 dunno Feb 28 03:46:27 sure you didnt use the wrong ipk by mistake? Feb 28 03:53:11 oh, I just realised what happened! Feb 28 03:53:47 I forgot to clean my unslung disk (it's conf partition) before using it for openslug, and it would have had a /usr/lib/ipkg directory, containing postinst scripts. Feb 28 03:53:55 ok, false alarm :-) Feb 28 03:54:08 heh Feb 28 03:54:26 hmm - guess I need to blow away this disk and start again ... Feb 28 03:54:29 The Turnip page was pretty explicit about starting with a clean disk. ;-) Feb 28 04:27:01 ok, back now with a clean disk and openssh Feb 28 04:28:14 now, where was I ? Feb 28 04:30:28 ... "How to ipkg install openssh when logged in via dropbear" Feb 28 04:31:07 yeah, did that by serial ... :-) Feb 28 04:31:46 heh. Cheater. ;-) Feb 28 04:32:28 that's right - python for ipkg-make-index ... Feb 28 04:44:33 now, which bb package would have "ar" in it ? Feb 28 04:45:02 need python, tar, ar for ipg-make-index (so far) Feb 28 04:45:44 binutils ... Feb 28 04:49:35 Thanks for your miau startup script. I pilfered it for mt-daapd. Feb 28 04:50:00 np, I pilfered it from lirc Feb 28 04:55:24 ipkg-make-index works on the slug Feb 28 04:55:31 hey [g2] Feb 28 04:56:03 <[g2]> morning Feb 28 04:56:10 <[g2]> ipkg-make-index ? Feb 28 04:56:29 <[g2]> Does that build the Packages index file ? Feb 28 04:57:19 yep Feb 28 04:58:01 I'm scping ipks up to /home/feed on prodslug, and wanted to make a Packages file there so I could put "src local file://home/feed" in /etc/ipkg.conf Feb 28 04:58:17 <[g2]> cool. Feb 28 04:58:37 Funny how that works. Feb 28 04:58:39 <[g2]> I just bb openslug-image and then cp the whole ipk directory Feb 28 04:59:03 <[g2]> can the src to the local dir Feb 28 04:59:09 needed python, tar, binutils, ipkg-utils Feb 28 04:59:19 I wish this would parse faster. Feb 28 04:59:24 <[g2]> actually, I've been cp the Packages to wher the web server is running Feb 28 04:59:41 <[g2]> dyoung, parse what ? Feb 28 04:59:47 <[g2]> the oe Feb 28 04:59:57 "handling bitbake files" Feb 28 05:00:08 <[g2]> beewoolie's parser takes 1 second Feb 28 05:00:15 [g2]: this is for new packages that are not in openslug-image Feb 28 05:00:18 <[g2]> 6 seconds for the full 2K packagees Feb 28 05:00:43 I guess I should just install it. Feb 28 05:01:02 <[g2]> dyoung, well it doesn't work with OE, its just POC Feb 28 05:02:44 now, to remember why I needed a local feed .... Feb 28 05:02:55 <[g2]> native build Feb 28 05:03:08 [g2]: disk ordering is not predictable Feb 28 05:03:48 <[g2]> Actually the testing I've done as long as you don't have a single disk in port 2 your probably be ok Feb 28 05:04:08 <[g2]> Although I didn't try two disks or two flashes Feb 28 05:04:09 you've been lucky Feb 28 05:04:43 two disks - one at boot, hotplugged the second. sda and sdb Feb 28 05:04:57 (disk1 for sda, disk2 for sdb) Feb 28 05:05:09 rebooted with both still plugged into the same ports. They swapped Feb 28 05:05:21 <[g2]> Does mount support that ssid thingy ? Feb 28 05:05:26 <[g2]> label Feb 28 05:06:15 dunno Feb 28 05:06:23 <[g2]> If you mount by label (as long as it's ext3) that'll work if the label mounting is supported Feb 28 05:06:28 <[g2]> via the mount command Feb 28 05:08:25 [g2]: sources dir for openslug will be ready soon Feb 28 05:08:37 do we need to remove the intel stuff? Feb 28 05:09:52 <[g2]> I'll need to check the license again Feb 28 05:11:24 <[g2]> If forget what exactly it was referring to, but there was an exception in the Intel license for something like significant improvements or enhancements of sw Feb 28 05:12:00 anyone looked at CSR 1.5 yet? Feb 28 05:12:05 <[g2]> Since were going from a 2.4 to at 2.6 