irclog2html for blob on 2002.04.26

01:21:29SammyYo everyone
05:02:47BZFlagRuss: you around?
05:06:16Russya
05:08:40Russew...you might formatting stuff k&r -i8 before commiting?
05:21:38BZFlagnah, just joined some lines.
05:21:59BZFlagload_kernel_test segfaults trying to fill the tables.
05:22:28BZFlagmy original version of your patch works fine.
05:22:53BZFlagutil/test/Makefile needs to be hacked to autograb the kernel path.
05:56:53Russdid you play with memset?
05:57:43Russit could be the buffer used in load_kernel_test is too small
05:57:51Russincrease it by 10
06:08:30BZFlaghmm... I use --pad on the mkfs.jffs2 line so it should already be padded.
06:08:50Russthe buffer in load_kernel_test.c is kinda small
06:10:15Russcvs update utils/test/load_kernel_test.c
06:17:06BZFlagstill seg faults.
06:22:35Russcan you send me the image?
06:22:37Russ(DCC)
06:35:11BZFlagcan you get it to work on an image?
06:35:47BZFlaghttp://rikers.org/tmp/root.jffs2
06:46:30Russunable to find magic
06:48:11erikmmorning, guys
06:50:51Russo, my bad
06:52:52Russok, I see what happened
06:53:02Russthe cramfs_memset function got replaced by memset
06:53:09RussI should have named it u16_memset
06:54:08erikmwow, you have been busy when I was asleep
06:54:20BZFlagmorning.
06:54:33BZFlagwhat did cramfs_memset do?
06:54:50Russset everything by a u16 instead of u8
06:55:16Russactually int, sorry
06:55:29RussI'll name it int_memset
06:55:45erikmBZFlag: wow, an ext2fs reader!
06:57:50BZFlagdisk.c? hmm. looks like the comment is wrong.
06:57:55Russtry that
06:58:25BZFlagoh, yeah. just says ms-dos partition. ;-)
06:59:00erikmBZFlag: I suggested seletz to look at libe2fs. this reader is smaller, but otoh libe2fs is more flexible
06:59:27erikmheads for breakfast
07:01:42mmatten_morning -_-
07:03:15BZFlagRuss: that did it.
07:04:45BZFlagis frustrated with automake and friends.
07:11:12BZFlagutil/test/Makefile.am needs help. it works, but it's not right. /me needs to read up on automake etc.
07:12:07Russdefinately
07:14:04erikmBZFlag: explain me what you want in a message to blob-cvs-commit and I can give some advice
07:19:49BZFlagthe make there should override gcc and ld command to get host binaries.
07:20:11BZFlagthe sources used to build that app are the same as sources in other directories.
07:20:25BZFlagit now makes symlinks to each of the sources it needs.
07:20:49BZFlagideally it would use them wherever they are and not have to have the symlinks.
07:21:57BZFlagthe maintainer mode does not like the variables: load_kernel_test_OBJECTS load_kernel_test_SOURCES EXTRA_load_kernel_test_SOURCES
07:22:38BZFlagthe two sources it thinks are unused, and the objects it says should not be defined.
07:31:42erikmRuss: there already is a memset in src/lib/memset.c
07:32:20erikmoh wait, this is int_memset
07:33:21erikmgoes to work
09:03:34seletzmorning
12:01:21seletzerikm: hi
12:19:55que tal, erikm
12:19:55erikmhi
15:01:49salut, seletz
15:01:49seletzprpplague: hi
15:02:23prpplagueseletz: morning
15:02:49prpplagueseletz: you use docbook?
15:04:19seletzprpplague: no, not yet. It's one of the things i want to have a look at.
15:04:39seletzprpplague: i use TeX tho.
15:05:11prpplagueseletz: i need to find something to help collect document fixes/howto's
18:26:52here is a pic, the silk screening isnt perfect, it is more of a cream than white, but it was just a test http://www.tuxboxproject.com/images/Jan16_01.JPG
18:26:52seletzerikm: still here?
