**** BEGIN LOGGING AT Thu Feb 10 23:59:57 2005 **** BEGIN LOGGING AT Fri Feb 11 01:04:24 2005 Feb 11 07:41:39 [g2], pong Feb 11 07:42:57 piiiiiing Feb 11 07:43:08 or should that be wrooooong since i'm not g2? Feb 11 07:43:41 that is the sound every submariner doesn't want to hear. Feb 11 07:43:49 Tiersten: or you could use "ding" Feb 11 07:43:59 ka6sox: hehe Feb 11 07:44:04 he removed that "feature" Feb 11 07:44:15 morning Feb 11 07:44:24 nah. it'd be more of a SCRAAAAAAAAAAAAAAAAAAAAAAPE CLUNK CLUNK WUUUUUSHHHHH.... :) Feb 11 07:44:33 ka6sox: hehe, yea, knowning that someone has a firing solution on you isn't too comforting Feb 11 07:44:46 uh huh Feb 11 07:45:06 I wouldn't know never having been in a submarine under fire ;) Feb 11 07:45:15 Tiersten, yes that would be a disconcerting sound. Feb 11 07:45:55 off to an early breakfast meeting....catch up soon. Feb 11 07:46:00 Bye Feb 11 07:50:19 <[g2]> ka6sox-away, ping Feb 11 07:50:31 You're 4 minutes too late! Feb 11 07:50:53 <[g2]> that happens :) Feb 11 07:51:04 <[g2]> So you're a submariner eh ? Feb 11 07:51:06 [g2]: he's done left for steak and eggs Feb 11 07:51:12 Not me... Feb 11 07:51:19 Although I could do with steak & egg... Feb 11 07:51:31 [g2]: i did some cross training when i was in the army Feb 11 07:52:45 A ground pounder in a sardine tin? :) Feb 11 08:02:59 Tiersten: hehe, i've been in all kinds of military units Feb 11 08:09:48 Cool Feb 11 11:18:17 [g2]: howdy. Feb 11 11:34:34 <[g2]> beewoolie, hey! Feb 11 11:34:56 <[g2]> how is the woolie master today ? Feb 11 11:34:57 Did you get your JTAGd slug? Feb 11 11:35:06 chilly Feb 11 11:35:10 <[g2]> dunno Feb 11 11:35:35 I coerced APEX into building with 2.95.3 and I've added a compact-flash/FAT driver pair. Feb 11 11:35:53 <[g2]> awesome Feb 11 11:36:00 I'm still looking for someone to belp me get APEX up on the slug. Feb 11 11:36:12 <[g2]> not that I'm interested in the 2.95.3 part, but FAT is handy Feb 11 11:36:26 <[g2]> I'M ALL FOR THAT Feb 11 11:36:44 <[g2]> helping get APEX running on the slug! Feb 11 11:36:54 :-) The 2.95.3 is for those whingers who don't know how to install more than one toolchain. Feb 11 11:37:10 <[g2]> doesn't make ANY sense to me Feb 11 11:37:34 <[g2]> my little mind remembers there were lots of ARM fixes in 3.3+ Feb 11 11:37:39 The FAT stuff is really handy. It can load the kernel and whatever else you need from a CF card. Getting USB working shouldn't be much more work. Feb 11 11:37:54 <[g2]> excellent Feb 11 11:37:56 Oh, yeah. the 2.95.3 version is somewhat bigger. Feb 11 11:38:42 <[g2]> lennert just identified a critical issue we were hunting and was causing the usbnet driver not to receive IP packets properly Feb 11 11:39:08 <[g2]> the big endain alignment trap is not working in the 2.6 kernel properly Feb 11 11:39:36 <[g2]> I'd guess there may be an outside chance that could effect APEX Feb 11 11:39:43 Interesting. Feb 11 11:39:46 Unlikely. Feb 11 11:39:58 APEX doesn't depend on the alignment trap. I have to code correctly. Feb 11 11:40:25 <[g2]> I was thinking more hw access alignmnet Feb 11 11:40:35 <[g2]> spelled properly of course :) Feb 11 11:41:03 It is possible. However, the code does run on the slug. Feb 11 11:41:21 I can tell that it all works because I've written it to flash and run it from there. Feb 11 11:41:35 What I don't know is what inits is RedBoot doing that I don't. Feb 11 11:42:13 It would be most helpful to get some memory dumps from a slug that is in the reset state. Feb 11 11:46:39 <[g2]> beewoolie, I'd think we could go through the initial Redboot setup and see what gets set Feb 11 11:46:57 <[g2]> or are you looking to handle resets better Feb 11 11:47:10 I've done that. I suspect that there is something buried elsewhere in the code. Feb 11 11:47:47 As the CPU isn't too complex, it should be easy to dump a couple of register files to see what's being done. I suspect that there is someting in the GPIO section. Feb 11 11:49:48 <[g2]> beewoolie, I don't have a BDI, I'm guessing you want to read out the default register values at power-on reset from the ixp420, correct ? Feb 11 11:50:07 The jtag tools can do it too. Feb 11 11:50:17 There is a readmem command or something like it. Feb 11 11:50:20 <[g2]> they should work then Feb 11 11:50:49 <[g2]> I'll find out today where my fatslug is Feb 11 11:51:17 <[g2]> dyoung and jacques are also jtag enabled with the jtag tools Feb 11 11:51:24 Actually, this is challenged by the jtag tools because I know that they change the system state. Feb 11 11:51:37 dyoung was helping, but he lost interest. Feb 11 11:53:30 <[g2]> well I'm very excited about APEX and JTAG/OpenJTAG it's the last stop to full control of the device Feb 11 11:53:48 FULL CONTROL Feb 11 11:53:56 <[g2]> :) Feb 11 11:53:57 i like that Feb 11 11:54:34 Honestly, I've been distracted with a milestone. I've been hacking at OPIE for a few days now. If I can get it working, I can take a short breather and look at the SLUG again. Feb 11 11:55:31 <[g2]> with the usbnet issues identified and a work-around available (and other fixes available too) the bootload is the last of the non open-source Feb 11 11:55:35 <[g2]> on OpenSlug Feb 11 11:55:46 <[g2]> and Redboot is open source Feb 11 11:55:53 ...sort of. Feb 11 11:56:05 <[g2]> just not very user friendly Feb 11 11:56:11 For it to be a useful open source component, there has to be someone willing to hack it. Feb 11 13:23:12 you might want to see if Connectiva's changes to skbuff are having any affect on alignment too Feb 11 13:23:56 kas11: is this about the USB driver? Feb 11 13:24:04 uh huh Feb 11 13:24:46 What I've read and understood is that the issue is simple. The USB driver isn't aligning it's skbufs which ethernet drivers (all/most) appear to do. Feb 11 13:24:51 The fix is a simple one. Feb 11 13:28:07 <[g2]> kas11, the alignment stuff is tested. lennert submitted two patches adapted from the 2.4 kernel for the 2.6 tree that resolves the issue Feb 11 13:28:34 <[g2]> also, I think there will be a usbnet patch that aligns the IP addr and some other stuff better Feb 11 13:29:07 <[g2]> I'll probably pull both patches into the 2.6.9 OpenSlug tree Feb 11 13:29:51 <[g2]> I'm not going to enable the alignment traps by default, but the code will be there to enable it at runtime and there are already counters one can check Feb 11 13:35:54 bewoolie, its not a matter if interest. Feb 11 13:35:57 Its a matter of time. Feb 11 13:36:05 And my complete lack thereof atm. Feb 11 13:36:45 dyoung: you're a lazy bastard...and so am i :-) Feb 11 13:37:08 dyoung: np. I'm distracted too. The good thing about this project is that it will get done when it gets done. Feb 11 13:39:33 [g2]: I was under the impression that the alignment trap is completely wrong on BE. Feb 11 13:40:05 [g2]: Perhaps that needs fixing too. Feb 11 13:41:37 <[g2]> beewoolie, it was/is until the latest patches are applied Feb 11 13:41:54 Ah, I didn't see the full scope of the patches. Feb 11 13:42:09 <[g2]> however, lennert has a set and the tested out on the OpenSlug Feb 11 13:42:25 <[g2]> s/and the/and they/ Feb 11 14:04:19 ~hug [g2] Feb 11 14:04:19 dyoung: I don't know, could you explain it? Feb 11 14:23:06 ~seen ka6sox Feb 11 14:23:06 ka6sox <~tking@adsl-66-159-198-116.dslextreme.com> was last seen on IRC in channel #openjtag, 6h 37m 11s ago, saying: 'off to an early breakfast meeting....catch up soon.'. Feb 11 14:23:22 ~seen ka6sox-away Feb 11 14:23:23 ka6sox-away <~tking@adsl-66-159-198-116.dslextreme.com> was last seen on IRC in channel #openjtag, 1d 18h 58m 20s ago, saying: 'I made it home....drove around the slide.'