**** BEGIN LOGGING AT Tue Jun 07 23:59:56 2005 Jun 08 00:14:49 perl testsuite finished, all passed except hostname which is because hostname isn't in /etc/hosts Jun 08 00:15:07 heh Jun 08 00:15:11 well..thats good. Jun 08 00:15:18 perl for OpenSlug...very nice Jun 08 00:15:21 yep, but I am concerned about the lfs tests Jun 08 00:15:34 they were all skipped, and they didn't use to be Jun 08 00:15:49 :( Jun 08 00:16:02 got up at 4am...wiped out now. Jun 08 00:16:43 any quick way to see if the feature is in there? Jun 08 00:16:45 `night ka6sox-zzzz Jun 08 00:17:00 I don't know of one Jun 08 00:17:38 testsuite alone took 74m24.220s Jun 08 00:18:52 wow Jun 08 00:18:56 okay niters. Jun 08 00:19:01 cya tommorrow. Jun 08 00:58:42 hello Jun 08 01:04:49 /dev/sda: Jun 08 01:04:49 Timing cached reads: 208 MB in 2.02 seconds = 102.97 MB/sec Jun 08 01:04:49 Timing buffered disk reads: 30 MB in 3.17 seconds = 9.46 MB/sec Jun 08 01:04:58 is it just me or is that improved a bit? Jun 08 01:05:28 jacques: what's that on? Jun 08 01:06:06 rwhitby-away, latest openslug Jun 08 01:06:22 Do we have a DNS proxy for openslug? Now that my internet router doesn't have custom firmware, I need AccessSlug to run a DNS server for the LAN .. Jun 08 01:06:53 caching proxy ? Jun 08 01:07:05 I use pdnsd on my local net Jun 08 01:07:12 that'd be fine, but not absolutely required. Jun 08 01:07:41 root@access:~# ipkg install perl Jun 08 01:07:41 Installing perl (5.8.4-r2) to root... Jun 08 01:07:41 Downloading http://ipkg.nslu2-linux.org/feeds/openslug/unstable/perl_5.8.4-r2_armeb.ipk Jun 08 01:07:41 Configuring perl Jun 08 01:07:41 root@access:~# perl Jun 08 01:07:42 perl: symbol lookup error: /usr/lib/libperl.so.5: undefined symbol: __udivsi3 Jun 08 01:07:43 root@access:~# Jun 08 01:07:52 did someone say we had a working perl on openslug? Jun 08 01:08:08 rwhitby-away, I built it native (ans said so :-) Jun 08 01:08:23 I have never been able to get the oe-built perl to actually work Jun 08 01:08:28 build yes, work no Jun 08 01:08:47 I wonce spent three straight days on it and only got a perl which could not load dynamic modules Jun 08 01:09:31 I just finished building 5.8.7 and testing it Jun 08 01:09:55 is pdnsd in OE ? Jun 08 01:10:16 hmm I haven't looked, if not ther emay be something else comparable Jun 08 01:10:35 hrw|work just put the flex patch back in, so flex builds in oe for arm* again Jun 08 01:10:59 ah, adns and dnsmasq Jun 08 01:11:08 those sounds promising Jun 08 01:26:34 jacques: can you pull into nslu2-linux? Jun 08 01:26:57 03jeanfabrice * 10unslung/ (13 files in 8 dirs): Control file now generated for dropbear, dhcp, ed, ffmpeg, flac, ghostscript Jun 08 01:28:41 has anyone try to run bind9 on the slug? Jun 08 01:28:55 jf-work: nice work :-) Jun 08 01:32:28 thx. I'm pushing them to wiley feed. Since IPK_VERSION has been pumped, can someone do the same for unslung to see if everything's ok ? (I don't know if jp30 will launch a compilation during its travel) Jun 08 01:42:55 03jeanfabrice * 10unslung/Makefile: Corrected type in wiley optware name ghostcript => ghostscript Jun 08 02:05:07 both adns and dnsmasq build for openslug .... Jun 08 02:06:28 looks like dnsmasq is what I wnat Jun 08 02:13:47 wonder why CIA is not reporting for me .... Jun 08 02:13:54 I added dnsmasq to openslug-packages Jun 08 02:15:10 damn emacs takes a long time Jun 08 02:16:09 well, not nearly as long as perl, but perl's in a league of its own Jun 08 02:18:24 maybe I'll get accessslug to do dhcp too with dnsmasq ... Jun 08 02:19:44 maybe rename it to LanSlug Jun 08 02:23:35 yay! I have emacs now Jun 08 02:24:21 do you have python to run ipkg-build? Jun 08 02:25:45 ah, nope I haven't built python yet (tho I have built it in the past on openslug native) Jun 08 02:26:24 I don't suppose python's in oe ... Jun 08 02:26:46 yeah it is hmmmm Jun 08 02:27:47 I'll try the oe way first Jun 08 02:29:32 weird, oe downloads python 2.4 to build 2.4.1 even tho there is a 2.4.1 tarball Jun 08 02:30:00 ooh, it's building 2.4.0 python-native ... also kinda weird Jun 08 02:32:57 even if it builds, it's a chore figuring out which os the 200 python ipks are necessary Jun 08 02:35:05 openslug feed is being updated now. any gotcha's we should be aware of, or should it all be sweet? Jun 08 02:35:27 (e.g. does the current openslug build work?) Jun 08 02:38:36 someone brave could try an ipkg update on openslug now .... Jun 08 02:39:29 I can start an openslug clean fresh build if you like, but it built fine last night Jun 08 02:41:16 if you have one installed, just do an ipkg update on it Jun 08 02:42:04 I'd rather not on this install.... this is my native devel install and it took me all day to get it setup Jun 08 02:42:22 oh ipkg update Jun 08 02:42:26 i thought you mean upgrade Jun 08 02:42:42 hmm, I setup my own local repos and commented out the main one Jun 08 02:42:57 what will the ipkg update prove? Jun 08 02:43:33 ok so there are only 49 python ipks Jun 08 03:01:19 I did mean upgrade :-) Jun 08 03:03:16 d`oh Jun 08 03:03:52 well that's why I made the multiple partitions, so I could have some "scratch" ones Jun 08 03:04:08 but I'm in the middle of building python native (as a test) right now Jun 08 03:04:11 I need more slugs Jun 08 03:04:15 starting DNS forwarder and DHCP server: dnsmasq... dnsmasq: cannot create DHCP packet socket: Address family not supported by protocol. Is CONFIG_PACKET enabled in your kernel? Jun 08 03:04:31 looks like we need to mod the openslug kernel ... Jun 08 03:04:35 hmm, I thought CONFIG_PACKET was enabled - ah maybe you need to load a module Jun 08 03:04:49 what's the kernel .config say? Jun 08 03:04:54 is it M ? Jun 08 03:06:03 what if you modprobe af_packet ? Jun 08 03:06:12 /lib/modules/2.6.11.2/kernel/net/packet/af_packet.ko Jun 08 03:08:15 don't exist on my openslug Jun 08 03:08:31 weird Jun 08 03:08:34 how old is it? Jun 08 03:09:10 ipkg install kernel-module-af-packet Jun 08 03:09:42 now it's running Jun 08 03:09:49 yay! Jun 08 03:10:39 strange that it's already on mine tho - I didn't install that module explicitly Jun 08 03:14:42 rwhitby-away, do we currently have an auto-pull into bk://nslu2-linux@nslu2-linux.bkbits.net/openembedded from oe-devel ? Jun 08 03:15:02 yep, every hour, unless one of jbowler's changes causes a conflict :-) Jun 08 03:15:12 ok thanks Jun 08 03:15:29 do you happen to know when? on the hour, 15 after ? Jun 08 03:15:40 dunno Jun 08 03:15:49 ka6sox would know Jun 08 03:16:08 jbower knows how to fix the conflicts, so that's cool now too ... Jun 08 03:16:16 ok, I'll ask him sometime Jun 08 03:16:26 my native python config is grinding along Jun 08 03:29:55 now if we could fix the "irq26: nobody cared" I would be very very happy Jun 08 03:49:51 looks like the auto-pull is sometime between 37 and 49 minutes after the hour Jun 08 04:07:56 Is anyone available to answer a few Q's on the openslug firmware? Jun 08 04:08:38 <[g2]> kared, sure I'm an openslug dev Jun 08 04:08:52 i've got an nslu2 running openslug now Jun 08 04:09:01 <[g2]> excellent! Jun 08 04:09:09 how simple is porting unslung images to openslug Jun 08 04:09:11 <[g2]> binary or build-from-source Jun 08 04:09:18 build from source Jun 08 04:09:28 <[g2]> double excellent! Jun 08 04:09:40 <[g2]> what packages do you want to run Jun 08 04:09:59 I actually built from source and installed it on my nslu2 without using the linksys firmware ounce Jun 08 04:10:28 <[g2]> kared, that's normal Jun 08 04:10:44 well im after a mail server, dns server and asterisk Jun 08 04:10:55 <[g2]> we're actually just in the final stages to replace the jffs2 from openslug for everyone Jun 08 04:11:29 <[g2]> I think there are a couple mail servers in oe like postfix and nail Jun 08 04:12:08 <[g2]> you can just change add the packages you need to the nslu2-symlinks or change your tree to the full oe packages directory and then just Jun 08 04:12:12 <[g2]> bb postfix Jun 08 04:12:14 <[g2]> or bb nail Jun 08 04:12:18 <[g2]> and I think bb bind Jun 08 04:12:34 <[g2]> then bb openslug-image (just to update the package index) Jun 08 04:12:50 asterisk is in oe too Jun 08 04:13:05 <[g2]> jacques, does it run on openslug ? Jun 08 04:13:11 that I do not know Jun 08 04:13:41 <[g2]> kared, you know I *really* could use from someone ? Jun 08 04:14:03 yeah? Jun 08 04:15:17 whats that Jun 08 04:16:04 <[g2]> sorry. Jun 08 04:16:31 <[g2]> The unslung package feed is based on the old tool chains and glibc build via crosstool Jun 08 04:17:28 python native openslug build took 57m42.848s Jun 08 04:18:17 <[g2]> It's on my to-do list to switch tool chain and try building all the packages for openslug Jun 08 04:18:38 <[g2]> we'd probably want to change some of the paths for the installs Jun 08 04:22:03 <[g2]> kared, you can add yourself to the openslug list at yahoo if you like http://groups.yahoo.com/group/nslu2-linux/database?method=reportRows&tbl=7 Jun 08 04:22:04 Yesterday i found a website that listed all the packages in oe with fairly long descriptions but i haven't been able to locate again, do you happen to know? Jun 08 04:22:26 I will Jun 08 04:23:30 <[g2]> I'll be back in a bit Jun 08 04:34:50 Back. Jun 08 04:35:14 Brought the virgin slug (never opened) into service as the new LanSlug running OpenSlug 1.12 Jun 08 04:35:27 :-) Jun 08 04:36:00 now to save the rootfs image from the flash disk onto the newly attached 80Gb drive, and then try an ipkg uprade on it .... Jun 08 04:36:49 can you dd from an ext3 partition to a file, and then expect it to work if you dd it back again later ? Jun 08 04:37:15 (when you're running rootfs from that ext3 partition, that is ...) Jun 08 04:37:20 openslug-nslu2-20050606105342.rootfs.jffs2 <-- what is this file? Jun 08 04:37:24 python is failing the tests related to berkeley db, investigating... Jun 08 04:37:37 kared: the rootfs jffs2 image Jun 08 04:37:58 contents of /dev/mtd4 Jun 08 04:38:05 ahh ok Jun 08 04:38:23 it's packed into the 8MB image file by slugimage Jun 08 04:38:32 rwhitby, you can if the fs isn't mounted when you dd it Jun 08 04:38:41 if i flash with the flashdisk.img, i dont need to flash the jffs2 to a drive seperatly Jun 08 04:38:51 kared: correct Jun 08 04:39:05 cool thanks clears it up for me Jun 08 04:39:21 jacques: guess I'll have to turndown before saving it away ... Jun 08 04:41:45 so what is openslug over open embedded? is it simply the configuration files, the symlinks for the known working packages and the openslug- specific packages? Jun 08 04:45:40 oe-built file seems broken Jun 08 04:45:45 investigating... Jun 08 04:48:28 kared, openslug is a dist in openembedded Jun 08 04:49:11 thanks for the help guys will be back tomorrow :) Jun 08 04:57:42 <[g2]> rwhitby-away, morning! Jun 08 04:58:37 <[g2]> rwhitby-away, FYI I tried dd'ing the rootfs.jffs2 into mtdblock4 yesterday and it didn't seem to take Jun 08 04:58:40 yeah, oe-built file is broken - native-built works Jun 08 04:59:13 popped that off stack, going back to investigating oe-built db breakage Jun 08 04:59:18 <[g2]> the dd completed but it seems data didn't get transferred Jun 08 05:02:32 hmm, the king of text based browsers still isn't in oe Jun 08 05:03:09 03jeanfabrice * 10unslung/ (9 files in 5 dirs): Control file now generated for iptables, ircd-hybrid, jove, vdr-mediamvp Jun 08 05:03:25 no more left. Jun 08 05:04:00 well, for wiley :/ Jun 08 05:04:10 :-) Jun 08 05:06:52 03jeanfabrice * 10unslung/sources/libbt/control: remove unused control file Jun 08 05:16:37 is iptables optware a classic package ? (means : do you know guys who play with it ?) Jun 08 05:18:32 03jeanfabrice * 10unslung/Makefile: Demote iptables for wiley. Doesn't built on build machine Jun 08 05:36:36 <[g2]> rwhitby-nslu2, nlsu2 ? Jun 08 05:38:04 As I'm now a happy unslung customer, I was wondering if I can be a happy openslug customer (using my slug in production) Jun 08 05:40:42 [g2]: rwhitby-nslu2 is the alternate when miau can't get rwhitby-away Jun 08 05:50:04 03jeanfabrice * 10unslung/make/ (ipkg.mk libbt.mk): Added zlib dependency and staging, bumped IPK_VERSION Jun 08 05:50:06 <[g2]> rwhitby, ahh Jun 08 05:50:40 <[g2]> rwhitby, do you attach and detach to the miau client ? Jun 08 05:51:23 <[g2]-away> bbiab Jun 08 05:51:28 dunno - Miranda just does what it does Jun 08 05:51:47 <[g2]-away> hmmmm... more stuff to learn about :) Jun 08 05:53:23 03jeanfabrice * 10unslung/make/ipkg.mk: remove ipkg.mk Jun 08 05:54:37 sorry for that :( I did not specify the libbt.mk filename when publishing so cvs took all the updated mk in my make/ dir Jun 08 06:08:36 damn, no tcl in oe either Jun 08 06:25:12 03jeanfabrice * 10unslung/make/dropbear.mk: dropbear.mk patched to avoid repetitive ipk rebuild each time a global compilation is launch. Bumped IPK_VERSION Jun 08 06:41:14 Just added ftpd-topfield to OpenSlug Jun 08 06:49:29 strange: http://www.cnn.com/2005/US/06/08/plane.body.ap/index.html Jun 08 06:54:43 rwhitby: Do I publish http://www.nslu2-linux.org/wiki/HowTo/RestoreWorkingIpkgFromJffs2 both to mailing list and FAQ & HowTo troubleshooting ? Jun 08 06:55:23 jf-work: yep - follow up the thread where I said someone would follow-up with instructions ... Jun 08 09:10:56 bye Jun 08 09:28:54 rwhitby: mount -o remount,ro /; dd if=/dev/mtdblock4 should work without problems. Jun 08 09:29:18 The inverse is tricky and may not be permitted because it relies on the jfffs2 not being read while the dd happens. Jun 08 09:30:37 I managed to reflash jffs2 (/dev/mtdblock4) yesterday on a non-jffs2 rootfs system using devio and filling with 0xff - i.e. I 'upgraded' the whole rootfs without upslug etc. Jun 08 09:32:09 Here's the command line: devio '<>/dev/mtdblock4;cp$;fb#$-255' (all seems to work fine - jffs2 is the name of the deploy/image file) Jun 08 11:19:57 <[g2]> long lost lennert :) hey! Jun 08 11:20:03 hello [g2] :) Jun 08 11:20:08 still alive Jun 08 11:20:24 <[g2]> and married right ? Jun 08 11:20:50 yes, yes Jun 08 11:20:54 that happened a while ago already, though :) Jun 08 11:21:05 three months on friday Jun 08 11:21:35 is it me or is the wiki slightly.. disorganised? Jun 08 11:22:46 like.. i want to find what all the led codes mean Jun 08 11:22:54 the info about that is spread out throughout the wiki Jun 08 11:23:06 the obvious page, LEDs, contains just one line: "LEDs are the Light Emitting Diodes found on the front of the NSLU2." Jun 08 11:23:39 the openslug LEDs page is even better "More than one LED." Jun 08 11:24:14 <[g2]> congrats on the 3 months Jun 08 11:24:21 [g2]: thanks Jun 08 11:24:56 <[g2]> the wiki needs some updates Jun 08 11:25:10 <[g2]> we're getting really close on a new OpenSlug release Jun 08 11:25:32 i was going to try openslug-1.12-beta.bin Jun 08 11:25:48 <[g2]> lennert, you can Jun 08 11:26:00 <[g2]> however, there some cool changes on the way Jun 08 11:26:01 sometimes my slug gets weird and starts outputting $T0a0f:67726f74;0d:0003dddc;#28 Jun 08 11:26:16 that's the gdb protocol or something Jun 08 11:26:19 [g2]: like what? Jun 08 11:26:49 <[g2]> Ok... there's a guy named jbowler that's been active with lots of stuff Jun 08 11:27:23 <[g2]> he's got a program that accesses block devices and does operations on the data inside Jun 08 11:27:49 <[g2]> it's used during our "turnup" proceedure which is significantly enhanced since the last beta Jun 08 11:28:35 <[g2]> in addition, he and I (just minutes ago) used it to replace the jffs2 partition from a non-jffs2 partiton Jun 08 11:28:49 <[g2]> the kernel can be replace with it also Jun 08 11:29:02 <[g2]> finally, I'm getting into the hardware business Jun 08 11:29:17 <[g2]> I've got a 64MB slug with serial, and JTAG Jun 08 11:29:29 wow, cool Jun 08 11:29:35 hey, my slug just spit out this: Jun 08 11:29:38 <[g2]> I'm testing the JTAG but the serial and 64MB have been running for a week Jun 08 11:29:40 iixxEEtthhAAccccRReeggiisstteerrQQMMggrrCCaallllbbaacckkss(()) ffaaiilleedd Jun 08 11:29:40 NNoo nneettwwoorrkk iinntteerrffaacceess ffoouunndd Jun 08 11:29:50 do you think it's fried somehow? Jun 