**** BEGIN LOGGING AT Sat Feb 21 02:59:57 2009 Feb 21 05:14:15 blogic ping Feb 21 05:14:54 you might want to talk to noz before ordering one. Feb 21 05:19:05 Hmm, Dec 04... and still no answer... http://forum.embedded-projects.net/viewtopic.php?pid=3139 Feb 21 05:21:05 From what I gather... Benedikt has a USB ARM debugger here on this page... but it's a ftdi based device... http://www.embedded-projects.net/index.php?page_id=256 Feb 21 05:23:18 but the usbprog device... noz got... here... http://www.embedded-projects.net/index.php?page_id=165 is not really ready for serious debugging. Noz hasn't been able to get it working on a XScale board. Feb 21 08:45:49 blogic: That's true, I haven't got the USBprog working with openocd. I think it's a problem in the adapter specific code in openocd, rather than a problem with the USBprog itsefl Feb 21 08:45:53 *itself Feb 21 08:47:05 urjtag doesn't have support for the USBprog, but there is tentative openwince (precursor to urjtag) support, so a port is not impossible Feb 21 08:47:50 So, by all means go ahead and buy one, but it'll be a work-in-progress (but I'll have someone to talk to about it!) Feb 21 10:53:19 noz: so which dongle works ? Feb 21 10:57:21 ah i get it Feb 21 10:58:54 so, how likley is it to get mips debugging if i have a ftdi based dongle Feb 21 10:58:55 ? Feb 21 10:59:00 or is that lso WIP ? Feb 21 11:06:12 afaik urjtag uses boundary scan to do the flashing, while openocd tries to use xscale's jtag debug Feb 21 11:06:30 and the latter requires separate srst and trst signals Feb 21 11:07:26 does usbprog provide that? Feb 21 11:16:07 no idea Feb 21 11:16:14 i will have to look into it in detail Feb 21 11:16:40 * blogic wisches the bdi was cheaper :) Feb 21 11:17:19 well, that was to johnrw i guess Feb 21 11:17:29 i don't think it applies to mips Feb 21 11:18:58 ah sorry Feb 21 11:20:02 so, how slow are the usbprog and the ftdi dongle ? Feb 21 11:20:18 i mena, what clock speed can you get on the jtag ? Feb 21 11:22:27 not too great with ftdi Feb 21 11:23:40 ok Feb 21 11:23:48 still, better than buying a bdi Feb 21 11:24:01 i normally only need it to get the kernel to a point where serial works Feb 21 11:24:10 after that the rest is a piece of cake Feb 21 11:24:35 and i recently found out that without jtag it takes days to weeks and with a bdi you need 1 night Feb 21 11:24:57 so if it ia slow jtag, well then it takes 2 night, which is still better than weeks Feb 21 11:25:08 but i am not telling you anything new i guess ;) Feb 21 11:25:38 days to weeks? o.o Feb 21 11:26:03 hmm Feb 21 11:26:13 i did the infineon danube last year Feb 21 11:26:36 there was a really crappy 2.4 kernel as a reference and we rewrot the whole code for 2.6.25 Feb 21 11:26:49 took 2 weeks to get serial going Feb 21 11:26:52 4 weeks for mtd Feb 21 11:26:59 and ... 8 months for pci :) Feb 21 11:27:11 so how did you work without jtag? Feb 21 11:27:25 ermm Feb 21 11:27:32 readying the code and datasheet Feb 21 11:27:46 and nbd helped me Feb 21 11:27:53 he is kinda a wizard Feb 21 11:28:16 but 2 weeks ago, we were working on the new ralink wisocs Feb 21 11:28:25 so let me guess... you flashed code directly to the chip and tried to run it blind? Feb 21 11:28:28 and it was a piece of cake as we had a bdi Feb 21 11:28:36 roxfan: yeah :) Feb 21 11:28:40 ouch Feb 21 11:28:43 until serial worked Feb 21 11:29:08 and then merged that as a hack into the kernel a EARLY_PRINTK Feb 21 11:29:17 the pci was more tricky though Feb 21 11:29:54 the infineon has a EBU that endianess swaps only 1 way under certain conditions. so the code needed a lot of work arounding to detect and handle that Feb 21 11:30:02 its an amazingly shitty design Feb 21 11:30:59 hacking kernel arch support is great fun Feb 21 11:31:00 ) Feb 21 11:31:57 and by looking at the nicks in the channel i guess i am not the only one here that thinks so Feb 21 15:05:28 last night, at the last, I finally got a combination that was able able to set a hard break in the kernel. I have an Olimex, OpenOCD(svn pulled less than 12 hours ago) and used the OpenWrt generated gdb. Feb 21 15:05:53 Olimex-Arm-USB Feb 21 15:09:28 johnrw: what baord ? Feb 21 15:11:04 On that USR8200 with a ixp425 chip(rare but they do exist, but it has the A0 stepping) Feb 21 15:11:35 that was the one with firewire on it ? Feb 21 15:11:44 Yes Feb 21 15:12:11 olimex has separate trst and srst Feb 21 15:12:29 That's how I set it up. Feb 21 15:12:33 so openocd can enter dbg mode on xscale Feb 21 15:13:36 I had always been able to set a hardware breakpoint in OpenOcd... just couldn't get gdb to work with OpenOCD... Feb 21 15:16:15 I had tried the "newer" svn openocd... but since it was such a work in progress... there was a lot of challenge in configuring it... because the docs were talking about the old openocd. They have changed it, alot. Feb 21 15:17:25 But I gave it another try last night... and it worked out of the make... using the ftd2xx library. Feb 21 16:22:22 blogic: I have a few of these usr8200's... and when I setup openocd, the settings that work on 1 board... don't work on another. Ie. On 1 board... I can use jtag speed 0(full speed), but on another board, i couldn't use full speed... so I clocked it down to jtag speed 16 in openocd.cfg Feb 21 16:23:38 The one that works at full speed is a ixp422 B0 stepping. The A0 ixp425 is the one that needs the 16 divisor. Feb 21 16:24:30 I know your looking for mips... but I am just showing how none of this is an exact science... Feb 21 16:25:01 unless of course, conservative timings are used, I guess. **** ENDING LOGGING AT Sun Feb 22 02:59:57 2009