. Feb 11 14:24:34 <[g2]> dyoung what's the hug for ? Feb 11 14:24:57 Getting the ball rolling on the usbnet issue. Feb 11 14:25:01 [g2]: do hugs have to be *for* something? Feb 11 14:25:15 I'm still a little confused why it worked for me. Feb 11 14:25:30 My pegasus device worked ok. Feb 11 14:25:46 <[g2]> pegasus may be a different driver Feb 11 14:25:57 Maybe I licked too many things and was hallucinating. Feb 11 14:26:08 <[g2]> beewoolie, hugs don't have to be *for* something, but it is nice to know why :) Feb 11 14:26:22 <[g2]> I've done that lately :) Feb 11 14:26:59 bewoolie, Do you prefer I change back to factory/unslung firmware for apex testing? Feb 11 14:27:23 Right now I have openslug with 2.6.9 in the JTAG enabled box. Feb 11 14:27:37 If you do, we can start by finding out if you can boot the firmware with APEX running from SDRAM. Feb 11 14:27:58 Seems to me that it's wise to reduce the number of variables. Feb 11 14:33:29 ok I'll replace my kernel this weekend. Feb 11 14:33:54 back later Feb 11 14:36:10 dyoung: ta Feb 11 14:39:56 <[g2]> dyoung, pegasus is a different kernel module Feb 11 14:42:40 so basically, I just got lucky. Feb 11 14:43:03 <[g2]> you could say that Feb 11 15:02:36 * beewoolie hums to a tune Feb 11 15:08:16 ~seen beewoolie Feb 11 15:08:16 beewoolie is currently on #openjtag (4h 46m 24s). Has said a total of 39 messages. Is idling for 0s Feb 11 16:32:37 prpplague: hey there. Feb 11 16:32:58 prpplague: I've put up a version of APEX that builds with 2.95.3 Feb 11 16:33:13 prpplague: It's also got CF and FAT drivers. Feb 11 16:50:46 beewoolie: lovely Feb 11 16:50:54 beewoolie: what about 2.4 kernel support? Feb 11 16:51:22 That one's tough since I don't have a setup to test it. Feb 11 16:51:25 beewoolie: i'll need to add mmc support and 8bit interface for the cs8900 Feb 11 16:51:44 beewoolie: oh ok Feb 11 16:51:48 Thew cs8900 is tough, too. I do have a board with it. Feb 11 16:51:54 It's one of the old kev's. Feb 11 16:52:20 Actually, 2.4 should really be a problem. Feb 11 16:52:34 You can just hard-code the command line and memory size in the startup. Feb 11 16:52:45 s/should/shouldn't/ Feb 11 16:52:59 That's what most of the vendors have done. Feb 11 16:53:14 Do you know anything about qt-embedded? Feb 11 17:00:47 beewoolie: yea doing the hardcoded command line is what i had planned to do Feb 11 17:01:05 beewoolie: no sorry, what little experience i've had with qt was no favorable Feb 11 17:01:14 beewoolie: i'm more of either gtk or fltk Feb 11 17:01:40 I need to get a demo running on one of these boards. OPIE is what I've been trying. Is there something else? Feb 11 17:02:57 beewoolie: i don't recall, however most of the zaurus ppl are qt ppl Feb 11 17:03:10 beewoolie: so i'd suggest asking in #oe Feb 11 17:03:17 beewoolie: memtest for apex? Feb 11 17:03:37 Haven't looked at the memtest. Feb 11 17:03:54 beewoolie: was planning to do some simple test program Feb 11 17:04:07 BTW, I added a README_overview file that explains some of how the program is put together. Feb 11 17:04:18 beewoolie: cool Feb 11 17:04:26 beewoolie: i've been out all day with the kids sick Feb 11 17:04:38 Priorities Feb 11 17:04:38 beewoolie: they have the double whammy, strep throat and the flu Feb 11 17:04:46 beewoolie: yea Feb 11 17:05:27 beewoolie: and although i don't feel bad, i tested positive for both strep and flu Feb 11 17:05:29 beewoolie: :( Feb 11 17:05:54 bad scene Feb 11 17:06:01 time for a sauna Feb 11 17:08:54 yea Feb 11 17:41:59 [g2]: Did you see this? http://trustedreviews.com/article.aspx?head=21&page=2163 Feb 11 17:46:51 <[g2]> beewoolie, what a wifi usb card ? Feb 11 17:47:01 <[g2]> or one that works at 20Mbs Feb 11 17:48:36 I haven't seen many of these. I thought that this one sounded pretty good. Relatively cheap. WPA. Quick. Feb 11 17:48:55 <[g2]> I've got one or two Feb 11 17:49:14 <[g2]> and 4 or 5 wrt's Feb 11 17:49:39 * nslu2-bricker mumble Feb 11 17:50:18 <[g2]> now a bricker in JTAG that's original :) Feb 11 17:51:01 [g2]: I missed part of it. It has a disk image of the drivers which means it may not be supportable on Linux. Feb 11 17:53:01 I keep making bricks. Feb 11 17:53:41 bewoolie, were there any salient changes between 1.0.19 and 1.0.21 ? Feb 11 17:54:58 (for the ixp42x target that is...) Feb 11 17:55:57 ~seen ka6sox Feb 11 17:55:57 ka6sox <~tking@adsl-66-159-198-116.dslextreme.com> was last seen on IRC in channel #openjtag, 10h 10m 3s ago, saying: 'off to an early breakfast meeting....catch up soon.'