**** BEGIN LOGGING AT Fri Jan 13 02:59:57 2006 Jan 13 05:45:02 heyho :) Jan 13 05:45:26 do you have any working patchset + config for a 2.6.15er kernel for the nslu2? Jan 13 05:47:00 Yep. Jan 13 05:47:12 its in the repo. Jan 13 05:47:23 are you familiar with the Master Makefile Jan 13 05:47:24 ? Jan 13 05:48:13 hmm nope, im not Jan 13 05:48:21 im using gentoo on my slug Jan 13 05:48:39 google for MasterMakefile. Press "I feel lucky" ;-) Jan 13 05:49:10 currently, i run a 2.6.14.3, (1mb in the nslu flas the rootfs and everithing else on the hdd) Jan 13 05:49:49 eh.. wrong Jan 13 05:49:59 only the kernelimage is in the flash Jan 13 05:50:07 and anything else on the hdd Jan 13 05:50:43 http://www.nslu2-linux.org/wiki/Development/MasterMakefile Jan 13 05:51:10 once you do the checkout, you can look at the patches and defconfig Jan 13 05:54:15 what do you mean with checkout? whouldnt a "make openslug-kernel" be enough? Jan 13 05:54:32 If you just want to look at it, just do make setup Jan 13 05:54:46 make openslug-kernel will make the kernel. Jan 13 05:55:02 if thats what you really want then yes. :-) Jan 13 05:55:37 but if youre after the internals, make setup will run the monotone checkout to bring the files from the database into human readable format. Jan 13 05:55:42 then you can inspect the, Jan 13 05:55:45 them. Jan 13 05:56:02 na, i dont wanna build the kernel :) just wanna see the patches Jan 13 05:56:10 thanks for your help, man! Jan 13 05:56:11 :) Jan 13 05:57:18 You will wantn to look in openembedded/packages/linux/nslu2-kernel/2.6.15 Jan 13 05:57:43 that is what i just wanted to ask you :) Jan 13 05:59:06 How did you get Gentoo on there ? Jan 13 05:59:44 without any black magic Jan 13 06:00:09 it was straight forward. you speak germen? Jan 13 06:00:43 My german is pretty rusty. I am aware of the Germal forum post though. Jan 13 06:00:56 You basically followed that howto? Jan 13 06:01:02 i wrote it ;) Jan 13 06:01:08 Woo! Jan 13 06:02:18 Would you mind translating the information and/or putting a link (someplace) on the nslu2-linux.org wiki ? Jan 13 06:02:24 when the final step is done (compile the ixp4xx drivers nativeley), then the howto is complete and could be translated into english Jan 13 06:02:41 yep, i can of course Jan 13 06:02:50 Ok thanks ! Jan 13 06:03:10 but ive 2 problems, wich make the whole thing .... some kind of crappy Jan 13 06:03:14 native compilation of the ixp4xx is another story. Jan 13 06:03:32 but maybe, you can help me ;) Jan 13 06:03:33 there is a whole set of patches for that to make it work under 2.6.15 . Jan 13 06:03:55 hmm Jan 13 06:04:04 for that you will look in openembedded/packages/ixp4xx Jan 13 06:04:29 vapier (a gentoo dev) wrote a ebuild (a gentoo package) for that ixp4xx Jan 13 06:05:02 so you can simply emerge it by "emerge ixp4xx" Jan 13 06:05:20 but this only works on x86, crosscompiling for arm Jan 13 06:05:33 it doesnt work on an arm for an arm Jan 13 06:05:49 I'm not at all familiar with how emerge works. Jan 13 06:06:06 I am only qualified to tell you where to find these patches. Jan 13 06:06:08 :-) Jan 13 06:06:55 the best way would be to have no extra package for that driver, but a patchset for the kernel, and let the user choose in make menuconfig if he wants to compile it with his kernel or not Jan 13 06:07:13 You cant. Jan 13 06:07:17 It must be a module. Jan 13 06:07:25 Oh you said with. Jan 13 06:07:27 I read in. Jan 13 06:07:31 yep. and where is the problem? Jan 13 06:08:22 just to have a "markable" selection in make menuconfig for the ixp4xx Jan 13 06:08:46 and when that kernel compiles fine, nativeley, almost everithing is done Jan 13 06:08:56 almost.. Jan 13 06:09:07 should be doable; just some additioanl stuff to get the ixp4xx sources and some extra Kconfig stuffs. Jan 13 06:09:31 there was a second point.. i wanted to tell you 5 minutes ago.. but i forgot it.. :D Jan 13 06:09:52 hmm damn Jan 13 06:10:15 aaahh ) Jan 13 06:11:22 that snippet, i got from a gentoo-dev: http://chilla.biz/projekte/gentoo-slug/files/prepimage Jan 13 06:11:42 my kernel didnt boot up, and he told me to use that snippet... i did, and it worked Jan 13 06:12:04 but i dont have any fucking idea, what this does :) Jan 13 06:12:27 and the person who gave me this, is not online these days Jan 13 06:13:40 that is a kernel prepend. Jan 13 06:13:45 Let me verify that. Jan 13 06:15:56 eek g2build monotone to view your kernelpatches :) Jan 13 06:18:48 I'm too tired to explore this fully. Jan 13 06:19:07 I'm pretty sure thats the bit to set the machine id properly. Jan 13 06:19:32 because the bootloader pretends the nslu2 is a ixdp425; while its really machine id nslu2 (597). Jan 13 06:19:41 so you trick it by prepending it. Jan 13 06:19:55 ah okay Jan 13 06:20:00 thanks for the info :) Jan 13 06:20:10 np Jan 13 06:20:19 setting the machine id is critical. Jan 13 07:04:13 morning Jan 13 07:04:54 damn, I always miss the good discussion. It seems I'm all alone in this freaking timezone. :( Jan 13 07:07:04 