irclog2html for blob on 2002.05.29

00:15:49prpplaguehowdy howdy
01:49:10Sammyhello all
03:12:24Russhey Sammy
03:12:28Russget some more shaking?
03:47:17SammyRuss:  as usual as that thing ...
03:49:05Sammybut before that , I am already shaking in my head from some beer ...
05:56:54Russtim, do you think andersee would want one of these?
05:56:57Russhttp://www.thinkgeek.com/stuff/apparel/592a.shtml
07:06:56mmatten_morning all
07:21:47Russcan't seem to fool bootldr into thinking that blob is a kernel image
08:37:26Sammynew platform ?
08:40:57mmatteneither i'm incredibly stupid, incredibly blind or somehow sqrt() doesn't work as i think it does
08:41:21mmattenon ARM that is
08:56:25mmattenis there some magic compiler switch there or why does it return what i throw in for calc? sqrt(9) returns 9 on my sys which is kinda stupid (the test prog works fine on my main pc)
10:48:52seletzhello
10:50:52erikmhi
10:52:19seletzerikm: I know you're very busy, but have you seen my generic io stuff i did for blob?
10:52:27erikmseletz: not yet
10:52:41seletzerikm: ok
10:54:02seletzerikm: my sysupd stuff works quite well. update from CF->blob flash partitions. Maybe sometimes you get around to look at it.
10:55:05erikmwill do
11:37:29seletzerikm: silly question: what could be the cause of sin() and sqrt() not working on ARM? (i have a libm, i have compiled in fpe in my kernel)
11:37:34seletzpuzzled
11:45:07erikmseletz: I don't know. those are the two obvious things
11:46:03seletzhmm
11:46:12seletzerikm: thanks anyway
14:30:27seletzprpplague: hi
14:31:01prpplagueseletz: morning
14:49:21prpplagueseletz: whats on your plate today?
15:11:20seletzprpplague: havin strange ln problems, cramfs update fun and all that :(
15:11:48seletzprpplague: ever head that ln -s is behaving very strange over NFS?
15:12:43seletzprpplague: over here "ln -s somefile link" does weird things. looks like the link name is not null-terminated.
15:13:21prpplagueseletz: yes i have had strange problems with ln over nfs
15:13:42seletzprpplague: ahh, really? Then its not me alone ...
15:13:54prpplagueseletz: part of our software uses ln functions, and we were not able to run it over nfs due to this problem
15:14:27prpplagueseletz: as too wether it was just me? don't know, it was just a test run and i didn
15:14:34prpplague't research it much
15:14:52seletzprpplague: what is uname -a saying on your system?
15:15:29prpplagueseletz: when i tested this, it was on rhat 6.2 2.2.14 kernel
15:16:15prpplagueits been awhile
15:33:15Russseletz: are you using fastfpe (sqrt problem)
15:33:43prpplagueRuss: howdy
15:34:33seletzRuss: no, the normal one
15:34:53seletzRuss: but meanwhile i guess its a libc/libm/gcc problem.
15:35:32seletzRuss: we're currently investigating...
15:35:45Russseletz: I know nwfpe has had issues in the past
15:35:49Russmorning prpplague
15:36:13seletzRuss: ah, ok. Then i'll forget to switch to it for speed reasons.
15:36:37Russseletz: nothing wrong with trying fastfpe
15:36:54mmattenthat'd be a rather hefty issue
15:37:06seletzRuss: ah, ugr. I get confused ....
15:37:19mmatteni mean sin and sqrt aren't _that_ uncommon :)
15:37:41seletzRuss: nwfpe is the standard one, right? the fastfpe is the one tweaked with no exceptions?
15:38:07RussI don't know if it has no exceptions
15:38:20RussI just no its designed to be faster at the cost of not being perfect
15:38:32seletzgot this knowledge(?) from config.help ....
17:10:13seletzupdating cramfs _again_
17:10:38Russ|werkit only takes a few seconds...
17:10:45seletzlol
17:10:47Russ|werk(at least with ftp)
17:11:17seletzRuss|werk: i dont know your system, but mine has a cramfs root
17:11:27Russ|werkso does mine
17:11:43seletzRuss|werk: hmm. Ok, how do you do it?
