| 00:09.34 | beewoolie | [g2]: Well, it looks like the MMU code works. |
| 00:10.00 | beewoolie | [g2]: According to Voodoo_Z, the boot time is about half, but that includes an upgrade to version 3.1 of openslug as well. |
| 00:14.49 | lennert | beewoolie: you _are_ using an initial 1:1 map, are you? |
| 00:14.59 | beewoolie | Yep. |
| 00:15.03 | lennert | ok, good |
| 00:15.05 | beewoolie | I don't protect anything. |
| 00:15.14 | beewoolie | All I'm doing is adding C and B bits to the SDRAM. |
| 00:15.21 | lennert | i know kyllikki and fluff were struggling with ABLE for days until they did the 1:1 |
| 00:15.23 | beewoolie | Even flash remains uncacheable. |
| 00:15.37 | lennert | apparently it's very processor-dependent what happens after the mmu enable cp write |
| 00:15.38 | beewoolie | It's silly to do anything else. IMHO. |
| 00:15.43 | lennert | agreed |
| 00:16.02 | beewoolie | It doesn't seem to be that different between the xscale and armv4. |
| 00:16.18 | lennert | what armv4 core you using? |
| 00:16.24 | beewoolie | arm922. |
| 00:16.41 | beewoolie | I have arm7 cores, too. All of them work with the same code. |
| 00:16.54 | beewoolie | The xscale has only a couple of new things. The cache flush routin is different. |
| 00:17.47 | lennert | well, because you do 1:1 |
| 00:17.53 | lennert | so there is no ambiguity |
| 00:18.06 | beewoolie | Yeah, but what else could anyone do? Even the kernel goes 1:1. |
| 00:18.15 | lennert | if you don't do 1:1 it's apparently cpu-dependent on which cycle the mmu 'starts working' |
| 00:18.23 | lennert | well, ABLE did non-1:1 for a while |
| 00:18.32 | beewoolie | Oh, that. Well, it's not wise to depend on that. |
| 00:18.42 | lennert | sure, agreed |
| 00:18.51 | lennert | 1:1 should never be a problem |
| 00:18.56 | lennert | on any core |
| 00:19.00 | beewoolie | Besides, there are clever ways to make that irrelevent. I do so in APEX when I relocate. |
| 00:19.28 | beewoolie | I put the code in both places and jump to the copy. |
| 00:19.28 | lennert | that should work |
| 00:19.33 | lennert | does APEX support ixp2000? :) |
| 00:19.41 | lennert | guess not, huh |
| 00:19.48 | lennert | perhaps i should try and add that |
| 00:19.51 | beewoolie | I just cannot see any reason to put the code that is running in any other place when the MMU is enabled. |
| 00:19.59 | beewoolie | I don't have one of those. |
| 00:20.14 | beewoolie | It probably takes very little to make it work. |
| 00:20.15 | lennert | enp2611 uses silly redboot |
| 00:20.23 | beewoolie | Most of the coding has to do with SDRAM init. |
| 00:20.25 | lennert | just the sdram config, i think |
| 00:20.26 | lennert | yeah |
| 00:20.31 | beewoolie | Yeah. Redboot is a bear to work with. |
| 00:20.39 | lennert | s/bear/bee-yatch/ |
| 00:20.44 | beewoolie | The only thing it has on me is support for the NPE. |
| 00:20.48 | beewoolie | lol |
| 00:21.03 | beewoolie | At some point, I'll get fed up and do the work. |
| 00:21.06 | lennert | the next-gen intel npu, the ixp23xx, has NPUs too :-/ |
| 00:21.17 | beewoolie | Still, it's the licensing that gets my goat. |
| 00:21.31 | lennert | there's _somewhat_ more docs than for the ixp4xx, but still wildly insufficient |
| 00:22.05 | lennert | "to get one's goat".. sounds like a soviet countryside proverb ;) |
| 00:22.42 | [g2] | beewoolie cool! |
| 00:22.58 | [g2] | how long does it take for Voodoo_Z |
| 00:23.02 | beewoolie | [g2]: I'm adding the notes to my wiki. |