darn I feel rusty with the whole OE thing already. Jan 13 07:07:48 I want to override the default R/DEPENDS (ex: removing reiserfs & others, adding my webcam drivers, etc..) Jan 13 07:08:19 Where is the best place to put this? I use to simply override the default in my local.conf but it doesn't seem to pick it up. Jan 13 07:08:29 Do I have to change the system openslug.conf?? Jan 13 08:42:41 <[g2]> AH velinp hello! Jan 13 08:43:03 hi, [g2] Jan 13 08:43:10 <[g2]> I'm glad you stopped back Jan 13 08:43:30 <[g2]> I missed you in #openjtag several days ago about the FATSLUG Jan 13 08:43:54 my mail is vel@netissat.bg; go offline? Jan 13 08:45:25 [g2], Is it normal for apex to take a few seconds caching JFFS2 fs on boot? I didn't notice any major speed improvement vs redboot. Jan 13 08:45:58 <[g2]> velinp I guess you didn't see that PM Jan 13 08:46:17 <[g2]> VoodooZ_Work yeah as JFFS2 is compressed Jan 13 08:46:17 PM? Jan 13 08:46:22 <[g2]> Private message Jan 13 08:46:28 <[g2]> your nick is nod registered Jan 13 08:46:32 <[g2]> s/nod/nick/ Jan 13 08:46:33 [g2] meant: your nick is nick registered Jan 13 08:46:59 I think so, but am a newbie on irc Jan 13 08:47:00 <[g2]> s/is nick/is not/ Jan 13 08:47:00 [g2] meant: your nick is nod registered Jan 13 08:47:33 <[g2]> velinp type "/msg [g2] foo Jan 13 08:47:34 [g2], darn. I guess I'm stuck with redboot for good then. do you use jffs2 or ext2 on the loft? Jan 13 08:48:01 [g2]: I did see you message, another tab Jan 13 08:48:18 <[g2]> velinp then type in that tab :) Jan 13 08:48:39 <[g2]> VoodooZ_Work I think ext2 loads faster Jan 13 08:49:04 [g2]: need to register; will take me a while Jan 13 08:49:24 ok. I guess I'll have to optimize openslug. Jan 13 08:50:34 <[g2]> VoodooZ_Work you just need a board right ? Jan 13 08:51:00 A board? Jan 13 08:51:22 loft you mean? Jan 13 08:51:29 <[g2]> Yeah, you just use the NSLU2 board right Jan 13 08:51:34 yes. Jan 13 08:52:06 <[g2]> and if you had a Loft board you would just use the board Jan 13 08:52:14 yes Jan 13 08:54:09 rwhitby, I'm thinking of going with your roboslug.conf idea. do I just clone openslug.conf basically and modify? Jan 13 08:56:28 [g2]: I've managed to free some time from uni to do some payed work, so I'll probably get a loft in februari/march :) Jan 13 08:57:07 <[g2]> koen cool! Jan 13 08:57:42 533mhz/62mb, right? Jan 13 09:04:04 <[g2]> 533/64MB Dual Ethernet, 3 MiniPCI Jan 13 09:04:25 <[g2]> 1 DB-9 Serial, one on board, JTAG Jan 13 09:04:33 <[g2]> I2C, 4-5 GPIOs Jan 13 09:04:42 <[g2]> running 2.6.15 Jan 13 09:05:09 <[g2]> I'm currently testing iptables/masq'ing Jan 13 12:38:59 damn, my power-supply died. Thank god I wasn't flashing my slug! :) Jan 13 12:40:34 VoodooZ_Work: you need to either copy/modify openslug.conf (if you want to remove things) or copy/modify ucslugc.conf and change 'slugos-btu' to 'slugos-bag' if you want to add things. Or you could just write it from scratch based on the existing conf files. Jan 13 12:41:28 yeah, I just copied openslug.conf from distro/ to roboslug.conf and modified stuff accordingly. Jan 13 12:41:45 The only problem is that it was checking all 3000 OE packages. Jan 13 12:42:20 I might be switching to ucslugc soon too so I guess I can just use the proper include. Jan 13 12:42:40 It makes it nice and tidy for my changes. me like! Jan 13 12:42:53 I'll try to flash it now... Jan 13 12:42:56 * koen|tv should try BTU soon Jan 13 12:45:26 * VoodooZ_Work is flashing... Jan 13 12:46:25 I wish I could use a longer nickname: it would be VoodooZ@Work_but_working_on_personal_stuff hehehehe Jan 13 12:49:28 yipee! Roboslug-1.0 is born! Jan 13 12:53:53 :) Jan 13 12:54:13 does is use the brain of a police officer? Jan 13 12:54:23 or from a decapitated snail? Jan 13 12:55:40 heheh Jan 13 12:56:40 cool. it seems the bootup time of openslug-3.1 is faster somehow. it's 45seconds total with the annoying Redboot delay. Not too bad and I haven't started cleaning yet. Jan 13 13:01:51 VoodooZ_Work: Howsitgoin. Jan 13 13:06:12 not bad. Jan 13 13:06:22 I loaded apex earlier and it worked. Jan 13 13:06:50 but it seemed to take as long as redboot to load openslug. Jan 13 13:07:01 It caches the jffs2 fs for a long time. Jan 13 13:07:05 [g2] said it was normal Jan 13 13:19:10 VoodooZ_Work: ah. I've got a fix for that. Jan 13 13:19:22 I haven't told [g2] about it. Jan 13 13:19:36 It looks like if I enable caches, I can get the jffs2 load to go faster. Jan 13 13:19:51 good! Jan 13 13:19:55 I cannot do anything about the algorithm because that's just the way that the filesystem works. Jan 13 13:20:20 like [g2] said I could just use ext2 rootfs but that would probably take more space right? Jan 13 13:20:27 There is code to enable the cache in the latest version of APEX. I haven't made the ixp do the MMU/cache trick. Jan 13 13:20:47 Yeah. it also isn't good to put ext2 into flash. It will work, but it isn't considered safe. Jan 13 13:20:57 yeah. Jan 13 13:21:07 So many little tasks... Jan 13 13:21:10 So you are saying i can't