**** BEGIN LOGGING AT Mon Mar 21 23:59:56 2005 Mar 22 11:33:55 hey Mar 22 11:35:31 ih Mar 22 11:35:39 8-p Mar 22 12:09:29 beewoolie: How's APEX doing? Mar 22 12:10:02 Pretty good as far as I can tell. Mar 22 12:10:09 I got two users on new platforms. Mar 22 12:10:19 They've vetted the ports I did. Mar 22 12:10:30 ...kinda blind if the truth be told. Mar 22 12:10:35 good to hear. Mar 22 12:10:46 I'm looking forward to trying it on the slug. Mar 22 12:10:48 I've got one more key feature to add before I call it stable. Mar 22 12:11:03 Yeah, me too. The problem is that I don't have working jtag on that platform. Mar 22 12:11:08 good. Mar 22 12:11:11 I've got some tests to run to determine the problem. Mar 22 12:11:20 Because without JTAG I can't really risk it before then. Mar 22 12:11:22 I have a cable, but it isn't reliable. Mar 22 12:11:29 You know it. Mar 22 12:11:30 ouch Mar 22 12:11:45 I don't know if it's the cable or my soldering job on the slug. Mar 22 12:11:54 hehehehe Mar 22 12:11:55 I worked better on a previous slug. Mar 22 12:12:05 But I salted that one good. Mar 22 12:12:34 How much time approx. would I shave off on bootup going from redboot to apex would you say? Mar 22 12:12:35 Also, we don't really know why APEX isn't booting the kernel. Mar 22 12:12:47 Have you tried openslug? Mar 22 12:13:01 Someone made some improvements there. It boots pretty fast. Mar 22 12:13:25 yeah, but redboot takes a good chunk I was told Mar 22 12:13:30 There are about 15 seconds of pauses in the boot process. Moreover, the decompression phase is slowr than it needs to be. Mar 22 12:13:34 And booting directly would be niuce. Mar 22 12:13:40 yeah. Mar 22 12:13:42 When I tried openslug is was really speedy. Mar 22 12:13:52 right now It's like 50+ with bluetooth. Mar 22 12:14:12 that was probably before they added the CSR stuff though. Mar 22 12:14:20 the real benefit of APEX will come through better flash management. Mar 22 12:14:30 CSR? what stuff was done? Mar 22 12:14:33 yeah. I heard it's pretty small too. Mar 22 12:14:56 oh, I meant made the ethernet drivers load by default. Mar 22 12:15:18 APEX should fit in a single block. The problem is that we think or wonder or fear (take you pick) that the Intel access library is necessary to initialize the NPEs. Mar 22 12:15:19 stock i think it's like 41 seconds. Mar 22 12:15:31 i see. Mar 22 12:15:49 Thankfully I don't use them for my robot so.. Mar 22 12:16:18 I think the gating factor is me having a working JTAG emulator. Mar 22 12:16:46 I've got a milestone in progress that, once done, means I can spend some quality time with the skug. Mar 22 12:16:47 yeah. that's pretty much a requirement. Mar 22 12:16:54 nice Mar 22 12:19:06 <[g2]> beewoolie how's APEX going ? Mar 22 12:19:19 [g2]: we were just talking about that. Mar 22 12:19:24 <[g2]> I got GigE running on the avila board Mar 22 12:19:55 <[g2]> so the PCI is working OK to start with :) Mar 22 12:20:03 I saw some of that. what got the driver to link? Mar 22 12:20:24 <[g2]> Oh, I just purge those calls Mar 22 12:20:27 <[g2]> purged Mar 22 12:20:43 beewoolie: hmm, you haven't got the one we sent? Mar 22 12:20:45 <[g2]> I've got a patch that's #ifdef FIXME Mar 22 12:20:53 beewoolie: jtag dongle that is? Mar 22 12:21:07 <[g2]> it's only two calls in e1000_ethtool.c Mar 22 12:21:16 <[g2]> and they're actually for loopback setup Mar 22 12:21:24 g Mar 22 12:21:32 prpplague: Nothing yet. Mar 22 12:21:38 prpplague: when did you send it? Mar 22 12:21:53 beewoolie: doh, i turned it over