18:27:45seletzhmmm, guess not.
18:28:13seletzwell, bye folks. Happy weekend!
19:22:44BZFlagwe should extend the fs support so we have --enable-ext2 --enable-vfat (perhaps just fat is enough)
19:23:07BZFlagread from ide an read from flash should both be able to use the same fs drivers.
19:24:41Russright
19:25:06RussI'm thinking we should move towards a model more like grub, but not use grub as a model because it does a lot of things wrong
19:26:54BZFlagagreed. (on both points ;-)
19:27:24BZFlagwhy does lxrbot quit and rejoin from time to time?
19:27:33BZFlaglog cycleing?
19:27:37Russwe should really lay out how we want to do it
19:27:48Russlog cycling? wha?
19:27:58BZFlagRuss: lxrbot
19:28:09BZFlaglxrbot: status
19:28:09lxrbotSince Fri Apr 26 21:30:12 2002, there have been 0 modifications and 0 questions.  I have been awake for 1 minute, 8 seconds this session, and currently reference 6733 factoids. Addressing is in require mode.
19:28:29well, lxrbot is my fellow bot. he knows a lot about the linux kernel and normally lives on #kernelnewbies.
19:28:29BZFlagibot: lxrbot?
19:29:32RussBZFlag: erikm has lxrbot restart every now and then because it sometimes can't tell when it is disconnected
19:34:19BZFlagah. happens with ibot from time to time too, but not often so we restart it by hand.
19:34:58BZFlagwould be nice to have the infobot ping chanserv or some service from time to time and restart if it does not get a reply.
20:02:58hi, seletz
20:02:58seletzhi
20:05:26prpplaguehowdy seletz
20:05:34prpplaguewhats up?
20:05:47seletz:)
20:08:19BZFlagseletz: you saw the ext2 code I added from brad/duck ?
20:08:38seletzBZFlag: yup
20:09:08BZFlagwhat are your plans? you are going to put a tarball in flash?
20:09:13BZFlagor on a raw cf?
20:09:24seletzBZFlag: i'll try that code when i verified that i read from the CF card is correct
20:10:05BZFlagthe board I'm on wants kernel from jffs2 and reflash from fat cf ide.
20:10:09seletzBZFlag: first I'll put a tar on the CF ("raw", no partitions)
20:10:30seletzBZFlag: then I'll move to FSystem code (FAT, ext2)
20:10:46BZFlagduck is now using that ext2 code to reflash his board.
20:11:30seletzBZFlag: I want a easy method for updating the complete OS of my board (easy for customers)
20:11:48BZFlaga file on fat cf would work?
20:11:50seletzBZFlag: First step is raw tar on a CF card.
20:12:32seletzBZFlag: yup, that's what i want. a file on a FAT FS (probably a TAR archive of blob, zImage, initrd.gz and cramfs)
20:12:49seletzBZFlag: final version, that is.
20:13:08BZFlagmore unit testing is always good, so read from tar might be nice. this board just wants the two things though. kernel in jffs2 in flash and reflash rootfs from a file on fat cf
20:13:46BZFlagI just want a file called "rootfs" or similar and other for "romimage" that would also include a new blob.
20:14:07BZFlaga direct memory dump of all flash, or just the rootfs partition.
20:14:15seletzBZFlag: Have some FAT code? My code is 400kB (yuck!)
20:15:28seletzBZFlag: What about having the whole bunch in one tar arch? Would be a nice interface...
20:16:10seletzBZFlag: note that the tar code is block-aware, that is it does not rely on flat mem access.
20:16:43BZFlagno small vfat code yet. though bootldr include read and write vfat code. might borrow from there.
20:17:02BZFlagseletz: my users won't even know what tar is.
20:17:08seletzBZFlag: plain FAT 8.3 would do fone too.
20:17:09BZFlagin most cases anyway.