enable that trick yet? Jan 13 13:21:14 Right. Jan 13 13:21:27 I need to add some code to the platform code for the IXP so that it can handle the MMU enable. Jan 13 13:21:44 that's ok for now. I'll be working on tweaking openslug in the meantime. Jan 13 13:21:50 Good. Jan 13 13:22:01 thanks for trying it. Jan 13 13:22:06 No problem. Jan 13 13:22:27 Just let us know when you make some changes or update your page as I keep checking it out. Jan 13 13:22:41 And I'll be happy to beta-test it for you. Jan 13 13:25:14 K. Jan 13 13:25:25 It isn't really hard, I just need to test it myself. Jan 13 13:29:19 probably just a matter of figuring out to init the MMU right? Jan 13 13:32:19 VoodooZ_Work: Yeah. I'm doing a very simple setup. There is no protection, all addresses are accessible. Jan 13 13:32:36 The only thing I'm using it for is the cachealbility/bufferability. Jan 13 13:32:53 I need to read the IXP programmers guide so that I can setup the tables correctly. Jan 13 13:33:09 ...and make sure I use the right CP15 commands to get things going. Jan 13 13:33:51 I recently added interrupt support, too. It makes the timer code more reliable and means that I can add interrupt handles for some of the IO if it makes the code easier. Jan 13 13:38:46 nice. Jan 13 13:41:09 VoodooZ_Work: Looks like it may be easier than I thought. Jan 13 13:41:18 They use the same MMU as the ARM standard. Jan 13 13:43:35 good stuff Jan 13 13:44:09 * VoodooZ_Work crosses his finger and hopes it will be real speedy Jan 13 13:46:29 do you want to try a build with the MMU enable? I mean just in RAM. Jan 13 13:46:43 VoodooZ_Work: It builds, so the only thing to see is if it runs. Jan 13 13:47:15 I *could* run a test, but it will not be until next week that I'll be able to bust out the hardware. Jan 13 13:47:37 hmmm Jan 13 13:47:40 mem=32M@0x00000000 Jan 13 13:47:49 didn't the loft have 64MB ram? Jan 13 13:47:52 of course! Jan 13 13:48:08 OK. I'll put it on my ftp server and you can download it... Jan 13 13:48:10 just give me the code Jan 13 13:48:17 the bin file? Jan 13 13:48:45 Right. Jan 13 13:48:50 ok. Jan 13 13:50:30 ftp://ftp.buici.com/pub/apex/apex-openslug-mmu.bin Jan 13 13:51:20 koen|tv: IIRC, the problem is getting redboot to change it's understanding of memory. Jan 13 13:51:34 ok. Jan 13 13:54:14 beewoolie: ah, I remember something like that from the fatslug stories Jan 13 13:55:18 beewoolie, seems faster but it fails: # copy jffs2:/boot/zImage 0x00008000 Jan 13 13:55:18 Caching jffs2 filesystem Jan 13 13:55:18 1128 directory nodes 3549 inodes nodes Jan 13 13:55:18 36 bytes transferred Jan 13 13:55:19 # wait 10 Type ^C key to cancel autoboot. Jan 13 13:55:21 Type ^C key to cancel autoboot. Jan 13 13:55:23 # boot Jan 13 13:55:25 Booting kernel at 0x00008000... Jan 13 13:55:27 $T0a0f:0026b394;0d:00214234;#53 Jan 13 13:55:34 That's a crash. Jan 13 13:55:44 can I type stuff? Jan 13 13:55:46 Does it cache the Jffs2 faster? Jan 13 13:55:58 seems to but hard to say without a full boot to time it Jan 13 13:56:15 I am probably not disabling the MMU properly. Jan 13 13:56:22 That should be easy to figure out. Jan 13 13:56:30 Let me check the code. Jan 13 13:56:36 that or the new openslug-3.1 beta is no longer compatible with apex. Jan 13 13:56:49 No. You aren't even getting there. Jan 13 13:56:56 As i tested apex with openslug-2.7 yesterday and now I'm using 3.1 Jan 13 13:56:59 you should at least see the kernel decompression. Jan 13 13:57:00 ok. Jan 13 13:57:08 true. Jan 13 13:57:15 Let me do some looking. Jan 13 13:57:38 can I reset from there or do I have to power down physically? (I'm working remotely.) Jan 13 13:57:55 worst-case I can try it once I get home. Jan 13 13:58:22 You'll need to cycle power. Jan 13 13:59:00 ok. Jan 13 14:00:04 This is where my new JTAG code would come in really handy. Jan 13 14:00:11 If it were done. Jan 13 14:00:18 hehe Jan 13 14:03:50 I think I found it. Jan 13 14:05:59 * VoodooZ_Work jumps up and down screaming Jan 13 14:07:45 let me know when the new load is ready and I'll download for testing. (I'm going home in less than 45 minutes.) Jan 13 14:10:12 VoodooZ_Work: OK. It looks like the problem is the data cache clean code. The routines for the ARM9 don't work on the XScale. Poop. Jan 13 14:10:23 darn Jan 13 14:17:53 VoodooZ_Work: Almost there, Jan 13 14:18:07 great. thanks for your time. Jan 13 14:20:55 OK, try another. Jan 13 14:21:01 Same location Jan 13 14:21:32 Turns out that the Xscale cache cleaning code is much simpler than the arm9 version. Jan 13 14:22:21 good. Jan 13 14:22:47 I've downloaded the file but I'll only be able to test it out once I get home (around 5pm est). Jan 13 14:23:06 I'll post here either way to let you know. Jan 13 14:23:57 OK. I'll check the logs if I'm not around when you get back online. Jan 13 14:24:00 Thanks for the help. Jan 13 14:24:28 perfect. Thanks for doing this. Jan 13 15:03:42 beewoolie, nope. didn't work Jan 13 15:03:51 VoodooZ: Same error? Jan 13 15:04:16 similar yes Jan 13 15:04:30 