**** BEGIN LOGGING AT Mon Dec 14 02:59:56 2009 Dec 14 08:26:13 juhosg: ping Dec 14 10:35:26 ryd * r18775 /trunk/target/linux/beagleboard/ (19 files in 8 dirs): (log message trimmed) Dec 14 10:35:26 Adding Beagleboard support - unfinished Dec 14 10:35:26 The beagleboard target is still unfinied: Dec 14 10:35:26 running: Dec 14 10:35:26 - MMC Card (incl. boot) Dec 14 10:35:26 - DSS2 graphic driver Dec 14 10:35:28 missing: Dec 14 11:03:00 acinonyx * r18776 /packages/net/xtables-addons/Makefile: [packages] xtables-addons: Move iptable modules to Network category Dec 14 13:48:04 nbd: you around by chance? I got some good news :) Dec 14 13:48:34 yes Dec 14 13:48:42 but not for much longer Dec 14 13:50:41 nbd: i got the ar8316 docs Dec 14 13:50:43 and ssdk Dec 14 13:51:32 under nda tho :/ Dec 14 13:52:31 with permission to write GPL code? Dec 14 13:52:38 and where did you get them from? Dec 14 13:54:05 nbd * r18777 /trunk/include/kernel.mk: do not download kernel sources when using an external kernel tree Dec 14 14:01:39 nbd: yeah i stated clearly that i'm making a gpl driver Dec 14 14:01:44 they also sent me gpl snippets Dec 14 14:02:33 nbd: i wrote their support an email, it just took them a week to reply Dec 14 14:02:54 as a new wiki user i guess i can't create wiki pages Dec 14 14:03:01 could someone create this for me Dec 14 14:03:03 http://wiki.openwrt.org/toh/mikrotik/rb450g Dec 14 14:09:48 russo: please add it into a subpage of inbox: http://wiki.openwrt.org/inbox/start Dec 14 14:10:50 oh okay, i see how its done now, danke Hauke Dec 14 14:17:02 alright, i'm off for a bit, i'll tty all later Dec 14 15:57:02 nico * r18778 /trunk/package/busybox/patches/430-uniq_memleak.patch: [package] busybox: remove uniq memleak patch, applet was fixed upstream in v1.15.2 (closes: #6354) Dec 14 16:27:32 xMff: I found where is the MAC address is stored in the ART partition Dec 14 16:27:53 but if I modify it the wifi driver doesn't work anymore Dec 14 17:37:19 solca: maybe it is checksummed Dec 14 17:49:49 jow_laptop: yeah, which algo they would use? md5? Dec 14 17:55:21 solca: no idea Dec 14 17:56:34 jow_laptop: ok, thx Dec 14 17:59:53 solca: crc is more likely than md5 imo but I do not know Dec 14 18:01:34 jow_laptop: ok, thk u, anyways is hard to know from where to where they are doing the checksumming Dec 14 18:08:52 nbd: hi Dec 14 18:09:29 nbd: from having a quick look at the docs, it seems like the only difference is the gigabit configuration in the 8316 Dec 14 18:14:38 russo: fyi I'm having problems with the rtl8306 Dec 14 18:15:41 solca: I doubt it's as complicated as crc Dec 14 18:15:51 solca: something like xor is more likely Dec 14 18:17:24 rtz: the thing is determine the start offset and the end to try various checksumming methods :( Dec 14 18:19:16 solca: pretty ugly Dec 14 18:19:28 solca: is you origianl firmware still working? Dec 14 18:19:44 rtz: yeah Dec 14 18:24:08 solca: you could try to change settings and compare the dumps Dec 14 18:24:25 solca: so you can at least find the postition of the checksum Dec 14 18:24:42 (if there are any changeable settings for this partition) Dec 14 18:25:39 rtz: the thing is that the ART partition in flash is the ath9k PROM stuff and doesn't change Dec 14 18:25:51 but I have not tried to change things Dec 14 18:26:14 good suggestion, will try