irclog2html for blob on 2001.12.05

01:53:21SammyYo :)
02:58:27Sammyknee how prpplague :)
02:58:55prpplaguewo han
02:59:53prpplaguehow
06:38:29asamianyone knows clart's flash spec?
07:11:13asamianyone knows clart's flash spec?
08:47:37asamianyone knows clart's flash spec?
09:37:28Russyes amrutha?
09:38:04SammyYo Russ
09:55:58Sammyecho " Russ "
09:56:37Sammyecho " still "
09:56:40Sammyecho " Kernel panic: No init found.  Try passing init= option to kernel. "
10:05:20amruthai need a little help regrading the blob
10:13:37Sammywhat help ?
10:21:08SammyYo erikm , you back ....
10:22:01erikmyes
10:23:22Sammyerikm : anything new ?
10:24:28erikmnot really
10:25:40Sammyohhh... 2.0.5 pre2 a little big ...
10:26:11Sammy35364 Dec  5 18:35 blob
10:26:41Sammybut I add some debug massage
10:56:06erikmyes, a bootloader with all features and debug messages can be quite large
11:04:20SammyDinner .......time
12:00:19seletzhi!
12:02:10skandaseletz: I need help regrading certain aspect of the blob
12:02:16skandacould  you please help me
12:02:28seletzskanda: what kind of help :)
12:02:42seletzskanda: i'll helf if i can
12:02:49seletzs/helf/help/
12:02:59skandaok
12:03:14skandaI am trying to implement a flash tester on the Assabet
12:03:31seletzskanda: ok. What's the problem?
12:03:47skandaAll I do is, start the blob, run a Diagnostic to test the flash integrity and move on
12:04:21skandasince I cannot test the flash while I am executing instructions from the flash itself, I try to branch to the SDRAM
12:04:47seletzskanda: ok.
12:04:58seletzskanda: is sdram initialized a that point?
12:04:59skandabut before that I try to dump the necessary flash testing code to the RAM and start executing instruction from the RAM
12:05:04skandayup
12:05:30skandainitializd as in setting up all critical parameters like SLFRSH
12:05:37seletzskanda: AFAIK blob already copies itself as soon as possible to ram and starts executing from it.
12:06:12skandareally, but I thought once you JFlash it, blob sits in the FLASH
12:06:14seletzskanda: all C code is executed from RAM in fact.
12:06:21skandaand executes code from there
12:06:32seletzskanda: yes. thats right.
12:06:46seletzskanda: blob is stored as a whole in flash
12:06:49skandaso how will it copy itself to the RAM
12:06:54skandaok
12:06:55seletzskanda: but consists of 2 parts
12:07:02skandaok
12:07:09seletzskanda: one part is pure assembly
12:07:17seletzskanda: one part is C code
12:07:41skandai understand now
12:08:02skandano but my idea goes something like this ok
12:08:06seletzskanda: the assembly part sets up SDRAM, initializes the serial port, checks memory areas to be used, and initializes a C stack. Then it branches to c_main().
12:08:21skandaok I get it
12:08:49skandabut I am trying to check the flash before C code is reached,
12:09:09skandaif a flash failure occurs, the blob should never set up a C stack
12:09:17seletzskanda: why?
12:09:44skandalets say as part of a routine check or as part of something like POST
12:09:50skandaok
12:10:33seletzskanda: why do you want to use assebmbly code for that? I'd rather use C as soon as possible.
12:11:12seletzskanda: even a POST routine can be easy made in C. In fact, diag has quite some bit of C code for that already.
12:11:55skandaseletz: ...right now..as diag, i have checked the sdram. next on my list is to check the flash.i have to use assembly for everything.
12:12:05seletzskanda: When your flash is completely hosed up and blob is corrupted, then you're out of luck anyway IMHO.
12:12:45seletzskanda: Why? Flash access is ready to use in Blob already, and those routines are pure C.
12:13:06skandaexactly:thats why I need to branch to the SDRAM thru Assembly, perform checks on a random blocks
12:13:25skandaand continue with normal boot if FLASH is okay
12:13:58seletzskanda: IMHO what you really want is a CRC checksum of the blob partition.
12:14:07skandayup
12:14:30seletzskanda: and a bit of assembly check of that checksum, stored in BLOB object.
12:14:39skandaok
12:14:53skandaseletz: how do i go about that?