kernel I think that's a significant enhancement, although I am not a laywer Feb 28 05:12:14 <[g2]> jacques, ! Feb 28 05:12:19 * [g2] hugs jacques Feb 28 05:12:26 safer to leave it out Feb 28 05:13:19 drat that took more time than i expected. nail is gonna have to wait till tommrow. Feb 28 05:13:26 night all. Feb 28 05:13:35 [g2]: did you intentionally want dropbear instead of openssh? Feb 28 05:13:41 <[g2]> night dyoung -- sweet dreams Feb 28 05:14:04 <[g2]> rwhitby, nope just that's the default Feb 28 05:14:10 dyoung and I would like to have openssh in the base firmware instead of dropbear Feb 28 05:14:20 <[g2]> fine by me for the switch Feb 28 05:14:30 for 1.6 .... Feb 28 05:14:38 (after beta release) Feb 28 05:15:27 [g2], I openslugged my new nslu2. Feb 28 05:15:46 <[g2]> dyoung, Congrats! Feb 28 05:16:11 like I took it out of the box made the network settings and flashed it per instructions. Feb 28 05:16:32 I'll finish following the turnip instructions tommrow... Feb 28 05:16:45 <[g2]> rwhitby, I think leaving it out (Intel firmware) would be prudent Feb 28 05:17:25 <[g2]> I don't know how OE pulls the sw and whether it prompts one about the license stuff Feb 28 05:17:46 it gets it direct from intel Feb 28 05:17:48 jacques, I started to look at csr-1.5.. we need some bits from it to load the npe startup stuff in apex. Feb 28 05:17:54 and kergoth makes sure that it's not mirrored Feb 28 05:18:10 but it was making my head spin, so I stopped for a few days. Feb 28 05:18:38 that vegetable name is gonna stick .... Feb 28 05:18:59 ;-) Feb 28 05:22:11 dyoung, yeah I know the feeling Feb 28 05:23:25 hmm there's a new gcc patch in oe - I wonder what it's about Feb 28 05:23:30 checking Feb 28 05:24:13 damn, no comment Feb 28 05:26:15 <[g2]> hey what's the standard unslung upgrade path ? Feb 28 05:27:04 ixp400AccessLibrary-1_4.zip Feb 28 05:27:04 ixp400LinuxEthernetDriverPatch-1_1.zip Feb 28 05:27:11 are those the two files I need to delete? Feb 28 05:27:24 redboot Feb 28 05:27:49 <[g2]> Those and there may be 1 other patch Feb 28 05:27:52 openslug has no web interface, so one must upgrade via redboot (to ensure that they can recover via redboot) Feb 28 05:28:03 ixp425_eth.c.patch ? Feb 28 05:28:11 isnt that gpl though? Feb 28 05:28:17 <[g2]> that sounds familiar Feb 28 05:28:19 <[g2]> :) Feb 28 05:28:21 the ethernet part was gpl'd Feb 28 05:30:52 [g2]: sources-openslug-1.5-beta in /home/openslug/firmware on nudi Feb 28 05:31:59 <[g2]> rwhitby, I'm assuming that's a statement Feb 28 05:32:09 yep Feb 28 05:32:22 also making a copy of the repo for you Feb 28 05:32:24 <[g2]> nod. thought so. Feb 28 05:36:29 ok, what's the timeline for release? Feb 28 05:38:44 <[g2]> Good Question Feb 28 05:39:17 <[g2]> We'll need to noodle on the IXP driver issue Feb 28 05:39:28 <[g2]> and possibly talk to gelloglue Feb 28 05:39:30 what's the issue? Feb 28 05:40:25 <[g2]> The binary can't contain the Ixp driver right ? Feb 28 05:40:52 <[g2]> So unless they have a usb-ethernet adapter the box is useless Feb 28 05:41:14 huh? Linksys firmware and Unslung include the ixp driver. Feb 28 05:42:19 <[g2]> Linksys is Cisco and I'm sure Intel if fine with them Feb 28 05:42:45 <[g2]> Unslung copies the kernel from the Linksys release right ? Feb 28 05:42:59 copies the ixp modules, yes Feb 28 05:43:27 <[g2]> So Unslung is just repackaging something the user already has from Linksys Feb 28 05:43:40 <[g2]> same identical binary Feb 28 05:43:52 <[g2]> OpenSlug is building a different binary Feb 28 05:44:05 <[g2]> OpenSlug has built a different binary Feb 28 05:45:43 ok, this has totally been off my radar. are you saying that we've gone to all this trouble