to shipping like 2 weeks ago Mar 22 12:21:58 beewoolie: i'll check on it Mar 22 12:22:36 <[g2]> so about APEX, I'd like to start using the JTAG header I've got on the device for APEX Mar 22 12:22:44 [g2]: you've got some cohones, hombre. Mar 22 12:22:57 [g2]: OK. Mar 22 12:23:08 <[g2]> what commenting out the call ? Mar 22 12:23:28 [g2]: uh huh. Mar 22 12:24:04 <[g2]> I was going to stub the routine and put an oops in but it was simpler to just try it Mar 22 12:24:17 <[g2]> it's not like it was in the main init routine or anything Mar 22 12:24:46 <[g2]> but it shows that nobody's insmoded the thing with the later kernels Mar 22 12:25:20 <[g2]> The device was responding to 6-8K pings a second Mar 22 12:25:54 <[g2]> it didn't measure the link rate, but there were more than 1M packets in no time and 2+GB of rx/tx data Mar 22 12:26:13 That would be what we call fast, I think. Mar 22 12:26:34 <[g2]> dunno how fast exactly Mar 22 12:27:22 <[g2]> the GigE card on the P4 does nearly 600Mbs TCP xfer as measure by ttcp on a 700MB file Mar 22 12:27:35 <[g2]> that's plenty fast for me at $29 buck for the card Mar 22 12:28:28 <[g2]> I'm looking for another adapter for the disk side Mar 22 12:28:35 <[g2]> the 3 choices are Mar 22 12:28:42 <[g2]> Same NEC chip Mar 22 12:28:46 <[g2]> as in the slug Mar 22 12:28:57 <[g2]> Firewire 400 and possibly 800 Mar 22 12:29:21 <[g2]> There's an Adaptec SCSI module that's got RAID support Mar 22 12:29:24 Seems like most of the HDs are both USB2 and firewire. Mar 22 12:29:31 which raid? Mar 22 12:29:53 <[g2]> I think just 0 1 Mar 22 12:30:05 <[g2]> I'm not excited about SCSI Mar 22 12:30:06 Keep in mind that most of those raid devices aren't really raid devices. Mar 22 12:30:27 At least that's true on the IDE world. Mar 22 12:30:50 <[g2]> well there's a bunch of stuff to check out. Mar 22 12:31:09 <[g2]> I'm really kinda curious why the PCI bug doesn't seem to show up with the GigE card Mar 22 12:31:49 <[g2]> not that I've stressed it super hard Mar 22 12:34:18 [g2]: good question. Mar 22 12:37:09 <[g2]> so back to APEX, I can make a connector, but I may just get an adaptor from the vendor Mar 22 12:37:44 Don't they use the standard ARM pinout/ Mar 22 12:37:50 <[g2]> probably Mar 22 12:38:09 <[g2]> it's a 14 pin connector 2x7 Mar 22 12:39:35 yeah. Mar 22 12:39:57 <[g2]> the pinout is documented Mar 22 12:40:08 <[g2]> does someone make a cheap cable ? Mar 22 12:40:19 What is on the other end of the cable? Mar 22 12:40:53 <[g2]> well they sell a wiggler like device for $99 that plugs into the PC and they've got sw Mar 22 12:41:03 <[g2]> that runs on the windows platform Mar 22 12:41:47 <[g2]> I could rig up some wire and use my own cable from digilent Mar 22 12:42:18 <[g2]> but I'd rather spend my time on APEX Mar 22 12:42:26 <[g2]> and other things Mar 22 12:44:37 <[g2]> maybe I'll play with it tonight, but anyway what's the state of APEX ? Mar 22 12:45:13 there hasn't been much change on the nslu2 front. I've done a lot to make it build better. Mar 22 12:45:31 The new config stuff is really great. I can say that because I didn't write it. Mar 22 12:46:15 <[g2]> where's ext2 at ? Mar 22 12:46:27 beewoolie: hehe Mar 22 12:47:12 [g2]: ext2 is in good shape. It supports symlinks, directories, and partitions. Mar 22 12:47:31 there isn't a directory list function, yet. That's part of the feature set I need to implement. Mar 22 12:47:32 <[g2]> cool Mar 22 12:47:44 beewoolie: i've been testing the sd/mmc code, but i'm not satisfied with the way it is now