**** BEGIN LOGGING AT Fri Feb 11 23:59:56 2005 Feb 12 00:00:03 maybe I was hallucinating. Feb 12 00:00:20 oh no I see Feb 12 00:00:24 Youre right. Feb 12 00:00:27 * dyoung hides Feb 12 00:00:36 * beewoolie grins Feb 12 00:01:35 So here's the crazy part. I'm going *exactly* the same steps as redboot at this point. Feb 12 00:04:20 The only real difference is that the CPU isn't already in bigendian mode. Feb 12 00:05:14 This is the stuff in mach-ixp42x/initilize.c ? Feb 12 00:05:19 yeah. Feb 12 00:07:43 spinner-nslu2.c ;-) Feb 12 00:08:14 You could make it do the kit thing... Feb 12 00:09:22 OK. I'm going to try this before I give it to you. Feb 12 00:09:39 ok Feb 12 00:17:34 rwhitby-away: you around? Do you know if there is a meaning to the 29th bit in the CPSR? Feb 12 00:17:45 There isn't anything in the programmer's manual Feb 12 00:18:37 beewoolie, super secret bit. Feb 12 00:18:42 :) Feb 12 00:18:59 That's what I'm afraid of. Feb 12 00:24:49 dyoung: you ready to try another? Feb 12 00:24:54 first rule of 29th bit is you don't talk about 29th bit Feb 12 00:25:03 what's a bit Feb 12 00:25:33 I dunno, but I'm ready to nibble a new apex.bin. Feb 12 00:25:52 beewoolie: nope, unless it's in the databook, I don't know anything about it Feb 12 00:25:54 OK. It's on the ftp server. Feb 12 00:26:13 rwhitby-away: I don't see it doing a text search on the PDF. Feb 12 00:26:28 it's not the bit which switches the flash memory mapping is it? Feb 12 00:26:42 No. That's in the EXP register file. Feb 12 00:27:05 CPSR stuff is usually core related. Bus synchronization and the like. Feb 12 00:27:28 halfway jtagged. Feb 12 00:30:19 Hmm Feb 12 00:30:26 What did we get? Feb 12 00:30:56 same Feb 12 00:31:06 power cycled it a couple times to be sure Feb 12 00:31:12 Refresh me. Feb 12 00:31:41 I read the leds as binary. The power-on is bit zero, amber is bit one, and so on. Feb 12 00:32:16 Are all on except for the ready light? Feb 12 00:32:24 If so, that's the CPSR for sure. Feb 12 00:32:25 the ready light is yellow. Feb 12 00:32:30 so from top to bottom.. Feb 12 00:32:49 yellow, (eth), green, green, green Feb 12 00:32:56 last green is the power light. Feb 12 00:33:16 Ah. Now I know what you're saying. We don't have control over the power led. Feb 12 00:33:47 There are only four we can control. Two to the left of the ethernet and two to the right. Feb 12 00:34:03 Are you sure that the green isn't on next to the red? Feb 12 00:34:04 Eh? Feb 12 00:34:06 what? Feb 12 00:34:09 I only have 4 leds Feb 12 00:34:12 and the power Feb 12 00:34:16 Not really. Feb 12 00:34:29 OH Feb 12 00:34:29 There are five in a cluster at the angle'd corner. Feb 12 00:34:33 I SEE what youre saying Feb 12 00:34:36 Okay Feb 12 00:34:38 Two of them are in the same bit of plastic. Feb 12 00:34:42 that would explain why its yellow. Feb 12 00:34:56 If you look really closely, you can see both elements. Feb 12 00:34:58 are both on? Feb 12 00:35:01 I dunno Feb 12 00:35:07 Lemme take it apart. Feb 12 00:35:09 one moment Feb 12 00:35:23 Ah. I hadn't considered that. Feb 12 00:36:37 Yes Feb 12 00:36:46 so green and red make yellow. ;-) Feb 12 00:37:10 Sorry for the confusion there. Feb 12 00:37:14 Ah. That's different. OK. That means it's the instruction that sets the CP15 controil word. Feb 12 00:37:18 np Feb 12 00:39:14 I think thats green,red,eth,green,green then Feb 12 00:39:34 I need to disassemble the file to make sure it's doing the right thing. Feb 