knowing that we cannot release the binary? Feb 28 05:47:10 <[g2]> No, I don't know that we cannot release the binary, but I don't know that we absolutely can Feb 28 05:47:56 hmm - a pretty big item to not be on the punchlist :-) Feb 28 05:48:12 <[g2]> I've state several times about checking the license agreement and it's been in the punch list Feb 28 05:48:19 <[g2]> agreements Feb 28 05:48:30 you're right - my bad Feb 28 05:48:34 <[g2]> maybe I could have been clearer about that Feb 28 05:49:01 I thought you were only talking about redistribution of the source (a-la what kergoth had to do to stop the source being mirrored) Feb 28 05:49:28 <[g2]> I wasn't very clear about it Feb 28 05:49:54 <[g2]> partly not to focus on the issue and scare ppl away Feb 28 05:50:13 dammit Feb 28 05:50:19 openslug build failed Feb 28 05:50:37 <[g2]> jacques, when did you last pull ? Feb 28 05:50:46 a few seconds before I started the build Feb 28 05:51:34 <[g2]> jacques, did you wipe tmp ? Feb 28 05:51:51 yes. Feb 28 05:51:53 <[g2]> I pulled last night after the kergoth gcc change and rebuilt Feb 28 05:52:07 <[g2]> but I did not wipe tmp Feb 28 05:54:45 I've rebuilt twice clean since then Feb 28 05:54:50 where's it failing? Feb 28 05:55:18 in glibc Feb 28 05:55:46 weird Feb 28 05:55:48 in the packaging stage Feb 28 05:55:50 OSError: [Errno 2] No such file or directory: '/home/packages/oe/nslu2/bb_openslug_default/tmp/staging/shlibs' Feb 28 05:58:06 well, looks like openslug is a source-only distro then .... Feb 28 05:59:09 <[g2]> rwhitby, we can certainly do binaries for usb-ethernet, usb-BT adapter and possibly usb-serial Feb 28 05:59:45 <[g2]> with the old hotplug usb-ethernet booted on first boot Feb 28 06:00:04 <[g2]> with an alias change it will do so with now Feb 28 06:00:46 <[g2]> We could easily ship the beta Intel free with known good adapters Feb 28 06:04:16 I've lost all interest in a binary now .... Feb 28 06:05:26 * [g2] comforts rwhitby Feb 28 06:07:26 night all Feb 28 06:08:30 <[g2]> night Feb 28 06:16:58 <[g2]> rwhitby-asleep, there is a 3rd possibility Feb 28 06:17:15 <[g2]> regarding the intel drivers Feb 28 06:17:45 <[g2]> on OpenWRT someone created wrapper around a 2.4 driver that ran on 2.6 experimentaly Feb 28 06:18:20 <[g2]> we could in theory write a wrapper around the 2.4 driver that ran on 2.6 Feb 28 06:18:49 <[g2]> it'd probably be a bunch of work and I'm not really interested in doing that, but it is an option Feb 28 09:53:46 [g2] where is the offending license? Feb 28 09:54:44 <[g2]> ka6sox-zzzz, I'm looking back a the licensing now Feb 28 09:55:35 stupid question...can we just use the linksys kernel (with -able style) and continue? Feb 28 09:55:42 <[g2]> I hope I'm wrong about the issue Feb 28 09:55:59 <[g2]> ka6sox-zzzz, We'd have to put wrappers around that stuff Feb 28 09:56:12 okay I saw that. Feb 28 09:58:55 I'm looking/searching for the licensing issue as well (have not been able to find it) Feb 28 09:59:27 ByronT is in the intel provided modules Feb 28 09:59:39 s/is/its Feb 28 10:01:20 for setting up CSR and NPE Feb 28 10:25:18 I think we're clean Feb 28 10:25:44 I just went to Intel's site and tried to download V1.4 device driver and was presented with the GNU GPL Feb 28 10:26:45 <[g2]> ByronT, I've been doing the same thing Feb 28 10:26:56 <[g2]> I noticed that the 1.4 stuff appears to be GPL Feb 28 10:27:13 1.4 device driver is published under GPL Feb 28 10:27:18 so maybe we are clean. Feb 28 10:27:19 <[g2]> however, the Readme.txt file when you unzip it has Intel's licensing info in it Feb 28 10:27:33 cruft Feb 28 10:27:45 <[g2]> We maybe clean.... And the licensing may have changed between 