irclog2html for #openjtag on 20050802

04:54.09*** join/#openjtag dyoung-away (~dyoung-aw@dyoung.nslu2-linux)
09:57.12rwhitby-awayguys, let's continue landslide's 128MB apex discussion here ...
09:57.39rwhitby-away[19:07] landslide: u know how i have a 128mb slug (4 x 256mbit).
09:57.39rwhitby-away[19:08] landslide: i set the bank size to 0x04000000 and enabled the second bank
09:57.39rwhitby-away[19:08] landslide: then compiled
09:57.39rwhitby-away[19:08] landslide: when i do version i get this
09:57.40rwhitby-away[19:08] landslide:   memory:
09:57.40rwhitby-away[19:08] landslide:           0x0 0x02000000 (32 MiB)
09:57.41rwhitby-away[19:08] landslide:           0x4000000 0x02000000 (32 MiB)
09:57.43rwhitby-away[19:08] landslide:   nor:    8MiB total  32B write buffer
09:57.45rwhitby-away[19:08] landslide:           region 0:  64 blocks of 131072 (0x20000) bytes
09:57.47rwhitby-away[19:08] landslide:   cpu: id 0x690541f1  ctrl 0xafa  cpsr 0x200000d3
09:57.49rwhitby-away[19:09] landslide: it still thinks they are 32MiB yet the second bank starts at 64MiB.
09:57.51rwhitby-away[19:09] landslide: i'm going to trace through the code and work out what's going on.
09:57.55rwhitby-away(for beewoolie to see in the logs)
10:02.41dyoung#openjtag. the place where all the funky hardware developments happen.
10:03.03rwhitby-awayheh, thx
10:08.59landslideok here..
10:09.33landslidei'm just debugging the drv-mem.c memory_scan() function.
10:09.54landslidewhere it does a mark and identify and counts the blocks.
10:10.01landslidei'm trying to work out what it does.
10:10.08landslidekeep in mind I might have faulty ram chips...
10:10.35landslidethey were removed from another board so got 2 doses of heat from the rework station
10:29.00landslidei think i need to load it into flash
10:29.38landslidesome of that drv-mem.c code uses the APEX_VMA_START/END which i don't understand why.
10:40.56*** join/#openjtag DaKa2 (~david@193.13.58.211)
10:42.16*** join/#openjtag NAiL (~repvik@repvik-active-pdpc.nslu2-linux)
11:12.58landslidehow many regions is memoryscan() supposed to find? I think there is a bug in the identify portion, it seems that once it gets past 0x2000000 it keeps finding the region end.. should there be only 1 region found?
11:13.24landslideor is this relying on some quirk or trick to find out how many ram chips there are?
11:14.08landslidesee i printed out the pointer and data and the data and pointer are the same until the second block of 16 mb.
11:15.30landslidethe data is then 0x02000000 less than the pointer.
11:16.37landslidepointer = pl ... data = *pl
11:30.49*** join/#openjtag DaKa2 (~david@193.13.58.211)
12:25.45*** join/#openjtag VoodooZ (~VoodooZ@CPE00c0f021712c-CM000f9fac823a.cpe.net.cable.rogers.com)
12:38.30landslidei'
12:38.35landslidei've had some success..
12:38.45DaKa2:)
12:40.11landslidei found out that if i did printf("SDR_CONFIG %lX", (unsigned long)SDR_CONFIG); from drv_mem.c  in function memory_init()
12:40.22landslideit returned 0x18, which is wrong..
12:40.49landslidei had forced it to SDR_CONFIG = SDR_CONFIG_CAS3 | 0x03;  in initialize.c
12:40.56landslidefor some reason it gets reset.
12:41.22landslideanyway i forced it again in memory_init() and all the memory is now available.
12:41.30landslidethe memory_scan() function worked fine.
12:41.38landslidefor both banks.
12:42.20landslideso anyway after reading SDR_CONFIG once i forced it in memory_init() it was 0x1B which is correct for my configuration (4x16Mx16)
12:44.51landslidenow i get this: ->
12:44.59landslide<PROTECTED>
12:45.00landslide<PROTECTED>
12:45.00landslideServices:
12:45.00landslide<PROTECTED>
12:45.01landslide<PROTECTED>
12:45.01landslide<PROTECTED>
12:45.02landslide<PROTECTED>
12:45.04landslide<PROTECTED>
12:49.14landslidethe ram is working fine, i've done numerous fills and dump checks in each bank.
13:19.57rwhitbyso does it work?
13:32.44landslidevery close.
13:32.49landslidei can't get the kernel to boot.
13:32.58landslidemind you, i'm running apex from ram.
