**** BEGIN LOGGING AT Mon Oct 30 02:59:57 2006 Oct 30 03:00:17 wheras it works ok on the supplied kernel. Oct 30 03:03:41 <[g2]> GPSFan bonnin++ runs fine Oct 30 03:05:15 great! trying to track down this rtc issue. Oct 30 03:07:27 [g2]: so adding the delays to the cf driver did the trick? Oct 30 03:09:25 <[g2]> GPSFan yeah Oct 30 03:09:43 <[g2]> I'll touch base with GW on Monday about that Oct 30 03:10:10 Hi, someone know if unslug can run on DNS-120 device ? Oct 30 03:10:10 <[g2]> there was data corruption without them Oct 30 03:10:59 <[g2]> that's what was causting the boot issues earlier today Oct 30 03:11:14 <[g2]> s/causting/causing/ Oct 30 03:11:14 [g2] meant: that's what was causing the boot issues earlier today Oct 30 03:12:59 <[g2]> Dsbeerf do you know what's inside the DNS-120 ? I that the dlink-box ? Oct 30 03:13:06 <[g2]> Is Oct 30 03:13:16 [g2]: good to hear about th ecf driver being more stable. with the supplied kernel which looks like it uses the same rtc driver hwclock works fine, it's the 2.6.19-rc3 build that doesn't work. both are BE Oct 30 03:13:27 Yes is a dlink but what i have to know what is inside the box what you mean ? Oct 30 03:13:41 <[g2]> Dsbeerf which processor Oct 30 03:13:56 Oh dunno i have to open the box ? Oct 30 03:14:18 <[g2]> Dsbeerf or look on google Oct 30 03:14:25 k thanks Oct 30 03:18:04 <[g2]> Dsbeerf http://www.dns120-linux.org/pmwiki/pmwiki.php?n=Info.Index Oct 30 03:19:02 yeah i found that one im reading it Oct 30 03:19:44 <[g2]> Dsbeerf google is your friend :) Oct 30 03:19:53 hehe yeah :P Oct 30 03:21:04 <[g2]> GPSFan ok so you don't have /dev/rtc setup on your rootfs Oct 30 03:21:42 <[g2]> GPSFan have you kicked the oscillator ? Oct 30 03:22:09 <[g2]> often it's not running as the RTC is in deep power down Oct 30 03:23:56 [g2]: /dev/rtc is there in both cases. it is even the same rootfs. the clock runs, and either kernel can read the date/time via the /sys/class interface. However hwclock --show doesn't work on 2.6.19 and does on 2.6.15-?? Oct 30 03:24:38 [g2]: there seems to be something wrong with the ioctl interface on 2.6.19. Oct 30 03:26:06 <[g2]> GPSFan ok I'll take your word for it on BE 2.6.19-rc3 Oct 30 03:26:39 <[g2]> hwclock --show Oct 30 03:26:40 <[g2]> Mon Oct 30 03:43:33 2006 -1.025484 seconds Oct 30 03:26:46 <[g2]> things run fine LE Oct 30 03:26:57 <[g2]> and it works fine on 2.6.18 Oct 30 03:27:02 <[g2]> LE Oct 30 03:27:33 <[g2]> what happens when you run with the --debug option ? Oct 30 03:28:04 * [g2] is guessing that's how your found the problem or strace. Oct 30 03:28:13 [g2]: same error message. Oct 30 03:28:45 # hwclock --debug Oct 30 03:28:45 hwclock 2.4c/util-linux-2.11a Oct 30 03:28:45 Using /dev/rtc interface to clock. Oct 30 03:28:45 Assuming hardware clock is kept in local time. Oct 30 03:28:45 Waiting for clock tick... Oct 30 03:28:46 hwclock: ioctl() to /dev/rtc to turn on update interrupts failed unexpectedly, errno=25: Inappropriate ioctl for device. Oct 30 03:29:24 <[g2]> ok Oct 30 03:30:11 oh well, almost time for zzz anyway. gotta get up early tomorrow. cul... Oct 30 03:30:46 <[g2]> GPSFan talk to dwery in the AM. He wrote the original driver Oct 30 03:30:57 <[g2]> and got it pushed upstream. Oct 30 03:31:50 <[g2]> joshin around ? Oct 30 03:32:08 [g2]: ok, I'm hoping it's just another be/le issue I might try building a 2.6.18 version just to have a reference. Oct 30 03:32:14 <[g2]> GPSFan I'd also check with joshin as he's running BE on gentoo Oct 30 03:32:18 <[g2]> with 2.6.18 Oct 30 03:32:45 does his ethernet work? Oct 30 03:32:53 <[g2]> think so Oct 30 03:33:21 <[g2]> but he has a slug so the RTC is different Oct 30 03:33:30 yeah, the 2.6.19 port needs some tlc. Oct 30 03:33:56 ah, I can't spare either of my slugs to experiment with. Oct 30 03:34:38 <[g2]> GPSFan later this week I may get around to running BE Oct 30 03:34:50 <[g2]> I'm working on setting up my regression lab Oct 30 03:35:20 <[g2]> today I was DHCPing an IP addr from RedBoot Oct 30 03:36:18 <[g2]> so I'll be able to see when the boxes reboot as they first ask for an IP from RedBoot, then again from Linux Oct 30 03:36:35 * [g2] needs to compare the bootp packets Oct 30 03:37:20 <[g2]> GPSFan it's not in the repo, but there's an eeprom notifer patch the uses the MAC in the EEPROM Oct 30 03:37:20 so redboot on the avila/loft supports giving out dhcp addresses? Oct 30 03:37:36 <[g2]> in RedBoot? sure. Oct 30 03:37:57 <[g2]> oops... giving out ? Oct 30 03:38:24 <[g2]> I'm running dnsmasq on another Loft that's giving out the addresses Oct 30 03:38:24 that would be good, this random mac address is not cool when you have an eeprom on board for that purpose. Oct 30 03:39:30 <[g2]> right. The proper upstream support of