20:17:30BZFlagyeah, 8.3 would be fine with me too.
20:17:46BZFlagand not have to worry about the microsoft vfat patent.
20:17:47seletzBZFlag: My users for sure dont know even what an archive is. I'll provide the archive images.
20:18:05BZFlaggood point
20:18:41seletzBZFlag: Note that all this is just thought as "here you have a file. Put it on your CF card. Plug it in. Power-on. wait. Done."
20:19:05BZFlagI'd add "hold down foo. and power on"
20:19:27seletzBZFlag: And that "put it on CF" is just drag'n drop on a laptop
20:19:38BZFlagright
20:19:46seletzBzFlag: yup, maybe. But maybe the archive is deleted on CF.
20:20:14seletzBZFlag: Our devices have CF for data storage (FAT FS)
20:20:35seletzBZFlag: so they could just stay where they are ...
20:21:10prpplagueseletz: a word of caution, i've had problem with initrd
20:21:26prpplagueseletz: and rootfs images coming from fat 8.3 cf
20:21:30seletzprpplague: what problems?
20:22:10seletzprpplague: my layout is: on-board-flash( blob, initrd, cramfs) CF( data )
20:22:28seletzprpplague: so the update would be CF => on-board flash
20:22:53prpplagueseletz: i've been using the flash utility from inhand electronics which reads from fat 8.3 formated cf cards
20:23:16seletzprpplague: opensource?
20:23:48prpplagueseletz: not all of it
20:24:05seletzprpplague: the FAT code?
20:24:07seletz:)
20:24:09prpplagueseletz: ya
20:24:31seletzHmmm, can you put it on sf?
20:25:06prpplaguei can check about it
20:25:26seletzBZFlag: Ok, so let us summarize:
20:25:43seletzBzFlag: we both want some sort of update CF => internal flash
20:26:05seletzBzFlag: we both want to keep it easy for customers (fool-proof)
20:26:42Russ|werkseletz: full cramfs code is smaller than tar code
20:27:01Russ|werkif someone wants small fat code, I'll make it for $$$ :)
20:27:29seletzwell, but i dont have cramfs on my CF card
20:27:44Russ|werkseletz: ?
20:28:30seletzI have a file on CF card (FAT) which contains an system update (blob, initrd, kernel, cramfsimage)
20:28:55Russ|werkwhat is a system update?
20:29:01Russ|werkoh, I see
20:29:07seletzATM i just want to leave away FAT and put a tar on the CF (raw)
20:29:13Russ|werkthe CF card is fat so it can be written by a win machine
20:29:18Russ|werkthen you put a tar on there with everything in it
20:29:24seletzyeah. Now you got it.
20:29:35Russ|werkI'm saying, why not put everything in a cramfs instead of a tar
20:29:46seletzwin machine = customers laptop
20:29:48Russ|werkor, if you really wanted simplicity, a table
20:29:59Russ|werkie:
20:30:22Russ|werkstruct dir_entry { char filename[20]; long size; long offset; }
20:30:24seletzyeah, i know. But tar is simple
20:30:42Russ|werkstruct root { long num_entries; }
20:30:47seletz(ok, it got some weird octal BCD stuff)
20:30:55Russ|werkroot, then dir_entries, then files
20:31:01Russ|werkmuch simplier
20:31:05seletzyeah, would work.
20:31:34seletzbut now i can do a "tar cvf /dev/hda blob initrd cramfs.image"
20:32:27seletzsimpler in terms of: smaller header, easier to decode.
20:32:53seletznot simpler in terms of: portable, re-implementing the wheel :)
20:33:04Russ|werkright, thats why I suggested cramfs
20:33:21seletz(You'd have to write an archiver etc.)
20:33:30Russ|werkcramfs is actually pretty similar to what I described, but with block lists
20:33:30seletzok, that apporach is ok
20:33:48seletzyou could do mkcramfs ...