12 00:41:34 Okay Feb 12 00:50:06 dyoung: Willing to try again? Feb 12 00:50:18 yep Feb 12 00:50:21 I put up another version. There was a very minor error in the code. Feb 12 00:50:33 I either pay the 6min penalty or the 20min penalty. Feb 12 00:50:33 I don't think it was the problem, but who knows. Feb 12 00:50:36 ok Feb 12 00:50:47 I also shrunk it a bit. Feb 12 00:51:02 I didn't want to make it too stripped in case it works. Feb 12 00:51:15 18k! BONUS. Feb 12 00:52:15 I can get it under 16384. Feb 12 00:52:46 To do this, it even picks another code path in the nor flash driver that is slower, but simpler. Feb 12 00:55:01 ok its done Feb 12 00:55:21 That seemed faster. Feb 12 00:55:26 same net. Feb 12 00:55:37 Really? It shouldn't be. Feb 12 00:55:37 (yellow),eth,green,green Feb 12 00:55:50 Hmm. That's wierd because I changed the LED codes. Feb 12 00:56:03 Apparently not getting that fa? Feb 12 00:56:04 far Feb 12 00:56:06 ? Feb 12 00:57:12 I'm really suspicious. Here's the code: Feb 12 00:57:24 _L(LEDf); /* Start with all on */ Feb 12 00:57:24 _L(LED8); Feb 12 00:57:58 There isn't even an option for 0xe which is what you're seeing. Feb 12 00:58:24 Okay Feb 12 00:58:41 okay let me try somethingi then Feb 12 00:58:47 this has always confused me Feb 12 00:58:52 What has? Feb 12 00:59:08 the endian-ness that you need to set in jtag tools Feb 12 00:59:20 I've been loading "endian big" Feb 12 00:59:22 I wonder if it hasn't been reprogramming the device. Feb 12 00:59:29 let me try change it to endian-small. Feb 12 00:59:29 Which is what you're supposed to do. Feb 12 00:59:37 You can check the data with the readmem command. Feb 12 01:00:11 I'm reading it now Feb 12 01:00:43 At this point, I would devolve this exercise into one where we truncate the code and write small program sequences that step the leds and show our execution progress. Feb 12 01:01:04 That's what I had to do to get it running in the first place. Feb 12 01:06:22 Okay Feb 12 01:06:26 endian "little". Feb 12 01:06:32 no leds. Feb 12 01:06:48 Not suprised. Was the read-out the same as the original? Feb 12 01:06:54 yes Feb 12 01:06:57 I thought that the NPE stuff doesn't work with LE. Feb 12 01:07:07 It doesn't. Feb 12 01:07:16 The CPU can run in little endian mode, tho. Feb 12 01:07:37 Really, most things don't care. It's only a problem for half-word and byte accesses. Feb 12 01:07:48 beewoolie, okay. Feb 12 01:08:29 So, I'd like to do a couple of short tests to see if we are really getting the results we expect. Feb 12 01:08:32 writing it back in in endian big. Feb 12 01:08:39 Okay Feb 12 01:08:55 I'm trying to get it down to something really small. Feb 12 01:11:08 I put up a file that's less than 10K. It should light just the disk 2 led. Feb 12 01:11:46 its called apex.bin ? Feb 12 01:11:51 I could probably get is smaller if it's still taking too long. Feb 12 01:11:54 yeah. Feb 12 01:12:25 Let's do a check. the MD5sum is bb6002ebbdc59e82811a22986bb47843 apex.bin Feb 12 01:13:00 rwhitby-800bucks: why not $1mil Feb 12 01:14:08 bb6002ebbdc59e82811a22986bb47843 apex.bin Feb 12 01:14:32 ;-) Feb 12 01:14:46 Red LED Feb 12 01:14:55 Right. That's the right one. Feb 12 01:15:01 but it aint the disk2 led Feb 12 01:15:18 I got that wrong. It is supposed to be LED2 which is the red one. Feb 12 01:15:23 its the ready/status Feb 12 01:15:24 okay Feb 12 01:15:31 So I wonder whats going on.... Feb 12 01:15:44 Let's do another. Feb 12 01:15:59 