that needs to be fully worked about. But this patch works fine and stuff the EEPROM addr in on the ifconfig up if it's not already sest Oct 30 03:39:34 <[g2]> set Oct 30 03:40:36 so that is not the 45-eeprom-notifierxxxxx patch in svn. Oct 30 03:41:37 <[g2]> I didn't know if blaster8 pulled it out Oct 30 03:42:51 <[g2]> GPSFan I think the mac patch is missiing Oct 30 03:43:12 <[g2]> GPSFan there are two cases Oct 30 03:43:21 [g2]: looking at tht 45 patch in svn, it doesn't grab the mac address Oct 30 03:43:24 <[g2]> 1) when you run the IXP drivers as a module Oct 30 03:43:39 <[g2]> 2) when you run built-in Oct 30 03:44:38 yeah, I can see issues developing differently in each case. Oct 30 03:45:53 <[g2]> what happens is the built-in runs way before the EEPROM subystem Oct 30 03:46:05 <[g2]> so the driver is up long before the notifier runs Oct 30 03:46:29 <[g2]> consequently, it thinks there's no platform MAC and uses a random one Oct 30 03:46:42 yes it would see tha mac address as undefined, all zeros/random or who knows what. Oct 30 03:47:01 <[g2]> right the test for all zeros causes the generation of the random mac Oct 30 03:47:30 which can then be replaced by a script that init's the interface Oct 30 03:47:41 <[g2]> in the case of the module, by the time it runs the EEPROM notifier has already filled in the value Oct 30 03:48:22 <[g2]> well a script works if you don't mind having random mac traffic on the net for a while and bouncing the interface Oct 30 03:49:17 <[g2]> personally, I did the "wrong thing" in the patch by checking when the device interface is brought up Oct 30 03:49:28 random mad addresses are not cool. Oct 30 03:50:16 <[g2]> so until the proper upstream is all worked out, I'll use the notifier + ixmac_open hack Oct 30 03:50:25 question: why isn't the eeprom/i2c system real early in the kernel boot to overcome this problem? Oct 30 03:51:26 <[g2]> it's a good but somewhat mute question Oct 30 03:51:41 <[g2]> s/mute/moot/ Oct 30 03:51:42 [g2] meant: it's a good but somewhat moot question Oct 30 03:51:54 yeah, I know, ours is not to reason why.... Oct 30 03:52:11 <[g2]> no I don't mean that Oct 30 03:52:55 <[g2]> I mean I'm not sure exactly how the module depenancies run. I'd guess on compiler linkage Oct 30 03:52:56 well, it's not in our hands to determine that kind of kernel startup sequencing. Oct 30 03:53:25 <[g2]> well for the Loft and possibly the Avila platform we might be able to tweak it Oct 30 03:54:00 <[g2]> but I haven't really seen anything in the kernel about module ordering (nor have I looked at all) Oct 30 03:54:03 IIRC ther is a list built that gets run through by /init/main.c Oct 30 03:54:52 <[g2]> GPSFan I know there's a list somewhere :) either explicitly or implicity Oct 30 03:55:40 I'll look at it tomorrow... uh.. maybe the next day. I've instrumented the kernel startup process before, so it shouldn't be too hard to stumble across again. ;>) Oct 30 03:56:55 <[g2]> GPSFan we'll I've got working patches for the other stuff Oct 30 03:57:42 <[g2]> that's actually the very next thing on my list now that the CF is running stable Oct 30 03:58:25 ok, good, well it's off to zz land cu tomorrow or the next day.. Oct 30 03:58:41 <[g2]> sweet dreams Oct 30 03:58:59 ;>) Oct 30 04:52:43 Any developers with ssl client certs in the house today? Oct 30 05:02:57 <[g2]> rwhitby you mean the ones you sent out ? Oct 30 05:03:45 yep Oct 30 05:03:56 * [g2] has one of those Oct 30 05:04:31 can you test http://cgiirc.nslu2-linux.org (our new developer-only IRC web gateway) Oct 30 05:05:20 It should ask you for your cert. Oct 30 05:08:13 <[g2]> rwhitby it doesn't it asks for a password Oct 30 05:08:13 <[g2]> did you want a https: ? Oct 30 05:08:13 http:// should redirect to https:// Oct 30 05:08:13 <[g2]> and there's mention that the cert is for limax not cgiirc Oct 30 05:08:24 yeah, I just need to update the server cert Oct 30 05:09:02 <[g2]> I get a request for a master password Oct 30 05:09:07 that's your browser Oct 30 05:09:59 that's your client browser private key storage password (FireFox calls it your master password) Oct 30 05:10:56 <[g2]> yeah that thingy Oct 30 05:11:27 it should get your real name from your certificate now too Oct 30 05:11:59 <[g2]> I set that like 6-9 months ago and dunno what it is Oct 30 05:13:32 <[g2]> rwhitby the cert setup for checkins and trac is nice Oct 30 05:20:13 rwhitby: i'm not sure i understand the question you asked me? Oct 30 05:28:29 [g2]: the server cert name mismatch should be gone now Oct 30 05:29:04 eno: we need a way to determine (at run-time, in an /opt/etc/init.d/SXXfoo script for instance) what optware target we are running on Oct 30 05:29:16 so we can do specific stuff for specific targets. Oct 30 