20:33:50seletzhmmm
20:34:09Russ|werkthat would compress things a bit more, I don't think it would give you a crc tohugh
20:34:11Russ|werker, though
20:34:17seletzwell, ok.
20:34:25Russ|werkyou probably want to do a crc or md5sum if you are including blob
20:34:26prpplaguei like the tar idea
20:34:53seletztar is useful anyways IMHO
20:35:14Russ|werkbut the code is so big...
20:35:49seletzjust do a xmodem download of an archive and flash every single file to different partitions
20:36:32seletzor even: have filenames match partition names
20:37:00seletzwell, the tar code isnt written with code size in mind
20:37:11seletzerm, wasnt
20:39:51BZFlagreads
20:42:29BZFlagseletz: Russ has a point though the cramfs code in in there and debugged.
20:42:43BZFlagso a cramfs on cf would be easier to get to as a "first step"
20:43:18Russ|werkBZFlag: actually its not
20:43:26Russ|werkthe current cramfs.c is very incomplete
20:43:30BZFlagI'm happy with partitions having names mapped in blob someplace, and then blob looks for those names on fat cf.
20:43:34Russ|werkbut I have some here that I wrote for grub
20:43:41BZFlagthat's what I want long term. no tar code.
20:44:06BZFlagRuss: I though you were using the cramfs code on aboard there?
20:44:14Russ|werkright, but it only loads /linux
20:44:59BZFlagah. true.
20:45:35BZFlagwell the reflash from cf would be files in the root too.
20:46:26BZFlagso /<partitionname> where the name is hopefully 8.3 or less.
20:47:05BZFlagblob,kernel,ramdisk params would all be fine. I'd like a "romimage" that is all of flash too.
20:47:29Russ|werkwell, you'd want to flash things like blob in one peice asap
20:47:41Russ|werknot erase everything, then start flashing blob
20:47:48BZFlagagreed.
20:48:06prpplaguenot to be a pest, but have you guys actually tried storing initrd and rootfs images on fat 8.3 before?
20:48:14BZFlagI'd like that to happen automagically for "romimage".
20:48:39BZFlagprpplague: yes on the zaurus, bt that uses angel monitor, not blob (yet)
20:50:01prpplagueBZFlag: hmm, i've done a couple of test where i transfered an initrd from ext2 to a fat8.3 then back to ext2, the resulting initrd was unusuable
20:50:40Russ|werkprpplague: I would guess something was wrong with the implementation
20:51:03Russ|werkok, cramfs booting on doc2000 x86: check
20:51:14prpplagueRuss|werk: implementation of what? the initrd?
20:51:29Russ|werknow I just have to figure out how to store config stuff on the doc200
20:51:32Russ|werker, 2000
20:51:41Russ|werkprpplague: fat reader/writer
20:54:12prpplagueRuss|werk: this was from one HD to another HD and back
20:54:16prpplagueRuss|werk: http://www.abcsinc.com/yopy/tcsx1-rootfs.tar.gz
20:54:42prpplagueRuss|werk: your welcome to look at the tcsx-1 rootfs for an example on how i handel configs
20:56:10Russ|werkdid you check your kern.log to see if there were any IO errors or DMA problems?
20:57:24Russ|werkuncormpresses fine for me
20:58:04Russ|werkthats a big root
20:58:32BZFlagprpplague: sounds like you have bugs someplace. I've done the same hundreds of times.
20:59:52Russ|werkI work with an image that is 1.3M compressed including kernel
21:03:54prpplagueBZFlag: ya thats what i thought but i never could trace it down
21:04:16prpplagueRuss|werk: thats the rootfs for our little thin client
21:04:54Russ|werkI guess X really weighs in
21:05:18Russ|werkis it a ro fs?
21:05:28prpplagueRuss|werk: ya something like 3mb for the xserver
21:05:35prpplagueRuss|werk: ya totaly ro
21:05:59prpplagueRuss|werk: except for when saving configs
21:06:02Russ|werkdid you know that bb init has a mechanism for chrooting once booted an running?