47d7dbfa82236520c753354d086f12c7 apex.bin Feb 12 01:16:50 writing Feb 12 01:18:19 no leds Feb 12 01:18:30 Interesting. Feb 12 01:18:46 Binary search. Feb 12 01:19:17 c6d4b0b13d4a561ea0b78581eae0b55d apex.bin Feb 12 01:19:19 Another. Feb 12 01:19:36 Aw heck. Wait. Feb 12 01:19:44 ok Feb 12 01:19:52 The GPIO mask was being cleared. Feb 12 01:21:47 4f68c0baa193ca30b5349dbfba381f92 apex.bin Feb 12 01:21:49 OK. Feb 12 01:22:13 writing Feb 12 01:22:36 BTW. If your slug gets salted during this process, I'll buy you a new one. Feb 12 01:23:08 dont worry about it Feb 12 01:23:13 I have faith. Feb 12 01:23:34 THis one has kinda been through hell already. Feb 12 01:23:57 Good thing this new flash is rated for 100K writes. Feb 12 01:23:57 (green),eth,green,none Feb 12 01:24:18 That's a 101 Feb 12 01:24:21 5. Feb 12 01:24:39 OK. Now it's getting past the CPSR instruction. Feb 12 01:24:43 (I dont remember the position of the gree/red led in the first lens) Feb 12 01:25:29 Green is bit zero. Yellow is bit one. Disk2 is bit two. and disk 1 is bit three. Feb 12 01:26:53 Okay that will help. Feb 12 01:27:09 2d60b4bfa3d2ae4167f7f35b93d75ff2 apex.bin Feb 12 01:27:46 writing Feb 12 01:27:52 I got that wrong. It isn't the CPSR that it passed, but the MMU disable command. Feb 12 01:28:53 so this is the magic bit 29 test. ;-) Feb 12 01:28:59 Right. Feb 12 01:29:15 I'm so lazy. Feb 12 01:29:34 I'm actually considering running an extension cord switch so I can sit here and power ccle it. Feb 12 01:29:51 1001 Feb 12 01:30:01 green,eth,none,green Feb 12 01:30:05 Then we're almost there. Feb 12 01:30:32 Cool. Feb 12 01:30:40 Have you made any changes though? Feb 12 01:30:56 It makes me wonder whats hanging up. Feb 12 01:31:10 Hard to tell. We're now going for the SDRAM. Feb 12 01:31:40 Remind me to bug you about SDRAM config later. Feb 12 01:32:19 2a57278d2a156a517c7ba1159e8a55bf apex.bin Feb 12 01:32:23 k Feb 12 01:33:01 writing Feb 12 01:34:19 1 Feb 12 01:34:28 That's the right answer. Feb 12 01:34:36 green eth none none Feb 12 01:34:47 Oh. Wait. There was a window over it. Feb 12 01:35:08 Ah. The problem is that it thinks it is running from SDRAM. Feb 12 01:35:17 That would be a serious problem. Feb 12 01:35:40 because it aint there anymore eh? Feb 12 01:35:47 no no no. I'm tired and I'm being forgetful. Feb 12 01:36:07 Yeah, it slike 130 am... Feb 12 01:36:10 dont forget to sleep.... Feb 12 01:36:15 It's right. It thinks it's running from flash. I need to let it go a little further. Feb 12 01:36:18 Soon. Feb 12 01:37:06 beewoolie: 800 bucks is how much in donations we have to spend on the cross-compilation development build machine for nslu2-linux Feb 12 01:37:09 I needed to know that it is properly detecting the flash configuration bit. So, it knows that it is running from flash locatred in low memory. Let me reread this code because it's never been executed. Feb 12 01:39:28 ad124c799bba8879b7f2ca66733ba9a1 apex.bin Feb 12 01:39:43 This time, it will either stay at 1 or move to something else. Feb 12 01:41:39 rwhitby-800bucks: Is there a goal? Feb 12 01:42:02 101 Feb 12 01:42:27 that's the right answer. Really. It's at the end of the SDRAM init. Feb 12 01:42:46 OK. Let's see how far it gets in the rest. Feb 12 01:42:57 So whats left then? Feb 12 01:43:16 I'm going to instrument a couple of places in the startup code before the serial driver is enabled. Feb 12 01:46:43 ad124c799bba8879b7f2ca66733ba9a1 apex.bin this one