wait... Jan 13 15:04:36 OK. Jan 13 15:04:50 i got:Caching jffs2 filesystem Jan 13 15:04:51 1128 directory nodes 3549 inodes nodes Jan 13 15:04:51 36 bytes transferred Jan 13 15:04:51 # wait 10 Type ^C key to cancel autoboot. Jan 13 15:04:51 Type ^C key to cancel autoboot. Jan 13 15:04:52 # boot Jan 13 15:04:54 Booting kernel at 0x00008000... Jan 13 15:04:56 $T050f:00008000;0d:00214200;#de Jan 13 15:05:03 booting kernel? Jan 13 15:05:14 I think I'll have to spend more time on this. I'll have to test on m hardware. Jan 13 15:05:36 but didn't it get further this time? booting kernel? Jan 13 15:05:40 I think that means that it is crashing at the first insruction Jan 13 15:06:06 I'm just wondering again if the new openslug is different somehow. Jan 13 15:06:08 I suspect that the cache isn't flushing properly, but I'll need to look more closely. Jan 13 15:06:18 ah Jan 13 15:06:28 darn Jan 13 15:06:46 I don't know the details, but I suspect that this $T050f:00008000;0d:00214200; means that it is failing at 0x80000 which is the start of the kernel. Jan 13 15:06:57 I can look up the GDB protocol. Jan 13 15:07:29 Wow, over my head big time. I never heard of GDB Jan 13 15:07:33 I'll let you know when I can look at it in more detail. Jan 13 15:07:42 GDB? GNU Debugger. Jan 13 15:07:58 I believe that that string is a GDB debug protocol emission. Jan 13 15:08:51 Too bad. I was hoping for an easy swish. Jan 13 15:13:23 yeah Jan 13 15:20:44 beewoolie, anyways, thanks for trying and let me know if you need a guinea pig to test it. :) Jan 13 15:21:04 K. I'm rereading the guide. I don't see it yet. Jan 13 15:25:15 let me know if I can help. Jan 13 15:31:31 beewoolie, hey! Jan 13 15:31:37 Yo. Jan 13 15:31:42 Still reading. Jan 13 15:31:50 I just reflashed my slug with the older openslug and apex now works! Jan 13 15:32:04 So my theory might be right after all. Jan 13 15:32:12 Perhaps they changed openslug 3.1 Jan 13 15:32:14 Huh? Jan 13 15:32:30 Are you saying that the MMU enable APEX works with the older openslug? Jan 13 15:32:36 yep. Jan 13 15:32:43 I'll try again but it boots. Jan 13 15:32:51 in fact I just logged in. Jan 13 15:32:52 Hmm. The only explanation I can see is that the kernel you are loading is invalid. Jan 13 15:32:59 ...in the new openslug. Jan 13 15:33:08 well, openslug-3.1 boots fine without apex. Jan 13 15:33:11 Does openslug 3.1 work with redboot? Jan 13 15:33:28 yep Jan 13 15:33:47 The thing is this. The only reason it wouldn't show the kernel decompression would be that the kernel as loaded was invalid. Jan 13 15:33:56 OK. Let's go back a step. Is it faster? Jan 13 15:34:19 it seemed faster but it went by too fast for me to time it. so I'll try again... Jan 13 15:34:40 OK. A measured time value will be helpful. Jan 13 15:35:32 the problem is I don't have a stop watch so it's not exact. Jan 13 15:35:44 Hmm. Jan 13 15:36:31 wow! Jan 13 15:36:42 There is one other thing that has bothered me. I don't cache the flash, so reads from flash are still kinda slow. Jan 13 15:36:49 what do you see? Jan 13 15:36:51 37secs from 'g 0x1000000' to openslug login prompt! Jan 13 15:37:01 What did it used to be? Jan 13 15:37:14 Or, better yet, how does that compare to the redboot boot. Jan 13 15:37:24 well, the last time i tried it I had a modified version but it was around 45-50secds Jan 13 15:37:38 Hey! What a minute. The redboot cannot read from jffs2, right? Jan 13 15:37:47 So, you are running a different kernel in these cases. Jan 13 15:38:10 Six seconds off isn't bad. Jan 13 15:38:17 I was hoping for something more. Jan 13 15:40:05 oops. that time was for redboot/openslug-2.7 Jan 13 15:40:05 VoodooZ: will you do another test for me? Jan 13 15:40:12 sorry I was on the phone... Jan 13 15:40:15 yep. Jan 13 15:40:36 I'd like to see the output of 'version'. Will you type ^c before it boots so that we can verify the system state. Jan 13 15:40:42 Then, run the version command. Jan 13 15:40:48 I'm sorry I can't really time exactly especially when counting just the caching. Jan 13 15:40:58 with openslug-2.7 or 3.1 Jan 13 15:40:59 VoodooZ: NP. Jan 13 15:41:13 I just care about the apex-1.3.13. Jan 13 15:41:20 ok. Jan 13 15:41:24 let me load it. Jan 13 15:41:25 I just want to make sure it is enabling the MMU. Jan 13 15:42:02 ok Jan 13 15:42:13 unless fedora core4 has a stopwatch built-in? Jan 13 15:44:02 ok. I did 'g 0x1000000' and CTRL-C immediately after. Jan 13 15:44:09 I got the apex> Jan 13 15:44:16 OK version Jan 13 15:44:40 APEX Boot Loader 1.3.13 -- Copyright (c) 2004,2005 Marc Singer Jan 13 15:44:51 want the rest too? Jan 13 15:45:10 Please. Jan 13 15:45:23 apex => mem:0x00200000+0xc2fc (49916 bytes) Jan 13 15:45:23 env => nor:128k+64k (no-write) Jan 13 15:45:23 *** No SDRAM init when APEX executed from SDRAM. Jan 13 15:45:23 memory: 0x0 0x02000000 (32 MiB) Jan 13 15:45:23 nor: 8MiB total 32B write buffer Jan 13 15:45:24 region 0: 64 blocks of 131072 (0x20000) bytes Jan 13 15:45:28 cpu: id 0x690541f1 ctrl 0x1aff cpsr 0x200000d3 Jan 13 15:45:44 