1.1 and 1.4 Feb 28 10:27:58 yeah. Feb 28 10:28:06 <[g2]> HOWEVER, we still need to sort it out and verify that it's correct Feb 28 10:28:09 I'll see what I can find out Feb 28 10:29:29 [g2] the only README that I get with either patch or device driver points to web Feb 28 10:30:47 <[g2]> ByronT, unzip ixp400AccessLibrary-1_4.zip Feb 28 10:36:34 that one I don't have... but I don't see a link for it either Feb 28 10:36:53 its in the other download Feb 28 10:41:04 I must be missing something because I only see the Linux integration patch and Linux ethernet patch Feb 28 10:41:31 can you link the download page? Feb 28 10:41:32 the linux integration patch talks about the library that *has* to be obtained from Intel Feb 28 10:42:43 and that it has to be compiled as a module to not violate the GPL. Feb 28 10:43:11 + The library can be compiled only as a module, as linking it Feb 28 10:43:11 + statically into the kernel would violate the GPL. The library module Feb 28 10:43:11 + will be called ixp400.o. Feb 28 10:43:22 (taint the kernel) Feb 28 10:43:51 (this is the csr stuff) Feb 28 10:44:55 <[g2]> Right that's what we are doing is compiling it as a module Feb 28 10:45:20 but can we distribut the compiled module is my question. Feb 28 10:45:47 <[g2]> That's what we are checking into Feb 28 10:46:09 <[g2]> What all the license agreements are for the interface SW and the binary stuff Feb 28 10:48:09 [g2] looks like v1.2 has a GPL wrapper Feb 28 10:48:32 from the 1.5 release Feb 28 10:49:03 <[g2]> ka6sox, it's quite possible Feb 28 10:49:06 flood: Feb 28 10:49:08 +# "Adapter" makefile Feb 28 10:49:08 +# Feb 28 10:49:08 +# This makefile serves as the interface between the linux kernel Feb 28 10:49:08 +# makefiles and Intel's xscale_sw makefile. It's purpose is to allow Feb 28 10:49:08 +# the 'make modules' rule to build the access library modules, and to Feb 28 10:49:09 +# get them included in the ramdisk image produced by the 'make Feb 28 10:49:11 +# ramdisk' rules. Feb 28 10:49:29 <[g2]> I think jacques and dyoung have looked at that a little Feb 28 10:49:34 and then it goes into a GPL compatible license Feb 28 10:50:28 okay this one looks clean... Feb 28 10:50:41 intel grants rights to redistribute. Feb 28 10:50:48 source and binary Feb 28 10:51:04 since its a module it doesn't taint the kernel. Feb 28 10:51:22 that's what I am reading as well... Feb 28 10:51:39 they are very careful to point it out. Feb 28 10:51:59 not so in v1.1... Feb 28 10:53:29 +# Intel hereby grants Recipient and Licensees a non-exclusive, worldwide, Feb 28 10:53:30 +# royalty-free patent license under Licensed Patents to make, use, sell, offer Feb 28 10:53:30 +# to sell, import and otherwise transfer the Software, if any, in source code Feb 28 10:53:30 +# and object code form. Feb 28 10:53:52 its a stupid patent license.. Feb 28 10:54:19 which could be revoked..but probably not. Feb 28 10:55:08 (all this before my first cup of coffee) Feb 28 10:56:47 ok... I've got to get some *paying* work done.... later all! Feb 28 10:57:47 laters Feb 28 10:58:41 is it worth doing a diff of the 2 patches? Feb 28 10:59:11 to see if anything significant has changed (other than the license) Feb 28 11:00:15 <[g2]> ka6sox, dunno, jacques and dyoung have looked at it. We've got lots of options and we need to check it out a little more Feb 28 11:00:43 this does look better (from a license perspective) Feb 28 11:00:52 but don't know what this breaks. Feb 28 11:09:27 ugh...the NPE driver is still encumbered. Feb 28 11:10:11 <[g2]> ka6sox, where do you see the reference to the NPE being encumbered Feb 28 11:10:44 + Note that this driver is Intel proprietary code and must be