08 11:29:53 <[g2]> You've got Echo turned on :) Jun 08 11:29:56 it was running with 5V on the serial for a while Jun 08 11:30:08 no, echo was off Jun 08 11:30:12 hmm Jun 08 11:30:29 it also spit out this: Jun 08 11:30:34 SSeettttiinngg ccoonnssoollee bbaauudd rraattee ttoo 5522 ffaaiilleedd Jun 08 11:32:46 <[g2]> are you using mincom ? Jun 08 11:32:53 <[g2]> minicom Jun 08 11:33:06 yes Jun 08 11:33:15 <[g2]> go toggle the echo Jun 08 11:33:27 "Local echo ON" it now says Jun 08 11:33:46 <[g2]> So to you ggggeeeetttt 4 chars ? Jun 08 11:33:51 it only does it once every few boots Jun 08 11:33:59 <[g2]> ah... Jun 08 11:34:06 i noticed that even before serial was working, only once every few boots it would actually boot Jun 08 11:34:10 <[g2]> do you have hw flow control turned off ? Jun 08 11:34:18 LEDs: http://www.nslu2-linux.org/wiki/OpenSlug/OpenSlugLEDsDuringBoot Jun 08 11:34:23 it would either get stuck in upgrade mode, or just hang Jun 08 11:34:31 [g2]: yes Jun 08 11:34:36 <[g2]> jbowler-away, meet lennert Jun 08 11:34:36 LEDs (more): http://www.nslu2-linux.org/wiki/OpenSlug/UseLedAndBuzzerOnOpenslug Jun 08 11:34:40 hi Jun 08 11:34:44 hi jbowler-away Jun 08 11:34:47 <[g2]> lennert meet jbowler-away Jun 08 11:34:49 jbowler-away: thanks for the links Jun 08 11:35:07 why would redboot say "No network interfaces found" ? Jun 08 11:35:26 <[g2]> lennert is a ass-kicker from the arm-linux ml and originally help with some key kernel / toolchain issues on unslung and openslug Jun 08 11:35:38 oh, and if i leave the redboot prompt idle for ~30 seconds, the entire slug also hangs Jun 08 11:35:49 No ethernet cable plugged in? Jun 08 11:35:55 [g2]: that's too much honor Jun 08 11:36:09 <[g2]> lennert, it's just the truth Jun 08 11:36:11 jbowler: it does have eth plugged in, to a 10/100 switch Jun 08 11:36:28 damn, now it hangs again Jun 08 11:37:34 this is bizarre Jun 08 11:37:39 RedBoot> ip_address -h 192.168.0.99 Jun 08 11:37:39 IP: 192.168.0.1/255.255.255.0, Gateway: 192.168.0.1 Jun 08 11:37:39 Default server: 192.168.0.99, DNS server IP: 0.0.0.0 Jun 08 11:37:41 so that's fine Jun 08 11:37:42 then i do Jun 08 11:37:45 RedBoot> load -r -v -b 0x01000000 openslug-1.12-beta.bin Jun 08 11:37:45 ** Error: Illegal command: "load" Jun 08 11:37:47 so then i type Jun 08 11:37:50 RedBoot> help Jun 08 11:37:52 hangs again Jun 08 11:38:00 <[g2]> lennert, you're missing the -l on the ip_addr Jun 08 11:38:22 [g2]: i'm looking at http://www.nslu2-linux.org/wiki/HowTo/RecoverFromABadFlash Jun 08 11:39:05 <[g2]> type this: ip_addr -l 192.168.0.1 -h 192.168.0.99 Jun 08 11:39:26 same output Jun 08 11:39:30 IP: 192.168.0.1/255.255.255.0, Gateway: 192.168.0.1 Jun 08 11:39:31 Default server: 192.168.0.99, DNS server IP: 0.0.0.0 Jun 08 11:39:46 RedBoot> load -r -v -b 0x01000000 openslug-1.12-beta.bin Jun 08 11:39:46 ** Error: Illegal command: "load" Jun 08 11:39:51 <[g2]> the -l sets the local IP address the -h is tftp server address Jun 08 11:40:10 yeah, i got that Jun 08 11:40:39 <[g2]> is openslug... at / on the tftp server ? Jun 08 11:40:43 at least it's consistent with what i saw before (when serial wasn't working) Jun 08 11:41:01 [g2]: it's in /tftpboot Jun 08 11:41:27 [g2]: i load the kernel on my other arm board in the same way though Jun 08 11:41:39 [g2]: ("load -r -m tftp -b 0xc208000 /zImage-enp2611") Jun 08 11:42:42 <[g2]> lennert, can you ping 192.168.0.99 ? Jun 08 11:43:07 <[g2]> I'd also try a different local IP addr like 192.168.0.2 Jun 08 11:45:20 i can: Jun 08 11:45:28 i mean, i can ping the slug from my x86 box Jun 08 11:45:34 phi% ping 192.168.0.1 ~ Jun 08 11:45:34 PING 192.168.0.1 (192.168.0.1) 56(84) bytes of data. Jun 08 11:45:34 64 bytes from 192.168.0.1: icmp_seq=0 ttl=64 time=13.0 ms Jun 08 11:45:45 problem is, redboot crashes after ~30 sec Jun 08 11:45:49 and it also stops pinging then Jun 08 11:46:01 it also seems to crash if i copy-paste something into minicom Jun 08 11:46:35 <[g2]> I'd start by separating the issues Jun 08 11:46:45 ok, i was going to type 'help' but after the 'e' it starts spitting out "$T050f:00020000;0d:0003dda4;#6f" Jun 08 11:46:46 <[g2]> 1) known good IP connectivity Jun 08 11:47:01 for as long as the slug is up, i can ping it from my x86 host Jun 08 11:47:29 <[g2]> do you have a cross-over cable ? Jun 08 11:49:26 i should have one in a box somewhere, why? Jun 08 11:49:48 <[g2]> some ppl have had problems with the switches they connect with Jun 08 11:50:21 <[g2]> it would remove the switch from the equation Jun 08 11:50:21 but would that cause my unit to say that it doesn't know the 'load' command and crash when i type 'help'? Jun 08 11:50:27 ok, i'll give it a try Jun 08 11:51:28 <[g2]> I've used the load command all the time Jun 08 11:51:35 is the ethernet led driven in hw or sw? Jun 08 11:51:42 <[g2]> I think that's hw Jun 08 11:52:04 <[g2]> the other 3 are sw Jun 08 12:11:57 what's your email again? Jun 08 12:12:02 i mean [g2] Jun 08 12:13:01 <[g2]> lennert are you aware of Marc Singer's APEX boot loader ? Jun 08 12:13:06 no Jun 08 12:13:46 <[g2]> it's been out for a little while Jun 08 12:13:53 <[g2]> I've replace Redboot with it Jun 08 12:14:26 <[g2]> it doesn't handle Ethernet in the boot loader, but it's tiny 16K and fairly easy to understand Jun 08 12:14:55 redboot is confusing Jun 08 12:15:03 i was looking into it for one of the ixp2000 platforms Jun 08 12:15:11 <[g2]> http://wiki.buici.com/twiki/bin/view/Main/ApexBootloader#IXP42x_and_the_Linksys_NSLU2_aka Jun 08 12:15:22 the source tree is huge, directories many levels deep Jun 08 12:15:35 <[g2]> personally, I think it's a mess Jun 08 12:15:53 <[g2]> it's a development project not a bootloader Jun 08 12:16:43 <[g2]> It think it's like MicroSoft... ppl keep using it because of enertia Jun 08 12:17:08 it's probably like.. Jun 08 12:17:15 companies using redboot hardly ever submit their changes Jun 08 12:17:36 but if they do, they're invariably ugly Jun 08 12:17:51 so either you merge something ugly, but gain functionality, or don't merge it and lose out Jun 08 12:18:01 asking a company to clean up their code isn't probably going to happen Jun 08 12:18:07 <[g2]> well APEX is clean and tight Jun 08 12:18:15 fortunately the linux kernel standards are somewhat stricter than that Jun 08 12:18:20 [g2]: i'll have a look at it Jun 08 12:18:33 [g2]: how do you load your kernel, from flash? Jun 08 12:18:34 <[g2]> are you familiar with USB ? Jun 08 12:18:53 not very much Jun 08 12:18:58 <[g2]> ok. Jun 08 12:19:03 <[g2]> here's the latest with openslug Jun 08 12:19:20 'nobody cares' etc? Jun 08 12:19:33 <[g2]> THAT's still open Jun 08 12:19:48 <[g2]> I meant the latest with re-flashing Jun 08 12:19:55 <[g2]> and openslug Jun 08 12:20:03 i'd love to fix it as a way of paying back you guys for the pain in the ass that i've been :) Jun 08 12:20:12 you mean new features? Jun 08 12:20:24 <[g2]> you haven't been a pain at all Jun 08 12:20:33 <[g2]> and I'd *really* love to see that one fixed Jun 08 12:21:05 i think i should be able to.. it's a software problem, you see :) Jun 08 12:21:08 <[g2]> with jbowler's latest changes one can reflash the jffs2 from openslug when not running on jffs2 (obviously) Jun 08 12:21:26 right Jun 08 12:21:40 so.. load kernel from flash, mount / from the external disk Jun 08 12:21:40 <[g2]> one can also reflash the kernel Jun 08 12:21:54 that's how most arm boards i know of do it, btw Jun 08 12:22:13 <[g2]> well APEX can almost load the kernel directly from jffs2 Jun 08 12:22:27 <[g2]> then we can lose the kernel partition Jun 08 12:23:00 a la grub and ext2 Jun 08 12:23:14 <[g2]> something like that Jun 08 12:23:38 <[g2]> the longer-term plan would be to have multiple rootfs's each with a kernel in it Jun 08 12:24:00 <[g2]> then you could have a failsafe with a complete local system Jun 08 12:24:42 ehm Jun 08 12:24:50 i think CUPS is fscking up my nslu2 Jun 08 12:24:58 cups periodically broadcasts the list of printers Jun 08 12:25:06 my nslu2 now crashed with this on the serial: Jun 08 12:25:13 Ethernet recv: Jun 08 12:25:13 0006EF8C: FF FF FF FF FF FF 00 04 23 46 E2 58 08 00 |........