17:12:07Russ|werkbusybox init accepts a signal that makes it kill everything an exec another program
17:12:36Russ|werkthe script then copies a statically linking flashing program to a ramfs where the image already resides
17:13:36seletzRuss|werk: hmm, so you update cramfs root while a kernel runs, with mountet cramfs??
17:13:42Russ|werk/proc and the other ramfs's are umount'd, piviot_root, chroot and exec flash program, flash program umounts old root (cramfs)
17:13:58Russ|werkflash program flashes cramfs and reboots
17:14:06seletzRuss|werk: way cool
17:14:30Russ|werkcould alternately remount root and re-exec /linuxrc
17:14:52seletzRuss|werk: have this flash proggy lying around somewhere? :)
17:15:01Russ|werkits pretty simple
17:16:08Russ|werkit uses a umount system call, mmaps the image, opens the mtd char device file (which has also been copied to the ramfs)
17:16:10BZFlagI don't bother unmounting the old root. I just remount it ro (for jffs2 etc) and sync;reboot after updating.
17:16:37Russ|werkBZFlag: works until you get a page miss
17:16:43seletzhmm, maybe i have a simpler case, my root is RO
17:17:06BZFlagif you have done the pivot_root you should not get a miss ever, should you?
17:17:08Russ|werkdoes the erase ioctl across the entire region
17:17:34Russ|werkBZFlag: if you are already piviot_rooting, why not umount
17:17:50Russ|werkthen writes it out, closes, and the reboot()
17:17:57BZFlaggood point, I just have not added that to the script.
17:18:38Russ|werkall you do is killall -HUP init
17:19:00Russ|werkand have a ::restart:/linuxex line in inittab
17:19:02seletzRuss|werk: ca::: in inittab?
17:19:09seletzah, i see
17:19:20seletzca is -INT
17:19:37BZFlagI am meaning to update the script assuming busybox, and have it copy busybox and needed libs into ram, pivot_root there and then camm busybox for everything. ie: "./busybox mount ..." and "./busybox dd ..." etc.
17:20:21BZFlagseletz: ca is non-busybox init. Russ is using busybox init.
17:20:38seletzohhh
17:20:41Russ|werkBZFlag: the way I'm doing it, I don't need busybox or libs in ram
17:20:41seletzsh***
17:21:03BZFlagRuss: but you need your binary instead. size?
17:21:05seletzdoes'nt busybox init have ca?
17:21:22BZFlagseletz: not really. feel free to add it. ;-)
17:21:43Russ|werkBZFlag: 18k
17:21:45seletzBZFlag: i need ca for system poweroff.
17:22:00seletzRuss|werk: 18k? holy Sh***, thats small
17:22:52seletzadds this to his todo list
17:23:47BZFlagseletz: why? "shutdown" is not enough?
17:23:58seletzRuss|werk: when i open a mtdblock dev, cant i just write() to it? Do i have to do some ioctl()?
17:24:35seletzBZFlag: no. I have a PIC which gives me an IRQ via GPIO18 for poweroff.
17:24:36BZFlagRuss: might be worth adding the source to the busybox tree someplace as example code?
17:25:36seletzBZFlag: i do a kill_proc( 1, SIGINT, 1) in my irq bh.
17:26:15seletzBZFlag: that way its just like a PC box: user pressed Ctrl/alt/del
17:27:24Russ|werkseletz: with mtdblock, you can just write
17:27:53Russ|werkseletz: but I use mtdchar, in case I'm writing a jffs2 image (because I'm purposely erasing the whole thing)
17:28:13seletzRuss|werk: ok. why do you a mmap()? To use memcpy()
17:28:28Russ|werkso I can use write, and not read/write read/write
17:28:47Russ|werkand also, I don't need to allocate a buffer that I'm going to copy stuff in and out of for no reason
17:28:52Russ|werkthe image in already in ram
17:29:21seletzRuss|werk: sorry if this is silly, why do you erase when using jffs?
17:29:41seletzRuss|werk: and not when using cramfs
17:29:54Russ|werkif you have old jffs(2) stuff lying around (past the end of your image), it will pick them up and use them
17:30:08Russ|werkcramfs doesn't do that
17:30:18seletzah, yes, i remember.