has all of the basics. Feb 12 01:47:06 wrting Feb 12 01:47:33 wait are you sure? Feb 12 01:47:43 Am I sure of what? Feb 12 01:47:46 the last 2 md5sums you pasted are the same Feb 12 01:48:04 Hang on. Feb 12 01:49:02 879a49f3da3605b89a2916df06131b46 apex.bin Feb 12 01:49:04 Nice catch. Feb 12 01:49:15 The compile failed. Feb 12 01:49:22 writing Feb 12 01:49:47 ta. Feb 12 01:49:56 Actually the file size caught my eye. Feb 12 01:50:08 What do you mean? Feb 12 01:50:19 in the window that I'm wgetting all this. Feb 12 01:50:21 Oh, about catching the mistaken build. Feb 12 01:51:25 111 I think Feb 12 01:51:42 7. Feb 12 01:51:49 (yellow),eth,green,none Feb 12 01:51:56 Really. And no console? Feb 12 01:52:22 nope Feb 12 01:52:27 Hmm. Feb 12 01:52:29 lemme double check the connection Feb 12 01:52:36 I unplugged it when I took it apart Feb 12 01:53:00 looks like its in there. Feb 12 01:53:23 After 7 means that it is running from SDRAM and it's going for the drivers. Is this a modified slug? Ram wise? Feb 12 01:53:56 no the ram is stock. Feb 12 01:54:40 is the serial stuff being properly initialized/set? Feb 12 01:54:45 (I wonder out loud) Feb 12 01:54:55 Let's find out which driver isn't loading. Feb 12 01:56:26 I have an idea that might short-circuit this process. Feb 12 01:56:43 short circuit? or shortcut? Feb 12 01:56:54 :-) Feb 12 01:57:27 At this point, it is probably the memory discovery code. That's usually what goes wrong. Feb 12 01:57:40 So, we can prevent any mishaps by being explicit. Feb 12 01:57:46 Ah thats what I was gonna bug you about Feb 12 01:57:59 SDR_CONFIG is set but never used AFAICT. Feb 12 01:58:28 Huh? Feb 12 01:58:38 in initilize.c Feb 12 01:58:48 so I did a serach on SDR_CONFIG on everything. Feb 12 01:58:54 and didnt see anyplace it was used. Feb 12 01:58:56 It's a register. Feb 12 01:59:01 oh Feb 12 01:59:02 heh Feb 12 01:59:08 that would explain it now wouldnt it... Feb 12 01:59:11 In the ixp42x.h Feb 12 01:59:55 What the memory driver does is scan the memory bank. It writes every meg a signature and then goes back and checks them. Wherever the signatures match addresses, it knows it has a valid 1M block. Feb 12 02:00:04 nice... Feb 12 02:00:18 This is important in the sharp processors because memory can be all over the place. Feb 12 02:00:22 Here, we don't really care. Feb 12 02:00:23 so this should work easily on a memory modified nslu2 Feb 12 02:00:28 Sort of. Feb 12 02:00:32 There are two kinds of mods. Feb 12 02:00:52 If we add chips, we'll have a second bank. That has to be handled separarately. Feb 12 02:01:09 If we put on larger chips, well we have to change the config anyway. Feb 12 02:01:18 I would be changing to the 4x8Mx16 configuration.... Feb 12 02:01:36 Right. That's four chips which means another bank. Feb 12 02:01:45 Another bank of 32MiB. Feb 12 02:01:49 Right Feb 12 02:01:57 okay well I'll bug you about it later. ;-) Feb 12 02:02:10 do you have one of these baby's yet? Feb 12 02:02:31 Not yet. Feb 12 02:02:47 ka6sox was gonna mod one for me, but I didnt get around to send ing it to him yet. Feb 12 02:04:51 fc9748dabb98755ad4f5c2752f32b503 apex.bin Feb 12 02:04:58 Let it rock. Feb 12 02:05:09 writing! Feb 12 02:05:58 nite all... Feb 12 02:06:03 nite Feb 12 02:06:25 catch ya from dizzyland. Feb 12 02:06:32 woo Feb 12 02:06:55 * ka6sox-away hacks while riding the Matterhorn Feb 12 02:07:00 Is that...like...hawaiian? Feb 12 02:07:22 nope...pure Californian