that, if anything changes that would be the entering point :) Dec 14 18:28:09 solca: how big is it? Dec 14 18:28:53 rtz: 64KB Dec 14 18:29:35 solca: could you upload a hexdump? Dec 14 18:29:59 solca: or binary Dec 14 18:30:13 I can create the hexdump myself :) Dec 14 18:31:55 rtz: I have 2 differents art.bin that differs because they are from different devices, each device ships with it's own art.bin, I spot the wifi MAC address but probably there is calibration data as well, there are not many differences as reported by vbindiff but can't find the f*ing checksum :( Dec 14 18:32:11 if you wanna try I can upload both ART partitions Dec 14 18:34:18 solca: I can't promise anything, but I would like to take a look Dec 14 18:37:25 rtz: http://home.galileo.edu/~solca/openwrt/art/ Dec 14 18:37:32 rtz: thank you! Dec 14 18:43:22 solca: where are the macs? Dec 14 18:44:19 offset 0x120C Dec 14 18:59:17 solca: are there other atheros nics in your router? Dec 14 19:00:18 rtz: nope, it's the Linksys WRT160NL, it has just one WiFi radio and two ethernet devices Dec 14 19:00:30 why? Dec 14 19:01:33 well, at 0x1000, there is also something looking like a MAC adress Dec 14 19:03:09 solca: you sure, you changed the right one? Dec 14 19:04:35 rtz: let me check... Dec 14 19:06:07 rtz: my second WRT160NL (art2.bin) have in the Atheros WiFI radio the MAC address 00:03:7F:BF:42:EF Dec 14 19:07:20 yeah, it's the same as the one at 0x120C at the flash Dec 14 19:07:39 dunno why there are others MACs addresses there Dec 14 19:17:48 solca: how is your perl? Dec 14 19:18:09 rtz: not good :( Dec 14 19:18:21 do you find something? Dec 14 19:18:40 solca: maybe Dec 14 19:18:49 give me a few moments Dec 14 19:18:52 ok Dec 14 19:19:51 solca: or how is your python? Dec 14 19:20:09 + or - Dec 14 19:20:17 what you need and I do it... Dec 14 19:22:08 solca: take a look at 0x1200 Dec 14 19:22:22 2 bytes for a length field and 2 bytes checksum Dec 14 19:24:14 try to calculate an xor (2 bytes wide) starting 0x1202 to 0x01eb8 (not inculding 1eb8) Dec 14 19:24:56 * solca looking Dec 14 19:26:44 solca: i'm talking about the ar8316 Dec 14 19:29:21 rtz: 1eb8-1200=0cb8 Dec 14 19:29:28 definitely length as you said Dec 14 19:30:26 russo: yeah but as you could base your driver on the ar8306 I though you would like to know Dec 14 19:30:44 ? Dec 14 19:30:49 hmm ar8306 or rtl8306? Dec 14 19:30:55 ar8316 Dec 14 19:30:58 atheros Dec 14 19:31:07 not realtek Dec 14 19:31:19 has anyone seen more than 35mbit with ath9k? Dec 14 19:31:22 ar8316 ~= ar8216 Dec 14 19:32:52 russo: my bad, I was talking about the RTL8306, sorry :( Dec 14 19:33:10 hmm, how i can force the ar8216 to load? modprobe? Dec 14 19:33:46 i'm hoping to learn something about kernel hacking ;) Dec 14 19:35:10 Acinonyx: https://forum.openwrt.org/viewtopic.php?pid=97604#p97604 Dec 14 19:35:35 Acinonyx: I used to have great performance in those days, if you could check what revision was in that date you could check for yourself Dec 14 19:40:13 thanks solca Dec 14 19:40:22 I am testing latest trunk Dec 14 20:58:20 ryd * r18779 /trunk/target/linux/ (23 files in 10 dirs): Make beagleboard as subtarget from omap35xx Dec 14 21:16:53 ryd: nice :) beagle board support... sweet! Dec 14 21:34:58 russo: thx Dec 14 21:35:21 ryd: guess what i'm gonna install on my beagleboard at work tomorrow ;) Dec 14 21:35:45 in other words, thank YOU ;) Dec 14 21:39:41 russo: it's not finished at the moment Dec 14 21:40:24 russo: what you wanna do with owrt on it? Dec 14 21:40:41 just play around with it Dec 14 21:40:50 we do l4/android stuff with it Dec 14 21:41:14 its just nice to have a nice portable router linux :) Dec 14 21:44:23 well - usb in not running at the moment - so no router Dec 14 21:44:46 but I will spent this weekend to get this running Dec 14 22:12:43 nbd * r18780 /packages/ (29 files in 6 dirs): move net-snmp from libs/ to net/ Dec 14 22:54:02 http://pastie.org/743323 <- ideas? Dec 14 22:57:32 does anybody here has a working reset button? Dec 14 23:09:16 rtz: yeah mine works, at least failsafe works Dec 14 23:10:01 rtz: xor csum shell script: Dec 14 23:10:03 csum=0; for i in $(od -v -A n -t x2 -w2 -j 0x1204 -N 0x0cb4 art1.bin); do csum=$(($csum ^ 0x$i)); done; printf "%4X\n" $csum Dec 14 23:10:32 but I don't get the same value :( Dec 14 23:15:09 od: cannot skip past end of combined input Dec 14 23:15:54 crap, wrong file Dec 14 23:18:52 rtz: xor csum is defines as the calculating the xor between two numbers and using that result for the next number right? Dec 14 23:28:07 solca: I think I have it, give me a minute to make it work Dec 14 23:33:36 ok, do it that way: Dec 14 23:33:47 replace the checksum wit 0x0000 Dec 14 23:34:01 make your changes to the mac adress and stuff Dec 14 23:35:12 calculate an xor checksum from 0x1200 length 0x0cb8 (include the size field) Dec 14 23:35:35 xor the result with 0xFFFF Dec 14 23:35:56 this should be your new checksum, write it to 0x1202 Dec 14 23:36:28 and be careful with the byte order, you have to swap it (maybe) Dec 14 23:36:50 ok, back to the reset button topic Dec 14 23:37:01 what platform and openwrt version are you using? Dec 14 23:46:48 solca: because I don't have the slightest clue how the reset button is supposed to work :/ Dec 14 23:48:27 rtz: u're da'man!!! Dec 14 23:48:41 that was simply genious!!! Dec 14 23:49:52 rtz: on the reset button: I have wrt54gl and wrt160nl Dec 14 23:51:39 afair any button in openwrt if pressed _when_ openwrt takes control from the bootloader will activate failsafe mode Dec 14 23:52:12 normally the power led blinks at reset when the bootloader have the control Dec 14 23:52:29 solca: are you running a 2.4 or 2.6 kernel? Dec 14 23:53:07 both Dec 14 23:53:18 wrt54gl with 2.4 and wrt160nl with 2.6 Dec 14 23:55:20 solca: the 160nl is a ar71xx right? Dec 14 23:55:58 rtz: will try later the csum stuff but it is simply amazing how you can deduce and read at the byte level ;) Dec 14 23:56:03 thank you!! Dec 14 23:56:14 this should be documented somewhere **** ENDING LOGGING AT Mon Dec 14 23:56:40 2009 **** BEGIN LOGGING AT Mon Dec 14 23:57:07 2009 Dec 14 23:57:20 solca: the 160nl is a ar71xx right? Dec 14 23:57:39 rtz: yep Dec 14 23:59:09 thanks :) Dec 14 23:59:18 I like to help Dec 15 00:00:59 but mostly, those things boil down to getting lucky Dec 15 00:15:22 solca: are you using trunc or a stable version Dec 15 00:36:18 uniq -w is segfaulting on me on wgt634u Dec 15 00:45:12 [florian]: ping Dec 15 00:47:42 <[florian]> rtz: ping ? Dec 15 00:48:29 [florian]: I have some patches for led and button support for the dc-230 Dec 15 00:48:38 <[florian]> rtz: cool Dec 15 00:49:32 <[florian]> rtz: where