. Feb 11 18:02:34 <[g2]> nslu2-bricker == ka6sox Feb 11 18:07:57 <[g2]> dyoung, did you brick a slug ? Feb 11 18:08:26 It's his IP but it doesn't mean it's him Feb 11 18:09:04 <[g2]> Tiersten, I figured that out via the CGI Feb 11 18:09:05 dyoung-web: whoever you are. Feb 11 18:09:13 I intentionally bricked it. Feb 11 18:09:20 dyoung-web: I've been adding loads of features and documentation. Feb 11 18:09:35 You intentionally bricked it? Feb 11 18:09:37 There isn't anything different in the startup process, however. Feb 11 18:09:37 Okay, I looked at the change log. Feb 11 18:09:40 You got nothing better to do for 40 minutes? :) Feb 11 18:09:54 I have been lax about updating the changelog. I forgot I had one. Feb 11 18:09:59 Yeha, I felt the need to pay the Penalty Fee to play. Feb 11 18:10:44 I made 1.0.21 using 3.4.3 now, see if that has a better result. Feb 11 18:11:17 If I split the redboot image into 2, I should be able to just flash write the first block right? Feb 11 18:11:41 Apex is small enough to fit in one block, so I'm only really screwing around with that block. Feb 11 18:11:56 Yes Feb 11 18:12:00 That'll reduce my penalty fee my a bunch. Feb 11 18:12:12 Down to just over 20 minutes Feb 11 18:12:59 Bewoolie, what time zone are you in? Feb 11 18:13:00 dyoung-web: right. You only need to flash the first block.. However, you need to make sure it's the right first block. Feb 11 18:13:15 the right first block? huh? Feb 11 18:13:18 PDT. Seattle Feb 11 18:13:35 I have my standard redboot neatly stored away, so I know which "right" first block it will be. Feb 11 18:13:43 since I know what i'm splitting... Feb 11 18:14:15 If you have redboot version X, then there are two halves of the melon. But if redboot Y is a cantaloupe and redboot X is a honeydew, the halves won't fit if you mix them up Feb 11 18:14:38 well yeah but he split the image in the first place so hopefully he'd use the correct portion... Feb 11 18:15:08 I'm not that silly I dont think....? Feb 11 18:15:19 dunno... UBESSA? :) Feb 11 18:17:01 UMESSA? Feb 11 18:17:18 Whats the M? Feb 11 18:17:38 I thought it was an anagram of assume, some sort of byte-swapping joke. Feb 11 18:17:48 Nope Feb 11 18:18:10 an acronym? Don't blurt it out yet. Feb 11 18:18:19 You won't guess it Feb 11 18:18:37 I usually don't, but that doesn't stop me from trying. Feb 11 18:18:55 I'd be very surprised it you did without hint Feb 11 18:18:59 s Feb 11 18:19:14 Is it an acronym? Feb 11 18:19:24 yes Feb 11 18:20:06 * beewoolie realy likes LVM Feb 11 18:23:18 <[g2]> dyoung-web, Can't you do a block erase and just flash the first 16K ? Feb 11 18:23:38 <[g2]> then the penalty is 20 / 8 right ? Feb 11 18:24:05 An erase block in the Strataflash in the NSLU2 is 128K Feb 11 18:24:13 You can't erase anything smaller Feb 11 18:24:25 Which means around 20 minutes to reprogram fully Feb 11 18:24:50 <[g2]> Tiersten, I said do a block erase which is a fast command Feb 11 18:25:06 <[g2]> and only program the first 16K bytes of the 128K block Feb 11 18:25:11 Yeah but the penalty is for when he's reflashing redboot Feb 11 18:25:20 16K is peanuts if he's just testing APEX Feb 11 18:26:09 That's why he is splitting his redboot image in half so he doesn't need to reprogram the second block Feb 11 18:26:12 <[g2]> I was wondering if it only takes a couple minutes to reflash APEX with the wiggler device Feb 11 18:26:40 Prob about 3 minutes or so Feb 11 18:26:56 <[g2]> nod. dd=if=/dev/mdtblock0 of=file bs=128K count=1 Feb 11 18:27:50 * [g2] would dd out the second half of the melon and put at least an MD5sum on it Feb 11 18:28:25 It should be fairly safe Feb 11 18:29:08 The penalty is in rewriting redboot when things dont work. Writing