Feb 12 02:07:30 :) Feb 12 02:07:31 nite all Feb 12 02:08:04 okay disk1 and disk2 l ights are dim, so they might be flashing real fast Feb 12 02:08:15 and the green status light is on. Feb 12 02:08:33 Huh? Feb 12 02:08:38 Let's see. Feb 12 02:08:50 I take it no serial. Feb 12 02:08:57 nope Feb 12 02:09:53 It might be yellow, but I dont think so. Feb 12 02:11:03 Are you sure about those leds? Feb 12 02:11:19 I'm sure the disk led's are dim. Feb 12 02:11:45 And the status green-yellow? Feb 12 02:12:51 If the red led is on, its dim Feb 12 02:13:26 It sounds like it's in a loop. I'm going to force the serial driver to init first. Feb 12 02:14:01 ok Feb 12 02:22:02 back 5 min. Feb 12 02:22:03 ff34b9fb6505accd072d55f36de72d7b apex.bin Feb 12 02:22:06 OK. Feb 12 02:22:09 oh I'll do that first. Feb 12 02:22:16 excellent. Feb 12 02:25:10 I'm thinking it might be good to support an early serial init the way that the kernel does. Feb 12 02:25:20 thats about the same. No serial, dim disk1/disk2 and possibly dim red with green Feb 12 02:25:59 OK. That's really wierd. The first init is the serial and it should be halted. Feb 12 02:26:22 It's sleepy by time for me. Feb 12 02:26:33 I'll think about a strategy tomorrow. Feb 12 02:26:34 I double checked the right image is burned. Feb 12 02:26:36 Okay Feb 12 02:26:37 sleep o it Feb 12 02:26:41 OK. Feb 12 02:26:45 Thanks for the help. Feb 12 02:26:47 I'll be back around around 11ish yourtime. Feb 12 02:26:50 thank YOU. Feb 12 02:26:52 am or pm Feb 12 02:27:03 I'm hoping we can use apex for the memory modified slug. Feb 12 02:27:06 AM. Feb 12 02:27:24 have a great night! Feb 12 02:27:25 OK. It ought to work just fine. I've got a board from sharp that has two banks of memory. Feb 12 02:27:42 Cool. Feb 12 02:28:17 ok see you tommrow then. Feb 12 02:33:59 had to stop at the office... Feb 12 02:34:05 what is the decision on the G4? Feb 12 02:34:15 None yet Feb 12 02:34:33 okay Feb 12 02:34:39 I'm happiliy disassembling the old Dozer Feb 12 02:34:47 catchya soon... Feb 12 02:34:51 poor webbie. Feb 12 02:34:53 its pretty sad, my manical laughing while doing this Feb 12 02:35:03 no this is the old desktop dozer Feb 12 02:35:07 : Feb 12 02:35:07 the one the Mini replaced. Feb 12 02:35:15 oh okay. Feb 12 02:35:25 Webbie is tossed on the side over there. Feb 12 02:35:36 ow oww owwww Feb 12 02:36:21 theres a lot of extra bits in this machine, I should use elsewhere Feb 12 02:36:23 "on the road again!" Feb 12 02:36:24 DVD Burner. Feb 12 02:36:32 DVD good Feb 12 02:36:32 SATA Disks... Feb 12 02:36:34 sigh Feb 12 02:36:40 I'm sad it broke. Feb 12 02:36:59 okay have fun roadie. Feb 12 12:04:57 dyoung-zzzz: I've posted a new version of the source, 1.0.24. This one has a config option, debug_ll, that enables console output early in the boot process. Feb 12 12:05:26 You'll have to build it. Feb 12 12:06:00 I've moved the CROSS_COMPILE macro from the Makefile to the config file. It doesn't, yet, allow an override from the environment. I'm not sure why. Feb 12 12:06:13 You can just change it in the configuration file for now. Feb 12 12:08:51 ...odd... I just say the behavior that would have been the same as what you saw with the dim LEDs. Feb 12 12:08:56 s/say/saw. Feb 12 12:09:08 s/see/saw/ Feb 12 12:09:10 :=) Feb 12 12:09:45 and now it's gone. Feb 12 12:09:53 hmm Feb 12 12:10:13 Anywho. It ought to look like this: Feb 12 12:10:21 RedBoot> g 0x1000000 Feb 12 12:10:21 AEreIS012M0kim3s