**** BEGIN LOGGING AT Tue Feb 21 10:59:56 2006 Feb 21 14:09:08 http://www.embeddedarm.com/epc/ts7300-spec-h.htm Feb 21 14:09:17 with a user-programmable fpga :) Feb 21 14:09:19 * lennert runs Feb 21 14:09:34 :S seen it Feb 21 14:13:06 still the same odd pc104 bus , the fpga part makes the board alot larger more like a 5.25" drive footprint rather than a 3.5" that the ts7200 is , if it was not for the fact that the ram on the fpga was realy for the video then it may be good for the jtag programmer Feb 21 14:14:45 2 ethernet and 10 rs232 ports though ? strange Feb 21 14:17:05 since the hat as been shelved the one i am working on atm would probabaly be a better option than that Feb 21 14:18:09 'hat' ? Feb 21 14:19:22 same 20Mhz ep9302 , with upto 4 64mb sdrams so a nice 256mb ram space , a 50 mhz bus not the slow pc104 thing and 124mm* 96mm pcb Feb 21 14:19:45 hat the pcmcia version of the aztag Feb 21 14:20:33 right Feb 21 14:20:50 200mhz rather not 20 Feb 21 14:21:00 you know we're rewriting the cirrus ep93xx code, right? Feb 21 14:21:14 yes Feb 21 14:26:55 http://www.whipy.demon.co.uk/SBC.pdf Feb 21 14:28:41 thers 4 ram and 2 flash chips , 2 rams on top and 2 on the underside , and one flash ontop and one on the under side Feb 21 14:34:15 theres a modified version of the aztag board that stacks ontop , up to 6 boards could be stacked to give 96 chans at 150mhz or 48chans at 300mhz for the logic analizer functions Feb 21 14:35:57 and should be able to sync the trigers between the boards Feb 21 14:39:02 the hat design wasshelved for 2 reasons , firstly a heat problem , that could only realy be solved by a type III pcmcia enclosure that is a custom item so too costly, second the cpld for the pcmcia cardbus would of needed 6 layers so more expence Feb 21 14:42:04 and given the £35 each for the cable , it would of easaly toped the £250 and be more like $400 each for 10's of including assembly .5mm pitch bga's are a real pita to do in a toaster oven Feb 21 15:03:43 indeed Feb 21 15:04:25 AchiestDragon, how close to the S3 is the 2C8? Feb 21 15:07:34 8256 LE's Feb 21 15:08:39 suprised they did not use the ep2c20 thats got dubble Feb 21 15:08:54 k Feb 21 15:09:41 you would think that as there using it to drive video one ethernet and about 6 rs232 ports that theres any LE's left for the user Feb 21 15:10:22 k Feb 21 15:11:25 I might go away as I have to check out a WET11 for a customer here. Feb 21 15:14:27 why they got the power plug placed in the middle of the front side of the board Feb 21 15:19:03 am I still here? Feb 21 15:21:22 :) Feb 21 15:27:57 waking kidlet and getting him moving Feb 21 16:02:24 lennert: ping? Feb 21 16:03:13 pong? Feb 21 16:03:43 You might be able to answer this :-) Feb 21 16:05:02 The DoC on the ds101 is on the ixp4xx exp bus (CS0), at 0x50000000. When ioremapped, it ends up at 0xc4855000. Trying to read 16-bit values from the ioremapped address gives me a page fault, while reading/writing 8-bit addresses work just fine... Feb 21 16:05:06 Got any idea why? Feb 21 16:11:08 the expansion bus is configured for 8 bit accesses? :) Feb 21 16:16:49 I get the same when I've configured CS0 for 16bit too... Feb 21 16:16:55 (I think) ;) Feb 21 16:17:17 how are you reading 16 bit values? Feb 21 16:17:32 did you use ioremap_nocache ? Feb 21 16:17:47 i.e. can you send me your code? :) Feb 21 16:18:23 right now I'm trying __raw_readw Feb 21 16:18:28 yes, using _nocache Feb 21 16:19:16 bit harder with the code part, since it's spread over several files ;) Feb 21 16:19:29 just send me everything Feb 21 16:19:36 i just got nand flash working on the ts7250 Feb 21 16:19:43 so i have a fairly good idea of what's involved now Feb 21 16:52:01 lennert: take a look at http://pastebin.com/565217 Feb 21 16:52:17 The hacked-together module I wrote to try to just detect the bugger Feb 21 16:52:34 the setup for CS0 is