apex is fast. Feb 11 18:29:19 <[g2]> just double checking Feb 11 18:29:26 dyoung-web: I don't recommend putting APEX in flash just yet. Feb 11 18:29:51 without some sort of debug support, you're not going to get much out of it. Unless it works the first time. Feb 11 18:29:52 <[g2]> in which case it's a tftp load penalty Feb 11 18:30:19 bewoolie, have you given up on UBESSA yet? Feb 11 18:30:22 The first test is to flash APEX into the config block and jump to it. Feb 11 18:30:34 dyoung-web: grr. No. I'm still thinking. Feb 11 18:30:48 <[g2]> or was dyoung bricking the first block to gather the reset info ? Feb 11 18:30:51 ooh, thats a good plan. Feb 11 18:31:00 dyoung-web: Then, we want to see if it can boot the kernel. Feb 11 18:31:18 Last time, we tried this, it didn't work. That was with the openslug firmware. Feb 11 18:31:39 Right, I was going to avoide breaking it all at the same time and do this incrementally. Feb 11 18:32:21 This latest brickage was kind of an accident. "Something Happened". Feb 11 18:32:41 trying to do stuff remotely. Feb 11 18:34:30 * beewoolie plugs in his slug Feb 11 18:35:29 * beewoolie is getting off his lazy little ass and is looking at the GPIO setup Feb 11 18:36:13 <[g2]> ~seen not-ka6sox Feb 11 18:36:14 not-ka6sox is currently on #openjtag (2d 5h 38m 31s) Feb 11 18:36:32 <[g2]> tking_, Feb 11 18:36:34 <[g2]> ping Feb 11 18:42:10 I think that this might be part of the problem. The GPIO settings are really different from the reset values. Feb 11 18:49:46 beewoolie, BTW, the guys down the hall do not have the xscale sw for the bdi. Feb 11 18:50:13 * beewoolie droops Feb 11 18:50:14 (and in fact their BDI is "lost") Feb 11 18:50:48 I have one more group to harass about that; but dont hold your breath. Feb 11 18:50:52 They just asked if I wanted today to pay mo money for continued SW updates. Feb 11 19:06:22 I think I figured out what broke my slug. I think I had set the GPIO lines such that it was forcing power-off immediately. Feb 11 19:14:04 * beewoolie finds another reason to resent OE Feb 11 19:16:33 dyoung-web: Are you going to try APEX again this evening? I've got a new version with some GPIO programming. Feb 11 19:28:28 ok, Yes I was planning to. Feb 11 19:28:45 Does this sound famliar to anyone? Feb 11 19:29:24 Boot a Linux install cd, (ie: it boots) then after the stage1 install process starts, it says "no cdrom" Feb 11 19:29:32 like some driver is getting confused. Feb 11 19:32:06 dyoung-web: Not seen that. Feb 11 19:44:09 The GPIO code changes don't appear to break anything. It runs from SDRAM and from flash. I'm still seeing a kernel dump at startup. Feb 11 19:53:18 <[g2]> dyoung-web, sounds familiar Feb 11 19:54:01 <[g2]> many install cd's uses stuff like syslinux which is really emulates a floppy Feb 11 19:54:05 <[g2]> on the cdrom Feb 11 19:54:43 <[g2]> so basically, the BIOS doesn't know about the CDROM and when the kernel boots the driver is missing for the CDROM also. Feb 11 19:55:02 <[g2]> knoppix is quite good with lots of older devices though Feb 11 19:57:52 Right, thats it. Feb 11 19:58:02 sysllinux knows about it but the kernel doesnt. Feb 11 19:58:09 No. Feb 11 19:58:20 syslinux is a DOS program. Feb 11 19:58:29 It uses the BIOS to get stuff done. Feb 11 19:58:43 As such, it's very reliable at the job it has to do. Feb 11 19:59:07 In other words, it is good about the CDROM if the BIOS is good about it. Feb 11 20:00:09 Okay so what I need is a kernel that knows about ich6 (the controller stuff on this board) Feb 11 20:00:17 yep. Feb 11 20:00:29 <[g2]> beewoolie, we are saying the same thing Feb 11 20:00:38 right. Feb 11 20:00:50 <[g2]> I'm running ich5 iirc Feb 11 20:01:17 it makes me wonder why I can boot/install redhat but I cannot boot/install mandrake. I cant imagine that rh has a patch that mdk doesnt./ Feb 11 20:01:22 ow my brain Feb 11 20:01:25 <[g2]> ICH5: chipset revision 2 Feb 11 20:01:52 <[g2]> heh Feb 11 20:02:13 <[g2]> actually I