13:34.00landslideeven when i revert to the same configuration as the wiki the kernel will uncompress then say 'crc checksum error'
13:36.58*** join/#openjtag cbpage (~cpage@ip68-6-39-177.sb.sd.cox.net) [NETSPLIT VICTIM]
13:37.11*** join/#openjtag landslide (~david@203-206-61-8.dyn.iinet.net.au) [NETSPLIT VICTIM]
14:01.47*** join/#openjtag ByronT (ByronT@ByronT.nslu2-linux)
15:41.11*** join/#openjtag prpplague (~billybob@72.22.131.116)
15:51.21*** join/#openjtag Jacmet (~jacmet@50C59F0E.flatrate.dk)
16:06.41*** part/#openjtag NAiL (~repvik@repvik-active-pdpc.nslu2-linux)
16:44.29*** join/#openjtag jbowler-away (~jbowler@jbowler.nslu2-linux)
18:02.41*** join/#openjtag beewoolie-afk (~beewoolie@florence.buici.com)
18:03.07beewoolie-afkprpplague: Hey.  I just missed your ping yesterday.
18:03.09beewoolie-afkdyoung: hey man
18:03.51prpplaguebeewoolie-afk: hey, just saying hello
18:03.54prpplaguebeewoolie-afk: nothing major
18:04.15beewoolie-afkprpplague: FYI.  The ethernet driver and IP stack for APEX are released.
18:04.19beewoolie-afkIt's kinda handy.
18:04.33beewoolie-afkwhat ethernet controller are you guys using?
18:05.57prpplaguecs8900 8-bit polled, i.e. the packet-wacker
18:06.40beewoolie-afkHmm.  I thought as much.
18:06.53beewoolie-afkI've got one on one of the KEV boards, so maybe I'll write the driver for that one, too.
18:07.05beewoolie-afkIt isn't much work since APEX is so simple.
18:07.14prpplaguebeewoolie-afk: simple for you, hehe
18:07.19prpplaguebeewoolie-afk: you wrote it
18:07.55beewoolie-afkPerhaps.  The interface for the ethernet drivers is pretty much three calls.  init, read, write.
18:08.16prpplaguebeewoolie-afk: lovely
18:08.25prpplaguebeewoolie-afk: i'll have another look
18:08.35beewoolie-afkv1.3.0
18:09.28beewoolie-afkdyoung: Are you listening?  If we have the IAL linked into APEX we could use the TFTP code on the slug.  Hint hint.
18:16.54*** join/#openjtag malfi (~malte@dsl-082-082-190-181.arcor-ip.net)
18:20.06prpplaguebeewoolie-afk: why not use the tftp code from rmk?
19:10.40*** join/#openjtag dyoung-away (~dyoung-aw@dyoung.nslu2-linux)
19:25.16*** join/#openjtag ep1220 (~NN@gate.epatec.at)
20:20.27beewoolie-afkprpplague: what code from RMK?  The tftp code isn't really the hard part.  In fact, the code isn't very interesting no matter how it is done.  The tricky part is integrating the protocol into the loader.
20:21.24prpplaguebeewoolie-afk: rmk release a bootloader tftp code awhile back
20:21.34prpplaguebeewoolie-afk: let me look to see if i have a copy handy
20:21.45beewoolie-afkKO.  It's kinda too late since it's already written.
20:22.01beewoolie-afkWhat we need is the IAL so that we can driver the MAC from APEX.
20:22.35beewoolie-afkprpplague: I'd be interested, tho, because I haven't added bootp or dhcp.
20:24.46prpplaguebeewoolie-afk: i can't seem to access our cvs server at the momment
20:24.53prpplaguebeewoolie-afk: i'll pull it down shortly
20:24.59beewoolie-afkthx
20:28.48rwhitby-asleepbeewoolie-afk: did you see landslide-zzzz's comments here earlier?
20:29.14beewoolie-afkrwhitby-asleep: Didn't.  Scrolling...
20:29.27beewoolie-afkLooks like it was before I was on.  Was it on this channel?
20:29.33rwhitby-asleepyep
20:29.48beewoolie-afkWanna give me the reader's digest version?
20:29.59rwhitby-asleephe's got a 128MB slug and was getting apex to run on it
20:30.07beewoolie-afkSuccessfully?
20:31.51rwhitby-asleephttp://pastebin.ca/19125
20:31.58beewoolie-afkthx
20:34.43beewoolie-afkAh.  One of the things that he didn't understand is that the SDRAM is *not* configured when APEX runs from ram.  It should be obvious, but most people haven't thought it through.