12:15:47seletzskanda: well, I'm not completely sure, but i guess someone already has written an assebly CRC32 routine for ARM. Maybe it's in the kernel somewhere.
12:16:49seletzskanda: To be honest, a CRC32 command that calculates CRC32 sums of arbitary Memory blocks is on my wish list for quite some time now.
12:17:11seletzskanda: Hmm, I'll talk to erik when he's back from lunch.
12:17:26skandaok, wow
12:17:35seletzskanda: He's a bit more into assembly code than myself ...
12:17:41skandaseletz:lemme clear my doubts..now..
12:17:59skandaonce the blob is being written..i have modified the blob to check the sdram.
12:18:20skandaso..u thimk checking the flash is meaningless..??
12:18:37skandasince the blob is written into the flash..
12:18:47seletzskanda: ok, let me explain my thoughts again.
12:18:55skandamy idea was..
12:19:08skandaone portion of flash is writrten with blob
12:19:48skandai needed to check the other blocks of flash..
12:19:48skandahow do i do that?
12:20:36skanda8well
12:20:50seletzskanda: ok, take this scenario for a start: I have a handheld-device built in large-scale numbers, delivered to customers who use them
12:21:04seletzskanda: Blob is used to boot linux. So.
12:21:05skandaok
12:21:09skandaright
12:21:43seletzskanda: The main concern now is: how does one make sure that blob does not get corrupted during a OS updade/FS update or similar, and
12:21:55seletzskanda: how does one detect if blob was corrupted.
12:22:13skandaonly at run time
12:22:23seletzskanda: AND how does one verify that those updates got correctly uploadfed.
12:22:49seletzskanda: so, if we had a CRC32 routine boild into blob, and if we had a
12:22:59skandaok
12:23:36seletzskanda: pre-calculated (compile-time) crc32 sum compiled into blob, then we could check taht every
12:23:41seletzskanda: startup. If it does'nt match, we refuse to boot. Service time, basta.
12:24:14seletzskanda: the crc32 routine checks whole mem blocks, so it is nearly 100% sure
12:24:23seletzskanda: thet if the sums match
12:24:29skandaso..the checksum is a constant right.
12:24:45seletzskanda: yes, say a 32 bit hex number.
12:25:15seletzskanda: up/downloads _could_ done the same way.
12:25:21seletzskanda: and more so:
12:25:44seletzskanda if one writes a flash area with new content (kernel update)
12:26:02seletzskanda: one could calculate the sum after the write and
12:26:18seletzskanda: make sure, that the content in flash is ok.
12:26:39seletzskanda: so, when your flash wears off and you get write errors/or
12:26:54seletzskanda: "loose" any bits, you detect them immediately.
12:27:25seletzskanda: agreed?
12:27:55skandaseletz:don't you think..sum varies for any new right?
12:28:33seletzskanda: yes, the checksum is different for every version of blob/kernel/whatever.
12:28:43seletzskanda: the trick is:
12:29:00skandasorry.. new write into the flash?
12:29:08seletzskanda: pre-calculate the sum of a known-good image
12:29:19seletzskanda: store that image into flash
12:29:40seletzskanda: re-calculate the checksum using the new content in flash
12:30:00skandaprovided the code remains the same for all the handhelds i write..
12:30:10seletzskanda: and compare. the sums (pre-calculated and the online-calculated) must match.
12:30:41seletzskanda: one could even store the sum in the same image, say the first 4 bytes of the image ist he sum of the rest
12:31:49seletzskanda: there's even a unix command: "sum"
12:32:27skandalemme4 understand..
12:32:32seletzskanda: all that packaging of an image could be done in a shell script of, say 4-8 lines.
12:33:06seletzskanda: on moment pls, let me get som caffeine ... :)
12:33:11seletzgetting coffe
12:33:19skandaokie..so i have an image..precalsulated...wuith a working flash..
12:33:39skandanext i load the next flash...calculate image length there and check right?
12:46:17seletzback
12:46:45seletzskanda: yes. right.
12:47:11seletzskanda: for every flash write you have to calc the checksum of the written flash area.
12:47:30seletzskanda: and compare the result with the pre-calculated checksum.
12:48:16seletzskanda: a dedicated flash-tester would write random blocks of data from memory (SDRAM) to
12:48:35seletzskanda: flash and does the same thing. Compare two checksums.