| 00:23.15 | beewoolie | [g2]: He says it took 31 seconds to get to the console prompt. |
| 00:24.19 | [g2] | from power on ? |
| 00:24.30 | beewoolie | Yea. |
| 00:24.42 | lennert | that's not very impressive... *ducks* |
| 00:24.55 | beewoolie | Note that openslug 3.1 doesn't put the kernel in jffs2. so all apex does is copy the kernel from flash and jump. |
| 00:25.01 | [g2] | lennert it isn't but it started at over 2 minutes :) |
| 00:25.11 | beewoolie | lennert: I agree, but most of the time is spent in booting the OS. |
| 00:25.33 | beewoolie | it used to take 30 seconds for the boot loader to get it's shit together. That is redboot. |
| 00:25.41 | [g2] | 20 |
| 00:25.44 | lennert | s/it's/its/ |
| 00:25.53 | lennert | redboot is kind of shite |
| 00:26.18 | lennert | i'm not very happy with it but my board comes with it |
| 00:26.35 | lennert | my strategy (for kernel code) is to trust the bootloader not more than i have to |
| 00:26.59 | [g2] | I'd like to replace it on the Loft, but since it all works it's not worth it yet |
| 00:27.05 | beewoolie | [g2]: I have an adorable habit of exaggeration |
| 00:27.25 | [g2] | beewoolie it was in the ball park |
| 00:27.29 | lennert | beewoolie: mine is worse |
| 00:27.31 | beewoolie | [g2]: Well, I think it would be more compelling if it had NPE support. |
| 00:27.44 | [g2] | the old boot time was around 40 seconds with APEX iirc |
| 00:28.03 | lennert | able is closed source so not an option |
| 00:28.10 | beewoolie | What is able? |
| 00:28.13 | lennert | redboot is open sourced but.. well.. ugh |
| 00:28.17 | lennert | ABLE is the simtec bootloader |
| 00:28.18 | beewoolie | yeah, I know. |
| 00:28.23 | lennert | it comes with simtec boards |
| 00:28.24 | beewoolie | Right. I saw that. |
| 00:28.29 | [g2] | and the Loft |
| 00:28.39 | [g2] | speaking of which |
| 00:28.42 | lennert | u-boot is not _too_ bad |
| 00:28.47 | beewoolie | I wrote APEX because blob was so bad as to be a hurdle for development. |
| 00:28.54 | [g2] | anyone got a xscale-elf LE tool chain around ? |
| 00:29.02 | beewoolie | I won't comment on uboot as it may tend to imcriminate me. |
| 00:29.12 | lennert | beewoolie: why so? :) |
| 00:29.13 | beewoolie | Not it! |
| 00:29.31 | lennert | beewoolie: death threats to the author, you mean? :) |
| 00:30.30 | beewoolie | :-) |
| 00:31.15 | beewoolie | Australians are nice. |
| 00:31.35 | lennert | what do australians have to do with it? :) |
| 00:42.29 | wookey_ | beewoolie: what was wrong with blob - I liked that loader |
| 00:42.41 | beewoolie | wookey_: Have you ever hacked on it? |
| 00:42.43 | wookey_ | nice and small and comprehensible |
| 00:42.46 | wookey_ | yes |
| 00:42.53 | beewoolie | comprehensible? |
| 00:43.01 | beewoolie | there is nothing comprehensible about automake. |
| 00:43.03 | wookey_ | with the diags separated out |
| 00:43.07 | beewoolie | Moreover, it's configuration sucks. |
| 00:43.21 | wookey_ | oh well, each to their own |
| 00:43.38 | beewoolie | It is hard to extend. It has ifdefs everywhere. It cannot be extended with drivers in a modular way. |
| 00:43.43 | wookey_ | we've been using bootldr recdently which is huge and confusing |
| 00:43.47 | beewoolie | Their command structure isn't orthogonal. |
| 00:44.04 | beewoolie | oops. |
| 00:44.11 | wookey_ | I wrote a review of available loaders about 3 years ago. probably time to look again |
| 00:44.35 | wookey_ | (I'd like to use something other than bootldr on ballloon3 if possible |