#F.X.. | Jun 08 12:25:13 0006EF9A: 45 00 00 AC 07 0F 40 00 40 11 1E 2A 0A 00 00 0A |E.....@.@..*....| Jun 08 12:25:13 0006EFAA: 0A 00 00 FF 02 77 02 77 00 98 2E 7D 39 30 30 65 |.....w.w...}900e| Jun 08 12:25:13 0006EFBA: 20 33 20 69 70 70 3A 2F 2F 70 68 69 2E 77 61 6E | 3 ipp://phi.wan| Jun 08 12:25:14 0006EFCA: 74 73 74 6F 66 6C 79 2E 6F 72 67 3A 36 33 31 2F |tstofly.org:631/| Jun 08 12:25:16 0006EFDA: 70 72 69 6E 74 65 72 73 2F 70 72 69 6E 74 65 72 |printers/printer| Jun 08 12:25:18 etc Jun 08 12:25:43 <[g2]> hmmm Jun 08 12:26:41 <[g2]> did the x-over cable help at all ? Jun 08 12:27:15 it didn't, but i used the same port Jun 08 12:27:42 <[g2]> you didn't plug it directly into the PC ? Jun 08 12:28:01 i did plug it directly into the x86 workstation, but still had the cups server running Jun 08 12:28:13 now trying without Jun 08 12:28:32 <[g2]> well that's probably running Fedora :) Jun 08 12:29:47 i wonder if it does the same with netbios broadcasts Jun 08 12:30:06 i still have ipv6 router advertisements on this subnet Jun 08 12:30:29 [g2]: yeah yeah, fedora is evil, i know :) Jun 08 12:30:52 * [g2] couldn't resist some fun :) Jun 08 12:31:05 i totally agree with the fedora haters about fedora, btw Jun 08 12:31:17 try getting an arm-specific patch in.. sheesh Jun 08 12:31:50 the fedora community which is really just a bunch of red hat employees have collectively decided that arm is not a supported architecture Jun 08 12:31:56 <[g2]> I haven't used fedora *at all* except for chrooting into your arm build Jun 08 15:23:06 if flex failing in openslug-packages for anyone else? flex-native is fine, but cross building the flex package fails as per yesterday Jun 08 15:23:12 s/if/is/ Jun 08 15:38:28 rwhitby: I pulled yesterday evening (gmt). My build completed successfully Jun 08 15:38:47 uhrm Jun 08 15:38:49 nm Jun 08 15:38:56 I did image, not packages. Jun 08 15:52:26 rwhitby NAiL: flex will not build correctly until you pull ChangeSet@1.3190.1.356, 2005-06-08 10:02:56+02:00 Jun 08 15:52:49 Prior to that change it last built correctly on May 26. Jun 08 15:53:42 From May 27 up to yesterday a cross-build was actually building the native (host) package! Yesterday this was partially fixed, but that broke the build. Jun 08 15:54:48 jbowler-away: is that changeset in nslu2-linux bk repo ? Jun 08 15:55:22 Yes, I think the number should match my number. If not the comment is: flex: with include.patch flex builds for arm(eb) target - credits goes to Dwayne Fontenot (jacques Jun 08 15:55:23 I thought jacques pulled that fix yesterday, and I am pulling from nslu2-linux but still getting the error. Jun 08 15:55:23 on #oe) Jun 08 15:55:34 Let me check my local repo. Jun 08 15:56:20 jacques seemed to say that he had pulled it... Check for packages/flex/flex-2.5.31/include.patch and the corresponding SRC_URI in the .bb Jun 08 15:56:25 yep, I have that changeset Jun 08 15:56:51 so openslug-packages completes for you now? Jun 08 15:57:57 back later Jun 08 15:58:03 Yes, though I've been having machine problems... I have work/flex-2.5.31-r2/install/flex/usr/bin/flex Jun 08 15:59:04 Oh, to fix the problem you must redo the do_configure task - i.e. rm -rf work/*/flex-2.5.31* or the configure won't re-run... Jun 08 15:59:24 (it was a Makefile.am patch) Jun 08 16:01:38 Um. Looks like OE just fixed the gcc 4.0.0 build. Now why does the darned cross compiler rely on a native mpfr? Who cares, it's just FORTRAN... Jun 08 16:02:20 Ok... I'm going to set off a gcc 4.0.0 build ;-) Jun 08 16:40:08 hi Jun 08 16:47:57 second time I wake up and find my ssh connection to my openslug aborted: Disconnecting: Protocol error: expected packet type 31, got 82 Jun 08 16:48:20 it's almost like there's a really long idle timeout Jun 08 16:52:53 * jacques starts the berkeley db testsuite again. Jun 08 16:58:01 nothing triggers "irq 26 nobody cared" like the berkeley db file operations tests :-\ Jun 08 16:59:51 I get one every few seconds Jun 08 17:00:59 Annoying :( Jun 08 17:01:36 NAiL, yeah :-( Jun 08 17:01:59 they seem to only be cosmetic tho - some ppl had said they noticed a delay when they got one, but I don't Jun 08 17:02:37 I notice a couple of secs delay when nobody cares during a filetransfer with samba Jun 08 17:03:53 hmm, I haven't gotten any diring file xfers, but while compiling or running testsuites it diesn't seem to affect anything Jun 08 17:04:28 I got quite a few while transferring all my mp3s to the the slug Jun 08 17:06:42 cool, seagate is shipping a 8GB 1" drive Jun 08 17:07:36 That should fit inside the slug :P Jun 08 17:08:03 it's the size they put in CF cards, ipod minis, etc Jun 08 17:11:04 It'd be nice to put inside the slug instead of the memstick I've got there right now Jun 08 17:12:38 hmm, looks like .45 to me: http://hardocp.com/image.html?image=MTExODI0NDY2M2pyQWVXRjdxNktfMV8xX2wuanBn Jun 08 17:12:49 musta been running windows Jun 08 17:14:05 hahaha Jun 08 17:28:39 How are you going to connect the CF disk to the NSLU2? Jun 08 17:28:43 If it's inside that is Jun 08 17:43:24 glc: feel free to check in your unslung /linuxrc changes ... Jun 08 18:09:22 I sent them to dyoung. I have had little luck checking in to bk. Jun 08 18:47:05 well, my last reason to run Unslung in the production environment has just disappeared. I got ftpd-topfield and puppy running consistently on OpenSlug. So now the Topfield and the 80GB disk enclosure are connected to the OpenSlug slug instead of the Unslung slug. Jun 08 18:52:08 welcome to the dark side Jun 08 18:53:13 rwhitby-web: I'm testing devio in an unslung build, assuming it's ok where would you like it? (packages/meta/unslung-image.bb?) Jun 08 18:53:49 yeah, it's a base tool, so unslung-image.bb is the right place Jun 08 18:53:56 <[g2]-away> yes the really dark side :) Jun 08 18:54:15 jacques: ipkg upgrade worked for openslug Jun 08 18:54:36 rwhitby-web, cool :-) Jun 08 18:54:57 So AccessSlug has been renamed HomeSlug, cause it manages home.rwhitby.net Jun 08 18:55:03 these damn db tests have been going for > two hours now Jun 08 18:55:17 Which BDB? (I.e. which rev?) Jun 08 18:55:19 <[g2]> rwhitby-web, did you see jbowler cmd for reflashing jffs2 from within OpenSlug in #openslug today ? Jun 08 18:55:21 and HomeSlug is a virgin slug (never been opened) Jun 08 18:55:35 jbowler-away, 4.3.28 Jun 08 18:56:27 [g2]: I did, but I'm waiting for the update-jffs2 script which can "Do The Right Thing"TM regardless of whether you give it a JFFS2 file, or the full 8MB image file :-) Jun 08 18:56:58 or maybe we have an update-firmware script which calls update-kernel and update-jffs2 Jun 08 18:57:27 <[g2]> I think that may be a little while or custom scripts Jun 08 18:57:35 <[g2]> the basic issue is compatibility with the old scripts Jun 08 18:58:06 jacques: I switched the ARM assembly mutex on in db_4.3.27, but you're presumably building an unslung version. Jun 08 18:58:25 [g2]: I don't understand the issue .... Jun 08 18:58:58 jbowler-away, this is build on openslug in native devel env Jun 08 18:59:10 from vanilla source Jun 08 18:59:14 There's no problem with updating the kernel and jffs2 (yes, I was intending it to be able to read from a full image). Jun 08 18:59:32 because I think the oe-built bdb is broken Jun 08 18:59:49 <[g2]> I'm talking about making the information in the stuff like /etc/network/defaults propagate properly and copying thinks like the ssh/dropbear keys Jun 08 18:59:54 db_4.3.27 wasn't, I've used it with openldap. Jun 08 19:00:18 <[g2]> rwhitby-web, does that clearify the issue more ? Jun 08 19:00:33 jbowler-away, hmm, well python sure diesn't like it :-\ Jun 08 19:00:38 [g2]: yes, but that's not too difficult, it's just a matter of making sysconfsetup correctly process the saved info. Jun 08 19:00:38 <[g2]> right now all that stuff is pitched and rebuilt from sysconf on a new image Jun 08 19:00:54 jacques: did you build 'db' or 'db3' into python? Jun 