05:29:36 e.g. for the fsg-3, you need to punch the firewall with iptables commands for certain programs. Oct 30 05:29:48 <[g2]> rwhitby the mismatch is gone Oct 30 05:30:54 ok, assuming we don't ever share binaries between optware targets Oct 30 05:31:38 during build time, we can echo $(OPTWARE_TARGET) > /opt/etc/somefile Oct 30 05:32:38 or echo "OPTWARE_TARGET=$(OPTWARE_TARGET)" > /opt/etc/somefile Oct 30 05:32:59 or prepend the above string in /opt/etc/init.d/SXXfoo Oct 30 05:33:20 then the string can check Oct 30 05:33:21 aha - so for an Sxxfoo script, you could do that echo to the file, and the cat the real file after it. Oct 30 05:33:25 snap Oct 30 05:33:38 s/the string/the script/ Oct 30 05:33:38 eno meant: then the script can check Oct 30 05:35:51 ok, I'll do that Oct 30 05:36:11 I'll also update the template Oct 30 06:06:27 03bzhou * r4334 10optware/trunk/make/libgmp.mk: libgmp: added libgmp-host-stage for swi-prolog Oct 30 06:07:07 03bzhou * r4335 10optware/trunk/make/swi-prolog.mk: swi-prolog: share host build Oct 30 06:50:48 03bzhou * r4336 10optware/trunk/make/perl.mk: perl: share host build Oct 30 06:54:55 g'nite Oct 30 09:12:27 exit Oct 30 15:18:01 03bzhou * r4337 10optware/trunk/Makefile: swi-prolog broken on mss and ts72xx Oct 30 15:37:43 <[g2]> quiet morning Oct 30 15:56:52 It's Monday... Oct 30 15:58:08 * barnseenio grunts Oct 30 15:58:27 dreaded bloody monday Oct 30 16:12:41 heh Oct 30 16:12:49 * NAiL just found his new gps coords Oct 30 16:12:55 easy to remember too Oct 30 16:13:37 63-24-24-24N, 10-25-25-25E *waits for icbm* Oct 30 18:46:19 hmz.. anyone running on an avila board? Oct 30 18:49:39 sorry.. only Loft here.. Oct 30 18:55:59 Kaloz: I'm running an avila. been working on various issues in 2.6.19-rc3 Oct 30 18:57:04 GPSFan: can you tell me what machtype does the redboot set on it? Oct 30 18:58:07 Kaloz: my redboot sends 245 (0xf5) as the machine type, I changed mach-types for avila from 546 to 245 as a temporary fix. Oct 30 18:59:07 bugger.. i knen that some vendor stupidity is playing tricks on me :p Oct 30 19:00:17 does anybody know a nice page that describes what needs to be configured when compiling redboot for xscale? Oct 30 19:00:20 g2 said the other day that he had some sort of "shim" that resolved that issue Oct 30 19:00:50 i have a wrv54g compatible board that runs this openrg boot loader crap, which i'd like to replace Oct 30 19:00:52 nbd: g2 has done it, he'd be the one to ask. Oct 30 19:00:57 ok, thanks Oct 30 19:01:15 [g2]? :) Oct 30 19:01:21 nbd: yes.. Oct 30 19:01:51 was just writing it to get his attention when he's back :) Oct 30 19:02:15 ;>) Oct 30 19:06:34 dwery: have you had any problems with the rtc? on loft? I'm able to read it via the /sys/class interface, but when I try to use hwclock--show I get: # hwclock --show Oct 30 19:06:35 hwclock: ioctl() to /dev/rtc to turn on update interrupts failed unexpectedly, errno=25: Inappropriate ioctl for device. Oct 30 19:07:36 GPSFan: sometimes hwclock is pesky... do you have UIE emulation turned on? Oct 30 19:07:36 dwery: I'll have to check. just a min... Oct 30 19:07:58 it's something like CONFIG_RTC_*UIE* Oct 30 19:08:29 dwery: no I don't. I'll turn it on and try again.. Oct 30 19:09:05 <[g2]> nbd ? Oct 30 19:09:45 hi. GPSFan told me you might have some experience with compiling redboot for xscale Oct 30 19:10:03 so i was wondering whether you could help me with it so i can get rid of this rgloader garbage Oct 30 19:10:04 <[g2]> nbd yeah I've got a full tree Oct 30 19:10:14 it's a wrv54g compatible board Oct 30 19:10:21 (gemtek) Oct 30 19:10:35 <[g2]> ok Oct 30 19:11:22 <[g2]> nbd running old or new 1.9x or 2.0x Oct 30 19:12:33 <[g2]> nbd do you have JTAG ? Oct 30 19:12:41 yep Oct 30 19:12:49 dwery: turning UIE emulation on worked like a charm, thanks.. Oct 30 19:13:15 i bricked this thing once, tried a few redboot binaries that didn't work, then flashed back to the old rgloader Oct 30 19:13:21 <[g2]> GPSFan congrats! dwery is THE MAN! Oct 30 19:13:55 [g2]: indeed, now if I can only get the eternet & madwifi running. Oct 30 19:13:58 GPSFan: it should work without it but sometimes hwclock is strange Oct 30 19:14:13 it should recognize that the device has no rtc interrupts and go back polling mode Oct 30 19:14:28 which hwclock --version do you have? Oct 30 19:14:51 dwery: it worked ok on the 2.6.15-? kernel that came with the avila, but not on 2.6.19-rc3. ..go figure.. Oct 30 19:14:54 can't help with ethernet... either new microcode or driver 's author required Oct 30 19:15:07 .15 did have my new rtc subsystem :) Oct 30 19:15:13 s/did/did not/ Oct 30 19:15:13 dwery meant: .15 did not have my new rtc subsystem :) Oct 30 19:15:17 # hwclock --version Oct 30 19:15:17 hwclock 