OK. CTRL is the key. Jan 13 15:45:53 uh? Jan 13 15:46:13 oh, you mean the ctrl number? Jan 13 15:46:33 hang on. otp Jan 13 15:50:42 rwhitby, jbowler, are you aware of anything that could have changed from openslug 2.7 -> 3.1 that would prevent apex from load the kernel image? Jan 13 15:51:03 <[g2]> beewoolie can you think of any reason Linux wouldn't run XIP ? Jan 13 15:51:29 what's XIP again. Jan 13 15:51:29 [g2]: you need to compile it for that, *and* you have to have support in the filesystem. Jan 13 15:51:37 Execute in Place Jan 13 15:51:42 VoodooZ: nope Jan 13 15:51:44 eXecute In Place Jan 13 15:51:52 ah. I thought i read that somewhere. Jan 13 15:51:55 <[g2]> and have JTAG support Jan 13 15:51:58 VoodooZ: The key is whether you are running the same kernel. Jan 13 15:52:00 rwhitby, ah. ok. Jan 13 15:52:14 Isn't the kernel in two places? Jan 13 15:52:20 not sure. Jan 13 15:52:31 One in flash, plain, and one in the jffs2 partition. Jan 13 15:52:37 redboot only reads the former. Jan 13 15:52:51 well, openslug 2.7 and 3.1 have different kernels so yes they are differnet: 2.6.12.2 to 2.6.15 Jan 13 15:52:51 If the data is different, you won't be testing the same kernel in both cases. Jan 13 15:53:03 oh! Jan 13 15:53:18 I just did the same flashing procedure I always did before. Jan 13 15:53:39 if somebody changed the way openslug works I don't know. Jan 13 15:54:08 If the layout is different, that could be an issue, too. Jan 13 15:54:10 but rwhitby would know i'm sure. Jan 13 15:54:26 What we need is the flash memory map for 3.1 Jan 13 15:54:30 layout is same Jan 13 15:54:42 kernel is only in flash Jan 13 15:54:45 VoodooZ: btw, the MMU is enabled on your slug. Jan 13 15:54:51 good. Jan 13 15:54:53 VoodooZ: how does it fail? Does APEX fail or does the boot fail? Jan 13 15:54:55 it seemed faster Jan 13 15:55:08 Ah. Then apex needs to read the kernel from flash instead of jffs2. Jan 13 15:55:11 I get the error i posted ealier Jan 13 15:55:22 jbowler: it crashes when it jumps to the kernel. Jan 13 15:55:26 wouldn't that be faster anyways? Jan 13 15:55:32 VoodooZ: Sure. Jan 13 15:55:37 MMU will still make it faster. Jan 13 15:55:44 nice Jan 13 15:55:51 is this something apex can do? Jan 13 15:55:53 That means we need to change where the kernel is copied from. Jan 13 15:55:55 Yeah Jan 13 15:55:56 VoodooZ: so the kernel is crashing, or APEX? Jan 13 15:56:27 jbowler: As I wrote, it is crashing when it jumps to the kernel. The first instruction of the kernel causes the system to fail. Jan 13 15:56:33 not sure. beewoolie what do think? Jan 13 15:56:52 The first instruction is a register-register instruction. Jan 13 15:57:22 I.e. if it's an instruction crash it isn't the first instruction in the kernel image. Jan 13 15:57:24 perhaps we just need to update the old openslug config for apex? Jan 13 15:57:46 VoodooZ: Right. Jan 13 15:58:00 But is it a trivial change? Jan 13 15:58:10 VoodooZ: as you can see fromthe start of apex, it cannot write it's environment. Jan 13 15:58:14 as in: does apex support loading the kernel from flash directly? Jan 13 15:58:20 You'll have to recompile to make the change permanent. Jan 13 15:58:28 not aproblem. Jan 13 15:58:30 Sure. That's the easy part. Jan 13 15:58:36 Hang on... Jan 13 15:58:44 cool Jan 13 15:59:40 you can just ship me a patch file too. Jan 13 15:59:40 You need to change the kernel region string in the configuration. Jan 13 15:59:50 env.c? Jan 13 16:00:03 No, you use the configuration tool. Jan 13 16:00:22 It says something about jffs2 right now. Change it to nor:0x60010+0xffff0 Jan 13 16:00:25 make xconfig Jan 13 16:00:31 yep Jan 13 16:01:11 searching... Jan 13 16:01:34 under which heading? Jan 13 16:01:48 Environment. Jan 13 16:02:00 Source region for kernel. Jan 13 16:02:57 ok. change it to nor:0x60010+0xffff0 Jan 13 16:03:09 rite. Jan 13 16:03:55 sorry, I'm dong this remotely so it's slow. Jan 13 16:04:04 that's it? Jan 13 16:04:07 NP. I made the change here. Jan 13 16:04:12 Do you want to try the binary? Jan 13 16:04:15 ok. Jan 13 16:04:19 with MMU? Jan 13 16:04:24 Right. Jan 13 16:04:43 Updated in the same place. Jan 13 16:04:44 I'll get the updated source later if it works. Jan 13 16:04:48 ok. Jan 13 16:04:55 If this works, I'll post a new version of apex. Jan 13 16:05:10 first I'll have to reflash openslug-3.1 ... Jan 13 16:05:16 NP. Jan 13 16:05:22 I'll be making a snack. Jan 13 16:05:58 * beewoolie eats Creme Brulee Jan 13 16:06:15 what was the ftp address again? (I'm home now) Jan 13 16:06:28 Creme-brulee? nice! Jan 13 16:06:45 * VoodooZ 's stomach makes noises Jan 13 16:07:30 ftp://ftp.buici.com/pub/apex/apex-openslug-mmu.bin Jan 13 16:10:25 ok, preparing to run... Jan 13 16:10:37 bang! it works! Jan 13 16:10:58 Nice. Jan 13 16:11:06 What was happening was the kernel copy was failing. Jan 13 16:11:08 I'll wait 'till it's booted and I'll try again for timing. Jan 13 16:11:13 Nice. Jan 13 16:11:20 i see. Jan 13 16:11:25 oops. Jan 13 16:11:27 error Jan 13 16:11:30 damn Jan 13 16:11:40 