think that happens a bunch. That RH backports stuff Feb 11 20:03:47 <[g2]> 915 mobo ? Feb 11 20:06:57 I figured out what the dump is that I'm seeing when the kernel decompresses. It's a normal kernel debug feature. What isn't clear is why the kernel doesn't do it when redboot boots. Feb 11 20:07:17 [g2]: do you have a copy of the SLUG kernel source handy? Feb 11 20:08:03 <[g2]> beewoolie, for a minute Feb 11 20:10:16 If some distro isnt using syslinux/isolinux for their bootcd, what would their other option(s) be? Feb 11 20:11:37 <[g2]> well the kernel on the install cd is different from the boot kernel Feb 11 20:15:09 hmm Feb 11 20:15:20 I think it might be conspiring to piss me off. Feb 11 20:15:48 <[g2]> beewoolie, what's up ? I'm about to rm -rf tmp and rebuild openslug from scrath Feb 11 20:15:51 <[g2]> scratch Feb 11 20:16:16 I am curious about the compressed head.S can you look at it. I clobbered mine to make space. Feb 11 20:17:02 dyoung-web: That isn't so uncommon. Syslinux is the most reliable solution I've found. Feb 11 20:17:46 The kernel switch trick is, also, very annoying. IIRC, Debian and Knoppix are a couple of the only ones that are relatively good about using the same boot kernel as the runtime kernel. Feb 11 20:18:28 <[g2]> beewoolie, I can e-mail you a copy if you like Feb 11 20:18:39 That would be fine. Feb 11 20:19:26 <[g2]> Umm.. I'm assuming OpenSlug 2.6 kernel is fine Feb 11 20:20:16 No. I need the original 2.4 kernel. That's the issue. I have the up-to-date kernel, but I don't know what wierdness they implemented in the 2.4 stuff. Feb 11 20:20:33 If you don't have it , don't sweat it. I'll look for the patch. Feb 11 20:21:15 <[g2]> ah.. I don't have 2.4 kernel around Feb 11 20:25:48 dyoung-web: I put up a version 1.0.22 that includes the GPIO initialization changes as well as some extra LED twiddling so that we can see where it stops, if it stops... Feb 11 20:31:20 beewoolie: oh i didn't know you were in seattle Feb 11 20:31:51 Does that mean you're definitely not going to use APEX, ever? Feb 11 20:31:55 ;-) Feb 11 20:32:22 beewoolie: hehe, i'm just jealous that you have mac-n-jack there Feb 11 20:32:44 beewoolie: if it compiles with 2.95.3 Feb 11 20:32:58 beewoolie: and i can hardcode the 2.4kernel command line options Feb 11 20:33:06 beewoolie: then i don't see any reason i can't use it Feb 11 20:33:17 If nothing else, I'd like to see if we can work out the SDRAM init issue. Feb 11 20:33:32 beewoolie: what issue is that? Feb 11 20:33:54 I thought that APEX wasn't getting to the command prompt on your target. I've assumed it was an SDRAM init issue. Feb 11 20:34:26 beewoolie: oh, i didn't debug it very far Feb 11 20:34:38 beewoolie: so i have no idea whether it was a sdram issue or not Feb 11 20:35:36 <[g2]> nite all Feb 11 20:35:38 So far, most of the problems I'm seen have been in one of two camps. Either SDRAM isn't getting INIT'd, or the MMU is on and Feb 11 20:35:41 [g2]: night Feb 11 20:35:43 [g2]: nite Feb 11 20:36:18 beewoolie: i'e got working settings for the sdram init Feb 11 20:36:25 beewoolie: so it shouldn't be a problem Feb 11 20:36:39 beewoolie: just need to look at the code Feb 11 20:36:47 I had some wierd problems on the SLUG because I didn't realise the MMU was on. Feb 11 20:37:13 That, and I had a couple of typos in the EXT init. Feb 11 20:37:17 yea having the mmu on can be a pain Feb 11 20:37:29 I just didn't realize that redboot was leaving it on. Feb 11 20:38:58 I just force it off, for now, but I'd rather not even have to do that. Turns out that the redboot exec command, which does disable the MMU, does work. It just takes a whole mess of parameters. Feb 11 20:41:31 yea, i'll grab the lastest apex on monday, and see if i can get going with it Feb 11 20:46:28 http://www-106.ibm.com/developerworks/library/pa-migrate2/?ca=dgr-lnxw01BootProcess Feb 11 20:47:02 back later. Eat now. Feb 11 20:47:37 prpplague: I saw that. It doesn't seem to be that