20:35.37beewoolie-afkThere is no way to know what should happen if we reconfigure the SDRAM controller when the running code out of RAM.
20:35.59beewoolie-afkIn other words, the SDRAM mode isn't being reset in APEX.  It's never being set.
20:36.54beewoolie-afkYou can also tell him that if he's having problems with the kernel decompressing, it is a good idea to checksum the uploaded file before asking the kernel to decompress.
20:37.44rwhitby-asleephttp://pastebin.ca/19126 <- some stuff at the start of this before the previous one, and a bit in the middle and end
20:37.49beewoolie-afkI suspect the problem is that the SDRAM isn't properly configured
20:38.35prpplaguebeewoolie-afk: hey, you know a good tool to convert arm asm into some psuedo code for readability?
20:38.47rwhitby-asleepbeewoolie-afk: he's in AU timezone, like me (but I'm up early)
20:38.50beewoolie-afkndisasm.  Look at my wiki.
20:39.13beewoolie-afkprpplague: I linked the disasm there some time ago.  wiki.buici.com
20:39.55beewoolie-afkrwhitby-asleep: Fair enough.  I'm reviewing the SDRAM init code just to see if there is something I can see.  Perhaps I should add a message at signon about the SDRAM *not* being initialized.
20:40.19rwhitby-asleepbeewoolie-afk: the OE compiler vs crosstool is intriging too (from the second pastebin)
20:40.34beewoolie-afkI haven't gotten all the way through yet.
20:41.06prpplaguebeewoolie-afk: cool thanks
20:42.43rwhitby-asleepbeewoolie-afk: I put apex into OpenEmbedded as a .bb file, but didn't test it.  When we buld using OE, it seems to produce a binary which gives GDB codes on the serial console.  Nobody knows why.
20:42.46beewoolie-afkrwhitby-asleep: The compiler stuff is odd.  As I've never seen any such failures, it's hard to imagine why.  I suspect that either a) they aren't using a big-endian compiler or b) there is something wrong with their config.
20:43.02beewoolie-afkrwhitby-asleep: That isn't what you think it is.
20:43.11beewoolie-afkThe gdb codes come from redboot.
20:43.26rwhitby-asleepah
20:43.31prpplaguebeewoolie-afk: hmm, that doesn't appear to provide anything more than just objdump
20:43.34beewoolie-afkApex has no GDB code at all.  It means that you are runnning APEX from RAM and it's crashing.
20:43.44beewoolie-afkprpplague: What are you looking for?
20:43.48beewoolie-afkC code?
20:44.13prpplaguebeewoolie-afk: not nessarly c, but something psuedo like that easier to read
20:44.35beewoolie-afkrwhitby-asleep: If you send me one of these errant versions of APEX, I'll see if I can see what is the matter.
20:44.46prpplaguei guess i'm slow with asm, takes me forever to step through the asm and write down what its doing
20:44.53beewoolie-afkrwhitby-asleep: It's important to set the MMU disable option in APEX so that it gets access to the whole address space.
20:45.01beewoolie-afkprpplague: Not sure what else there is.
20:45.28rwhitby-asleepbeewoolie-afk: I'll get dyoung to send you a failing binary
20:45.33beewoolie-afkK.
20:46.00rwhitby-asleepI haven't run Apex myself yet (lack of time), so I'm only relaying messages at the moment :-)
20:46.43rwhitby-asleepbut I'm keen to get it into our standard builds, as we need it for DebianSlug
20:47.23beewoolie-afkrwhitby-asleep: I'll take a look as soon as I get a binary.  The elf file is the most useful.
20:48.20beewoolie-afkrwhitby-asleep: BTW, there may be an error in the code.
20:49.03beewoolie-afkIt looks like I'm assuming that the CPU clock is 133MHz.  At least, that's the fundamental frequency for the SDRAM refresh clock. I need to make sure that a deunderclocked slug has the right refresh clock.
20:49.48rwhitby-asleepbeewoolie-afk: I might be able to build a binary for you now to look at
20:49.54beewoolie-afkk.
20:50.15rwhitby-asleepcan you /msg me an email address to send it to?
20:54.48rwhitby-asleepbeewoolie-afk: just the apex.bin file?
20:54.55beewoolie-afkThe apex file is better.
20:55.23beewoolie-afkI added some code to the ixp init that will exclaim that SDRAM has not been initialized if it runs from SDRAM.
20:57.54rwhitby-asleepbeewoolie-afk: sent - I haven't tried this one, so I dunno if it works or not.
20:58.22beewoolie-afkI assume you're building it with the OE compiler.