12:57:57skandaok
12:58:01skandasorry for the delay
12:58:19seletznp
13:02:55skandathanx a lot seletz, will look into it right now
13:03:28skandabut I will certainly need more help from you in the course of my implementation
13:03:31skandaok
13:04:15seletzskanda: yes, ok.
13:04:36seletzok folks, i'have to boot win2000 (eeeek!) now, so long ....
14:18:06seletzhi again
14:19:44seletzSammy: hi!
14:20:12Sammyseletz : knee how
14:25:27Sammykaun nee ge wa : erikm , seletz , Russ,
14:26:04seletz"woa heng mang" is all i can say in chinese ...
14:30:43Sammyohh...I can't image that  word , seletz what's that mean ?
14:30:54seletz"i am very busy"
14:30:59seletzsort of.
14:31:18seletzperhaps "hen mang"
14:31:29seletzdont know how to write.
14:31:42Sammyoh, I know ,
14:32:19Sammylet me show you ....want moment
14:40:48Sammyseletz: check here http://211.75.21.252/chinese.jpg
14:44:24Sammyand both the same to pronounce , but second line most use on your sex time ( you know ) "a long long time ....".
14:47:59Sammychange place ...
14:48:04Sammylater
14:56:37seletzcool. learning chinese on #blob :)
15:00:12prpplaguemonring all
15:00:27seletzprpplague: evening!
15:00:34seletz:)
15:18:39sammy_wmsis back again ...
15:20:54sammy_wmsprpplague: some new words you can check out in http://211.75.21.252/chinese.jpg
15:29:32prpplaguecool
15:56:01sammy_wmsremember someone say "tea" is good for health ... suggest everyone can try it :)
15:56:22prpplagueerikm: hey, i'm still a little new to cvs, i'm using the anonymous cvs access to get blob, there isn't anything i could do to screw up the cvs server is there?
15:57:34prpplaguegets his cup and heads to the kitchen for a cup of china green tea....
15:57:41seletzprpplague: no, as long as you not doing wild commits of hacked code nothing bat can happen ...
15:58:51prpplagueseletz: no commits on my part
15:59:22prpplagueseletz: except to a crazy house maybe
15:59:22seletzso nothing bad can happen. ;)
16:00:37prpplagueseletz: it took 3 weeks to convince our winblows sysadm to open up ssh and cvs ports on the firewall
16:02:45seletzoh my. same over here.
16:14:40prpplagueseletz: winblows sysadm - "i've never heard of this cvs, i'll have to research it make sure its not a security risk" in the mean time the entre sales department is down due to gone.scr
16:16:16seletzthey never hear about anything other than M$ news. There, of course, do no secutity risks in M$ software exist. Only secutity issues in "non-compatible" software.
16:16:57seletzso, it ends up in everything non-M$-compilant is a secutity hazard per se.
16:19:03prpplagueseletz: i've been pounding our local republican hq about bush's choice of an M$ guy for computer security post
16:20:11seletzprpplague: what do you expect of someone who making a man as ashcroft responsible for law&order ...
16:20:55prpplagueseletz: well, i'm normally a big bush supporter however, his choices regarding technology have been terrible
16:22:13seletzprpplague: My guess is that this "unamerican" open-source software movement will be banned sooner or later.
16:22:55prpplagueseletz: ya, i see a battle coming soon over open-source
16:23:27prpplagueseletz: i'm a law abiding citizen, but i'll never use m$ products ever again
16:23:53prpplagueseletz: even if its against the law to use open-source
16:26:42seletzprpplague: No offense meant, but i always find it funny how an country which has "freedom" on its banner such restrictive laws. I mean freedom of talk etc.
16:28:11seletzprpplague: _and_ in such a erm, quite funny way, too. One is _not_ allowed to say the f-word, but one can legally deny the holocaust. phew, thats's what i dont understand.
16:28:13prpplagueseletz: i agree, but you have to understand that vast majority of ppl in the U.S. are uneducated, which means complicated issues like open-source can really be mis-understood a large group of ppl
16:28:35seletzprpplague: agreed, big point.
16:29:15prpplagueseletz: and most people believe that the U.S. is a democracy, its not, its a representative democracy
16:29:44prpplagueseletz: the founders never believed that the average citizen would be educated enough to properly debated an important topic
16:30:27prpplagueseletz: thats why elect an official who sound be trustworthy enough to educate him/herself on the topic and vote his mind
16:31:08seletzprpplague: Ah, so we don't have a democracy, but a olegarchy (dont know the word in english): There rule few who are "educated" enough to rule.