informative. Feb 11 20:48:59 beewoolie: i haven't actually read it yet Feb 11 23:03:18 bewoolie around Feb 11 23:04:43 I have a working 1.0.22 Feb 11 23:04:59 and it boots a factory class kernel without incident. Feb 11 23:05:46 I tried making it with the oe toolchain but it doesnt run. Feb 11 23:09:06 Rushin Draggin? Feb 11 23:09:13 yeah... Feb 11 23:09:16 thats right. Feb 11 23:09:33 dyoung: still here Feb 11 23:09:35 That was a LONG breakfast! Feb 11 23:09:43 The Russians are here! Feb 11 23:09:44 ~seen ka6sox Feb 11 23:09:45 ka6sox <~tking@adsl-66-159-198-116.dslextreme.com> was last seen on IRC in channel #openjtag, 15h 23m 50s ago, saying: 'off to an early breakfast meeting....catch up soon.'. Feb 11 23:10:08 What compiler version is the OE toolchain? Feb 11 23:10:33 BTW, I've put up another rev. It adds a light show when the boot loader is busy doing something. Feb 11 23:11:22 dyoung: So, I've changed the boot code. It is worth another shot at writing APEX to the first block to see if it boots. Feb 11 23:11:36 3.4.3/2.3.2 I believe. Feb 11 23:11:38 dyoung: it will step through the leds as it performs the startup. Feb 11 23:11:52 Thats in .22 ? Feb 11 23:11:55 hmm. And you're sure it is a BE chain? Feb 11 23:12:03 Its supposed to be. Feb 11 23:12:08 1.0.23 Feb 11 23:12:23 But I havnt diddled it very much. I knew the other one would work, so I'm using it Feb 11 23:12:30 Does the OE built one set any of the IDs? Feb 11 23:12:30 3.3.5/2.2.5 Feb 11 23:12:40 It doesnt boot. Feb 11 23:12:45 It makes gook. Feb 11 23:12:54 What about the LEDs? Feb 11 23:13:05 g 0x1000000 Feb 11 23:13:05 $T0a0f:00000010;0d:0003bdf8;#a1$T0a0f:00000010;0d:0003bdf8;#a1$T0a0f:00000010;09 Feb 11 23:13:09 that kinda gook. Feb 11 23:13:16 Which usually indicates it didnt work. Feb 11 23:13:20 Oh. That's an alignment error Feb 11 23:13:23 if I hit return, it makes more gook. Feb 11 23:13:27 Or an invalid memory access. Feb 11 23:13:33 it's the GDB stub. Feb 11 23:13:51 Yeah, I'll work that out later. I know the other one works, so I'll stay with that for now. Feb 11 23:13:53 RedBoot installs exception vectors. any exception goes to that. Feb 11 23:13:57 k. Feb 11 23:14:15 Lemme go grab/make 23. brb Feb 11 23:15:42 dyoung: pong Feb 11 23:16:06 ka6sox-draggin: eve' Feb 11 23:16:17 eve` Feb 11 23:16:29 Do we have a schematic? Feb 11 23:19:18 Okay done. Feb 11 23:19:27 Built and running in ram. Feb 11 23:19:35 Excellent. Feb 11 23:19:36 2 green leds Feb 11 23:19:43 Which ones? Feb 11 23:19:50 status and ethernet Feb 11 23:20:08 You should actually have a green, red, blank, green pattern. Feb 11 23:20:45 I didnt do anything yet; just load with xmodem, and g 0x10.... Feb 11 23:20:48 That would be 7 to you ant me. Feb 11 23:21:15 Oh. Wait, I'm wrong. Feb 11 23:21:24 did you get the light show? Feb 11 23:21:36 I dunno, I didnt look, its way over there. Feb 11 23:21:55 I can repeat this experiment and observe. Feb 11 23:21:58 one moment. Feb 11 23:22:01 OK. With this version of the code, it does a little LED magic while it loads data from flash to ram. Feb 11 23:22:10 Wait! Feb 11 23:22:15 Did I get you? Feb 11 23:22:22 nope too late Feb 11 23:22:35 OK. I was going to explain the easy way. Feb 11 23:22:36 not to worry I'll be in the same place shortly Feb 11 23:23:16 You can restart APEX when APEX is running by "g 0x00200000" Feb 11 23:23:26 Neat! Feb 11 23:23:29 You just to the start of the loader and it will do it again. Feb 11 23:23:33 I feel like I'm driving KITT! Feb 11 23:23:42 I guess you saw the lights. Feb 11 23:23:54 Super Pursuit Mode! Feb 11 23:24:05 OK. I think that the next step is to try to write it to the sysconfig space. Feb 11 23:24:14 From APEX, do these: Feb 11 23:24:22 I anticipated that and backed up the sysconf. Feb 11 23:24:26 du nor:0x40000 Feb 11 23:24:31 (cool) Feb 11 23:24:51 You should see the sysconf memory Feb 11 23:25:01 I get a display of the beginning of sysconf Feb 11 23:25:31 Hang on. I'm going to