17:30:20BZFlagRuss: why does using char device help?
17:30:40Russ|werkBZFlag: since I'm erasing the whole thing anyway, why bother with mtdblock?
17:31:05Russ|werkalso, it allows the use of mtdblock_ro if you are using cramfs
17:31:06BZFlaghmmm... I suppose.
17:31:13Russ|werk(which still oopses on me...)
17:32:21seletzBZFlag: is there a cleaner way to do a poweroff switch? I didn't find one ...
17:33:01seletzRuss|werk: thanks..
17:35:08BZFlagseletz: not that I can think of, nope.
17:36:44seletzRuss|werk: man, that simple if you know how to do it...
17:41:36Russ|werkseletz: ya, I also took trollftp and stripped it to the bone
17:41:45seletzRuss|werk: so there's no magic restriction on pivot_rooting() more than one time
17:42:08Russ|werkso it basically only accepts a send, and if it gets a valid image, it just signals init
17:42:09seletzRuss|werk: ya, trollftp. That one i found too. Dont you use wget?
17:42:46Russ|werkI'm talking enduser flash updates
17:42:49seletzRuss|werk: that crazy? cool. automaic update initiated from a host.
17:43:07Russ|werkand its really easy to put the perl ftp module in a make file
17:43:12Russ|werkso you do make image ftp
17:44:07seletzRuss|werk: well, now my simple CF update/install method looks quite clumsy. :(
17:44:45seletzRuss|werk: I thought about sending CF disks to customers (with updates on it).
17:45:08Russ|werkwell I used to do ramdisks, so ftp updates to flash were very easy, since root was mounted on a ramdisk
17:45:20Russ|werkso when I switch to root mounted on cramfs, I wanted the same thing
17:46:35Russ|werkcf update wouldn't hurt if they screw root though
17:47:04seletzRuss|werk: works, btw. From blob. Its in CVS.
17:48:04seletzRuss|werk: system3.c::sysupd(). waiting for a review to integrate it fully into automake/conf.
17:48:35Russ|werkon my arm boards, I put a second cramfs root that just has a very minimal kernel and a ftpd daemon with the ability to flash root
17:48:52Russ|werkso if it gets screwed, you can hold down a button on power on, and it will boot to that root instead
17:49:26seletzRuss|werk: i have a initrd where i can init=/rootshell
17:49:32Russ|werkcosts me 393k
17:49:42Russ|werker, 384k
17:50:48seletzRuss|werk: i guess you basically have a own init? Or du you use busybox with scrips?
17:52:39Russ|werkI use the statically linked ftpd as init
17:52:41seletzneeds some time to digest Russ' ideas :)
17:53:05Russ|werkand use the kernel to configure the interfaces (ip on the command line)
17:53:56prpplagueany of you guys looked at the code for mingetty?
17:54:09seletzprpplague: no, sorry
17:55:09prpplagueseletz: i've got a strange cursor problem on mediaGX1 board
17:55:17seletzRuss|werk: so you merged the trollftp, your flash.c and bingo. cool. That's the way to go.
17:55:24seletzprpplague: cursor? mingetty?
17:56:16seletzprpplague: oh, ok, sorry.
17:56:26seletzneeds a break
17:56:49prpplagueseletz: ya the cursor stays white ignoring all ansi color codes, until you switch between mingetty screens, then the cursor works fine until you log out
17:56:50seletzSIGOVERFLOW, so to say :)
17:57:48seletzprpplague: ugh, sorry. I'm not (yet) into terminals that much.
17:58:12prpplagueseletz: yes its very strange
17:58:21prpplaguedl's mingetty source
17:58:54seletzprpplague: i'd try to issue some echo's from shell with the ESC codes you wand.
17:58:59seletzs/wand/want
17:59:03Russ|werkI just use busybox getty spawning busybox sh
17:59:30prpplagueseletz: ya thats what i did for debugging
17:59:40Russ|werkI need a decent telnetd though
18:00:03seletzhmmm, there was an announce recently of a minimal telnetd...
