irclog2html for blob on 2002.02.18

14:23:43Sammygoes home...
15:19:30prpplaguehowdy all
15:19:36prpplagueanything fun happening today
15:19:58erikmhello prpplague
15:34:37prpplagueerikm: hows life on campus?
17:43:54seletzhi folks
17:44:58seletzmoved filally to his new office
17:47:16prpplaguehowdy seletz
17:47:24seletzprpplague: hi
17:47:36prpplagueseletz: ever get your irq hdw problems straight?
17:47:41prpplaguewb erikm
17:49:04seletzprpplague: well, i moved to 2.5.2-rmk6, there the irq subsystem was rewritten. Now i get other errors, (read: random lockups). I now try to get a neponset board running with 2.5.2-rmk6. I definitely mess something up. sigh.
17:49:50seletzprpplague: right now i try to get used with mutt (hi erik...) and sendmail/procmail conf. grrr.
18:43:25erikmhmm, more people over here, just like I expected
22:14:19BZFlagerikm: hmm... If I have the memory I might very well want chkmem in blob.
22:14:31erikmBZFlag: no
22:14:46BZFlagno?
22:14:59erikmBZFlag: blob is a bootloader, diag is for diagnostics
22:15:25erikmBZFlag: testing memory is clearly diagnostics
22:15:39BZFlagfor most installs I agree. But say I have a system with 256k flash blocks. why not fill up the 256k with usefull stuff?
22:15:40seletzBZFlag: erik will probably point you to the motto.txt file now ;^)
22:16:26erikmBZFlag: chris hoover pointed it out correctly: it's a matter of policy
22:16:28BZFlagdump is clearly a diagnostics tool. but it has been _very_ useful to have in blob.
22:16:49erikmBZFlag: I left that in because its a good tool to debug blob
22:16:53BZFlagditto for peek and poke.
22:16:58erikmditto
22:17:57seletzerikm: I don't want to argue now, but consider someone wants a POSt routine _every_ startup, and that POST should do a memory test?
22:17:59BZFlaghmm. I can see that chkmem etc should be in diag by default, but I do not agree that removing the possibility of having them in blob is a good thing.
22:18:17erikmI already had reports from people implementing their own wicked dump implementation, so I'd rather have something that works well
22:18:51erikmseletz: we're already doing a memory test that determines the size of the memory
22:19:00erikmseletz: that's also the same as a PC BIOS does
22:19:32seletzerikm: yup, agreed. I thought of the general concept: POST is diagnostics. Highly boadr dependant, too.
22:19:36erikmBZFlag: ok, today it's chkmem in blob. tomorrow it is chk-ethernet-card. where do we draw the line?
22:20:01BZFlagwith --enable-stilly-misfeatures ;-)
22:20:08BZFlager silly.
22:20:56BZFlagimho the line is drawn based on your --enable-foo options at compile time.
22:21:21seletzerikm: How would you implement a POSt which should, say, check ETH, check some weird on-board HW, and then startups linux?
22:21:29BZFlaga system with small flash blocks will draw a much different line than a system with large blocks. that's the real line.
22:22:13erikmseletz: just assume the ethernet card works. if it doesn't it's time to load diag and figure out *why* it doesn't work
22:23:00BZFlagunless of course the only load method you compiled in was tftp. ;-)
22:23:02erikmseletz: no need to put diagnostics code into blob for the 0.001% of the cases it doesn't work
22:24:02erikmBZFlag: I think it's a good idea to have at least one kind of serial download method in blob
22:24:08BZFlagwell, I'd say "there is no sense in wasting the rest of the initial flash block by filling it with 0xff's"
22:24:51BZFlagerikm: serial. agreed when there is a serial port available. the vercel will not have one.
22:24:54erikmBZFlag: so cat diag after blob and put it in the same flash block
22:25:14seletz_yuck_
22:25:17erikmBZFlag: in that case there is no interaction possible with the booat loader at all
22:25:34BZFlaghaving all the commands in one binary would clearly allow more commands to fit.
22:25:56erikmBZFlag: but how would you like to give those commands if there is no serial port?
22:26:35BZFlagerikm: I'm looking at a tftp server mode to blob and considering udp console access. thoughts?
22:27:17erikmdon't you think it's easier to spend $2 on a RS232 tranceiver?
22:27:24seletzBZFlag: like in Redboot? (yuck-bleah)
22:27:29BZFlagThat's just an option. The one I do plan to do is button activated ide flashing.
22:27:43BZFlagerikm: I do yes. vercel does not.
22:27:56BZFlag. o O ( ugh. )
22:28:17erikmBZFlag: it will make the product slightly more expensive, but the support costs will go down a *lot*
22:28:37erikmBZFlag: after a LED, a serial port is the printf() of debugging
22:28:48BZFlagerikm: I agree completely. unfortunately I have been unable to convince the customer so far.
22:29:20erikmBZFlag: how many products do they want to make?
22:29:24BZFlagto the extent that they are pondering paying us 70k+ to add usb console access to blob.
22:29:48seletzerikm: man, i even had to solder LEDs on my board because the original hw design had no led ... :(
22:29:55BZFlagabout one new platform every 3 months.
22:30:28BZFlagheh, vercel has no leds either. I use the backlight as my indicator.
22:30:35seletz:)
22:30:37erikmBZFlag: sorry, I mean how many vercels do they want to sell?