later in the boot: Jan 13 16:11:59 Unless the MMU thing messes it up? Jan 13 16:12:14 ixp400_eth: Use MAC address 00:04:5a:0f:94:ae for port 0 Jan 13 16:12:14 Unable to handle kernel paging request at virtual address d897e1a0 Jan 13 16:12:14 pgd = c1f9c000 Jan 13 16:12:14 [d897e1a0] *pgd=00000000 Jan 13 16:12:14 Internal error: Oops: f5 [#1] Jan 13 16:12:14 Modules linked in: ixp400_eth ixp400 Jan 13 16:12:16 CPU: 0 Jan 13 16:12:18 pc : [] lr : [] Tainted: P Jan 13 16:12:20 sp : c1f65ad0 ip : 3178c868 fp : 00000080 Jan 13 16:12:22 r10: 00000000 r9 : bf07f47c r8 : c1f65ae8 Jan 13 16:12:24 r7 : 09ad1dc0 r6 : 49ad1e19 r5 : d897e160 r4 : 00000001 Jan 13 16:12:26 r3 : 09ad1e00 r2 : 09ad1dc0 r1 : d897e160 r0 : 00000001 Jan 13 16:12:28 Flags: nzCv IRQs on FIQs on Mode SVC_32 Segment user Jan 13 16:12:30 Control: 39FF Table: 01F9C000 DAC: 00000015 Jan 13 16:12:32 Process ifconfig (pid: 1123, stack limit = 0xc1f64194) Jan 13 16:12:34 Stack: (0xc1f65ad0 to 0xc1f66000) Jan 13 16:12:53 Doubtful that the kernel copy is broken. Remember it is decompressed before being executed. Jan 13 16:12:56 some page of hex dump and: Backtrace: invalid frame pointer 0x00000080 Jan 13 16:12:57 Code: bf0095f0 bf00978c bf0093bc e92d4070 (e5912040) Jan 13 16:12:57 <0>Kernel panic - not syncing: Aiee, killing interrupt handler! Jan 13 16:13:05 The bootloader doesn't do much. Jan 13 16:13:20 Does it boot with re boot? Jan 13 16:13:26 s/re/red Jan 13 16:13:29 s/re/red/ Jan 13 16:13:29 beewoolie meant: s/red/red Jan 13 16:13:33 grr. Jan 13 16:14:14 let me try... Jan 13 16:14:15 Remember, this same boot loader worked with the older version of openslug. Jan 13 16:14:43 I wonder if there is something else awry. Jan 13 16:15:31 nope. It booted fine under redboot. Jan 13 16:15:40 want me to try again with apex? Jan 13 16:16:00 Hang on. Are you running apex from flash or from ram? Jan 13 16:16:11 from ram of course. Jan 13 16:16:20 OK. Let me check something else first... Jan 13 16:16:24 ok. Jan 13 16:16:38 worst-case we could try without MMU. Jan 13 16:18:09 Let me get you a patch. Jan 13 16:19:52 you think it's a compiler issue? Jan 13 16:20:03 No. I think I'm not sure. Jan 13 16:20:14 If you have a patch, you can try things a little easier. Jan 13 16:20:19 yep Jan 13 16:21:00 OK. There is a patch file on the ftp server., apex-1.3.13.patch. It is applied with -p0 in the root of the project. Jan 13 16:21:40 ok, does that affect the xconfig change I made earlier? Jan 13 16:22:11 You'll want to do make openslug_config and work from there, Jan 13 16:22:26 You'll need to set the compiler prefix, for example. Jan 13 16:22:40 all this after applying the patch of course right? Jan 13 16:22:48 Yeah. Jan 13 16:22:55 The first thing to try is with the MMU disabled. Jan 13 16:23:20 ok. Jan 13 16:25:16 Oh. I know what the problem may be. Jan 13 16:25:52 I am willing to bet that APEX will work when it isn't being run from RedBoot. Jan 13 16:26:38 really? Jan 13 16:26:57 is the environ change already in the patch? Jan 13 16:28:10 i meant the kernel source change from earlier. Jan 13 16:28:46 Yep. Check it to make sure. Jan 13 16:29:38 yep. it's in there. Jan 13 16:30:07 You mean if apex was written to flash? Jan 13 16:31:59 trying again with my load. Jan 13 16:32:56 ok. booted fine using my load. Jan 13 16:33:02 with all defaults. Jan 13 16:36:31 beewoolie, is the MMU option you added the one under xconfig's 'General Setup'/'Enable MMU when APEX boots'? Jan 13 16:36:33 OK. So running of SDRAM, apex works fine Jan 13 16:36:38 Yeah. Jan 13 16:36:47 Hmm. Jan 13 16:36:59 It is still having problems shutting the MMU down, then. Jan 13 16:37:34 I'm not sure anymore as the make xconfig has the MMU option checked. Jan 13 16:37:43 oops Jan 13 16:37:56 There are two. One to disable on boot. The other to enable while apex runs. Jan 13 16:38:01 sorry, I just saw the 'disable MMU when apex boots' one right under it. Jan 13 16:38:08 strange . Jan 13 16:38:10 So, which one is disabled. Jan 13 16:38:44 they are both enabled. so no MMU right? Jan 13 16:39:16 No. Both enabled means that the MMU is running when APEX runs. Jan 13 16:39:32 ok. then why is it working then? Jan 13 16:39:34 The disable option is there to cope with the fact that RedBoot doesn't disable the MMU Jan 13 16:39:38 don't know. Jan 13 16:39:43 Sounds like the code works OK. Jan 13 16:39:44 wow. I'm confused. Jan 13 16:39:57 How is the speed? Jan 13 16:39:59 probably just the fact that I changed compiler then. Jan 13 16:40:09 let me do a full timing... Jan 13 16:40:09 It has to be faster than redboot. Jan 13 16:42:41 wow! Jan 13 16:42:47 31secs!!!!! Jan 13 16:43:00 As compared to the RedBoot time of 45 seconds? Jan 13 16:43:03 (I jus remembered that my cell phone has a built-in stopwatch) Jan 13 16:43:17 yes, although the new openslug is a bit faster Jan 13 16:43:26 yipee! Jan 13 16:43:33 OK. I'm going to release it. Jan 13 16:43:36 that's half the time! Jan 13 16:43:43 Thanks! Jan 13 16:43:44 Really? Jan 13 16:43:53 I thought that