21:06:26Russ|werkyou can signal init, chroot to a ramfs, umount the rootfs, then flash it
21:07:43prpplagueRuss|werk: that fs was done about two years ago and was my first attempt at an embedded system
21:08:07Russ|werkah, well embedded stuff has come a long way
21:08:19prpplagueRuss|werk: yes it has
21:08:29prpplagueRuss|werk: and so has my education
21:08:55Russ|werkI also have an mkcramfs that works with the device table that mkfs.jffs2 uses, but it has a couple bugs I need to work out (non-fatal)
21:09:23prpplagueRuss|werk: http://www.linuxdevices.com/articles/AT3193512578.html
21:09:38Russ|werkya, I saw that
21:10:04Russ|werkbe pretty usefull for MLS terminals (realestate thing)
21:10:12prpplagueRuss|werk: lol i did get a good plug for it "Anyone trying to build their own embedded Linux device would do well to study how Advantage put the TCSX-1's software together"
21:10:53Russ|werkthe one thing I would say is its probably a waste if ram=flash
21:11:10Russ|werka ratio of 4 to 1 or 2 to 1 is probably better
21:14:03BZFlagRuss: it's just a DOM though, you could replace it with anything you want.
21:14:05prpplagueRuss|werk: yes very wasteful, i've gotten a newer version on 2.4 kernel and using all bb and buildroot
21:14:17prpplagueRuss|werk: yes including DOC
21:15:26Russ|werkprpplague: any good ideas on saving configurations to a DOC?
21:15:45Russ|werk(ie, revision stampted so you can use the most recent one that has a good crc)
21:17:55prpplagueRuss|werk: are you talking stuff like ipaddress,passwords,iptables rules, that kind of stuff?
21:18:12Russ|werkand dhcp leases (saved every 2 - 3 hours)
21:18:26Russ|werkdoing it on NOR flash is easy
21:20:19prpplagueRuss|werk: well, what i like to do is use a small partition that i store the configs in a .conf file. when things change, the data is writen to a file in temp, the partion is re-mounted rw the file is copies, and then remounted ro
21:20:37Russ|werkwhat fs?
21:21:00prpplagueRuss|werk: i've used both jffs and ext2
21:21:23Russ|werkon NAND flash?
21:24:06prpplaguei assume you are refering to using it on DOC, ya it works fine
21:25:41prpplaguei've used the same tcsx-1 rootfs with DOC simply subsituting the /dev/hda with /dev/nftla and a modified lilo
21:30:55Russ|werkhmm...I have trouble trusting that
21:32:22prpplagueRuss|werk: why?
21:32:53Russ|werkbecause a lot of gc and wear leveling goes on
21:33:26Russ|werkwhat if some sector is getting erased that is very important for the fs, and power is lost
21:36:02prpplaguethats just it, the only section that is at risk is the /configs , nothing in the os
21:38:11prpplagueRuss|werk: we have around 2500 units out since november of 2001 using the DOC2000 on this fs with less than a .0002% failure rate
21:39:56prpplagueRuss|werk: my bad .0004%
21:40:09Russ|werkright, but I plan on writing the config every 2-3 hours
21:41:01prpplagueRuss|werk: ya that would present a considerable more difficult problem
21:41:47prpplagueRuss|werk: what kind of memory will you have on the x86 board?
21:42:58prpplagueerikm: all hale the arrival of mr.blob
21:43:09erikmheh
21:43:11erikmbows
21:43:48Russ|werkprpplague: doc2000
21:44:20prpplagueRuss|werk: what about ram?
21:44:43Russ|werkjust 72pin
21:45:13prpplague16mb?