obtained Feb 28 11:10:45 + from Intel Corp. See the README files that accompany the IXP400 Feb 28 11:10:45 + software or Linux Ethernet Driver distribution for details. Feb 28 11:10:56 v1.2 Feb 28 11:11:11 <[g2]> Where is that at ? Feb 28 11:11:52 ixp400LinuxEthernetDrive-1_2.patch Feb 28 11:12:12 ixp400LinuxEthernetDriver-1_2.patch Feb 28 11:12:24 <[g2]> line # ? Feb 28 11:12:41 21 Feb 28 11:14:10 * ka6sox wonders if there is any other device that sorta looks like this one and has a GPL driver that could be modified. Feb 28 11:15:23 <[g2]> ka6sox, that also predates the 1.4 release iirc Feb 28 11:15:30 yes Feb 28 11:15:32 it does Feb 28 11:16:15 <[g2]> so the question is does the GPL acknowledgement on the dl page override or just cover the non-Intel stuff Feb 28 11:16:52 fsf question? Feb 28 11:17:40 <[g2]> actually I think PJ (of Groklaw fame) was part of the Legal firm supporting OpenSource devs Feb 28 11:18:11 yeah...but she doesn't render legal opinions.. Feb 28 11:18:35 she has been very careful not to even have the appearance of that. Feb 28 11:18:36 <[g2]> I think if we gather the information together, it's like $50 a year as a developer or open-source organization and we could probably get real (and very qualified IMHO) advice Feb 28 11:18:54 that is money well spent. Feb 28 11:19:35 <[g2]> however, I think we have a bit of due-diligence to do first Feb 28 11:19:57 isn't that we're doing now? Feb 28 11:20:09 <[g2]> exactly. Feb 28 11:20:18 <[g2]> I didn't say we weren't doing it Feb 28 11:20:41 its a "click through" eula that we are being asked to sign up for. Feb 28 11:21:10 <[g2]> I think we did Feb 28 11:21:49 and most companies are more concerned about enforcing the "click through" eulas over written ones Feb 28 11:22:08 the question is is Intel one of those companies... Feb 28 11:22:55 <[g2]> Well we need to read through the doc and references and then ask gelloglue and possibly dsaxena their thoughts Feb 28 11:23:08 <[g2]> and also checkout the 1.5 version Feb 28 11:24:26 1.5version? Feb 28 11:25:57 <[g2]> Yes we're currently using the 1.4 version of the Intel CSR sw Feb 28 11:26:09 <[g2]> there's a version 1.5 Feb 28 11:26:13 yea.. Feb 28 11:26:28 thats where i've been reading the stuff I pasted in from. Feb 28 11:26:40 the first one is the CSR and the second is the NPE. Feb 28 11:26:57 (from V1.5 of the release Feb 28 11:27:08 <[g2]> So your saying that 1.5 still has the bit about the NPEs Feb 28 11:27:08 the patches are called 1.2 Feb 28 11:27:12 <[g2]> nod Feb 28 11:27:13 yes Feb 28 11:27:47 the readme points back to the website (as bryont said) Feb 28 11:27:56 and the website says GPL. Feb 28 11:28:16 but I suspect that is for the "non-proprietar" bits only. Feb 28 11:28:44 <[g2]> right and without those, the driver doesn't work Feb 28 11:28:55 yes..the "library" Feb 28 11:30:29 I go back to the question is this a eepro100 stripped down? or something completely different. Feb 28 11:30:46 and is there any docs related to the registers in the NPE. Feb 28 11:31:22 s/is/are Feb 28 11:37:25 <[g2]> the NPE is IP Feb 28 11:40:31 okay so the use the API's to control what can be done. Feb 28 11:40:32 :P Feb 28 11:45:44 this goes to an article that I read last night on /. about Bios. Feb 28 12:01:05 Any problems with me asking Intel Developer Services to clarify distribution? Feb 28 12:01:29 uh...that is a good question. Feb 28 12:02:32 <[g2]> ByronT that may work. Feb 28 12:02:54 <[g2]> gelloglue I'm sure would have a bunch of insight to the answer Feb 28 12:08:51 the dice have been thrown.... Feb 28 12:09:17 <[g2]> Craps is usually good on the first roll Feb 28 12:18:41 :) Feb 28 12:20:13 I've asked