redboot was about 50 seconds. Jan 13 16:44:17 well, when compared to my old openslug 2.7 setup which averaged almost 55 secs with everything. Jan 13 16:44:24 Hmm. Sure. Jan 13 16:44:32 Thanks for making this possible. Jan 13 16:44:37 well, the openslug load plays a big role too but those are just quick times anyways. Jan 13 16:44:43 No! thank you! Jan 13 16:45:07 Imagine once I have everything optimize in my scripts! Jan 13 16:45:20 I'm tempted to flash it now. Jan 13 16:45:43 do you think It's safe? How easy would it be to load new openslug images after? Jan 13 16:45:44 If I were you, I'd try reobooting several times. Jan 13 16:45:49 yeah. Jan 13 16:46:09 As long as you can get to apex, it's easy to reflash the openslug firmware. Jan 13 16:46:15 yeah, but worst-case apex would still be able to update itself right? Jan 13 16:46:21 how easy? Jan 13 16:46:22 Of course. Jan 13 16:46:38 Just checking because I've been doing the redboot sequence forever now... Jan 13 16:46:40 Well, you have to download the image to SDRAM. Jan 13 16:46:51 ok. xmodem xfer... Jan 13 16:46:52 Once it's there, you erase flash and copy from SDRAM to flash. Jan 13 16:46:55 Right. Jan 13 16:47:02 does it do zmodem too? Jan 13 16:47:10 No. Code is too big. Jan 13 16:47:16 so 3 commands total? Jan 13 16:47:20 I've thought about adding it anyway, but it hasn't happened/. Jan 13 16:47:22 Yeah. Jan 13 16:47:25 Really simple. Jan 13 16:47:31 nah! don't bother. xmodem is good enough. Jan 13 16:47:37 In fact, you can download just redboot and flash that. Jan 13 16:47:48 zmodem is faster. That would be someting. Jan 13 16:47:54 i see. Jan 13 16:48:04 and it has auto-start. hehehe Jan 13 16:48:09 Only I really care about it at that point. Jan 13 16:48:46 i guess it would be pretty slow loading a full openslug image over serial. Jan 13 16:48:58 but then again, it's not something I'm planning on doing every day once it's stable. Jan 13 16:49:20 Right. Jan 13 16:49:32 As I write, you can download just reboot and reflash that portion. Jan 13 16:49:37 I'll have to figure out the address for the loading commands later... Jan 13 16:49:38 In fact, you only need to rewrite one block. Jan 13 16:50:09 what? to flash apex? Jan 13 16:50:23 Both. APEX only needs one block. Jan 13 16:50:39 yeah but it would overwrite part of redboot. Jan 13 16:50:45 You can just erase that block, flash apex. If it fails, reload that first block from the openslug image. Jan 13 16:50:52 Only the first block. Jan 13 16:50:56 redboot takes two blocks. Jan 13 16:51:15 why? is redboot part of the openslug image? Jan 13 16:51:46 Should be the first thing in the image. Jan 13 16:52:00 I suppose it depends on the way they distribute it. Jan 13 16:52:07 It has been there in the past. Jan 13 16:52:10 oh. I forgot that. Jan 13 16:52:15 yeah. Jan 13 16:52:21 oh well. enough for one day. Jan 13 16:52:32 I'll look into that later this weekend. Jan 13 16:52:40 Thanks again for all your support. Jan 13 16:52:44 NP. Jan 13 16:53:04 later Jan 13 16:53:18 night. Jan 13 16:53:26 Nite. Jan 13 17:23:46 hey :) Jan 13 17:23:57 what is the currect nslu-kernel in openslug? Jan 13 17:25:07 i see your patchset for 2.6.15, but im wondering, wich rc you are using Jan 13 17:30:11 <[g2]> fx3|chilla 2.6.15 was out last week :) Jan 13 17:30:26 <[g2]> kernel.org is your friend :) Jan 13 17:30:58 <[g2]> 2006-01-03 Jan 13 17:31:55 oops :) Jan 13 17:32:10 thought, you use a rc* Jan 13 17:32:11 :) Jan 13 17:32:31 cool Jan 13 17:33:15 lets getz ready to rumble with full nativeley built gentoo system including a 2.6.15er kernel on my slug :)) Jan 13 17:33:26 <[g2]> well it was until the real thing came out Jan 13 17:33:28 damn typos.. :) Jan 13 17:35:22 im going to do a gentoo ebuild called "embedded-sources" wich includes a patched linuxkerne. It contains a useflag (duno if you know, what it is) called "+nslu2" wich activates the nslu-patchset from openembedded Jan 13 17:35:40 eh openslug Jan 13 17:35:43 or whatever Jan 13 17:37:04 <[g2]> fx3|chilla do you know vaiper/SpanKY ? Jan 13 17:37:10 a bit Jan 13 17:37:30 <[g2]> he's running gentoo on the slug and will be on my board called the Loft Jan 13 17:37:52 vapier did the ixp4xx ebuild (beneath of man others) and spanky helped me alot to get gentoo runign on my nslu2 Jan 13 17:38:10 <[g2]> vaiper == spanky Jan 13 17:38:34 nah Jan 13 17:38:38 <[g2]> yeah Jan 13 17:38:40 vapier != spanky Jan 13 17:39:02 <[g2]> vapier == SpanKY Jan 13 17:39:40 vapier != spanky Jan 13 17:39:42 :) Jan 13 17:39:44 :P Jan 13 17:39:56 <[g2]> you may be interested in a Loft also as it has a 533MHz and 64MB Jan 13 17:40:03 <[g2]> maybe not Jan 13 17:40:09 <[g2]> just fyi Jan 13 17:40:21 a "loft"? hmm the hardware seems to be expenive Jan 13 17:40:46 <[g2]> not nearly as expensive as time Jan 13 17:40:51 fx3|chilla: you can get the combined patchset for nslu2, nas100d and loft in cvs.sf.net, nslu project, "kernel" module Jan 13 17:41:07 (if you want one-stop shopping without having to go through OE and monotone) Jan 