08 19:01:30 jbowler-away, uhhh.... I dunno Jun 08 19:01:59 I didn't give it any special configs Jun 08 19:02:02 jacques: there's 8million bb files in packages/python, but none of them have 'db3' and the one I looked at used 'gdbm' Jun 08 19:02:22 this is a python built from vanilla source native Jun 08 19:02:49 the testsuite fails every test involving db/bdb Jun 08 19:02:58 jacques: ok, then it gets whatever you installed into staging... Very difficult to know what configure selects ;-) Jun 08 19:03:00 and finally aborts with a segfault on the dbm test Jun 08 19:03:08 [g2]: I guess the question is what stuff do we need to preserve (assuming rootfs is already on an external disk) from one flash to the jffs2 to the next. Jun 08 19:03:21 hey dyoung-web Jun 08 19:03:40 BIAB Jun 08 19:03:45 <[g2]> rwhitby-web, I don't think we need to preserve anything on the jffs2 Jun 08 19:03:46 Hi. Jun 08 19:03:55 [g2]: sysconfsetup needs to change... I think maybe it just looks for a tar archive of the old config in /boot. Jun 08 19:04:09 I was gonna push glc's changes to linuxrc, but I sorta got distracted. Jun 08 19:04:10 <[g2]> I think it's as simple as reflash the jffs2 and then turnup *without* the -i Jun 08 19:04:53 [g2]: I think we do. I want the new jffs2 to be able to boot with not disks attached, and preserve any changes to the passwd file and /linuxrc, and have the right modules in it for when I boot the next time from external rootfs Jun 08 19:05:29 [g2]: it's reflash then 'turnup flash' then reboot. Jun 08 19:06:06 I can live without the passwd and /linuxrc (cause turnup can do that), but we need to preserve network settings (especially if I want to access the slug via some dongle rather than the built-in ethernet) Jun 08 19:06:22 <[g2]> rwhitby-web, the boot issue is a good one Jun 08 19:06:32 <[g2]> You really have 2 choices Jun 08 19:07:00 <[g2]> 1) separate the kernel and all the libs as a package and allow them to be copied over to the jffs2 partition Jun 08 19:07:07 At some point in the future, I want to be able to plug a zd1211 802.11g usb dongle into a newly flashed openslug slug, and have it get on the network automatically based on network settings that are stored in sysconf Jun 08 19:07:33 re 1) do you mena kernel and modules, rather than kernel and libs? Jun 08 19:07:43 <[g2]> 2) use the stuff in the jffs2 partition to overwrite the external storage Jun 08 19:08:14 re 2) I see the external storage as already being updated by "ipkg upgrade". Jun 08 19:08:16 <[g2]> I mean kernel and modules in /lib/modules/.... Jun 08 19:08:32 nod on kernel and modules in /lib/modules Jun 08 19:08:51 <[g2]> re re 2) if that's the case then the jffs2 is already updated also right ? Jun 08 19:08:57 If a reflash mounts /dev/mtdblock4 afterwards and copies the /lib/modules tree onto the current rootfs it will boot ok. Jun 08 19:09:05 <[g2]> because the /lib/... matches the ipkgs Jun 08 19:09:11 <[g2]> it's from the same build Jun 08 19:09:34 right. I see the workflow as 1) ipkg upgrade 2) update-firmware Jun 08 19:09:51 <[g2]> I think you missed my point Jun 08 19:10:06 <[g2]> which was the rootfs.jffs2 would match the ipkg upgrade Jun 08 19:10:15 agreed. Jun 08 19:10:26 <[g2]> then there's no need to copy that stuff over Jun 08 19:10:45 yes there is - you need to copy it *from* external rootfs *to* jffs2 Jun 08 19:11:10 you need to update the kernel, and the jffs2 Jun 08 19:11:16 for the kernel, it's easy Jun 08 19:11:16 <[g2]> no... you can just replace the jffs2 partition with the one built from OE Jun 08 19:11:39 no, cause then you need to go through turnup again. Jun 08 19:11:52 I'm not going to go through turnup every time I do an ipkg upgrade Jun 08 19:11:52 <[g2]> you have to do turnup anyway Jun 08 19:12:05 why? Jun 08 19:12:10 <[g2]> you *only* need to do turnup for a jffs2 update Jun 08 19:12:42 <[g2]> and actually, you only need to copy the /linuxrc Jun 08 19:12:52 for the jffs2, you can choose either to copy stuff, or reflash jffs2 Jun 08 19:13:06 for the former, you don't need to turnup. for the latter, you do Jun 08 19:13:21 <[g2]> ok.. ground rules Jun 08 19:13:36 <[g2]> the first was that we are always doing the ipkg update from the non-jffs2 partition Jun 08 19:13:46 <[g2]> you are talking about a special case in the first item Jun 08 19:16:57 <[g2]> btw... the bigger fish to fry is that we don't update the package index automatically to make all thing match until we bb openslug-index Jun 08 19:17:26 Hmm. Let's make sure we're on the same page. Jun 08 19:17:47 <[g2]> ok.. sounds great Jun 08 19:17:54 <[g2]> shoot Jun 08 19:17:54 The situation I am discussing is when you have a running openslug system, and you are running from a non-jffs2 partition. Jun 08 19:18:07 <[g2]> ok great Jun 08 19:18:16 We developers do some stuff in BK, and we push that stuff to the openslug feed Jun 08 19:18:31 Therefore there is now pending stuff for an ipkg upgrade to deal with Jun 08 19:18:32 <[g2]> ok great Jun 08 19:19:02 that pending stuff can be simple package updates which don't affect boot (in which case conffiles handles most things for you) or they can be boot-affecting updates Jun 08 19:19:16 <[g2]> nod Jun 08 19:19:41 in either case, the ipkg system only touches the non-jffs2 partition and does nothing on the flash unless we have a postinst or user command to do something after ipkg upgrade has completed Jun 08 19:20:11 <[g2]> yes this is vanilla ipkg update Jun 08 19:20:22 upgrade, not update Jun 08 19:20:25 <[g2]> from a non-jffs2 partition Jun 08 19:20:37 yes. Jun 08 19:20:41 <[g2]> yes upgrade not update Jun 08 19:21:11 we currently have a user command "update-kernel" which will take a new kernel image from /boot on the non-jffs2 partition, and will flash it into the kernel partition in the internal flash memory. Jun 08 19:21:28 it is deficient, cause it should also update any associated kernel modules Jun 08 19:21:58 there is other stuff involved in booting openslug before the pivot_root is executed Jun 08 19:22:48 all that other pre-pivot stuff has to be copied from non-jffs2 partition to jffs2 partition to be able to say that you have completed a full ipkg upgrade (as opposed to the vanilla ipkg upgrade) Jun 08 19:23:22 <[g2]> ok I've got two points Jun 08 19:23:38 <[g2]> 1) update-kernel is broken and incomplete Jun 08 19:23:38 the pre-pivot stuff includes kernel, kernel modules, /linuxrc, contents of /boot, and networking settings. Probably other stuff too as we get more adventuruous in booting from non-ethernet devices Jun 08 19:24:15 <[g2]> 2) the rootfs.jffs2 already as everything except the very minor turnup tweaks Jun 08 19:24:42 re 2) and networking settings Jun 08 19:25:01 There are two approaches, either reflash the whole jffs2 and restore the config or upgrade the pre-pivot stuff only (no config changes required). Jun 08 19:25:22 is update-kernel broken in ways other than the fact that it uses dd instead of devio, and that it doesn't copy the associated other files? Jun 08 19:26:04 So far as I can see it isn't broken - it does update /dev/mtdblock2 fine. Jun 08 19:26:06 jbowler: agreed. I want to be able to do both approaches in different situations. Jun 08 19:26:41 jbowler: I don't think dd is as guaranteed in the face of errors compared to devio Jun 08 19:27:21 <[g2]> I dd'ed the rootfs partition yesterday and it appeared to work, but failed on the reboot Jun 08 19:27:34 * [g2] was saddened by that :( Jun 08 19:27:35 If someone pulls the plug (e.g. power cut) while the reflash is in progress it's back to RedBoot/APEX - that's the most likely error ;-) Jun 08 19:27:58 Anyway, it's something of a moot question because it can be made to work for both mtdblock2 and mtdblock4 Jun 08 19:28:37 right, we know how to flash stuff properly. Jun 08 19:28:53 I believe the two approaches I gave are, effectively, completely equivalent (on a non-jffs2 root system). Jun 08 19:29:08 ... because they both make the pivot work with the new kernel. Jun 08 19:29:15 The question