do I find them? Dec 15 00:49:46 [florian]: give me few minutes Dec 15 00:49:57 but there is another problem Dec 15 00:50:28 do you know, which gpio pin is used for the reset button with your devices? Dec 15 00:50:48 <[florian]> it used to be 0, but I am not sure anymore Dec 15 00:51:03 well, it's 6 for me Dec 15 00:51:23 <[florian]> at some point we need dynamic device detection Dec 15 00:51:46 [florian]: should I make it sitecom specific or put the 6 for all devices there and hope for the best? Dec 15 00:53:38 <[florian]> let's make it sitecom specific Dec 15 00:53:50 <[florian]> I will try to come up with a device detection mechanism Dec 15 00:56:38 [florian]: I could implement something, but I don't really know how Dec 15 00:57:03 maybe compiling different files for different subtargets into the kernel Dec 15 00:57:28 [florian]: openwrt.pastebin.com/m326f4c Dec 15 00:57:49 I admit it's kinda ugly Dec 15 00:58:32 [florian]: and there is another problem. If I try to plug in an usb device, it get this error: http://openwrt.pastebin.com/m3a13b241 Dec 15 00:58:34 <[florian]> rtz: I would rather we have a singe kernel for each of them in the end Dec 15 00:59:07 but usb did work some time ago, and I have no idea, what changed :/ Dec 15 00:59:46 <[florian]> rtz: kernel update? Dec 15 01:00:05 from .9 to .10 :/ Dec 15 01:01:04 <[florian]> yeah, doubt so Dec 15 01:03:14 [florian]: writing a switch driver will take longer, I may have to make changes to libphy to support multiple phys connected to one mac Dec 15 01:04:51 <[florian]> rtz: multiple phys ? Dec 15 01:05:03 <[florian]> rtz: do not you have a single phy to which you send commands? Dec 15 01:06:53 <[florian]> rtz: your switch should already be supported, r6040 does not hook into the phylib yet, but I plan on doing so Dec 15 01:07:13 [florian]: I already have some (non-working) code for that Dec 15 01:07:31 [florian]: but it's difficult Dec 15 01:07:58 first problem is, the mdio registers of the macs both control the same bus Dec 15 01:08:03 <[florian]> right Dec 15 01:08:13 <[florian]> that's a problem Dec 15 01:08:14 there are 6 phys on that bus Dec 15 01:08:23 <[florian]> oh they splitted up one phy per port Dec 15 01:08:37 yes, pretty much every switch works that way Dec 15 01:09:55 phy 6 is actually a mac, connected to one of the macs in the rdc Dec 15 01:10:16 <[florian]> usually you got two macs and you can "talk" to the switch using a specific phy address Dec 15 01:10:28 <[florian]> oh so it's a mac Dec 15 01:10:41 <[florian]> I mean, it has a kind of "control" port in the end Dec 15 01:10:50 phy 5 is connected to the other mac of the rdc Dec 15 01:10:53 right **** BEGIN LOGGING AT Tue Dec 15 02:08:00 2009 **** BEGIN LOGGING AT Tue Dec 15 02:11:45 2009 Dec 15 02:14:52 rtz: ath9k loads fine now with the modified MAC address and the correct checksum! Dec 15 02:17:20 solca: :) Dec 15 02:23:09 rtz: now the interfaces do not work on a 10Mbps switch, just on 100Mbps Dec 15 02:23:28 do you think this is forcing them to 100Mbps? Dec 15 02:23:30 + ar71xx_eth0_data.phy_mask = 0xf; Dec 15 02:23:30 + ar71xx_eth0_data.speed = SPEED_100; Dec 15 02:23:30 + ar71xx_eth0_data.duplex = DUPLEX_FULL; Dec 15 02:31:20 solca: yes, probably Dec 15 02:31:42 the phy and switch logic is really ugly :/ **** ENDING LOGGING AT Tue Dec 15 02:59:57 2009