setup myself... Feb 11 23:25:55 ok Feb 11 23:26:19 OK Feb 11 23:26:44 Right. You should see the h[net work] and so on. Feb 11 23:26:53 First we erase Feb 11 23:27:07 erase nor:0x40000 # That's four zeros Feb 11 23:27:16 Then we write APEX there, Feb 11 23:27:31 k, its blown away. Feb 11 23:27:48 You need to look at the output of the 'ver' command to get the apex=>data Feb 11 23:27:53 version tells m mem:0x00200000+0x5398 Feb 11 23:28:09 OK. Mine is different. Probably from a different compiler. Feb 11 23:28:13 for you... Feb 11 23:28:18 right. Feb 11 23:28:27 copy mem:0x00200000+0x5398 nor:0x40000 Feb 11 23:28:51 Once it is written, you have to 'look' at it so that the memory is in the right state. Feb 11 23:28:56 du nor:0x40000 Feb 11 23:29:03 You will now see the same thing at Feb 11 23:29:09 du 0x50040000 Feb 11 23:29:14 All good? Feb 11 23:29:22 yeah ee 11 0f .... Feb 11 23:29:58 OK. Now Feb 11 23:30:03 g 0x50040000 Feb 11 23:30:08 It should start again. Feb 11 23:30:28 Right. But I missed KITT. ;-) Feb 11 23:30:48 Just go to the same address again :-) Feb 11 23:31:08 When you're ready, we can write it to the first flash block and give it a whirl. Feb 11 23:31:26 Neat. Feb 11 23:31:31 so we think this will work? Feb 11 23:31:47 Well, I have made a change. I wasn't doing the GPIO setup. Feb 11 23:31:59 If there is something else, I'll have to delve deeper into redboot. Feb 11 23:32:03 I'm ready; but if theres more testing; we need to be prepared for the 20min penalty. maybe 30 since I didnt split redboot yet. Feb 11 23:32:22 Well, I'm getting close to my limit for the day. Feb 11 23:32:28 It's close to midnight. Feb 11 23:32:42 Okay I'll flash it. One moment. Feb 11 23:32:44 If this doesn't work, the important result will be the LED output. Feb 11 23:33:52 BTW, did I tell you what I think I did to my first slug. The salted one? I think I figured out what went wrong. Feb 11 23:34:08 Hmm Feb 11 23:34:16 erase nor:0 Feb 11 23:34:21 Hmm...what? Feb 11 23:34:24 OK. Feb 11 23:34:36 copy mem:0x00200000+0x5398 nor:0 Feb 11 23:34:37 right? Feb 11 23:34:40 Right. Feb 11 23:34:40 reset Feb 11 23:34:43 heh Feb 11 23:34:45 ...and Feb 11 23:34:56 Uh oh. Feb 11 23:35:00 I have a single yellow LED Feb 11 23:35:01 What do the LEDs show. Feb 11 23:35:11 The first one? Not the red one? Feb 11 23:35:13 lemme try power cycle it again. Feb 11 23:35:43 yel/grn/grn/grn Feb 11 23:36:15 On mine, it goes green, yellow, green green with a space left for the ethernet. Feb 11 23:36:20 Are you seeing all ON? Feb 11 23:36:39 Yes, but my ethernet is plugged in Feb 11 23:36:54 so you're seeing all five lights? Feb 11 23:37:09 from top to bottom including the pwoer l ed yes. Feb 11 23:38:21 OK. That's really early. The first thing it does it set the GPIO lines. then it turns on all LEDs. then it does several CP instructions, then it sets the flash memory timing. Then it sets led1 on. Feb 11 23:38:35 * rwhitby-away joins to watch the fun :-) Feb 11 23:38:43 :) Feb 11 23:38:49 Whew. so, that means that either the CP instructions are failing...or there is something wrong with th EXP timing code. Feb 11 23:39:01 Hi rw? Feb 11 23:39:07 s/?/./ Feb 11 23:39:42 Okay so where we are is it can run from flash; but for some reason it doesnt execute on power up? Feb 11 23:39:45 The part that mystifies me is that this code works when executed from redboot. So, what about the exp timing could break that? Feb 11 23:39:52 Right. Feb 11 23:40:41 If you are game, we can update the code to get more precise information about which instruction is going awry. Feb 11 23:40:54 It would mean only a 10 minute flash time. Feb 11 23:40:57 Or so. Feb 11 23:41:06 ooh. Feb 11 23:41:13 20k should go fast. Feb 11 23:41:20 2min. Feb 11 23:41:25 yeah I'm gamem for that. Feb 11 23:41:31 Let me make a few edits. Feb 11 23:41:57 okie. Feb 11 23:42:01 * dyoung gets a beer Feb 11 23:42:35 dyoung: do you want a source file or an apex.bin file or what? Feb 11 23:42:47