| 00:44.37 | beewoolie | I wrote APEX to cope with these things. It's smaller, simpler, has a consistent command syntax. |
| 00:44.46 | wookey_ | URL? |
| 00:44.53 | beewoolie | it doesn't have all sort of quirks that blob had. for example, it never modifies itself. |
| 00:44.57 | beewoolie | wiki.buici.com |
| 00:45.08 | wookey_ | I know a lot of people use uboot so it has wide support |
| 00:45.34 | wookey_ | whathardware does it cover? |
| 00:45.50 | wookey_ | (APEX, that is) |
| 00:46.15 | beewoolie | wookey_: APEX is supported on a IXP42x and the sharp processors. |
| 00:46.16 | wookey_ | I need a login for that wiki |
| 00:46.24 | beewoolie | You must be on a Macintosh. |
| 00:46.29 | wookey_ | nope |
| 00:46.34 | wookey_ | firefox on debian |
| 00:46.46 | wookey_ | on a thinkpad |
| 00:46.46 | beewoolie | really? It doesn't need one when I browse from other sites. |
| 00:46.52 | beewoolie | try to ignore the login. |
| 00:47.16 | wookey_ | ah yes - cancel it twice and I get through - odd |
| 00:48.38 | beewoolie | I don't like the wiki software I started with. I switched to mediawiki, but the site isn't running because it crashes the slug. |
| 00:49.38 | beewoolie | wookey_: Oh yeah, and blob is a PITA to build. |
| 01:03.29 | ka6sox-office | lennert: ping? |
| 01:03.40 | lennert | ka6sox-office: `pong |
| 01:03.50 | ka6sox-office | was somebody looking for me? |
| 01:04.01 | ka6sox-office | I saw you mention my name in vain :) |
| 01:04.50 | lennert | i think the bulgarian guy, velin something |
| 01:04.54 | lennert | he wanted to fatten his slsug |
| 01:05.00 | lennert | someone said you did that before |
| 01:05.44 | ka6sox-office | its a VERY hard proceedure and very delicate. |
| 01:05.54 | ka6sox-office | not something I want to repeat :( |
| 01:05.59 | lennert | if he's still there, maybe give him some hinys? |
| 01:06.01 | lennert | hints |
| 01:07.41 | beewoolie | ka6sox-office: he asked for email addresses. I told him he needed to lurk. |
| 01:08.43 | ka6sox-office | okay |
| 01:08.51 | ka6sox-office | I know a few bulgarians. |
| 01:08.57 | ka6sox-office | (software guys) |
| 01:15.56 | *** join/#openjtag vmaster_ (i=vmaster@p549B6EFA.dip.t-dialin.net) |
| 01:16.40 | beewoolie | ka6sox-office: Apparently, I was wrong about ep1220's board. There is nSRST line. |
| 01:20.49 | lennert | i speak a bit of bulgarian! |
| 01:22.18 | ka6sox-office | beewoolie-afk, I thought so! |
| 01:22.25 | ka6sox-office | he and I discussed it at length. |
| 01:26.35 | lennert | g'nite * |
| 01:27.04 | ka6sox-office | nite lennert |
| 01:31.15 | [g2] | sweet dreams of JTAG |
| 01:31.41 | ka6sox-office | yes |
| 02:28.06 | *** join/#openjtag ByronT (n=byron-po@nslu2-linux/ByronT) |
| 02:40.17 | beewoolie-afk | ka6sox-office: I'm still waiting on a resistor. The 487Ohm. |
| 02:41.00 | beewoolie-afk | ka6sox-office: I suppose that I could figure out a combination of the ones I've got to get that value. |
| 02:50.16 | ka6sox-office | parallel or Serial..it all works! |
| 03:17.03 | beewoolie-afk | right. I just have to figure out what makes sense. 2*243, 3*162... I don't have *every* other value. |
| 03:34.32 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
| 03:40.57 | *** join/#openjtag ka6sox (n=ka6sox@nslu2-linux/ka6sox) |
| 03:41.18 | ka6sox | morning |
| 03:54.20 | beewoolie-afk | ka6sox: did you just wake up? |
| 04:03.37 | ka6sox | something like that! |
| 04:19.13 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
