| 03:03.20 | *** join/#openjtag jkilb_ (n=jkilb@p5B209DE4.dip0.t-ipconnect.de) [NETSPLIT VICTIM] |
| 03:47.39 | sn9 | drath: ping |
| 03:47.51 | *** join/#openjtag aproemel (n=aproemel@p57BD1231.dip0.t-ipconnect.de) |
| 04:32.11 | *** join/#openjtag jkilb_ (n=jkilb@p5B20957F.dip0.t-ipconnect.de) |
| 05:12.49 | sn9 | rwhitby: do you by any chance know whether drath is around? |
| 05:13.37 | rwhitby | sn9: idle for 15 hours |
| 05:14.02 | sn9 | 15 hours ago, his dial-up reconnected |
| 05:15.37 | sn9 | urjtag works great for mips ejtag, but it doesn't know what to do with a feroceon, so i thought i'd try the new feroceon support in openocd |
| 05:15.53 | sn9 | i think i've discovered why it's not working |
| 05:16.21 | sn9 | it wouldn't be too complicated to fix |
| 05:17.23 | sn9 | apparently, at least some of these marvell boards deactivate the embedded ice completely whenever trst is deasserted |
| 05:29.22 | sn9 | it might need a brand new reset_config defined |
| 07:08.00 | *** join/#openjtag drath_ (i=vmaster@p5B07E32B.dip.t-dialin.net) |
| 07:33.53 | sn9 | drath: are you there? |
| 07:34.22 | drath | hey sn9 |
| 07:34.54 | sn9 | things don't seem to be going well with the feroceon support |
| 07:34.59 | sn9 | in openocd |
| 07:35.15 | sn9 | the target just won't halt |
| 07:35.20 | drath | it was contributed by Nicolas Pitre - I've never seen a Feroceon device nor a datasheet for one of them |
| 07:35.58 | drath | what do you mean here: "apparently, at least some of these marvell boards deactivate the embedded ice completely whenever trst is deasserted" |
| 07:36.12 | sn9 | that was i guess |
| 07:36.18 | sn9 | s/i/a/ |
| 07:36.28 | drath | well, if TRST is asserted, no JTAG communication is possible at all |
| 07:36.54 | sn9 | you mean deasserted, right? |
| 07:37.17 | drath | mhh, to me, an active low signal is asserted when it's 0 |
| 07:37.27 | drath | nTRST is active low |
| 07:37.45 | sn9 | but it's negated, so i was thinking the other way |
| 07:38.26 | drath | well, whatever - if nTRST is low, then JTAG is held in reset, and no communication is possible |
| 07:38.45 | drath | if it is transitions from high to low and back again, the state machine should be reset |
| 07:39.27 | drath | you could of course try without nTRST - in that case the OpenOCD will try to use a JTAG command sequence to reset the TAP |
| 07:40.11 | sn9 | i tried forcing req_trst to 1 |
| 07:41.26 | sn9 | that holds it in reset, right? |
| 07:48.34 | sn9 | drath: i tried a reset_config of "none" too |
| 07:49.29 | sn9 | could the non-halting have something to do with the "unknown EmbeddedICE version" error? |
| 07:50.08 | drath | it could - but Marvell changed so many things from the 926EJ-S design the Feroceon is based on, it's impossible to tell |
| 07:50.24 | drath | what JTAG interface do you use? |
| 07:50.30 | drath | "wiggler" style |
| 07:50.31 | drath | ? |
| 07:51.01 | sn9 | for the moment, the wiggler clone from diygadget.com |
| 07:51.43 | drath | did you try a higher jtag_speed (it's a divisor)? |
| 07:52.26 | sn9 | i could try. it'll take a minute or two |
| 07:53.48 | drath | gotta go to the office, i'll be back in irc in about 20 minutes |
| 07:57.47 | sn9 | drath: how high should i go? |
| 08:01.28 | sn9 | i just tried 128, no difference |
| 08:12.13 | drath | sn9: someone reported in the forum that it works a lot less reliable if serial terminal or network is attached |
| 08:12.27 | drath | sn9: maybe a problem with signal levels, ground, whatever |
| 08:13.52 | sn9 | i am using usb power, not Vcc |
| 08:15.09 | sn9 | nothing is plugged in besides power and jtag |
| 08:16.37 | sn9 | drath: what is the basic benefit of the EmbeddedICE version being "known" to openocd? |
| 08:21.29 | drath | there are differences in the register layout, and differences in the supported functionality |
| 08:22.11 | drath | for example some support monitor mode debugging, where are breakpoint/halt request is handled in abort mode, rather than halting the target |
| 08:22.18 | drath | that /could/ be a problem for you |
| 08:22.29 | drath | but it could also be something completely different |
| 08:27.24 | sn9 | since /something/ obviously worked for Nicolas Pitre, i think this is a very likely explanation |
| 08:28.12 | sn9 | is there an easy way to force specific versions? |
| 08:40.04 | drath | the feroceon code could overwrite the autodetected version |
| 08:42.17 | sn9 | hmm |
| 08:46.51 | sn9 | drath: looks like the problem is not feroceon-specific: http://forum.sparkfun.com/viewtopic.php?t=8325 |
| 09:16.31 | sn9 | drath: i'm going to try forcing version 7 instead of 8 |
| 09:25.58 | sn9 | the version detected is 0, not 8, and forcing to 7 has no effect |
| 09:26.24 | sn9 | comms control is 0x18 |
| 13:40.24 | *** join/#openjtag prpplague (n=dave@mail.americanmicrosystems.com) |
| 15:41.21 | *** join/#openjtag pleemans (n=peter@d51A5E76A.access.telenet.be) |
| 17:49.34 | *** join/#openjtag ce_geek (n=ce_geek@roger-blanchette-4.umelst.maine.edu) |
| 20:12.09 | sn9 | drath: upon closer examination of the code, feroceon.c handles all the embeddedice-version stuff locally, so that's not the issue. there's just no EICE_DBG_STATUS_DBGACK. is it possible that srst is absolutely required for a halt? |
| 22:42.37 | drath | sn9: normally you can do without nsrst, you just can't reset |
| 22:43.07 | sn9 | but is it possible that it might be required? |
| 22:43.39 | drath | i haven't seen a target that requires it yet |
| 22:44.38 | sn9 | i think it might be a problem that the cpu gets a chance to execute garbage code before openocd makes a connection |
| 22:45.54 | sn9 | background: i accidentally told u-boot to reorganize the flash partitions in such a way that the linux kernel tried to mount where u-boot lived as jffs2 |
| 22:46.28 | drath | sn9: that could be a problem |
| 22:46.46 | sn9 | mount.jffs2 automatically invokes mkfs if it doesn't find its magic |
| 22:47.09 | sn9 | that's why i'm asking about srst |
| 22:47.35 | drath | sn9: you can only recover from such a problem if you have nsrst, if nsrst isn't tied to ntrst (some consumer electronics ties them together), and if your target supports debug out of reset (lpc2000 for example doesn't) |
| 22:48.17 | sn9 | feroceon.c puts major restrictions on debug requests |
| 23:03.29 | *** join/#openjtag flyback_ (n=user@c-24-131-197-176.hsd1.pa.comcast.net) |