Intel Developer Services for guidance (not as an official rep for nslu2-linux.org) - just as someone wanting to develop a distro, and needed guidance on the distribution Feb 28 12:20:43 ByronT, very good. Feb 28 12:21:57 supposed to get response in 1 business day Feb 28 16:58:42 pb applied a change which fixed my glibc packaging problem from last night Feb 28 17:03:21 "morning" jacques Feb 28 17:03:29 hi ByronT :-) Feb 28 17:04:04 <[g2]-away> I'm finding some interesting things about OpenSlug Feb 28 17:04:08 just in time for me to be called away... (birthday party for niece) Feb 28 17:04:19 <[g2]-away> go EnJOY Feb 28 17:05:25 [g2]- I will have another - virgin - drive to hook up with OS by Thursday... (for another OS install instruction set) Feb 28 17:05:42 <[g2]> ByronT, cool Feb 28 17:05:55 <[g2]> things seem to be picking up Feb 28 17:06:09 no response from Intel as of yet Feb 28 17:06:23 <[g2]> hey we're only 6 hours in right ? Feb 28 17:06:34 <[g2]> 5 ? Feb 28 17:07:07 yeah... but I'm the impatient sort - as if everyone here hasn't noticed that... ;) Feb 28 17:09:45 <[g2]> jacques is there a way to clear the dmesg buffer Feb 28 17:10:03 not that I know of Feb 28 17:10:16 why would you want to do that? Feb 28 17:10:30 lol Feb 28 17:10:32 dmesg -c Feb 28 17:10:38 I guess lots of ppl want to do that Feb 28 17:10:51 -c Clear the ring buffer contents after printing. Feb 28 17:10:51 <[g2]> COOL thx Feb 28 17:10:58 <[g2]> it worked Feb 28 17:11:13 i learned something too :-) Feb 28 17:11:33 What are we learning? Feb 28 17:12:00 I learned that there is indeed a way to clear the dmesg buffer Feb 28 17:12:03 dmesg -c Feb 28 17:12:11 <[g2]> How to HAMMER OpenSLUG :) Feb 28 17:12:32 wow, i gotta remember -c. Feb 28 17:13:02 [g2], ?? Feb 28 17:13:08 can it handle? Feb 28 17:13:18 <[g2]> nope not yet :) Feb 28 17:13:29 <[g2]> I've got some facinating data Feb 28 17:13:34 what did you do to it? Feb 28 17:13:59 <[g2]> Ok.... Feb 28 17:14:24 <[g2]> So I setup thttpd on OpenSlug and I've got it pointing to the sources directory on the hd Feb 28 17:14:34 <[g2]> it's booted and running off the hd as root Feb 28 17:14:52 <[g2]> that good timing Feb 28 17:15:17 <[g2]> so Most of the files are pretty small < 20MB Feb 28 17:15:33 <[g2]> however, a few are a little bigger > 30 MB Feb 28 17:15:41 <[g2]> so guess what ? Feb 28 17:16:08 the interrupt hack dont work? Feb 28 17:16:35 <[g2]> dyoung-web, It's not the network that having problems Feb 28 17:17:24 <[g2]> so anyway, the redhat file is like 18MB and I can wget if repeatedly at nearly line rate 10MB Feb 28 17:17:28 int26: nobody cared ? Feb 28 17:17:32 <[g2]> yep Feb 28 17:18:08 <[g2]> However, when I pull the linux kernel at 36MB we start to swap the rate falls down to < 1MB and even .5MB Feb 28 17:18:45 <[g2]> interesting thing is a minute or two can go by after the xfer and the rate on a small file (18MB) is still 1MB Feb 28 17:18:59 <[g2]> 20-30 minutes later it's back at 10MB Feb 28 17:19:43 <[g2]> maybe 5 or 10 minutes later (the drive spins down in 5 minutes) Feb 28 17:20:05 thats pretty bizzare. Feb 28 17:20:25 what if you turn off swap? Feb 28 17:20:32 <[g2]> 500 Feb 28 17:20:44 <[g2]> the big files don't transfer http error 500 Feb 28 17:20:49 <[g2]> Internal error Feb 28 17:21:36 Interesting. Feb 28 17:22:12 maybe thttpd is doing something stupid with memory allocation/utilization ? Feb 28 17:22:23 and youre wgetting from the same machine? or another client? Feb 28 17:22:42 <[g2]> just the pc connected to OpenSlug Feb 28 17:23:34 What if you rsync the same from from the pc (ie: bypassing thttpd) Feb 28 17:23:38 or even scp Feb 28 17:24:23 nc Feb 28 17:26:56 <[g2]> ok