Mar 22 12:48:27 <[g2]> beewoolie, so on a ext2 partition it could load the kernel from /boot/zImage ? Mar 22 12:49:11 Yep. Even if it has symlinks in the path. Mar 22 12:49:24 <[g2]> awesome Mar 22 12:49:25 prpplague: What's wrong with your MMC code? Mar 22 12:50:00 beewoolie: not as clean as it should be Mar 22 12:50:19 <[g2]> beewoolie, there a way to specify the kernel to APEX ? Mar 22 12:50:24 <[g2]> is there Mar 22 12:50:28 Of course. Mar 22 12:50:34 <[g2]> from the cmdline Mar 22 12:50:43 <[g2]> or config Mar 22 12:50:46 What are you wondering? yeah. You can do it there, but that doesn't make sense. Mar 22 12:51:02 Specifying the kernel is really about copying it to SDRAM. Mar 22 12:51:20 <[g2]> nod. Mar 22 12:51:20 There is a config option to set the source region for the kernel. Mar 22 12:51:40 ...and to copy it on startup. Mar 22 12:51:48 <[g2]> I talking grub like pick a kernel from the cmdline Mar 22 12:51:57 <[g2]> during testing or recovery Mar 22 12:52:03 I'm going to add aliases that will make that sort of thing easier. Mar 22 12:52:20 You can always copy the kernel by hand or set the startup command in the environment. Mar 22 12:52:50 If you want a menu select, it'll have to be an extension. APEX doesn't expect there to be a user at the console. Mar 22 12:53:15 <[g2]> we'll need to change that at least for development :) Mar 22 12:53:53 What do you want to see? Mar 22 12:54:04 More importantly, what do you want to see that APEX doesn't do? Mar 22 12:54:24 <[g2]> I'll have to review the commands and run the latest from RAM Mar 22 12:54:40 <[g2]> it's probably all there Mar 22 12:55:09 since you can set the kernel region in the environment, I'm inclined to leave it as is. Mar 22 12:55:19 but I'm open to suggestions. Mar 22 12:55:47 <[g2]> right now you expect the kernel to be in its own partition correct ? Mar 22 12:55:56 No. Mar 22 12:56:14 What do you mean? Mar 22 12:56:19 <[g2]> ok the Kernel region is the memory area the kernel gets copied to Mar 22 12:56:28 <[g2]> in mem Mar 22 12:56:40 No. The kernel LMA is the address where the kernel is copied. Mar 22 12:56:47 The region is the source for the copy. Mar 22 12:57:07 Technically, the LMA is also a region, but that's a sneaky side-effect of the way that APEX works. Mar 22 12:58:32 <[g2]> Ok from my memory there was nor:... and mem: Mar 22 12:59:08 <[g2]> actually, I can just check the code Mar 22 13:00:10 The default driver is the mem driver. So, 0x00008000 is really the same as mem:0x8000 Mar 22 13:02:45 <[g2]> the code I've got is old 1.1.1 Mar 22 13:03:19 <[g2]> I'l pull the latest, but in that code CONFIG_KERNEL_LMA=0x00008000 defines the LMA Mar 22 13:03:22 I'm working on a 1.2.5. There's an error in the config handling. Mar 22 13:03:34 That's the memory address. tight. Mar 22 13:03:39 s/tight/right/ Mar 22 13:04:04 <[g2]> CONFIG_ENV_REGION_KERNEL=nor:0x60010+0xffff0 Mar 22 13:04:24 <[g2]> defines the kernel source Mar 22 13:04:36 That's the one to change. Mar 22 13:05:15 <[g2]> and how would one specify /boot/bzImage in the middle of a compressed ext2 partition ? Mar 22 13:05:36 <[g2]> or non-compressed for that matter ? Mar 22 13:05:40 We don't yet have compressed ext2 partitions, yet. Mar 22 13:05:45 Oh. Mar 22 13:06:02 We don't yet support ext2 in flash. That's the other thing I need to do. Mar 22 13:06:12 It will look something like this: Mar 22 13:06:46 nor:2m+6m|ext2:/boot/zImage Mar 22 13:07:02 I'm going to add two features. Mar 22 13:07:07 <[g2]> that looks cool Mar 22 13:07:16 There