21:45:44Russ|werk8M is enough
21:46:27Russ|werkgetting mobo, power supply, CPU, and 16M ram for $25, then putting it is a custom cube case
21:46:50prpplagueRuss|werk: cool
21:46:55Russ|werkadd a couple nics, a 2M doc on an ISA card, and its set
21:47:54prpplagueRuss|werk: both nics 10/100?
21:48:43Russ|werkyup
21:48:49Russ|werkRTL8139
21:49:15Russ|werkthe SA1100 design is much smaller and more reliable, but the CS8900A's are only 10Mbit
21:49:29Russ|werkwith the x86 version, we get to put in other things besides nic cards though
21:49:33prpplagueRuss|werk: i could send you tcsx-1 board, but you'd only have one 10/100 and one 10
21:49:42Russ|werk(56k, a usb card for a 3com ISDN modem, etc)
21:50:43Russ|werkI'd really like to move everything over to the embedded arm due to reliability problems
21:50:53prpplaguenods
21:51:17prpplaguei know the "reliability problem" all too well
21:51:36Russ|werkaltough a custom power supply/cmos battery combo may be able to solve that
21:51:42Russ|werkbut would add considerable cost
21:52:53prpplaguewhich mobo?
21:53:30Russ|werksome P120 ast thing
21:53:52prpplagueRuss|werk: btw, you still want to play with that camion unit? i still have not been able to get any source, so your welcome to play with it
21:54:01Russ|werkcourse
21:54:20Russ|werkit may not end up being that difficult to RE
21:54:39prpplagueRuss|werk: ok then, sometime next week, i'll ship it and a tcsx-1 board
21:54:46prpplagueRuss|werk: want anything else/
21:54:47prpplague?
21:55:11Russ|werkif you have any devel docs at all to go along with the camion
21:55:56Russ|werkI might be able to get the tcsx reselling as a mls terminal (telnet terminal basically)
21:56:06prpplagueRuss|werk: thats the deal, not a darn thing, i can't even get a return email
21:56:38Russ|werknot much you could do short of flying out to korea I suppose
21:56:50Russ|werkor trying to call the developers at their home
21:57:55prpplaguelol, well, i've ftp'd to their server and left a ton of messages, sent about 1000 faxes, and flooded their email server, so guess the next is calling them at home
21:58:14prpplagueits a real shame too, its a cute little unit for the price
21:58:53prpplaguei wanted them to make a dev board without the case
21:59:17Russ|werkoh, my mailing address is changing in 2 weeks
22:00:08prpplagueRuss|werk: i;ve got a bunch of arm+net toys too
22:00:22Russ|werkprobably would want to mail it to 4121 E Vista Dr; Phoenix, AZ 85032
22:00:52prpplagueRuss|werk: ok noted
22:01:07prpplaguemoving to az?
22:01:26Russ|werkI'm in AZ
22:01:36Russ|werkI'm graduating and they are kicking me out of the dorm
22:01:49prpplagueRuss|werk: ohh, for some reason i thought you were chicago
22:08:21prpplagueargh, server is down in hawaii, guess i need to go and fix it
22:09:17erikmprpplague: lucky guy!
22:10:24prpplagueerikm: not really, i don't like hawaii, my area is the bahamas
22:11:34Russ|werkthey need that remote robot arms thing for techies like the have for surgeons
22:12:46erikmRuss|werk: a remote reset button is quite easy to build
22:13:55Russ|werkright, but that doesn't fix a crashed hd
22:17:41prpplagueRuss|werk: sssssssshhhhhhhh, i'd never get to go anywhere
22:17:56Russ|werkis it on the big island?
22:20:51prpplagueRuss|werk: honolulu
22:21:00Russ|werkbummer
22:21:39prpplagueRuss|werk: doesn't matter they're back up, looks like a seagate dat went bad and was shorting out the scsi bus
22:22:20Russ|werkouch
22:23:05prpplagueohh well, guess i'l head to beer store instead
22:23:13prpplaguebbl

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with infobot logs, split per channel and by date, etc.