a straight cp -R will do it Feb 28 17:32:37 <[g2]> See http://www.nslu2-linux.org/wiki/OpenSlug/NobodyCaredIrqErrs Feb 28 17:35:57 [g2] is that booting with redboot or apex? Feb 28 17:36:03 <[g2]> redboot Feb 28 17:36:13 Are you in a position to try it with apex? Feb 28 17:36:14 * [g2] reboot without ixp (untained kernel) Feb 28 17:36:30 oh wait you cant. Feb 28 17:36:32 nevermind. Feb 28 17:36:36 <[g2]> soon Feb 28 17:37:24 if you boot with apex in ram can does the npe get initilaized? Feb 28 17:37:45 What I'm wondering is if there is a discrepancy in CPSR. Feb 28 17:37:56 <[g2]> I get the same thing with an untained kernel Feb 28 17:38:06 using USBNet? Feb 28 17:38:16 <[g2]> no networking :) Feb 28 17:38:20 <[g2]> I've got serial Feb 28 17:38:21 oh gotcha. Feb 28 17:38:59 ok in that case, can you try load boot openslug using apex from ram? Feb 28 17:39:19 <[g2]> I could but I don't think it matters Feb 28 17:39:47 <[g2]> APEX doesn't reset the hw Feb 28 17:40:14 <[g2]> If it was flashed it'd make a difference Feb 28 17:44:01 <[g2]> http://bhhdoa.org.au/pipermail/ck/2004-October/001422.html Feb 28 17:50:24 [g2] around still? Feb 28 17:50:31 <[g2]> nod Feb 28 17:50:41 wanna try some crazy stuff? Feb 28 17:51:26 <[g2]> dyoung-web, like ? Feb 28 17:51:45 modprobe-ing ehci-hcd with parameters. Feb 28 17:52:12 <[g2]> except that it's compile in right now :) Feb 28 17:52:24 <[g2]> compiled Feb 28 17:52:25 log2_irq_thresh = 1 for instance Feb 28 17:52:43 <[g2]> can I set that from /proc ? Feb 28 17:53:08 Not sure./ Feb 28 17:53:21 probably not? Feb 28 17:55:10 hmm, did we put it in the kernel didnt we. Feb 28 17:55:12 drat. Feb 28 17:55:46 I'm not sure how you twiddle that at compile time. Feb 28 17:56:05 oh thats what you just said. I wanst paying attemtion... Feb 28 17:57:32 <[g2]> static int log2_irq_thresh = 0; // 0 to 6 Feb 28 17:57:32 <[g2]> module_param (log2_irq_thresh, int, S_IRUGO); Feb 28 17:57:32 <[g2]> MODULE_PARM_DESC (log2_irq_thresh, "log2 IRQ latency, 1-64 microframes"); Feb 28 17:57:43 yeah, i know. Feb 28 17:57:52 <[g2]> you want it to be 2 Feb 28 17:57:56 its in the ehci startup code. Feb 28 17:58:04 <[g2]> it's easy to change Feb 28 17:58:11 the default is 0. Feb 28 17:58:18 <[g2]> I can change it Feb 28 17:58:27 <[g2]> takes 30 seconds Feb 28 17:58:34 <[g2]> and 2 minutes to reboot Feb 28 17:58:38 it does this later on Feb 28 17:58:40 temp |= 1 << (16 log2_irq_thresh); Feb 28 17:58:47 theres a plus in there Feb 28 17:58:53 16 plus log2.... Feb 28 17:59:04 give that a try. Feb 28 17:59:07 if I' Feb 28 17:59:26 m reading this right, it'll do *something*. ;-) Feb 28 18:03:53 <[g2]> flashing Feb 28 18:06:29 <[g2]> rebooting Feb 28 18:07:52 <[g2]> touched .sda2root rebooting Feb 28 18:12:08 <[g2]> dyoung-web, maybe a little better but still generated one Feb 28 18:12:26 what you change it to? Feb 28 18:12:31 <[g2]> 2 Feb 28 18:12:41 try something higher? Feb 28 18:13:35 <[g2]> I think it's ignoring an IRQ Feb 28 18:14:11 is it ignoring it, or is it getting past the watchdog timeout? Feb 28 18:14:46 what is the effect of the missed irq ? Feb 28 18:14:51 <[g2]> I think the soft_irq timer is set and the dog bites Feb 28 18:15:16 <[g2]> the ehci_watchdog Feb 28 18:15:22 if its easy to change, I'd almost change it to 6 and see what happens. Feb 28 18:19:27 <[g2]> dyoung-web, that may help but I don't think that identified the root cause of the issue Feb 28 18:19:57 Its making interrupts faster than ehci can handle them. Feb 28 18:20:30 <[g2]> the disk ? Feb 28 18:20:45 the disk or whatever usb device. Feb 28 18:21:05 I'm guessing btw.... Feb 28 18:21:08 <[g2]> How can it possibly do that ? Feb 28 18:21:42 I'm