2.4c/util-linux-2.11a Oct 30 19:16:24 <[g2]> nbd so what's the plan ? I'm guessing you don't have JTAG debugging or you could just debug the Redboot Oct 30 19:16:41 <[g2]> how far does it get ? The "+" is damn early Oct 30 19:16:42 i've never used jtag for debugging Oct 30 19:16:45 dwery: thanks anyway, I'll keep plugging along. maybe [g2] can help with the madwifi issues. Oct 30 19:16:58 GPSFan: months ago madwifi was working quite fine Oct 30 19:17:01 <[g2]> GPSFan what's the issue with madwifi ? Oct 30 19:17:02 on my nas100d Oct 30 19:17:02 the last image that i tried warned about missing flash settings Oct 30 19:17:04 then hangled Oct 30 19:17:07 hanged Oct 30 19:17:33 and i didn't compile it myself, so i guess it needs to be configured for the target Oct 30 19:17:39 dwery: I'm sure it was, this is a 2.6.19 issue, several in fact. Oct 30 19:17:50 <[g2]> nbd well that's what boot loaders do :) Oct 30 19:17:51 [g2]: what toolchain and redboot source would you recommend? Oct 30 19:18:20 i want to build my own and configure it for the target to see if it works out better Oct 30 19:19:51 <[g2]> nbd I'm using ./gcc --version Oct 30 19:19:51 <[g2]> gcc (GCC) RedHat-3.3 20021213 Oct 30 19:19:51 <[g2]> as a toolchain Oct 30 19:20:01 [g2]: first issue is that the recomended version of madwifi: madwifi-ng-r1503-20060415 wants include/linux/config.h in various places. 2.6.19-rc3 has removed it. Oct 30 19:20:05 <[g2]> I haven't played with other toolchains Oct 30 19:20:25 <[g2]> GPSFan start with r 1710 Oct 30 19:20:50 <[g2]> that's before the wackelf stuff Oct 30 19:21:05 where can i get it? the intel site only links to a weird libc4-linked one that i can't get to work Oct 30 19:21:16 <[g2]> it's from RedHat Oct 30 19:21:18 [g2]: I tried madwifi-ng-r1770-20061025 nd it's complaining about xscale-be-elf.hal.o uses VFP instructions, whereas /home/kenm/GW2348-4/buildroot/build_armeb/madwifi-ng-r1770-20061025/ath_hal/ath_hal.o does not Oct 30 19:21:34 <[g2]> GPS you need to run the wackelf tool Oct 30 19:21:51 [g2]: wackelf?? Oct 30 19:22:00 GPSFan: madwifi/tools/whackef iirc Oct 30 19:22:05 <[g2]> to clean up the VFP stuff in the ELF headers Oct 30 19:22:18 <[g2]> yeah Kaloz is pointing in the right direction Oct 30 19:22:18 Kaloz: ah, I'll rtfm.. Oct 30 19:25:17 <[g2]> GPSFan do you have a punchlist for BE ? Oct 30 19:25:22 the wackelf.c is in hal/public Oct 30 19:26:07 <[g2]> GPSFan and you've built you own cross toolchain right ? Oct 30 19:26:18 [g2]: not yet, been tracking down my mistakes first. so far the major thing that would be on that list is the ethernet. Oct 30 19:26:26 [g2]: yes Oct 30 19:26:53 list=not yet. built gcc etc=yes Oct 30 19:26:59 ;>) Oct 30 19:39:49 [g2]: Kaloz: wackelf -fpa xscale-be-elf.hal.o set the correct bits to allow it to compile. Oct 30 19:40:31 <[g2]> GPSFan nod. great! Oct 30 19:42:05 <[g2]> GPSFan I forget which GCC/binuilts are you using ? Oct 30 19:43:16 [g2]: gcc 3.4.5 binutils 2.16.1 Oct 30 19:45:08 <[g2]> GPSFan ok thx. I'll probably go straight to 4.1.1 and 2.6.16 or mabye 17 Oct 30 19:45:52 <[g2]> Kaloz your just building the XScale toolchain from the openwrt svn repo right ? Oct 30 19:45:53 * Kaloz now goes for .17 with gcc 4.1.1 and binutils 2.17 - now that he know that gateworks fsck'ed up the machtype Oct 30 19:45:59 [g2]: yeah Oct 30 19:46:29 <[g2]> Kaloz it's a very trivial 3 line patch to shim any machine type Oct 30 19:46:59 <[g2]> nslu2-linux prepends that on about 5 or 6 arches Oct 30 19:47:06 [g2]: what's the best way to get started with configuring redboot for this target? copy the ixdp425 stuff to a new target and modify the parameters if necessary? Oct 30 19:47:40 [g2]: i know, just i though it's correctly set up on the avila :p Oct 30 19:48:25 <[g2]> Kaloz the BSP from Gateworks comes with RedBoot source iirc Oct 30 19:48:41 <[g2]> which means there's a 1 line change for the Machine ID Oct 30 19:48:48 i don't have the bsp, just a board.. so no jtag and the other toys for it Oct 30 19:49:15 <[g2]> I think all there boards have a JTAG header Oct 30 19:49:20 <[g2]> s/there/their/ Oct 30 19:49:21 [g2] meant: I think all their boards have a JTAG header Oct 30 19:49:29 yeah, but it's no standard :) Oct 30 19:49:42 s/no/not/ Oct 30 19:49:42 Kaloz meant: yeah, but it's not standard :) Oct 30 19:50:00 s/s/s/s/s/s/s/ Oct 30 19:50:02 <[g2]> it's a production board Oct 30 19:50:04 :p Oct 30 19:51:04 <[g2]> Kaloz is there a header that's not installed ? Oct 30 19:51:12 <[g2]> and where's the HW porn pics ? Oct 30 19:52:11 [g2]: it's a standard avila 2348-2 or what (the one with only 2minipci slots) Oct 30 19:52:34 [g2]: but probably you could use some hw porn about the pronghorn metro i have here :P but i have no camera Oct 30 19:53:51 [g2]: now that's a nice board for sure :) Oct 30 19:54:24 [g2]: http://www.adiengineering.com/php-bin/ecomm4/productDisplay.php?category_id=27&product_id=85 Oct 30 19:57:44 <[g2]> Kaloz umm that's not really any different than the Loft's I've had for 14 months Oct 30 19:58:10 [g2]: are there any specific board dependent settings that i should check out in this source tree? Oct 30 19:58:16 <[g2]> except I went with the USB 2.0 host controller instead of the 4th minipci Oct 30 19:58:51 <[g2]> nbd do you have a working kernel you've ported for the board ? Oct 30 19:59:03 yes Oct 30 19:59:15 <[g2]> ok then you should know the hw layout Oct 30 19:59:51 yeah Oct 30 19:59:55 <[g2]> you can read the current setting for flash/memory controller/gpio either from the bootloader or by mapping them in via Linux right ? Oct 30 20:00:33 i have a kernel tree that works, but i can't boot it because the boot loader is fscked and refuses to boot anything Oct 30 20:00:57 the boot loader itself works, but it crashes every time it tries to load a kernel image Oct 30 20:01:23 <[g2]> your RedBoot boot loader ? Oct 30 20:01:29 no, the original one Oct 30 20:01:35 rgloader Oct 30 20:01:43 i don't have a redboot image yet Oct 30 20:02:18 <[g2]> how does the kernel crash when loaded ? Oct 30 20:02:29 <[g2]> does the decompress run ? Oct 30 20:02:34 no Oct 30 20:02:46 i flash a kernel image, tell the boot loader to load it and boom, it's dead Oct 30 20:02:52 with known working images Oct 30 20:03:32 <[g2]> well that doesn't make much sense (the known working images part) Oct 30 20:03:39 i know Oct 30 20:03:53 i can't figure out what's wrong with this thing Oct 30 20:04:21 <[g2]> I mean there are the following posibilities Oct 30 20:04:55 <[g2]> 1) the rgloader does something like the sercomm loader on the nslu2-linux as in checking a trailer and stop the boot Oct 30 20:05:05 <[g2]> 2) the image isn't really flashed properly Oct 30 20:05:15 <[g2]> 3) the wrong mach id is pass to the kernel Oct 30 20:05:29 <[g2]> s/pass/passed/ Oct 30 20:05:29 [g2] meant: 3) the wrong mach id is passed to the kernel Oct 30 20:05:30 when the wrong mach id is passed, shouldn't it still print the decompression stuff? Oct 30 20:05:40 btw. the boot loader image is from a backup of the original flash Oct 30 20:05:43 <[g2]> you usually get the first line Oct 30 20:06:04 <[g2]> ndb re the backup Oct 30 20:06:05 when i run boot -s 2, it doesn't even get to printing the newline after i hit enter Oct 30 20:06:21 so it dies fast Oct 30 20:06:34 <[g2]> if you can't copy out, change, and copy back and have it work the same when then the bits on the same in the flash Oct 30 20:06:56 it wasn't a full backup, probably Oct 30 20:07:00 i saved the mtd partitions from linux Oct 30 20:07:09 and there's a small area on the flash that could be missing in this backup Oct 30 20:07:39 although i don't remember actually overwriting that when bricking the box Oct 30 20:07:39 <[g2]> right we has the same issue with the sercomm trailer Oct 30 20:07:39 <[g2]> and a neutered RedBoot Oct 30 20:07:50 hence, i just want to get rid of this crappy loader and put a self compiled redboot on Oct 30 20:08:23 <[g2]> sure, believe me I fully understand Oct 30 20:08:54 * [g2] hasn't touched the nslu2's for over a year Oct 30 20:09:43 so, can you give me a list of things that i need from the running system? Oct 30 20:09:49 <[g2]> nbd this is for the wrv54g right ? Oct 30 20:09:53 yes Oct 30 20:10:01 well, a fully compatible board Oct 30 20:10:13 <[g2]> the gemtek Oct 30 20:10:16 yes Oct 30 20:11:03 <[g2]> ndb if I was in you shoes here's what I'd do Oct 30 20:11:10 <[g2]> I'd boot the bootloader Oct 30 20:11:29 <[g2]> then with JTAG break in to the processor and dump out the config regs via JTAG Oct 30 20:11:57 <[g2]> specifically, the flash/memory controller/GPIO/serial settings Oct 30 20:11:59 can openwince jtag do that or do i need something else? Oct 30 20:12:09 Kaloz: ping? Oct 30 20:12:17 because i have that one working Oct 30 20:12:21 <[g2]> I doubt it supports XScale Oct 30 20:12:32 i used it to flash the boot loader Oct 30 20:12:42 so it does support xscale Oct 30 20:12:55 <[g2]> no it support the chip BSDL file Oct 30 20:13:01 <[g2]> s/supports/ Oct 30 20:13:01 ah, ok Oct 30 20:13:04 i'm still new to jtag stuff Oct 30 20:13:17 <[g2]> they are just twiddling the individual line to the flash Oct 30 20:13:41 <[g2]> it probably takes like 40 minutes to reflash Oct 30 20:13:55 <[g2]> reflash RedBoot Oct 30 20:14:03 yeah Oct 30 20:14:08 so what else can i use? Oct 30 20:15:13 <[g2]> I've got a different adapter we are bringing up Oct 30 20:15:39 <[g2]> it does like 4Mhz Oct 30 20:15:57 <[g2]> it talks to the Avila board from Linux Oct 30 20:17:09 