18:00:10prpplagueRuss|werk: there is a new small telnetd that was possed on linuxdevices.com
18:00:32seletz http://www.pengutronix.de/software/utelnetd_en.html
18:02:53seletzRuss|werk: that URL is about utelnetd "A small telnet daemon ..."
18:03:11seletzRuss|werk: did'nt try it yet, though
18:03:54Russ|werkin the changelog, it says there is a telnetd in busybox
18:03:55OK, seletz.
18:03:55seletzibot: utelnetd is at http://www.pengutronix.de/software/utelnetd_en.html
18:04:23Russ|werkit must mean the netxit-tiny telnetd
18:04:39Russ|werkI've been using that, but it has some serious issues with negotiation
18:04:49Russ|werkit takes responses as data, and echos it back
18:05:06Russ|werkwhich the host takes as further commands, and echos more responses...
18:05:17Russ|werkso you have to use nc
18:06:07prpplagueRuss|werk: worst cause you can use the standard telnet and run it in debug mode, which is what we do on our thin clients
18:06:22prpplagues/telnet/telnetd
18:06:42prpplagueRuss|werk: that way you don't need to have inetd
18:07:12Russ|werkya, its from the netkit-tiny telnetd
18:07:24Russ|werkI'm guessing they fixed it
18:07:42seletzRuss|werk: you refer to utelnetd?
18:07:57Russ|werkyes
18:08:18seletzuahhrg
18:08:28seletzgets tired
18:09:05seletzok, i'm leaving for today. No more cramfs uploads. sigh.
18:10:00seletzRuss|werk: thank you again, you gave me some very cool new ideas to think of ....
18:10:09seletzgoing home
18:10:49seletz_awayhave a nice day...
18:11:33Russ|werk9.4k, nice
18:12:07prpplagueseletz_away: later
18:35:21Russ|werkhmmm
18:35:32Russ|werkutelnetd not quick working for me
18:35:36Russ|werk"All network ports in use!"
19:36:42prpplagueRuss: usually that indicated the you are missing the ttyp? and ptty?
20:08:23BZFlagtuxscreen has inetd and telnetd both from netkit-* in the buildroot.
20:17:02prpplagueBZFlag: you got any experience with geode processors?
20:19:45BZFlagyep, but I just did x86 stuff on them
20:25:11prpplagueBZFlag: have you ever run into any cursor color/control issues on the console using a geode?
20:26:59prpplaguethe cursor stays white ignoring all ansi color codes, until you switch between mingetty screens, then the cursor works fine until you log out
20:28:21BZFlagprpplague: hmm. nope. tried the bb reset command? that's worked for me on other systems.
20:28:55prpplagueno, not familiar with it
20:29:00prpplaguechecks man page
20:30:37BZFlagprpplague: as in the reset command in busybox.
20:31:22prpplagueBZFlag: ya i followed
20:31:37prpplagueBZFlag: looks like just a function of tset
20:33:15prpplagueBZFlag: thanks i'll have a look
20:50:02BZFlagk
21:07:28Russ|werkhas anyone seen this?
21:07:29Russ|werkhttp://www.tmi.com.tw/pp-TC5299J.html
21:29:40prpplagueRuss|werk: wow
21:29:52prpplagueRuss|werk: am i reading that right?
21:30:18prpplagueRuss|werk: i could hook that up to the sa-1110 pcmcia with just a little glue?
21:31:21prpplagueBZFlag: thanks for the help that didn't solve the problem but i was able to track it down to a bogus bios
21:33:22Russ|werksimilar to the smc chip
21:33:37Russ|werkthe cirrus 8900A is preferable, since it has a built in MII
21:33:43Russ|werkbut its only 10baset
21:34:08Russ|werkeveryone says it can't be done due to a number of issues, but I want to see what it would take to drive PCI devices
21:34:38Russ|werkcould really push a lot of data with a 32bit bus at 40Mhz
22:33:02prpplagueRuss|werk: seen this - http://www.thinplanet.com/products/generic.asp?f=PRnumber&k=s&v=PR47427
22:33:13prpplagueRuss|werk: might make a nice test platform
23:13:31Russ|werkheh
23:13:36Russ|werker

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.