| 05:55.31 | *** join/#openjtag nslu2-log (n=nslu2-lo@nslu2-linux/dyoung) [NETSPLIT VICTIM] |
| 05:55.34 | *** join/#openjtag wookey__ (n=wookey@217.147.92.89) |
| 05:56.06 | *** join/#openjtag DaKa (n=david@nslu2-linux/daka) [NETSPLIT VICTIM] |
| 05:56.16 | *** join/#openjtag ka6sox-office (n=tking@netblock-66-159-209-52.dslextreme.com) [NETSPLIT VICTIM] |
| 05:59.42 | *** join/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
| 06:00.51 | *** join/#openjtag toi (n=pleemans@d5152D12D.access.telenet.be) [NETSPLIT VICTIM] |
| 06:03.00 | velinp | hi all; ka6sox: you here? |
| 07:13.25 | ka6sox | velinp, ping? |
| 07:13.40 | velinp | hi |
| 07:13.51 | ka6sox | howdy...wazzup? |
| 07:14.49 | velinp | well, smd told me that you have maybe done a memory upgrade, replacing chips on an NSLU2; curious how strong the chips are glued |
| 07:15.11 | ka6sox | not bad but the traces are TINY. |
| 07:15.23 | ka6sox | I've been sucessful in one out of 4 attempts. |
| 07:16.05 | velinp | too bad; maybe I should go the additive way |
| 07:16.21 | ka6sox | might be the "only" way..its not fun however. |
| 07:17.04 | ka6sox | just curious..what do you want to do with the slug? |
| 07:17.23 | velinp | i dedicated it to armeb-buildd (wouter) |
| 07:17.54 | ka6sox | ah...is it online now? |
| 07:18.15 | velinp | yes; but ssh is firewalled; enable for you? |
| 07:18.34 | ka6sox | we have only keys enabled. |
| 07:18.41 | ka6sox | (on bob and wendy) |
| 07:19.14 | velinp | true; wouter had some problems, so he stopped the bdd; can't find hime since :) |
| 07:19.36 | ka6sox | ah....same on Bob...been idle for 8 days now. |
| 07:20.39 | velinp | wouter seems busy with fosdem, maybe the shine loft, too; mailed him 09.01 and 13.01 - no reply |
| 07:21.11 | ka6sox | okay last I spoke to aba he was looking for him too. |
| 07:21.39 | velinp | hope wouter will soon put efforts on bdd's |
| 07:22.22 | ka6sox | ya |
| 07:23.04 | ka6sox | I'm still working on getting something bigger (like a 64MB Nas100D running so we can eliminate some of the cc1plus swapping. |
| 07:23.31 | velinp | 64MB do not make a difference? |
| 07:24.00 | ka6sox | I haven't tried...I would need to ask [g2] about 64MB. |
| 07:25.31 | velinp | I have an idea: if I piggiback 2 larger chips (256 mbit instead of 128), could it be possible to use, say 96MB? |
| 07:27.31 | ka6sox | they have to be exactly the same size. |
| 07:27.48 | ka6sox | (the memory management only can deal with banks of the same size) |
| 07:30.22 | velinp | as far as I read the 256mbit specs, it seems in case the second set is larger, using part of it should not be a problem; or am i wrong? |
| 07:32.01 | ka6sox | the memory bank arraingement requires each bank to be identical. |
| 07:32.48 | ka6sox | so uneven banks will only respond as the smaller of the 2 banks. |
| 07:33.06 | ka6sox | (otherwise you will try to address memory that doesn't exist. |
| 07:35.01 | velinp | so if the second set is larger, it will only be partially useable, but will work? |
| 07:35.15 | ka6sox | yes |
| 07:35.27 | ka6sox | you have to configure the ram for 2 banks of the smaller chips. |
| 07:36.04 | velinp | when there are 2 banks, the MMU interleaves them? |
| 07:36.46 | ka6sox | I think that they are contiguous. |
| 07:36.52 | ka6sox | so the are "stacked" |
| 07:36.55 | ka6sox | not interleaved. |
| 07:38.01 | velinp | so, as the stock chips are selected by cs_1 (and are maybe the higher part), if hdw is configured for 128MB, but kernel is told to only use 96MB, it might work? |