Kaloz: Did you have a fsg-3? Booting a 2.6 kernel? :) Oct 30 20:17:36 [g2]: got madwifi compiled for 2.6.19-rc3. the modules all load and iwconfig can config the ath0 interface. will test out operation after lunch.. Oct 30 20:18:00 [g2]: so there's no way to do that with a simple buffered lattice cable? Oct 30 20:20:44 NAiL: i don't have an fsg :) Oct 30 20:21:15 ah, no... someone else on the openwrt forum had one and you mentioned Xscale. Oct 30 20:21:24 * NAiL has a teflon mind Oct 30 20:22:16 <[g2]> nbd sure it's sw that talks to the debugging side of the chip Oct 30 20:22:30 <[g2]> nbd stop by #openjtag Oct 30 20:24:30 NAiL: i have a linksys rt042 (bricked) and avila 2348-2, a gateway 7001 and a pronghorn metro :) Oct 30 20:27:45 ah Oct 30 20:28:16 * NAiL has two NSLU2's, an FSG-3, N2100 and one WRT54GS (again) ;) Oct 30 20:30:02 ah, yes, i have a n2100 now, too :) Oct 30 20:30:02 NAIl has yet to test my latest patches on them all :D Oct 30 20:36:49 I know Oct 30 20:42:39 hmm Oct 30 20:57:11 <[g2]> Kaloz nbd you guys haven't played with the RV series at all have you ? Oct 30 20:58:12 dwery: did you or blaster8 look at the via_sata issue? Oct 30 20:58:17 (if you know) Oct 30 20:58:42 NAiL: nope. what issue? Oct 30 20:59:06 support for the controller in the fsg-3 Oct 30 20:59:24 what's missing? Oct 30 21:00:00 It's an VT6421. Not quite sure what the problem was again Oct 30 21:00:15 ah yes. The PATA bit of the driver was missing Oct 30 21:00:22 * [g2] remembers an endian issues with that Oct 30 21:00:38 <[g2]> well relating to the G600 Oct 30 21:00:48 <[g2]> and a via chip Oct 30 21:01:14 <[g2]> dwery we had the G600 working with 2.6.16 and the patched code back then Oct 30 21:01:37 it's the same problem as the G600? Oct 30 21:02:50 [g2]: DMS-G600 was working when I tested it months ago... Oct 30 21:03:15 99-sata_via-add-pata.patch does not work? Oct 30 21:03:45 <[g2]> dwery nod it worked for me too Oct 30 21:03:57 <[g2]> I haven't tried the latest on the G600 Oct 30 21:05:05 does it apply to 2.6.18/19-rc3? Oct 30 21:07:28 don't know... never tried Oct 30 21:07:41 (I don't have a serial cable yet, so I can't boot the kernel on the fsg-3) Oct 30 21:10:09 * NAiL goes to bed Oct 30 21:10:11 nite all Oct 30 21:11:36 nite Oct 30 21:20:48 hmm Oct 30 21:20:52 cgiirc? Oct 30 21:22:06 <[g2]> sweet dreams NAiL Oct 30 21:29:51 [g2]: This is interesting, I can get the madwifi driver to associate with an ap, and scan and find several ap's but I can't ping. If I use the Gateworks supplied kernel/driver, things work ok. So I wonder if it's the same issue as with the ethernet. Oct 30 21:36:05 GPSFan: did you get an IP via dhcp or it statically assigned? Oct 30 21:38:34 dwery: ip is static. Oct 30 21:41:03 dwery: hold on, the pings think they want to go out the ethernet. Oct 30 21:41:42 now it works ok!! thanks for jaring my sanity.. Oct 30 21:42:16 :-D Oct 30 21:43:29 ok, madwifi issue resolved. Oct 30 21:45:27 dwery: that's the latest rev: madwifi-ng-r1770-20061025 Oct 30 21:45:41 good. will try it in the next few weeks Oct 30 21:52:21 morning all Oct 30 21:52:39 hey rwhitby Oct 30 21:53:52 hey barnseen1o Oct 30 22:14:26 dwery: did you have anything to do with the hwmonitor driver AD7418? Oct 30 22:14:39 maybe. Oct 30 22:14:46 :-D Oct 30 22:15:00 if it burned your house, not. Oct 30 22:15:07 anything else, I'am the author :) Oct 30 22:15:47 ;>) it works, it's just that the gateworks driver outputs 5050 and new driver outputs 50500 for the same temperature. Oct 30 22:16:08 mm Oct 30 22:16:16 same thing for the temp max and max_hyst. Oct 30 22:16:44 I guess we should check how sensors read that value Oct 30 22:17:32 those values are cooked by the driver to read in deg C 4500 being 45C Oct 30 22:17:50 that's from how the show temp program displays things Oct 30 22:18:01 I think there's a standard somewhere that should be followed.. Oct 30 22:18:05 let me see Oct 30 22:19:07 the unit should be millidegree celsius Oct 30 22:19:54 so the old driver is wrong Oct 30 22:20:29 dwery: the voltage reads ok, in4 reads out a hex # which when multiplied by 23.1 (from the Gateworks manual) makes 12980mv or ~13V which is what I'm putting in. Oct 30 22:21:11 good Oct 30 22:21:24 dwery: 50500 milidegrees is 50degrees :>)) Oct 30 22:22:57 isn't that right? Oct 30 22:24:20 dwery: that's perfectly correct. Oct 30 22:24:28 ok Oct 30 22:24:29 :) Oct 30 22:24:53 s/50degrees/50.5degrees Oct 30 22:28:31 +/ Oct 30 22:40:07 <[g2]> GPSFan I've got an old OpenSlug BE toolchain lying around on another computer Oct 30 22:40:37 <[g2]> I was going to build my mini jffs2 rootfs that can ping and run the ethernet driver Oct 30 22:40:47 <[g2]> BE style Oct 30 22:41:09 [g2]: great, then maybe we can figure out what's