I'm focusing on is how do we handle the two situations: 1) reflash the whole jffs2 and restore the config, and 2) upgrade the pre-pivot stuff only (no config changes required). Jun 08 19:29:52 <[g2]> there are over 900 files in the OpenSlug jffs2 over 1000 if you count /dev Jun 08 19:30:14 <[g2]> you've got a bunch of combinations to get right Jun 08 19:30:30 <[g2]> compare that to maybe 6 config files ? Jun 08 19:30:43 It's easy to do (2) by cp /lib/modules - and make sure the jffs2 doesn't fill up. It may be no longer possible to boot with jffs2 root though. Jun 08 19:31:09 It's easy to do (1) but with the restriction that the boot time config files need to be known. Jun 08 19:31:26 I agree with [g2], (1) is likely to be more reliable... Jun 08 19:32:09 <[g2]> I think we should focus on the config files that need to propagate Jun 08 19:32:51 It's only /etc/network/interfaces and /etc/module* Jun 08 19:32:53 <[g2]> for me that's the dropbear key, /etc/default/sysconf and /linuxrc or a re-turnup Jun 08 19:33:16 I think we need to focus on two things: 1) as [g2] says, the config files that need to propagate across a complete jffs2 reflash, and 2) the pre-pivot files that need to be copied from upgraded external rootfs to jffs2 Jun 08 19:33:23 <[g2]> and also my ntp.conf file Jun 08 19:33:46 [g2]: isn't that on the external rootfs? Jun 08 19:33:48 [g2]: none of those files are used! Jun 08 19:34:04 right. Jun 08 19:34:18 we need to be clear to not include files that are only accessed after pivot_root has executed Jun 08 19:35:14 There is another problem, which is getting turnup -i to be more sensible about the key and config files, but that's really a separate issue. Jun 08 19:35:15 the assumption being that your jffs2 configuration is just enough to allow you to ssh in and turn turnup. everything else is on the external rootfs and the upgrade of that is handled by ipkg conffiles Jun 08 19:36:13 so ntp.conf doesn't make the cut, whereas dropbear keys do Jun 08 19:36:20 <[g2]> rwhitby-web, that's not your premise Jun 08 19:36:37 pardon? that *is* my premise Jun 08 19:36:44 <[g2]> your premise was you wanted to pull the extrnal device and have stuff still be bootable and similar right ? Jun 08 19:37:08 bootable, just enough to run turnup to get the external device online again Jun 08 19:37:32 <[g2]> I'm confused Jun 08 19:37:33 not similar (apart from pre-pivot similarities, like getting networking from a usb dongle) Jun 08 19:38:09 rwhitby-web: it isn't necesary to return to flash after a reflash (I thought it was before but I can see that it isn't really necessary) Jun 08 19:38:12 <[g2]> first we would talking about getting the jffs2 to match the upgraded external fs Jun 08 19:38:27 <[g2]> now you are talking about just enough to re-turnup Jun 08 19:38:47 [g2]: you misunderstood Jun 08 19:39:03 <[g2]> please clarify for me Jun 08 19:39:20 <[g2]> or are there two separate issues Jun 08 19:39:25 I was always only talking about getting the jffs2 to match the upgrade external fs for anything which is relevant before pivot_root Jun 08 19:39:54 I will have heaps of other stuff installed on external rootfs which will never go anywhere near the jffs2 Jun 08 19:40:02 <[g2]> that's what you said and that is very important from a booting perspective Jun 08 19:41:00 <[g2]> however many of the files we discussed like /etc/default/sysconf and /etc/network/interfaces only come into play LONG after the pivot root Jun 08 19:41:08 I expect to have a jffs2 which is the base openslug firmware (with enough modules to allow networking to be done from a number of different places), and some simple config changes like passwd, dropbear keys, networking settings, and turnup settings. Jun 08 19:41:54 ok, two situations Jun 08 19:42:06 <[g2]> ok that's clear to me Jun 08 19:42:24 1) ipkg upgrade, run update-jffs2-config-files, reboot and that's it - you're back onto the external rootfs Jun 08 19:42:50 2) ipkg upgrade, run update-jffs2-flash-stuff, reboot, ssh in, turnup, reboot Jun 08 19:43:33 the first situation is for an incremental upgrade, where I don't care that the non-boot portion of the jffs2 is out of date (but still sufficient for everything pre-pivot) Jun 08 19:43:38 <[g2]> re 1) no need to reboot you can just mount the jffs2 partition update Jun 08 19:43:50 the second situation is the full jffs2 upgrade Jun 08 19:44:39 re 1) the reboot is to confirm that you've updated the config files and kernel and kernel modules properly - you need to reboot to get the new kernel Jun 08 19:44:41 <[g2]> re 2) you can do the same thing, the *real* question is you don't *know* that it really worked unless you pull the external device Jun 08 19:45:02 no, you've lost me on that one Jun 08 19:45:40 You don't know it worked until the reboot. If you try to mod the jffs2 in place you can't be sure turnup flash will work ever again. Jun 08 19:45:47 what does "same thing" mean. start again with the two situations and paraphrase back to me what you understand by those two situations. I don't think we're on the same page with them yet Jun 08 19:46:14 <[g2]> ok.. here goes Jun 08 19:46:24 1) is where you have a kernel upgrade, and you want to start using that new kernel without having to re-turnup Jun 08 19:46:43 2) is where you want to update your jffs2 contents to the latest and are prepared to have to run turnup Jun 08 19:47:51 1) could also be a new kernel module upgrade for a kernel module which is involved in pre-pivot activities Jun 08 19:48:11 1) can also be a linuxrc-related script upgrade Jun 08 19:48:45 I don't want to have to reflash and re-turnup just because jbowler does a small change to a script in /boot Jun 08 19:49:00 (reflash being flash the whole jffs2) Jun 08 19:49:19 <[g2]> right with 1 you copy all the modules, linuxrc, and possibly any other file (like device files) and reflash the kernel with the zImage Jun 08 19:49:26 yes Jun 08 19:49:41 Hum, it's not necessary to redo the turnup... How about this. Jun 08 19:49:49 This sequence of operations seems (to me) to satisfy all the requirements: Jun 08 19:49:49 1) Save the existing config from 'Flashdisk' (/initrd), that's: Jun 08 19:49:49 a) /etc/default/sysconf Jun 08 19:49:49 b) /etc/network/interfaces Jun 08 19:49:50 c) any files configed on the flashdisk from a known list, including passwd Jun 08 19:49:50 and the dropbear key. Jun 08 19:49:51 1) Overwrite the MTD partitions 'Kernel' and 'Flashdisk' with new images. Jun 08 19:49:53 2) Mount 'Flashdisk' and restore the files from (1) Jun 08 19:49:55 3) Reboot Jun 08 19:50:23 It just comes right back into the external rootfs, but with a nice consistent flash. Jun 08 19:50:24 oooh - that's good. Jun 08 19:50:56 yep, that covers both my situations nicely Jun 08 19:51:04 It's a refinement, or restatement of an idea [g2] came up with. Jun 08 19:51:26 Because we never need to hack SysConf it works with APEX. Jun 08 19:51:30 <[g2]> by Flashdisk you mean the jffs2 partition right ? Jun 08 19:51:37 yeah Jun 08 19:51:41 Right - cat /proc/mtd Jun 08 19:51:49 yes, working the same for redboot and apex is essential Jun 08 19:52:11 I.e. so long as the partitions have the correct name the data will go to the right one (I don't use /dev/mtdblock) Jun 08 19:52:12 <[g2]> this will dissapear with APEX Jun 08 19:52:14 and using partition names is good too, cause they will change for an optimised apex Jun 08 19:53:10 [g2]: will APEX be able to boot and load an NFS rootfs over my 802.11g usb dongle ? Jun 08 19:53:39 <[g2]> Redboot can't boot from WIFI Jun 08 19:53:53 <[g2]> you can nfs mount a root with turnup Jun 08 19:54:03 I expect even with APEX we will still retain the current /linuxrc in jffs2 structure to allow for complex non-jffs2 rootfs access scenarios like that one Jun 08 19:54:07 <[g2]> regardless of bootloader right ? Jun 08 19:54:29 <[g2]> I think the /linuxrc is a good thing Jun 08 19:54:42 <[g2]> I'm talking about the kernel partition going away Jun 08 19:54:51 right, so the scheme that jbowler outlined