16:31:20prpplagueseletz: unfortunately, our representatives have become all to concerned with popular vote rather than educated vote
16:31:48seletzprpplague: btw, much the same is getting on more and more over here.
16:32:18prpplagueseletz: well, its a little different that an olegarchy, because a representative democracy is accountable to the people, olegarchy is not
16:33:01seletzprpplague: ok, your point.
16:34:16prpplaguesorry for the off topic conversation folks
16:34:27erikmreturns
16:35:09seletz(hmm, the topic is still 'bout russ needing coffee :D )
16:37:22erikmwhat still wonders me the most about the US is the complete lack of interest about what's going on outside of it
16:37:45seletz*nod*
16:38:26seletz_and_ the wide spread believe that the US is always best (no offense meant, tough)
16:39:12erikmyou only see something about a foreign country on TV when: a) the president visits it; b) there is a war going on; c) it's struck by a major disaster
16:40:13prpplagueyes, most ppl in the U.S. barely can name their own state capital much less understand where another country is
16:41:01prpplaguei had to make a trip to canada a few months ago, one of our secretaries asked "isn't that where all those religious wacko's live?"
16:41:08seletzsorry folks, i lost my time, so im off now .... :(
16:41:14seletzSIGGIRLFRIEND pending ....
16:41:18prpplaguelol
16:41:20erikmseletz: heh :)
16:41:48prpplagueseletz: you better hurry before it turns into a kernel OOPS
16:42:08seletz_oopsing:D
16:42:30erikmhands seletz kgirlfriendoops
16:43:59prpplagueerikm: ya right, like thats going help debug a girlfriend oops
16:44:11prpplagueerikm: might as be using an m$ product
16:46:21erikmheh :)
16:49:24prpplaguewahoo, got cvs working behind the firewall
16:58:08prpplagueerikm: hmm, so is the xmodem support working in blob?
17:04:34erikmprpplague: not yet
17:08:44prpplagueerikm: ok
17:08:53prpplagueerikm: i'm just happy about the cvs
17:11:15erikmprpplague: heh, me too :)
17:23:47sammy_wmsAmm....here is some problem , with kernel panic : <- he can't find my init ? why ?
17:24:18erikmsammy_wms: there are usually clues before the kernel panic
17:24:44sammy_wmsbut it's the last massage...
17:26:03erikmsammy_wms: that's what I said
17:27:40sammy_wmsmy console  root=/dev/mtdblock3 init=/linuxrc
17:27:56sammy_wmsand I am sure have linuxrc in /
17:29:23sammy_wmsbut the last massage still show me kernel panic : can 't find init in root , tell me to add command , why ?
17:31:30sammy_wmseven I change init to /sbin/init , it's still can't find , but the file I sure must be there ....
17:31:32erikmsammy_wms: there are usually clues *before* the last line. something like "can't root fs" or "ramdisk too large"
17:32:42sammy_wmsoh, it's can't root fs ...
17:38:48sammy_wmsmy blob size = 0x40000 ,blob param = 0x40000 ,kernel = 0x100000 and all the other's give it to ramfs
17:39:21sammy_wmsand flash is 28F128J3A 2 chips ,
17:40:00sammy_wmsand kernel partation size is all the same with this ...
17:41:07erikmare you using jffs2 or ramdisk?
17:41:21sammy_wmsand flash map = ox40000 from 0x00000000 ~ 0x02000000
17:41:27sammy_wmsI use jffs2
17:41:34erikmon what platform?
17:41:41sammy_wmspuppy
17:42:08erikmdo the kernel and blob agree about the flash partitioning?
17:43:29sammy_wmswhen kernel boot , the MTD partition show 0,1,2,3, like I define , is this ?
17:44:24sammy_wmsI also change the sa1100-flash.c to match the blob partition ....
17:44:56sammy_wmsI mean the same size ...
17:45:57sammy_wmsand the jffs2 file image also use " -e 0x40000 -p " option ...
17:48:04sammy_wmsand also use (1)try modify from your lart-jffs2-1.0 file use busybox (2) made by my self without busybox ...
17:49:06sammy_wmsbut the kernel still hang on that panic massage ......why ?