will be the | pipe for cascading drivers. Mar 22 13:07:27 Then, I'm going to add aliases so that it is easier to specify these. Mar 22 13:07:33 Thus, we can say Mar 22 13:07:42 alias part2 nor:2m+6m Mar 22 13:07:59 alias root part2|ext2 Mar 22 13:08:01 then Mar 22 13:08:11 the kernel will be root:/boot/zImage Mar 22 13:09:03 <[g2]> or alias root nor:2m+6m|ext2 Mar 22 13:09:10 right. Mar 22 13:09:15 depends on your needs. Mar 22 13:09:37 It'll be a simple text substitution, so it can be used for commands, too. Mar 22 13:09:57 alias erase_flash nor:128k Mar 22 13:10:11 ...er... alias erase_flash erase nor:128k Mar 22 13:10:34 I also want to add confirmation to the erase command. Mar 22 13:10:40 <[g2]> so basically alias name .... Mar 22 13:10:53 <[g2]> confirmation is important Mar 22 13:10:55 Just as people have come to expect. Mar 22 13:12:18 <[g2]> ok so pipe is the key missing feature for that support Mar 22 13:12:27 ...and an option to override the confirmation: erase -f ... Mar 22 13:12:32 Right. Mar 22 13:12:44 Pipe is the crucial feature. Aliases make it nice. Mar 22 13:13:24 I just came up with the syntax a day or so ago. I've been having a hard time forming the feature so that it is clear, alias-able, and sufficient. Mar 22 13:13:31 I think that the pipe works well. Mar 22 13:14:16 <[g2]> can the pipe handle something like Mar 22 13:14:51 <[g2]> nor:2m+6m|ext2,ext2.gz,jffs2:/boot/bzImage ? Mar 22 13:15:13 <[g2]> similar to rootfstype Mar 22 13:15:14 What are you trying to say? Mar 22 13:15:32 You want a wildcard? Mar 22 13:15:36 ssh dyoung; init 6 Mar 22 13:15:46 :-p Mar 22 13:15:55 <[g2]> I asking whether you need to know aproari Mar 22 13:16:05 <[g2]> I asking whether you need to know apriori Mar 22 13:16:30 <[g2]> hey dyoung-zzzz Mar 22 13:17:56 That's an interesting question. Presently, yes. I'll have to think about that. It requires selecting the driver at runtime. The way I've been looking at it, there would be no way...let me think about it. There may be something that does work. Mar 22 13:18:14 I'm thinking that it would be like this: Mar 22 13:18:15 <[g2]> Ok. You do great work Mar 22 13:18:21 [g2] Axoimtek. Mar 22 13:18:25 http://www.axiomtek.com.tw/product_detail.php?model_num=AX8931A&majorcat=Industrial+PCs&subcat=Rackmount+Peripherals+%26+Accessories Mar 22 13:18:40 Theres a US Importer in Texas IIRC. Mar 22 13:18:52 <[g2]> I've got a tw importer already :) Mar 22 13:18:58 flashpart|(ext2,fat,ext2gz,jffs2):/boot/zImage Mar 22 13:19:21 <[g2]> dyoung-zzzz, I could live with IDE for a start Mar 22 13:19:35 <[g2]> I'd prefer SATA, but IDE's ok Mar 22 13:20:00 [g2]: I posted 1.2.5 to fix a problem with the config strings. Mar 22 13:20:04 <[g2]> that runs linux right Mar 22 13:20:14 <[g2]> beewoolie, thx. I'll pull that tonight and build it Mar 22 13:20:21 I found a bunch of MiniPCI stuff by accident a few years abck when I was looking for CPCI (compact PCI) stuff; but all my links are... lost in a huge bitbucket Mar 22 13:21:04 dyoung-zzzz: What's the white header for? Mar 22 13:21:20 * [g2] was looking at that too Mar 22 13:21:45 Aux Power i would guess. Mar 22 13:25:27 Or, maybe a LED. Yeah, iuts probably for a LED. Mar 22 13:43:40 dyoung-zzzz: that makes sense to me. Mar 22 13:55:47 <[g2]> dyoung-zzzz, THX for the axiomtek link btw Mar 22 14:06:32 That wasnt the one i was thinking of last night though; that one is still spinning around in my brain. Somewhere. **** BEGIN LOGGING AT Tue Mar 22 14:36:25 2005 **** ENDING LOGGING AT Tue Mar 22 23:59:56 2005