will still be required with APEX, cause stuff will still need to be preserved in the jffs2. Jun 08 19:54:56 <[g2]> and APEX looking the kernel from within the jffs2 Jun 08 19:55:04 <[g2]> that's what beewoolie working on right now Jun 08 19:55:17 ok, yes, the need to flash the kernel partition separately goes away. nothing else goes away Jun 08 19:55:40 <[g2]> well actually you then have 2 jffs partitions Jun 08 19:55:41 "this will dissappear from APEX" was a bit unspecific :-) Jun 08 19:56:03 <[g2]> so you don't need to save off and restore because you can ping-pong between the partitions Jun 08 19:57:07 hmm - ok, that one is a bit too complex and out-there in time for me to consider it now, but I don't disagree. Jun 08 19:57:09 <[g2]> I imagine not long after booting from the kernel in jffs2 there will be a new options to turnup Jun 08 19:57:13 <[g2]> turnup jffs2 Jun 08 19:57:55 <[g2]> where it simply copies to the "other" jffs2 partition Jun 08 19:58:31 Here's a useful page ;-) http://www.nslu2-linux.org/wiki/OpenSlug/DevelopmentNotes Jun 08 19:58:34 <[g2]> I'd imagine turnup memstick /dev/mdtblock6 -t jffs2 would work now Jun 08 19:58:48 [g2]: how will you tell apex which partition to boot from? Jun 08 19:58:49 [g2]: Howdy. Jun 08 19:58:57 <[g2]> hey! Jun 08 19:58:59 * beewoolie is catching up Jun 08 19:59:04 <[g2]> wb Jun 08 20:00:02 I gotta go to lunch, but will read the log when I get back .... Jun 08 20:01:24 Are we talking about retaining the redboot partition table? Jun 08 20:01:32 ...in order to have partition names? Jun 08 20:02:09 [g2]: I've finished my jffs2 partition parser and path-to-inode routines. Jun 08 20:02:17 [g2]: All I need to do is stuff the code into APEX. Jun 08 20:02:24 <[g2]> currently the turnup procedure which initializes and copies the jffs2 to the external device uses partition names not numbers Jun 08 20:02:38 Interesting. Jun 08 20:02:46 I've never looked at the partition table format. Jun 08 20:03:04 <[g2]> the devio program I think looks in /proc/mtd Jun 08 20:03:06 I was hoping we could drop it since there isn't a good way to find them. Jun 08 20:03:28 <[g2]> funny you should mention that :) Jun 08 20:03:29 Oh. So you don't mean that APEX will read them. Jun 08 20:03:44 <[g2]> Right not yet Jun 08 20:03:58 I can see how it could be useful. Jun 08 20:04:05 <[g2]> I was thinking of moving the partition table to the APEX config block though Jun 08 20:04:31 <[g2]> I had verified with jbowler that he used the name not the number Jun 08 20:05:07 Ah, that would be fine. Jun 08 20:05:25 <[g2]> my thinking was for the APEX version to have Block 0 as APEX, Block 1 APEX config, then the rest 1 or 2 jffs2 partitions Jun 08 20:05:33 We could even do the simplest thing which is to use the Kernel command line format for specifying the partitions. Jun 08 20:06:18 <[g2]> right, but it's probably trivial for devio to move the partition table Jun 08 20:06:28 I'm not familiar with devio. Jun 08 20:06:39 <[g2]> it's jbowlers tool Jun 08 20:06:58 BTW, jffs might be kinda big. It's looking like 8K in x86. Jun 08 20:07:03 Ah. Jun 08 20:07:26 Presently APEX config is not portable between APEX versions. Jun 08 20:07:28 <[g2]> it's kinda like a sed for block devices based on RPN commands Jun 08 20:07:39 Hmm. Jun 08 20:08:18 <[g2]> devio '<>/dev/mtdblock4;cp$;fb#$-255' Jun 08 20:08:50 <[g2]> that copies the jffs2 file to the mtdblock4 partition and for the lenght of the file and fills the rest with 0xFF Jun 08 20:08:53 Oh. Interesting. Jun 08 20:09:09 <[g2]> that's how we reflashed the jffs2 partion Jun 08 20:09:14 <[g2]> partition Jun 08 20:09:38 <[g2]> it worked splendidly today :) Jun 08 20:09:51 Nice idea. Jun 08 20:10:12 It's great when folks take the time to solve problems well. Jun 08 20:10:18 <[g2]> nod. Jun 08 20:10:30 ...so we don't have to go back and rewrite them... Jun 08 20:10:50 <[g2]> jbowler coded if for the openslug-init parsing. I told him to split it out as it's own tool Jun 08 20:11:02 <[g2]> nod. Jun 08 20:11:25 <[g2]> so now it's a oe package that can be used on any platform Jun 08 20:12:24 <[g2]> hey can I bug for 5-10 minutes about my other big problem ? Jun 08 20:12:47 Sure. Jun 08 20:13:18 <[g2]> reboot and shutdown don't usually work properly from the pivoted rootfs Jun 08 20:13:44 <[g2]> I'm looking into the code but I'll bet you're alot more familiar with it Jun 08 20:13:58 <[g2]> I see two things. Jun 08 20:14:13 <[g2]> 1) on shutdown -h we don't really halt Jun 08 20:14:27 I'll tell you one thing. I believe we should be unmounting the original. Jun 08 20:15:27 <[g2]> Ok... so what you are saying is that we are not really pivoted properly Jun 08 20:16:10 I'm not positive that that's stopping us. Still, I believe it is good hygiene to unmount the redundant root. Jun 08 20:16:30 <[g2]> we do Jun 08 20:16:32 <[g2]> it' Jun 08 20:16:40 <[g2]> it's on /initrd which is unmounted Jun 08 20:17:01 let me log in... Jun 08 20:17:38 Oh. I take it back. On my openslug system, there is no redundant root. Jun 08 20:18:03 Sorry. Continue. Jun 08 20:18:08 <[g2]> you are out-of-date and are not running on the external device Jun 08 20:18:15 <[g2]> ok... Jun 08 20:18:44 <[g2]> so I think are arch_reset we something goes bad Jun 08 20:18:55 <[g2]> a) we are jumping to 0 and just restarting Jun 08 20:19:12 Is this only with APEX? Jun 08 20:19:16 <[g2]> b) the on-chip reset isn't working as intended Jun 08 20:19:29 <[g2]> identical with APEX or RedBoot Jun 08 20:19:44 In apex, I reboot using the watchdog. Are we doing it that way? Jun 08 20:20:13 <[g2]> *IXP4XX_OSWK = IXP4XX_WDT_KEY; Jun 08 20:20:13 <[g2]> /* write 0 to the timer register for an immediate reset */ Jun 08 20:20:13 <[g2]> *IXP4XX_OSWT = 0; Jun 08 20:20:13 <[g2]> *IXP4XX_OSWE = IXP4XX_WDT_RESET_ENABLE | IXP4XX_WDT_COUNT_ENABLE; Jun 08 20:20:21 ...and halt should just be a matter of hitting the GPIO lines. Jun 08 20:20:24 <[g2]> WDT Watch Dog Timer ? Jun 08 20:20:33 Looks like the same thing I do. Jun 08 20:20:37 <[g2]> nod on the halt Jun 08 20:20:44 do we know that this code is being executed? Jun 08 20:21:02 <[g2]> that was the difference between the a) and b) above Jun 08 20:21:16 <[g2]> the a) code just cpu_reset(0); Jun 08 20:21:25 I don't quite get what you mean in the code above. Jun 08 20:21:50 <[g2]> the cpu_reset (0) ? Jun 08 20:22:01 Well, what are a) and b)? Jun 08 20:22:10 <[g2]> static inline void arch_reset(char mode) Jun 08 20:22:10 <[g2]> { Jun 08 20:22:10 <[g2]> if ( 1 && mode == 's') { Jun 08 20:22:10 <[g2]> /* Jump into ROM at address 0 */ Jun 08 20:22:10 <[g2]> cpu_reset(0); Jun 08 20:22:11 <[g2]> } else { Jun 08 20:22:13 <[g2]> /* Use on-chip reset capability */ Jun 08 20:22:15 <[g2]> /* set the "key" register to enable access to Jun 08 20:22:17 <[g2]> * "timer" and "enable" registers Jun 08 20:22:19 <[g2]> */ Jun 08 20:22:21 <[g2]> *IXP4XX_OSWK = IXP4XX_WDT_KEY; Jun 08 20:22:32 <[g2]> a) is the cpu_reset(0) Jun 08 20:22:39 <[g2]> b) is the dog biting Jun 08 20:23:43 I don't think that using a jump to zero is a good idea. I'd have to look at the code in more detail. I don't recall what the 's' indicates. Jun 08 20:25:21 <[g2]> All I know for sure is something doesn't happen properly after here Jun 08 20:25:26 <[g2]> printk(KERN_EMERG "Power down.