at the bottom, it's actually done in the machine setup Feb 21 16:53:51 (there's also a $100 bounty for whoever gets the DoC working on the DS101 ;) Feb 21 16:55:13 the working 2.4 sources are available, but they're messy. Feb 21 17:50:53 i don't have a ds101 ;) Feb 21 17:53:05 ok, so which access segfaults? Feb 21 17:55:27 NAiL: _why_ are you remapping 0x500000001 ? Feb 21 17:55:33 NAiL: try mapping 0x50000000 Feb 21 17:55:37 NAiL: right now it's just mapping zero Feb 21 17:55:45 NAiL: and you're writing over your first page of ram Feb 21 18:13:06 NAiL: ping? Feb 21 18:54:50 lennert: not 01, 0L Feb 21 18:55:01 or, 0l Feb 21 18:55:29 the first access segfaults on 16-bit access Feb 21 18:55:30 right, hard-to-read font Feb 21 18:55:57 16bit read or write? Feb 21 18:56:04 read Feb 21 18:56:05 you _are_ configuring cs0 correctly? Feb 21 18:56:14 well Feb 21 18:56:17 the first read is unaligned Feb 21 18:56:22 junk += ReadDOC_(virtadr,0x1fff); Feb 21 18:56:32 0x1fff is not an aligned address Feb 21 18:56:34 try 0x1ffe Feb 21 18:58:22 ok, uploading new kernel now Feb 21 18:58:28 takes a while with ymodem :P Feb 21 18:58:43 I *think* I'm configuring CS0 correctly Feb 21 18:58:50 look at the defines at the bottom Feb 21 18:59:46 *IXP4XX_EXP_CS0 = DS101_EXP_CS0_INIT; Feb 21 18:59:53 That is done during the board setup Feb 21 19:00:15 (this setup is identical to the 2.4-setup that works) Feb 21 19:00:54 but 2.4 can read and write okay.. Feb 21 19:01:09 yes Feb 21 19:01:16 it didn't panic now Feb 21 19:01:23 but it still doesn't detect the chip Feb 21 19:01:28 (chipID is 0x0) Feb 21 19:01:30 one step further :) Feb 21 19:05:09 Any ideas on other steps to take? ;) Feb 21 19:05:19 The detection sequence should be correct Feb 21 19:05:31 (according to the old 2.4-driver that worked) Feb 21 19:05:41 (which happens to be a msys-driver) Feb 21 19:05:56 (which happens to be a *mess*) Feb 21 19:08:22 btw, regarding the cpld... There's nothing in the original kernel that appears to do anything with it Feb 21 19:08:41 so I'm not sure if it's connected to the flash at all Feb 21 19:09:02 the ixp can boot off the DoC, since it pushes redboot to memory itself Feb 21 19:09:12 "it" = the DoC Feb 21 19:09:56 (and redboot has a hacked-on driver to allow it to load the first partition for the kernel) Feb 21 19:14:05 dwery: lennert gave me the obvious answer to the 16-bit kernel panic Feb 21 19:14:12 19:57 < lennert> the first read is unaligned Feb 21 19:14:21 doh! Feb 21 19:14:25 yes Feb 21 19:14:31 I was pretty sure lennert was your man! Feb 21 19:14:46 (still not detecting it tho) Feb 21 19:15:05 ask lennert :D Feb 21 19:15:11 heh Feb 21 19:15:20 I'm kinda bugging him Feb 21 19:15:29 eheh Feb 21 19:18:45 unfortunately, neither of you have a ds101 to test on :( Feb 21 19:22:28 $$$$$ Feb 21 19:23:00 :-D Feb 21 19:24:33 Currently, I am unemployed, so buying two ds101's are actually more than my monthly budget Feb 21 19:25:54 (ie, more than I get paid) Feb 21 19:27:22 otherwise, sending a ds101 or two would've been high priority ;) Feb 21 19:31:14 eheh :) Feb 21 19:31:22 we'll have to work out the problem remotely Feb 21 19:33:41 did you send the artop patch upstream btw? Feb 21 19:34:07 not yet Feb 21 21:20:02 NAiLzZz: if you get me a shell on your ds101 i can maybe debug it remotely Feb 22 04:23:15 ka6sox-office: I have some sad news Feb 22 04:24:01 You all might get this in the log. Feb 22 04:24:17 I've gotten my JTAG application working to the point that it is useful for flashing. Feb 22 04:24:44 It's still got a few bugs, but that won't stop me from releasing a first version. Feb 22 04:24:57 The sad news is that the FTDI performs far below expectations. Feb 22 04:25:43 I still need to implement the generic boundary-scan code so that I can test flash programming via the scan chain. Feb 22 04:26:29 using the ARM core, I'm getting about 