17:51:38sammy_wmseven Russ had tell me to change the eraseblock size to build the jffs2 file , and I did it to , but I can't use 0x20000 ,right ?
17:52:39sammy_wmsand even 0x10000.... right ?
17:54:18sammy_wmsso choose 0x40000 to be the eraseblock maybe right,...
17:54:46sammy_wmserikm: have any idea ?
18:00:54erikmfirst try to do without a /linuxrc
18:01:31erikmand it also won't work if you still have support for ramdisk in your kernel
18:03:28sammy_wmswhat's rootfs should I load , ramdisk or jffs2 ?
18:04:50sammy_wmsif I load jffs2 , it's will show me can't find ramdisk image...(but the kernel I always have choose to support)
18:05:33erikmJFFS2 *only* works if you *disable* support for ramdisk
18:06:29sammy_wmsOh, my god , so I must into menuconfig to disable the ramdisk support ?
18:06:50sammy_wmsto let the jffs2 work ?
18:08:03erikmyes
18:09:49sammy_wmsAmm? but in lart , I remember it's sounds like both of three all have choose ....<support ramdisk><MTD support><jffs2 filesystem support>
18:10:47sammy_wmswhy is lart can work ?
18:11:50sammy_wmsor because in configure for lart that mtd have depand choose for 28F160 flash chips , so it's can work ?
18:13:17sammy_wmsor I got wrong remember ?
18:14:38sammy_wmserikm: can you check all the massage I type on here , am I do something wrong or lost anything on it ?
18:15:16erikmis /linuxrc executable?
18:15:33erikmdoes init=/bin/sh give you a shell prompt?
18:16:30sammy_wmsthe kernel always scream to want the init ....
18:17:19sammy_wmsI had try init=/bin/sh init=/sbin/init init=/linuxrc ....but no one can pass into _login_
18:19:21sammy_wmsafter your check , if you say no problem , then this maybe the double check , because Russ had help me check once before (Thanx) and help me fix some erase block problem...
18:19:41erikmdid you build this ramdisk yourself?
18:19:57sammy_wmsbut , the _HANG_ problem ,still not gone...
18:20:30erikmdid you build this ramdisk yourself?
18:20:45sammy_wmsyeap,
18:20:59erikmsammy_wms: so did you generate /etc/ld.so.cache?
18:21:37sammy_wmsoh oh , what's that ?
18:21:53erikmthe cache file for the run-time linker
18:22:07erikmI found out that it doesn't work *without* that file
18:23:13sammy_wmsis that file can only copy it into the director or must build by some step ?
18:23:33sammy_wmsif only use copy , that's file must be there ...
18:23:48Russsammy_wms: remember, a jffs2 fs is not a ramdisk
18:23:48erikmyou have to generate it.
18:24:14sammy_wmshow to do it ?
18:24:39sammy_wmsso use copy , won't work right ?
18:26:54erikmsammy_wms: ldconfig
18:31:02sammy_wmsis now reading some page to get this "lost space"...
18:44:14sammy_wmserikm: if my ld.so.cache is the problem , that libc.so.5 also my problem too right ?
18:44:56erikmsammy_wms: libc.so.5 is considered obsolete
18:47:52sammy_wmsoh, so after I ldconfig that dir , also rebuild jffs2 again to load , right ?
18:48:49erikmyes
18:50:02sammy_wmsok, this maybe the last hope...hope kernel not barking again...
18:56:03sammy_wmssun will get up few hours later, must catce some time to sleep...
18:56:16sammy_wmsthanx erikm :)
18:56:43erikmit works?
18:56:44sammy_wmsnight all my friends ...
18:57:15sammy_wmsoh. the board is in my office , so I must try later...
18:57:43sammy_wmsso Sammy will try that ...
18:58:55sammy_wmsnight :)
18:59:34sammy_wmsz.z.z
18:59:39erikmnicght
18:59:44erikmnight, I mean
19:16:53erikmprpplague: you were looking for gregkh? he's right now on #kernelnewbies and prolly also in #usb
19:20:54prpplagueerikm: whats that?
19:21:02prpplagueerikm: about the xmodem stuff?
19:21:33erikmprpplague: about the USB stuff
19:22:11erikmprpplague: or was it seletz who had questions about usb?
19:24:07prpplagueerikm: thats was seletz
19:25:03erikmprpplague: ok. (/me was a bit jet lagged yesterday %)
19:25:40prpplaguenp

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.