20:58.28rwhitby-asleepyep
20:58.40rwhitby-asleepthat's the automatically built one (bb apex)
20:58.46rwhitby-asleep(bitbake apex that is)
21:05.54beewoolie-afkrwhitby-asleep: would you mind sending the .config file?
21:08.12rwhitby-asleepsent
21:08.55beewoolie-afkrwhitby: It looks like the optimization level is different.
21:08.59beewoolie-afkWhat compiler are you using/
21:09.01beewoolie-afk<PROTECTED>
21:10.01rwhitbygcc 3.4.4
21:10.09beewoolie-afkYou've also set a ramdisk.  Is this for unslung?  We don't need a ramdisk for openslug.
21:10.40rwhitbythis is just for general nslu2 - we haven't sorted out the right config yet
21:10.53beewoolie-afkShouldn't matter.
21:10.53rwhitbyjust trying to get to the apex console at this stage
21:11.30beewoolie-afkLet me see about the compiler version.  It's possible that there is some sort of new bug in the compiler.
21:11.36beewoolie-afkI'm using 3.4.2
21:12.33beewoolie-afkI'll have to build a new compiler to test this.  
21:12.44beewoolie-afkYou know what that means.
21:13.01rwhitbyI can tar up the cross directory and send it to you if you like
21:13.34beewoolie-afkI'd rather build one just so I can track the pedigree.
21:13.38rwhitbyok
21:13.57beewoolie-afkBesides, I'm going to test with the LH processors first and those are LE.
21:14.13rwhitbyok
21:14.18beewoolie-afkThanks for the help.
21:14.31rwhitbyI'll also send you the run.do_compile script which OE uses to compile it
21:16.24rwhitbysent.  Thank *you* :-)
21:20.22beewoolie-afkAlright.
21:21.24rwhitbyanything else that can help?
21:26.52beewoolie-afkI'll let you know.  It'll be later today when I can look at this again.
21:27.04beewoolie-afkTrying to get an audio codec working ATM.
21:27.48landslide-zzzzjust heading into work.. i scrolled back, and understand the problem now.. :)
21:27.58landslide-zzzzi'll have a look at it when i return home tonight.
21:27.59beewoolie-afklandslide-zzzz: hey man
21:28.07landslide-zzzzthanks beewoolie... :)
21:28.12beewoolie-afkcheers
21:28.30beewoolie-afkrwhitby: BTW, does the BB reference the openslug_config?
21:28.46beewoolie-afkIt should do something like this:  make openslug_config
21:28.52beewoolie-afkwhen it is doing the setup.
21:32.42beewoolie-afklandslide-zzzz: Do you get the chip config now?  the 4x16Mx16 is 4 banks, 16M per bank, 16 bit access per chip
21:33.26prpplaguebeewoolie-afk: what does a "bx r0" do when r0=0x590000 ?
21:34.30beewoolie-afkI thought that the BX instruction was for switching to thumb mode.
21:34.47prpplaguethats what i read as well
21:35.08beewoolie-afkSays, tho, that when the low bit is 0, it is *not* a thumb branch.
21:35.23beewoolie-afkThat's a simple jump, but I couldn't tell you why it is being used.
21:36.02prpplagueok so if its not in thumb its going to do a simple jump to 0x590000 ?
21:36.14beewoolie-afkThat's what I'd guess.
21:36.17beewoolie-afkIt's an odd address.
21:36.22beewoolie-afkOdd as in wierd.
21:36.44prpplaguebeewoolie-afk: hehe yea they did alot of weird stuff on the juicebox
21:37.01beewoolie-afkHm..  Back to the juicer.
21:37.09beewoolie-afkJuiceroni
21:37.15beewoolie-afkthe jucierator
21:37.21prpplagueyea, trying to take short break
21:37.38prpplaguebeewoolie-afk: i've got so much work to do, and i'm about a week behind schedule
21:37.49prpplaguebeewoolie-afk: and to add to it, i'm totally unmotivated today
21:38.02beewoolie-afkNAP TIME!
21:40.16landslide-zzzzbeewollie: does that mean i can't access it?
21:40.33landslide-zzzzi configured the SDR_CONFIG register (memory selection) to 0x03.
21:40.41landslide-zzzzalong with the CAS3 bit.
21:45.05landslide-zzzzl8r
22:56.50*** join/#openjtag landslide (fwuser@61.29.47.66)
22:57.28landslidebeewoolie: what were you saying about the memory configuration?
22:57.46landslideis my 4x16M16 not going to work because of the 'word size'?
23:54.24*** join/#openjtag Tiersten (tman@Tiersten.nslu2-linux)

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with blootbot logs, split per channel, etc.