going on . Oct 30 22:41:19 <[g2]> I should be able to run the same .19-rc3 build Oct 30 22:41:44 <[g2]> I just need to build the jffs2 BE and tweak that JFFS2 parameter in the .config Oct 30 22:43:11 <[g2]> GPSFan I think the temps are in 1/10ths of degrees Oct 30 22:44:10 [g2]: the chip datasheet says the resolution is .25C/bit of a 10bit value inside. Oct 30 22:46:10 <[g2]> GPSFan I remember it's a 10bit range over is it -100C to +100C ? Oct 30 22:47:49 [g2]: +/-128C 10 bit 2's comp Oct 30 22:48:35 I was sort of supprised at it reading 50C. the board gets a bit warmer than I expected. Oct 30 22:49:11 esp the Sipex TO220 chip near the CPU Oct 30 22:56:02 <[g2]> GPSFan that's right near the chip Oct 30 22:56:08 <[g2]> s/chip/cpu/ Oct 30 22:56:09 [g2] meant: GPSFan that's right near the cpu Oct 30 23:07:16 yes, it's right near the cpu, probably the regulator for the core voltage. Oct 31 00:03:12 <[g2]> GPSFan ping Oct 31 00:04:12 <[g2]> GPSFan what's CONFIG_CPU_BIG_ENDIAN set to in you config file ? Oct 31 00:11:24 [g2]: CONFIG_ARCH_SUPPORTS_BIG_ENDIAN=y Oct 31 00:11:24 CONFIG_CPU_BIG_ENDIAN=y Oct 31 00:15:48 <[g2]> GPSFan ok thx and what's the option for the rtc ? Oct 31 00:17:02 miniflood: CONFIG_RTC_LIB=y Oct 31 00:17:02 CONFIG_RTC_CLASS=y Oct 31 00:17:02 CONFIG_RTC_HCTOSYS=y Oct 31 00:17:02 CONFIG_RTC_HCTOSYS_DEVICE="rtc0" Oct 31 00:17:02 CONFIG_RTC_DEBUG=y Oct 31 00:17:03 # RTC interfaces Oct 31 00:17:05 CONFIG_RTC_INTF_SYSFS=y Oct 31 00:17:09 CONFIG_RTC_INTF_PROC=y Oct 31 00:17:11 CONFIG_RTC_INTF_DEV=y Oct 31 00:17:13 CONFIG_RTC_INTF_DEV_UIE_EMUL=y Oct 31 00:17:16 # RTC drivers Oct 31 00:17:17 # CONFIG_RTC_DRV_X1205 is not set Oct 31 00:17:19 # CONFIG_RTC_DRV_DS1307 is not set Oct 31 00:17:21 # CONFIG_RTC_DRV_DS1553 is not set Oct 31 00:17:23 # CONFIG_RTC_DRV_ISL1208 is not set Oct 31 00:17:25 CONFIG_RTC_DRV_DS1672=y Oct 31 00:17:27 the debug one isn't really neccessary Oct 31 00:18:16 also make sure your kernel command line has: rtc-ds1672.probe=0,0x68 Oct 31 00:24:34 <[g2]> GPSFan normally just pastebin > 4 lines Oct 31 00:24:44 <[g2]> but thanks Oct 31 00:26:31 [g2]: didn't know what the local pastebin threshold was, I will in future.. Oct 31 00:29:48 showvolt Oct 31 00:29:57 oops wrong window. Oct 31 00:33:37 03bzhou * r4338 10optware/trunk/ (Makefile make/swi-prolog.mk): swi-prolog: sharing host build is tricky, only share libgmp host build Oct 31 01:05:57 03bzhou * r4339 10optware/trunk/make/py-nose.mk: py-nose: upstream upgrade to 0.9.1 Oct 31 01:35:45 Hmm - optware procmail still uses /etc/procmailrc - need to change that to /opt/etc/procmailrc so I can use it as the local mail delivery agent of the fsg-3 and get foo+bar@domain.com addressing going into separate IMAP folders. Oct 31 01:40:09 <[g2]> GPSFan still up ? Oct 31 01:41:07 yep.. Oct 31 01:41:39 [g2]: about to go watch a movie, but still here for a bit. Oct 31 01:42:57 rwhitby, [g2]: is there any word from upstream on how to proceed with reading the MAC address from flash ? Oct 31 01:42:57 <[g2]> GPSFan ok I'm up in my jffs2 BE system Oct 31 01:43:27 dumfrac: not a single word from Deepak Oct 31 01:43:33 <[g2]> I've to two minor issues, one is PCI init and the second is the Intel NPE driver doesn't seem fully up Oct 31 01:43:51 [g2]: you have no pci bus? Oct 31 01:43:54 thanks - Martin and I are trying to figure out what to do for etch Oct 31 01:44:16 <[g2]> GPSFan one sec Oct 31 01:44:34 rwhitby: also, any resolution on the issue of putting the NPE microcode in flash (i.e. which partition, etc.) Oct 31 01:44:55 <[g2]> GPSFan pastebin coming.... and don't let me hold you up for the movie this is all just FYI Oct 31 01:45:28 <[g2]> http://pastebin.ca/230037 Oct 31 01:46:24 dumfrac: nope Oct 31 01:46:51 <[g2]> I think there may be an endian option for the ehci/ohci controllers Oct 31 01:46:57 I need to look at RedBoot code today to see if it retains non-FIS, non-config data in the last block Oct 31 01:47:14 If it doesn't, then we just go with a separate "Microcode" partition. Oct 31 01:47:21 [g2]: you using the latest patches from svn? Oct 31 01:47:42 rwhitby: thanks - I'll wait to hear from you then Oct 31 01:48:37 <[g2]> GPSFan dunno what's the "latest" it's an -r 538 plus some minor tweaks Oct 31 01:49:11 dumfrac: the other avenue is to find some use case which cannot be done in early userspace (i.e. netconsole debug of early userspace code) Oct 31 01:49:16 <[g2]> I've got a different defconfig, the #33 patch is enabled, the ixp drivers are builtin Oct 31 01:49:43 if you can do a proof-of-concept of something like that, it would be an effective way to prove to upstream that the kernel needs to read MAC address and microcode from flash, not early userspace. Oct 31 01:50:29 rwhitby: ok, I'll look