22:30:38BZFlagwell, so software accessible leds
22:31:00BZFlagthey are looking at 100k a month or so.
22:31:11erikmBZFlag: 100k? what's it supposed to do?
22:31:22seletz100k a MONTH ?
22:31:28BZFlagit's the teen pda/game box etc.
22:31:49BZFlagyes. perhaps 200k units/month if things go well.
22:31:57seletzphew
22:32:17seletzBZFlag: i hope you get licensing $$ .... :)
22:32:27BZFlagyep. that's in the plan.
22:33:10BZFlagceo of vercel is co-inventor of the furby. This will see similar distribution channels.
22:33:44erikmBZFlag: but ok. 100k a month. lets assume 1% returns to facture with a problem. let's assume 1% has a problem with the low level stuff. that's 10 per month.
22:33:53seletzhas gone quiet in envy ... :)
22:34:20BZFlagok.
22:34:59BZFlagthey would likely try reflashing from ide, if that fails, send a replacement.
22:35:04seletzhmmm. 1% of 1% is a bit low IMHO
22:35:09erikmBZFlag: let's assume that it takes 1.5 hours to figure out what went wrong and repair it. if the thing has a serial port it gets done in 30 minutes
22:35:22BZFlagor usb console access.
22:35:48erikmBZFlag: so that's 10 hours times whatever a service technician costs
22:36:33BZFlagagreed. cept that usb console get's them the same thing provided blob/usb hardware is ok.
22:36:51erikmBZFlag: so does it get an USB plug?
22:37:01BZFlag2. one host and one client.
22:37:07seletzBZFlag: how do you flash in case of a complete system crash?
22:37:09erikmBZFlag: USB client is broken
22:37:13erikmseletz: jtag
22:37:32seletzBZFlag: you have jtag cons but no serial?
22:37:39BZFlagjtag will be on the board but will require disassembly to get to it.
22:37:52BZFlagI'm trying to get that fixed, and likely will.
22:38:12seletzok, so the'll probably not flash at all, because of labor costs, no?
22:38:19seletzs/labor/labour/
22:38:34BZFlagjtag pins will be available on the board edge near the cf slot. cut a small hole and you are in. dev model may have that opening already.
22:38:42erikmBZFlag: you can do the same for serial. just have four connections (gnd, rxd, txd, +3.3V), add an external board with MAX232 and you're set
22:38:57BZFlagthat's exactly what I'm pusing for,
22:38:58erikmBZFlag: doesn't cost them anything extra, makes life a lot simpler
22:39:09BZFlagI just don't think it likely to happen.
22:39:21erikmBZFlag: and you can drop the b0rken USB client in that case
22:39:24BZFlagI'll know for sure around the end of march.
22:39:42seletzerikm: why is USB client broken?
22:39:54erikmseletz: because the SA-1110 is broken
22:40:12BZFlagseletz: there are just a ton of ugly work arounds required.
22:40:15erikmseletz: it's in the intel errata
22:40:25seletzerikm: the packet size issue?
22:40:34erikmBZFlag: the USB state machine can completely lock up
22:40:44seletz_ugh_
22:40:47BZFlagyep.
22:41:03erikmBZFlag: the sa1110 is slightly better than the sa1100, but still considered too broken to be really useful
22:41:06seletzis supposed to get USB client working somhow lather this year
22:41:23erikmseletz: it works as long as it's the only device connected to the host
22:41:29BZFlagzaurus and ipaq both have it working fairly well.
22:41:49BZFlagyes, as the only client.
22:41:57erikmas long as you keep within the constraints of the intel errata
22:42:15erikms/keep/stay/
22:43:01erikmseletz: easiest way to work around it is to put an extra USB host controller in your PC and use it exclusively for sa-11x0 devices
22:44:01seletzerikm: ok, I'll keep that in mind.
22:44:24seletzmakes a note to read the erratas when time comes :(
22:47:51erikmoops, people wanting to mount BGAs themself on the LART list :(
22:48:36seletz:)
22:48:51seletzwas on arm-kernel too
22:48:55erikmand we already discouraged them mounting QFP themself :)
22:49:18erikm(actually, I hand soldered a couple of LARTs :)
22:49:27seletzeeeek!
22:49:37erikmbut that's not something we want others to know
22:49:39seletzruns away in horror
22:49:57erikmseletz: a good soldering station helps a lot
22:50:08erikmseletz: and a micro reflow soldering tip
22:50:14erikmseletz: and *lots* of flux
22:50:46seletzerikm: man, i sure am not too bad in soldering, but that ... ~:^/
22:51:13RussBZFlag: don't put a RS232 transceiver onboard
22:51:31seletzugh
22:51:32RussBZFlag: make the serial signals available, as well as 3.3V and GND
22:51:41erikmRuss: *nod*
22:51:47RussBZFlag: then put a RS232 transceiver on your board connector
22:52:05seletzSIGGIRLFRIEND
22:52:12seletz:^/
22:52:13RussBZFlag: can even put diagnostic LEDs on there, if you are looking to save an extra penny
22:52:13erikmRuss: that's actually how I added a third RS232 receiver to a couple of LARTs
22:52:18seletzbye folks
22:52:23erikmbye seletz
22:56:00erikmgoes zzz
22:56:01erikmbye
23:38:39BZFlaglags...
23:40:22BZFlagRuss: agreed. I'm just fighting for getting at the connection at all. had not planned to ask for rs232 levels.

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.