| 07:40.05 | ka6sox | you would only get 64MB even if you used 32MB chips for the stacked set. |
| 07:40.17 | ka6sox | the memory manager can't handle different sized banks. |
| 07:40.42 | ka6sox | if you configure it for 128MB it will fail as you have a 32MB chunk off address space with NO RAM present. |
| 07:42.46 | velinp | true; i thought if the non-existent 32MB are at the end, kernel could avoid using them |
| 07:44.46 | ka6sox | I don't think that is the way the memory unit works..but I could be wrong. |
| 07:45.11 | ka6sox | everything I've read is that you *have* to have identical sized banks. |
| 07:46.34 | velinp | ok, thanks. I will try to talk my hdw friend to piggyback larger chips, read the specs, and come back about apex/redboot; I think it might not work beyond the 64MB, but 64 should work. |
| 07:47.37 | ka6sox | 64 will most definately work. |
| 07:47.38 | ka6sox | :) |
| 07:48.27 | velinp | so, maybe no luck with >64, but no harm <=64? |
| 07:52.00 | ka6sox | yes...I think that is the most reasonable cours. |
| 07:52.03 | ka6sox | er course |
| 07:53.41 | velinp | thanks, ka6sox; I go into readonly mode |
| 07:57.27 | ka6sox | its midnight here and I must sleep... |
| 07:58.03 | velinp | good night to you; it's 10AM here in BG |
| 07:58.21 | ka6sox | I was in .al last summer for a month |
| 07:58.45 | velinp | al == Albania? |
| 07:59.04 | ka6sox | yes |
| 07:59.52 | velinp | if you happen to come in BG, I would be glad to meet you; good night, sir :) |
| 08:00.15 | ka6sox | we were all set to help a girl get to Sofia but she met her father in Milan so we didn't have to. |
| 08:00.58 | ka6sox | night sir...cya laters. |
| 08:01.12 | velinp | buy, and thanks again |
| 08:01.22 | velinp | s/buy/bye/ |
| 08:01.43 | ka6sox | thanks purl |
| 08:01.46 | ka6sox | :) |
| 08:44.30 | *** join/#openjtag [g2] (n=Tom@cpe-066-057-008-035.nc.res.rr.com) |
| 09:54.19 | *** join/#openjtag NAiL (n=repvik@nslu2-linux/pdpc.active.repvik) |
| 13:41.27 | *** join/#openjtag nslu2-log_ (n=nslu2-lo@nslu2-linux/dyoung) |
| 15:53.49 | *** join/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
| 15:55.10 | *** join/#openjtag key2 (n=key2@gob75-2-81-56-64-17.fbx.proxad.net) |
| 15:55.17 | key2 | hi |
| 15:56.05 | key2 | lennert ? there ? |
| 16:18.01 | key2 | someone has ever ran a linux on a Xilinx XC2VP30 ? |
| 17:01.11 | *** join/#openjtag bullet (n=bullet@182.63.203.62.cust.bluewin.ch) |
| 17:48.57 | *** join/#openjtag cpage (n=cpage@netblock-68-183-40-90.dslextreme.com) |
| 18:25.58 | [g2] | key2 I've got a BlackDog with a Xilinx PPC core on it running Sarge |
| 18:39.01 | velinp | ka6sox: ping |
| 18:52.27 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
| 19:43.14 | key2 | g2: how many PPC is there in your Xilinx ? |
| 19:43.42 | key2 | g2: did you manage to do somethin interesting with it ? |
| 19:50.51 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
| 20:08.19 | *** join/#openjtag Tiersten (n=tman@nslu2-linux/Tiersten) |
| 20:58.04 | [g2] | key2 I think there's just one hardcore |
| 20:58.39 | [g2] | I haven't played with it a whole lot as I'm very busy with Lofts and stuff |
| 20:58.48 | [g2] | JTAG etc... |
| 20:59.44 | [g2] | I've been following softcores and hardcores on FPGAs for several years and it's just getting cheaper and more possible all the time |
| 21:00.23 | [g2] | I think that in 90% of the situations, they don't make sense from a cost side versus a specific chip and smaller FPGA/CPLD |