into that - thanks Oct 31 01:50:56 dumfrac: excellent. thanks! Oct 31 01:51:28 what's in 2.6.19 is 1 day old, that's what I used, one didn't apply properly (78 I think) but Idon't think that one is important for loft/avila. Oct 31 01:51:43 <[g2]> GPSFan looks like the mtd notifier isn't loading the RedBoot NPE code Oct 31 01:53:26 yep, after the mtd lists redboot, there should be some npe activity indication that it found the microcode. Oct 31 01:54:04 [g2]: you also have this which I don't see: # Oct 31 01:54:04 mtd: partition "FIS directory" doesn't end on an erase block -- force read-only Oct 31 01:54:04 # Oct 31 01:54:04 0x00fff000-0x01000000 : "RedBoot config" Oct 31 01:54:38 what version of redboot? Oct 31 01:55:04 03rwhitby * r4340 10optware/trunk/ (make/procmail.mk sources/procmail/config.patch): procmail: Changed /etc/procmailrc to /opt/etc/procmailrc Oct 31 01:55:12 <[g2]> GPSFan that's normal Oct 31 01:55:13 * GPSFan thisks that should have only been 2 lines :( Oct 31 01:55:43 the force read-only is when you have fis dir and config in the same block Oct 31 01:56:08 <[g2]> rwhitby actually there's an option in the kernel to override that Oct 31 01:56:40 <[g2]> if you want to be able to write to the RedBoot/FIS dir/config partitions Oct 31 01:57:02 [this is what my bootup looks like: http://pastebin.ca/230051 Oct 31 01:57:24 [g2]: really, for the non-erase-block size scenario too? (I know about the option to normally make the special whole blocks read-only or writeable) Oct 31 01:58:08 <[g2]> rwhitby I haven't played with the FIS directory/config area Oct 31 01:58:15 <[g2]> from Linux that is Oct 31 01:58:27 <[g2]> I've often replaced RedBoot though Oct 31 01:58:54 <[g2]> when I was doing all my APEX testing on the Slug I must reflashed Redboot 50 to 100 times Oct 31 01:59:17 <[g2]> s/must/must have/ Oct 31 01:59:21 [g2] meant: when I was doing all my APEX testing on the Slug I must have reflashed Redboot 50 to 100 times Oct 31 02:00:12 <[g2]> now that I've got through the FIS dir structure, the config seems pretty straight forward Oct 31 02:00:18 <[g2]> the two things I think about are: Oct 31 02:00:34 <[g2]> 1) full checksum support which isn't currently in there Oct 31 02:00:50 <[g2]> 2) fallback boot options Oct 31 02:00:55 [g2]: I ran a quick test on the madwifi driver running 802.11b I got 4Mbps one way, very simple file copy test. no errors. Oct 31 02:01:24 <[g2]> GPSFan I get 3.43M _B_ Oct 31 02:01:39 <[g2]> on the A/G band Oct 31 02:01:51 <[g2]> it's nearly line rate for 54Mbs Oct 31 02:01:53 I was only using 11b Oct 31 02:01:57 <[g2]> nod Oct 31 02:02:02 <[g2]> seems slow Oct 31 02:02:17 <[g2]> well let's see, not that slow Oct 31 02:02:26 I think 6mbps is max realistic. Oct 31 02:02:36 <[g2]> as about 5.5 I'd guess Oct 31 02:02:58 <[g2]> and depending where you are living (boondocks versus in the city) Oct 31 02:03:43 <[g2]> I've got around 30 AP in the B/G band probably within a 30m radius Oct 31 02:03:52 [g2]: I let the board run for an hoour or so and scanned it with my ir thermometer. got the cpu at 47.5C, the temp sensor read out 48.5C. Oct 31 02:03:57 <[g2]> heh... 0 in the A band :) Oct 31 02:04:18 <[g2]> GPSFan Gateworks builds nice hw :) Oct 31 02:04:19 [g2]: I can set up an a access point sometime tomorrow. Oct 31 02:04:51 <[g2]> GPSFan you're running glibc right ? Oct 31 02:05:23 <[g2]> GPSFan I've got scripts if you need any Oct 31 02:05:38 yes and no, what I'm testing with right now is the gateworks rootfs, tweaked on a microdrive. so I guess it's glibc. Oct 31 02:05:53 what I normally use is uClibc Oct 31 02:06:04 <[g2]> sissy :) Oct 31 02:06:22 why you say dat mon? Oct 31 02:06:36 <[g2]> I'm just teasing Oct 31 02:06:56 <[g2]> we're pretty endian/libc agnostic around here Oct 31 02:07:33 <[g2]> for a while we had all 4 major variants in production Oct 31 02:08:03 lol. well I & the wife are off to watch a movie. cul... I'm only using the gateworks rootfs 'cause it worls (mostly) with their supplied kernel. Oct 31 02:08:04 <[g2]> openslug BE/glibc debianslug/ LE/glibc and ucSlugc/ BE/LE uclibc Oct 31 02:08:33 * [g2] buys some virtual popcorn for you and the wife Oct 31 02:08:38 <[g2]> have a great time! Oct 31 02:08:40 I guess my slugs are out of date. they have been running about 2 years on unslung. Oct 31 02:09:06 <[g2]> yummy V 3.x ? Oct 31 02:09:54 Firmware Version: V2.3R29-uNSLUng-5.5-beta Oct 31 02:11:04 <[g2]> ahh ok, I thought you ment you were running the version from two years ago and hadn't upgraded Oct 31 02:11:18 <[g2]> anyway go have fun at the movies! Oct 31 02:12:00 I upgraded regularly for a while. ran one of the first beta's I guess I'm in the Yahoo logs somewhere. **** ENDING LOGGING AT Tue Oct 31 02:59:57 2006