15.6 ms/byte. And that's without checking the result of the programming operation. Feb 22 04:26:52 If I add the check for the programming result, it's about 18.6 ms/byte. Feb 22 04:28:19 What's killing this is the latency of the USB bus. A buffered write requires four transactions of the embeddedice core: program, count, data load, data write, confirm. Feb 22 04:28:39 The data load shouldn't be a problem since we don't have to synchronize the core with the memory system. Feb 22 04:28:44 Still, it's not very efficient. Feb 22 04:29:06 The reason this is so sad is that the Xlinix cable stomps it all over the place. Feb 22 04:30:11 ouch! Feb 22 04:30:58 this doesn't sound good at all! Feb 22 04:33:01 It handles the same algorithm to the tune of 56K in 31s. About 571us/byte. Feb 22 04:33:15 yikes. Feb 22 04:36:33 Part of this comes from the fact that the FTDI protocol is sub optimal, but I don't think that that is the real cost. Feb 22 04:36:46 I believe that the latency of the USB interface is the worst of it. Feb 22 04:36:48 usb framing and latency? Feb 22 05:14:14 I don't have any visibility into the real cause. Feb 22 05:14:23 I've read over some of the USB programming docs. Feb 22 05:14:40 It has a distinctly ATM feel. Feb 22 05:14:49 ah... Feb 22 05:14:59 I've found that there are oddities in the way that packets are accepted on the other end. Feb 22 05:15:18 Specifically, I was sending 64 byte chunks, but found that this broke the FTDI protocol. Feb 22 05:15:31 I had to do back to sending everything and letting the kernel do the blocking. Feb 22 05:15:33 everything I've read about USB indicates that short packets are killer. Feb 22 05:21:44 the largest packet on this device is 64 bytes. Feb 22 05:22:25 ewwww. Feb 22 05:22:29 ugly Feb 22 05:30:52 I think that that is a limitation of the USB protocol and not really the fault of the device. Feb 22 05:31:07 On one device I've looked at, the largest packet is 128 bytes. Feb 22 05:31:25 overhead is killer on packets that short Feb 22 06:42:32 lennert: I can try getting network to work, so you can take a look at it. Feb 22 06:42:59 * dyoung-away breaks into NAiL 's DS101 Feb 22 06:46:12 hmm Feb 22 06:46:17 I'd like to see you do that Feb 22 06:46:34 considering it is impossible right now Feb 22 06:46:55 If it aint got no net, I would have to use my telepathic powers to intercept your EM Waves. Feb 22 06:47:16 Or if its not plugged in to power, I would have to.. uh.... Feb 22 06:47:18 hmm Feb 22 06:47:25 cant th ink of a good one, sorry. youre safe. :-) Feb 22 06:47:29 Well, I powered it up now Feb 22 06:48:00 but you'd have to get through the wrt54gs, then my slug, onto my laptop and use minicom as 00t Feb 22 06:48:38 So Hardcore. :-) Feb 22 07:00:36 of course, you'd have to figure out my IP too :-P Feb 22 07:01:12 * NAiL ssh's to a server and uses irssi from there ;) Feb 22 07:31:17 argh Feb 22 07:31:24 I can't get ixp networking up though Feb 22 07:32:05 http://pastebin.com/566346 Feb 22 10:34:24 NAiL: okay Feb 22 10:35:11 NAiL: looks odd, is that on the ds101? Feb 22 10:35:20 yeah Feb 22 10:36:58 Dunno what makes it not work on the ds101, since it works on the lude and slug. Feb 22 10:37:08 It's got the same PHY as the lude Feb 22 10:38:53 I guess I could connect the DS101 with USB to my gateway-slug when I get home Feb 22 10:39:13 That way you could atleast get a terminal through minicom Feb 22 10:41:08 would you have gcc and stuff though for testing? Feb 22 10:41:53 Hmm, not yet. I could zmodem necessary packages over though Feb 22 10:42:03 right, okay Feb 22 10:42:11 i'll be on the way to work in some 10 minutes Feb 22 10:42:14 let me know Feb 22 10:42:32 I won't be home for a few hours, so I won't be able to do it quite yet Feb 22 10:42:39 I'll tell you when I get something working Feb 22 10:44:30 great, thanks **** ENDING LOGGING AT Wed Feb 22 10:59:57 2006