| 21:12.28 | *** join/#openjtag ulf_k (n=ulf_kypk@p54BDC0A8.dip0.t-ipconnect.de) |
| 21:30.51 | *** join/#openjtag ByronT (n=byron-po@nslu2-linux/ByronT) |
| 23:27.58 | *** join/#openjtag beewoolie-afk (n=beewooli@206.124.142.26) |
| 23:31.35 | beewoolie-afk | Tiersten: ping? |
| 23:33.24 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
| 23:33.37 | *** join/#openjtag ulf_k_ (n=ulf_kypk@p54BDADF6.dip0.t-ipconnect.de) |
| 23:42.11 | lennert | hello beewoolie |
| 23:47.45 | beewoolie | lennert: |
| 23:47.49 | beewoolie | Hey. |
| 23:47.57 | beewoolie | I have another circuit question. Care to try it? |
| 23:48.09 | lennert | sure :) |
| 23:48.15 | beewoolie | ;-) |
| 23:48.17 | lennert | it's like a quiz |
| 23:48.19 | beewoolie | http://wiki.buici.com/twiki/pub/Main/SimpleJTAG/jtag-schematic.pdf |
| 23:48.27 | beewoolie | that's the schematic for the JTAG dongle I built. |
| 23:48.31 | beewoolie | It looks like it works OK. |
| 23:48.48 | lennert | 403 Forbidden |
| 23:48.59 | beewoolie | I don't understand that. |
| 23:49.05 | beewoolie | You can fetch it from another place... |
| 23:49.07 | lennert | (with wget) |
| 23:49.27 | beewoolie | ftp://ftp.buici.com/pub/jtag/jtag-schematic.pdf |
| 23:49.35 | lennert | firefox also gives a 403 |
| 23:49.42 | beewoolie | From ftp? |
| 23:49.44 | lennert | maybe you don't allow lithuanian IPs :) |
| 23:49.47 | lennert | trying ftp now |
| 23:49.57 | beewoolie | I need to look into that. It seems to work sporadically. |
| 23:50.15 | lennert | okay, got it |
| 23:50.47 | beewoolie | OK. So the transisitor drives the nTRST line. |
| 23:51.05 | lennert | i see that |
| 23:51.15 | beewoolie | Trouble is, the signal is really sloppy. |
| 23:51.38 | beewoolie | Based on my inspection of the circuit, I believe that the transistor is inverting the signal. |
| 23:51.40 | beewoolie | Is that right? |
| 23:52.04 | beewoolie | Or, is it just amplifying it? |
| 23:52.23 | lennert | let me check.. |
| 23:52.29 | beewoolie | I wonder because I'm driving that signal low and the TAP is working. |
| 23:53.31 | lennert | well |
| 23:53.42 | lennert | the base has a pulldown |
| 23:53.51 | beewoolie | I suppose I can test it OK. |
| 23:53.53 | lennert | so normally the base is at GND and it won't be conducting |
| 23:53.54 | beewoolie | Right. |
| 23:54.04 | beewoolie | Hmm. So, it's just amplifying. |
| 23:54.08 | beewoolie | I wonder why this is working. |
| 23:54.12 | lennert | well |
| 23:54.15 | beewoolie | There are really two questions. |
| 23:54.21 | lennert | you have to write an 1 to pull nTRST to ground |
| 23:54.28 | lennert | i assume that nTRST is normally pulled high |
| 23:54.31 | lennert | i.e. it's inverting |
| 23:54.35 | lennert | (i think :) |
| 23:55.21 | lennert | what signal is sloppy? |
| 23:57.03 | beewoolie | The output of that transistor. |
| 23:57.18 | beewoolie | I need to get a power strip so that I can look at it again... |
| 23:57.46 | lennert | i'm looking at a 2n3904 datasheet which says the prop delay is 35ns and the rise/fall are 35/50 ns respectively |
| 23:58.07 | lennert | does that match what you saw? |
| 23:58.24 | beewoolie | lennert: What I've been concerned about is whether or not it is driving the line, or simply grounding it when the base is ground. |
| 23:58.30 | beewoolie | Let me try again... |
| 23:58.45 | lennert | when the base is driven _high_, the nTRST line is shorted to ground |
| 23:58.57 | lennert | when the base is undriven (pulldown), the nTRST line floats |