13 17:41:32 <[g2]> nod I'll let vapier know about that too Jan 13 17:41:46 rwhitby: thanks for the tip! (but i already installed monote etc. today to get your patchsets) Jan 13 17:42:12 fx3|chilla: we're also moving to use the ixp4xx-kernel dir in monotone, and deprecating nslu2-kernel and nas100d-kernel Jan 13 17:43:17 rwhitby: i dont know how you exactly handle it in oe, but in gentoo, we have several *-sources wich contain vanilla-sources + a little patchset Jan 13 17:43:36 and i wanna make a ebuild called "embedded-sources" with several useflags Jan 13 17:44:04 e.g. the useflag "nslu2" activates the nslu2 patchsets for he sources Jan 13 17:44:06 and so on Jan 13 17:44:12 so you're going to support more targets from OE over and above the nslu2? Jan 13 17:44:19 (and other nslu2-linux related targets) Jan 13 17:44:34 yes and no Jan 13 17:44:48 i.e. is embedded-sources going to be able to build gentoo for the Zaurus? Jan 13 17:45:10 for the first time, i wanna build a package, containing vanilla-sources, wich includes some patches for embedded-systems Jan 13 17:45:42 for example the no nls-patch (wich makes make menuconfig working on uclibc systems wich dont have a libintl.h) Jan 13 17:46:04 and the user can activatesome useflags Jan 13 17:46:15 maybe "+nslu" Jan 13 17:46:35 the he gets the nslu2 patchset (all the ixpblablastuff) Jan 13 17:47:25 and maybe, there will be some other small devices with a working patchset wich will be implemented Jan 13 17:47:55 maybe "+zaurus" as a useflag, wich means that all the zaurus patches are applied Jan 13 17:50:17 so that a user can get complete kernelsources working for his embedded system, by just adding his devicename to /etc/portage/package.keywords and typing "emerge embedded-sources" Jan 13 17:50:35 the he can do a make menuconfig/make && make modules install Jan 13 17:50:40 and has a working kernel Jan 13 17:50:58 without setting up a crappy crossbuildplatform or sth. else Jan 13 17:57:15 but for now, im going to bed :) n9 Jan 13 17:57:17 n8 Jan 13 18:00:05 night fschmitt Jan 13 18:00:11 oops, night fx3|chilla Jan 13 18:00:48 * rwhitby wonders how many days an emerge will take on the nslu2 ... Jan 13 18:28:28 beewoolie-afk: FYI, I got more weird crashes on shutdown similar to the last ones but I doubt it's apex. Not sure. openslug-3.1 is fairly new too so it could be that. Jan 13 18:44:07 beewoolie-afk: FYI, I just tried the 1.3.13 tarball and there's a typo in your code on line 155. There's an extra " at the end which breaks the build. Jan 13 18:44:24 VoodooZ: Oops. Jan 13 18:44:44 Possibly. Perhaps you should try without apex and see how it goes. Jan 13 18:44:47 not a big deal but just wanted to let you know. Jan 13 18:45:18 Yeah, and like I said it crashed later on so I'm unsure what's going on. Jan 13 18:45:35 I'll have to stress-test the openslug load by itself first I guess. Jan 13 18:46:24 It's kinda unlikely that it's apex, but there's no way to know unless it crashes with redboot. Jan 13 18:46:34 beewoolie-afk: Is there any problems with disabling the kernel command line override option? I'm just thinking it might have change sufficiently in openslug 3.1 to cause problems. Jan 13 18:46:48 I don't know what you mean? Jan 13 18:46:51 yeah, that's why I'll need to test it more. Jan 13 18:47:17 there's a config option to override the kernel command line Jan 13 18:47:58 what if the one apex uses is different from the new openslug? Unless i don't understand properly? Jan 13 18:47:59 I've updated the source archive to fix the typo. Jan 13 18:48:04 good. thanks Jan 13 18:48:21 Those options have to do with the way that apex decides how to set the command line for thekernel. Jan 13 18:48:30 If the kernel accepts atags, apex does it only one way. Jan 13 18:48:39 Older versions of the kernel used to ignore the command line. Jan 13 18:48:47 I don't think that's the case anymore. Jan 13 18:48:59 ok. I'll look into more later. don't worry about it. Jan 13 18:49:23 alright, later... Jan 13 18:49:33 Wait. Jan 13 18:49:41 what? Jan 13 18:49:43 The option you are talking about is simply this. Jan 13 18:49:53 Each arch has a default command line in the source code. Jan 13 18:50:00 This option lets you set the command line explicitly. Jan 13 18:50:03 You want to do that. Jan 13 18:50:22 no, I meant to disable it as it's enabled by default it seems. Jan 13 18:50:26 It's kinda binary. Once the kernel boots, you can be sure that apex has done it's job. Jan 13 18:50:38 Right. You need it for openslug. Jan 13 18:50:44 It was created for openslug. Jan 13 18:51:03 Enjoy the movie. Jan 13 18:51:11 ok. Jan 13 18:51:20 I'll try things tommorow. Jan 13 18:51:35 btw, can I disable JFFS2 now that we don't use it? Jan 13 19:23:02 VoodooZ_Movie: Yeah. I'll do that in the source as well. Jan 13 19:23:17 VoodooZ_Movie: It isn't tight on flash space, so it isn't critical. Jan 13 22:46:36 * NAiL suddenly loves Ubuntu **** BEGIN LOGGING AT Fri Jan 13 22:55:06 2006 **** ENDING LOGGING AT Sat Jan 14 02:59:58 2006