not sure but that seems to be what that gook is saying... Feb 28 18:21:51 <[g2]> I'm thinking there's a bug in the nesting of the interrupts Feb 28 18:22:46 <[g2]> I think we are waiting for I/O to complete and a timer int goes of and it's not happy about it Feb 28 18:23:21 that would have the same net result too yes... Feb 28 18:24:00 is it worse with zillions of little files vs one giant file? Feb 28 18:24:17 we need a knob to control the speed. Feb 28 18:24:36 so we know what speed something starts getting unhappy. Feb 28 18:24:56 <[g2]> that's the NEVER speed Feb 28 18:25:10 <[g2]> It's NEVER supposed to get unhappy like taht Feb 28 18:25:22 Right, but it does, so furthur investigation is required. Feb 28 18:25:52 and if you posit that its a timer issue, knowing the rate at which it starts getting u nhappy might help track down the source. Feb 28 18:26:19 <[g2]> the timer is from the stack dump Feb 28 18:27:13 <[g2]> when we are getting interruped from the idle state Feb 28 18:27:59 oh well, those were my $0.02 for today. Feb 28 18:29:59 <[g2]> well THX Feb 28 18:30:06 <[g2]> I appreciate your support Feb 28 18:32:06 what seems pretty clear is the dog bites because nobody card. Feb 28 18:32:09 cared. Feb 28 18:32:47 <[g2]> /* ehci could run by timer, without IRQs ... */ Feb 28 18:32:47 <[g2]> ehci_work (ehci, NULL); Feb 28 18:33:14 <[g2]> now that'd be interesting ..... Feb 28 18:33:33 <[g2]> IRQs for Net, but polled EHCI Feb 28 18:33:35 hahahah swap the timers of the eth and usb. Feb 28 18:34:19 :-\ Feb 28 18:34:26 This all sounds so familiar... Feb 28 18:34:32 <[g2]> I think the usb queues up a bunch of requests Feb 28 18:35:04 <[g2]> So a 1K freq wouldn't be so bad Feb 28 18:35:07 I was working with a 56k DSP that handled framed data. one of the classic error messages was similar to what we're seeing here. Feb 28 18:35:27 "Cant handle data 'cause data keeps coming!" Feb 28 18:35:53 <[g2]> I'm wondering if this got better with 2.6.11-rcx Feb 28 18:36:14 <[g2]> I'm gonna post a message to the arm-linux ml with the dump from the untained kernel Feb 28 18:36:35 <[g2]> no ixp driver loaded Feb 28 18:36:53 hey... Feb 28 18:37:05 anybody home? Feb 28 18:37:24 dyoung-web: ping Feb 28 18:37:48 bbiab...dinner Feb 28 18:37:59 uh... Feb 28 18:38:01 okay. Feb 28 18:38:31 I haven't seen any mail.... Feb 28 18:38:41 I wont have time to look at it today. Feb 28 18:38:52 okay Feb 28 18:38:53 np Feb 28 18:39:07 bbiaf...dinner (eyetalian Feb 28 18:45:03 dont forget to feed the dog. Feb 28 18:52:30 <[g2]> nite all Feb 28 18:58:42 so what's the latest on licensing? Feb 28 19:02:52 I just skimmed over that. Feb 28 19:02:58 I just skimmed over that. Feb 28 19:03:00 when I looked they seem to have made the CSR available Feb 28 19:03:08 but the NPE is a stickler. Feb 28 19:03:15 But basically I think its doable with later csr's Feb 28 19:03:29 unless we do a wrapper for the NPE> Feb 28 19:05:18 even with the 1.5 release. Feb 28 19:05:43 bryont asked about the gpl clickthru that we had to agree to before we could download the patches. Feb 28 19:05:51 (asked Intel) Feb 28 19:19:53 back from dinner Feb 28 19:43:39 hmmm...missed dyoung-web Feb 28 19:57:58 Okay, *now* i've seen everything. Feb 28 19:58:05 http://www.savetoby.com/ Feb 28 19:58:44 I'm totally flabberghasted. Feb 28 20:11:49 they've raised $16K too ... Feb 28 20:13:19 man I hate not having a log in here Feb 28 20:15:30 yeah...I thought nslu2-log was here. Feb 28 20:15:51 hmmm... and where can I find those logs? Feb 28 20:16:10 don't say it ... Feb 28 20:16:15 I'll search the wiki!!!!! Feb 28 21:23:30 $16K on diet FOOD!?!?!?! **** ENDING LOGGING AT Mon Feb 28 23:59:56 2005