Feb 12 12:10:27 APEX Boot Loader 1.0.24 -- Copyright (c) 2004,2005 Marc Singer Feb 12 12:10:32 and so on, Feb 12 12:11:01 What I just saw was the capital S repeating over and over. Feb 12 12:11:16 That would have mimic'd the dim LED behavior. Feb 12 12:11:50 I'm not as smart as you guys. I don't have a logger for this channel. Feb 12 12:12:38 I'll be back in a few hours and I'll check on the channel. Feb 12 12:14:39 BTW, the codes are mnemonic. Apex, Entry, Relocate, entry, Init, Services, 0, 1, 2, Memory Service, Bank0, mark, identify, Service 3, services. Feb 12 12:14:46 At that point, the system had better be running. Feb 12 12:15:12 ttfn Feb 12 12:46:34 kay, I found our loop. Feb 12 12:46:56 AErAErAErAEr .... forever and ever. Feb 12 13:10:07 ...decided not to make that errand anyway.... Feb 12 13:10:52 that looks like a problem with accessing SDRAM. The system will reset when there is a fault and... I have an idea. Feb 12 13:13:51 I'm staring rather blankly at relocate.c Feb 12 13:14:01 need more caffiene or something. Feb 12 13:15:27 dyoung-zzzz: I think I understand the problem. I believe I've misunderstood the exp_cfg0 register. Feb 12 13:16:12 I think what is happening is that I'm attempting to write the loader to SDRAM before disabling the flash remapping. I thought that it was only mapping a portion of the flash. That looks not to be true. Feb 12 13:17:20 so, I should be able to fix this with...get this...a one line change. Feb 12 13:17:28 Cool! Feb 12 13:18:13 Ah heck. I can't do it in one line. The compiler needs a little help. Feb 12 13:18:43 back 5 min Feb 12 13:18:46 k Feb 12 13:31:05 dyoung-zzzz: Back? Feb 12 13:31:14 yeah back now Feb 12 13:31:47 2fc17e02d46882dd41d98d171e8a7d62 apex.bin Feb 12 13:32:02 Try this one. Feb 12 13:34:23 writing Feb 12 13:34:38 * beewoolie-away busts out the snare again Feb 12 13:34:52 BTW, are you on the big island? Feb 12 13:35:04 Nope, Crowded Oahu. Feb 12 13:35:21 Home of all the potholes in the world. Feb 12 13:35:32 Oh. wait. I think I got that wrong. Is oahu the island with honolulu? Feb 12 13:35:45 Yeah. Feb 12 13:35:48 ...and thanks for finding homes for those potholes. Feb 12 13:35:53 Big Island is the one with the active volcano's Feb 12 13:36:15 Did you know that the army base there auctions off lots of stuff. Often for cheap? Feb 12 13:36:36 "Cheap" is relative. Feb 12 13:36:45 I think everything is over priced now Feb 12 13:36:45 I bought a logic analyzer for $90. Feb 12 13:36:55 Minimum bids are $35. Feb 12 13:37:01 Not here Feb 12 13:37:07 Which isnt a LOT; but I used to get pallate loads of stuff for $4. Feb 12 13:37:11 Things often go for $1 Feb 12 13:37:27 I wonder if we're talking about the same place. Feb 12 13:37:31 In 1998-ish, I made a ton of money. Feb 12 13:37:35 ah Feb 12 13:37:40 buying and selling DRMO stuff. Feb 12 13:38:06 That's it. They don't have a min bid as far as I can tell. Maybe things have changed again. There is a private firm handling the auctions. Feb 12 13:38:06 I think the best was the Watkin-Johnson Recievers. Feb 12 13:38:13 What's that? Feb 12 13:38:23 I stopped going when the private company took over. Feb 12 13:38:28 because of the minimum bids Feb 12 13:38:38 W-J Recievers are real high end radio stuff. Feb 12 13:38:41 Odd. I wonder if they've changed their policy. Feb 12 13:38:48 Nice. Feb 12 13:38:51 My $35 Investment turned into $1900 in a matter of 3 days. Feb 12 13:38:56 Wow. Feb 12 13:39:04