\n"); Jun 08 20:25:26 <[g2]> machine_power_off(); Jun 08 20:25:26 <[g2]> unlock_kernel(); Jun 08 20:25:26 <[g2]> do_exit(0); Jun 08 20:25:26 <[g2]> break; Jun 08 20:26:09 <[g2]> that's the power-off case Jun 08 20:26:14 OK. Looks like it should be straightforward to debug. We see the power down message and then it doesn't do the right thing. Jun 08 20:26:48 <[g2]> right. I'm not really worried about the power-off, cause the GPIO will for-sure power off the box Jun 08 20:27:00 Oh. And is that working? Jun 08 20:27:29 I don't see that you are claiming that it doesn't. Jun 08 20:27:39 <[g2]> the power-off doesn't Jun 08 20:27:52 Neither works. Jun 08 20:27:54 <[g2]> shutdown -h locks up Jun 08 20:27:57 <[g2]> nod Jun 08 20:28:04 Does halt work? Jun 08 20:28:14 Guess I can try... Jun 08 20:28:26 * [g2] trying Jun 08 20:29:13 <[g2]> heh.... same issue as shutdown -h Jun 08 20:29:33 <[g2]> Deactivating swap... Jun 08 20:29:33 <[g2]> Unmounting local filesystems... Jun 08 20:29:33 <[g2]> umount: forced umount of ramfs failed! Jun 08 20:29:33 <[g2]> [fatal] IXETHACC:ixEthAccPortDisable: ixEthAccPortDisable failed port 0 (state = 4) Jun 08 20:29:33 <[g2]> ixp425_eth: eth0: BUG: ixEthAccPortDisable(0) failed Jun 08 20:29:34 <[g2]> Power down. Jun 08 20:29:36 <[g2]> usb 2-1: new full speed USB device using ohci_hcd and address 2 Jun 08 20:29:38 <[g2]> usb 2-1: not running at top speed; connect to a high speed hub Jun 08 20:29:40 <[g2]> SCSI error : <0 0 0 0> return code = 0x70000 Jun 08 20:29:46 <[g2]> then more usb messages and then it's locked up hard Jun 08 20:30:01 <[g2]> need to power cycle from the brick not the button Jun 08 20:30:21 <[g2]> ok... the button actually worked there but it usualy doens't Jun 08 20:30:40 Hmm. Jun 08 20:30:46 Failed for me, too. Jun 08 20:30:54 I don't have a console so I cannot see the messages. Jun 08 20:30:57 <[g2]> but that's the power-down case I'm not too worried about that Jun 08 20:31:38 Do you get the same messages on reboot? Jun 08 20:31:48 <[g2]> no Jun 08 20:31:58 <[g2]> it looks like it tries and then locks up Jun 08 20:32:42 I'd see about forcing the reboot code to use the watchdog. I wouldn't want it to jump to zero. A simple edit of the if() should suffice. Jun 08 20:35:48 <[g2]> Hmmmm... Jun 08 20:36:04 I wonder if the 's' means sleep. Jun 08 20:36:11 <[g2]> It's working right now but usually doesnt' Jun 08 20:36:18 <[g2]> probably Jun 08 20:37:07 <[g2]> I'd think w for warm-start Jun 08 20:37:48 Methinks remove the distinction. Only one sort of restart. Jun 08 20:40:05 <[g2]> hmmm. Odd it's worked that last 3 times in a row Jun 08 20:40:25 I wonder if it has to do with use. Jun 08 20:40:46 <[g2]> I'll let it up for a while and use it some more Jun 08 20:41:19 <[g2]> thx you're a real problem solver :) Jun 08 20:41:45 <[g2]> I just talk to you and it starts working in fear of your powers Jun 08 20:41:52 %/) Jun 08 20:42:26 <[g2]> Ok... that's enough for now.... Jun 08 20:42:38 <[g2]> I've got to look at the LED think for ep1220 Jun 08 20:42:46 <[g2]> he may be ready soon Jun 08 20:42:58 Whats the ep1220? Jun 08 20:43:45 <[g2]> in openjtag he's just about readay to dl directly into the mini I cache Jun 08 20:44:10 Ah. What did you resolve about data reads in the code? Jun 08 20:44:18 PC relative data reads. Jun 08 20:44:23 <[g2]> the LED thing is the program that just sets the LED and loops Jun 08 20:44:30 <[g2]> I haven't checked yet Jun 08 20:44:52 <[g2]> I figured I could look at the assember and make it go Jun 08 20:45:00 <[g2]> load immediate ... Jun 08 20:45:10 That's the only way. Jun 08 20:45:16 <[g2]> old-skool Jun 08 20:45:26 You have to load addresses as multiple immediates. Jun 08 20:45:35 <[g2]> yeah Jun 08 20:45:40 Should work then. Jun 08 20:45:55 old-skool --> short bus Jun 08 20:46:08 <[g2]> I can't imagine it being more than a couple dozen instructions Jun 08 20:46:43 <[g2]> unless I've got to setup all the GPIO block Jun 08 20:47:04 Nah. All you need to do is set bits in two registers. Jun 08 20:47:11 Output enable and output value. Jun 08 20:47:21 Even the timer is really easy. Jun 08 20:47:26 <[g2]> that's what I thought and was hoping for Jun 08 20:47:26 Sounds Fun. Jun 08 20:47:37 <[g2]> perfect timing I was thinking about you Jun 08 20:47:52 <[g2]> dyoung-web, do you run nfs as a server on Openslug ? Jun 08 20:47:56 I'm still digesting what y'all were discussing for the past 2 hours. Jun 08 20:48:14 I did before. Right now I dont have anything with openslug on it. Jun 08 20:48:31 that was the whole reason for getting that nfs-utils stuff working. Jun 08 20:48:36 <[g2]> do you have to do something special for the nfsd ? Jun 08 20:48:39 Ah. I'd like to use it. Jun 08 20:48:42 Is it in OE? Jun 08 20:48:46 <[g2]> nod. Jun 08 20:48:56 "something special" like ... ? Jun 08 20:48:57 <[g2]> I ran the client from the OE nfs-utils Jun 08 20:49:48 I dunno; I can go through it again later. I might hvae been running it before I had it in OE/. Jun 08 20:50:01 I should double check that anyways. Jun 08 20:50:25 <[g2]> I've got it all setup Jun 08 20:50:32 what doesnt it do? Jun 08 20:50:39 <[g2]> can you talk me through it ? Jun 08 20:50:56 <[g2]> the PC doesn't see the service from the nfs server Jun 08 20:51:02 make a /etc/exports, then exportfs -a it. Jun 08 20:51:12 <[g2]> ah... Jun 08 20:51:14 then start mouted. Jun 08 20:51:17 mountd. Jun 08 20:51:22 and nfsd Jun 08 20:51:33 <[g2]> I made the /etc/exports but didn't exportfs -a it Jun 08 20:51:38 and you may or may not have to mokey around with lockd. Jun 08 20:51:50 <[g2]> I did all the other suff Jun 08 20:51:53 <[g2]> stuff :) Jun 08 20:52:12 ok cool, hope it works. Jun 08 20:53:57 mt non-jffs2 partition doesn't have /boot Jun 08 20:53:59 my Jun 08 20:54:28 <[g2]> jacques, how did you create it ? Jun 08 20:55:46 turnup disk -i /dev/sda1 Jun 08 20:56:40 <[g2]> hey... my memstick doesn't have it either :) Jun 08 20:57:03 <[g2]> where's jbowler when you need 'em :) Jun 08 20:57:14 :-) Jun 08 20:57:22 <[g2]> open a slugbug :) Jun 08 20:57:27 <[g2]> LOL Jun 08 20:58:20 why is it needed tho? Jun 08 20:58:25 later boyz. Jun 08 20:58:36 K G Dog. Jun 08 20:58:47 <[g2]> gentoo2 root # mount 192.168.123.117:/ /mnt/nfs Jun 08 20:58:47 <[g2]> mount: RPC: Program not registered Jun 08 20:59:14 <[g2]> I did get a couple error on the exporfs Jun 08 20:59:20 and mountd is running on .117 ? Jun 08 20:59:33 <[g2]> maybe I missed mountd Jun 08 20:59:48 <[g2]> now it is :) Jun 08 21:00:00 <[g2]> OHHHHHHHHHH Jun 08 21:00:24 * [g2] HUGS dyoung-web Jun 08 21:00:43 was that preemptive because youre about to tell me its busted? Jun 08 21:00:56 <[g2]> no Jun 08 21:01:00 <[g2]> :) Jun 08 21:01:05 Cool. ;-) Jun 08 21:01:12 <[g2]> I"m *so* happy to be working with sane ppl Jun 08 21:01:19 over four hours and the bdb testsuite is still chugging Jun 08 21:02:25 <[g2]> dyoung-web, you big dog... and TCP works too Jun 08 21:02:43 what mde it go? Jun 08 21:02:48 <[g2]> the mountd Jun 08 21:02:55 <[g2]> and the exportfs Jun 08 21:03:03 Cool. Jun 08 21:03:07 <[g2]> THX Jun 08 21:03:11 now, you may or may not need to fiddle with lockd. Jun 08 21:03:13 <[g2]> bewoolie will be pleased Jun 08 21:03:35 <[g2]> well it seemed to be working Jun 08 21:03:45 if you mount from gentoo2 without the nolock option, you will probably need to have lockd running on the openslug. Jun 08 21:04:03 <[g2]> lockd was started automagically Jun 08 21:04:09 <[g2]> on the openslug Jun 08 21:04:14 <[g2]> Oslug Jun 08 21:06:31 <[g2]> nite all Jun 08 21:06:44 nite Jun 08 21:06:48 <[g2]> thx dyoung-web Jun 08 21:06:50 <[g2]> and all Jun 08 21:21:19 dyoung-web: got the DS18B20's today as samples. The USB dongle is back-ordered Jun 08 21:21:46 can't do much with it without the DS9490B though ... Jun 08 21:22:40 Cool. Jun 08 21:22:45 I never got around to ordering my samples. Jun 08 21:22:50 But I did buy a car finally. Jun 08 21:25:11 crap Jun 08 21:25:22 bdb testsuite wants too much memory Jun 08 21:26:20 hmm, it's still going Jun 08 21:32:50 rwhitby-away: where have you gone? Jun 08 21:33:22 Hmm - looks like I need that UPS .... Jun 08 21:33:59 :-\ Jun 08 21:34:14 power glitched three times here yesterday Jun 08 21:35:23 and HomeSlug doesn't have AutoPowerOn yet, so it won't come back up again .... Jun 08 21:36:13 back later. Jun 08 22:52:29 jacques: /boot - you're right, it isn't needed so it isn't copied ;-) Jun 08 22:53:16 and you don't need /boot/zImage* either on the jffs2 - that saves 800k. Jun 08 22:54:45 jbowler-away, yep, not until we get jffs2 support in APEX :-) Jun 08 22:56:07 of course, then we can get rid of the kernel partition Jun 08 22:59:36 egrep -s Kernel /proc/mtd && rm -f /boot/zImage* **** ENDING LOGGING AT Wed Jun 08 23:59:56 2005