Hi *, I just committed a bunch of new code to CVS.
Originally, I intented to do this within a few commits, but when reviewing the diff, I noticed both implementations already collided into too many places, so dissecting them into two commits would have been a tedious work (as a lot of conflicts had to be repaired afterwards when preparing the second commit). The first part of this commit is the import of Erik Walthinsen's STK500v2 protocol implementation (finally). Brian sent it to me, and I did only some minimal cleanup on it (see commit message below). The code still causes a couple of compilation warnings (to be cleaned up later), but I could test it against an STK500v2 here, and it does work well. It causes one run-time warning on my version of the STK500v2 I've got access to: avrdude: stk500v2_getsync(): can't communicate with device: resp=0x01 I verified Erik's code behaves as the STK500v2 specs say, so it's obviously the STK500v2 returning a different response code here. Everything else works fine though, so I suspect a problem with the documentation. Someone else might want to report that to Atmel. The second part of the commit implements JTAG ICE mkII programming support for avrdude. Yes, that (in)famous new JTAG ICE that used to be unsupported by any opensource tools by now. ;-) I've suddenly got access to such a device, and the one who gave it to me was really eager to see Linux/Unix support for it (so they can drop Windows out of their toolchain). After a bit of consideration, I've chosen avrdude as the platform to use for the implementation. First, I know avrdude quite a bit already, being an occasional contributor to it back since the days when it was Brian's privately maintained avrprog baby. Second, after Russell Shaw's statement that he's considering to add direct JTAG ICE support to GDB (he's already done mkI, and currently exploring mkII) it became apparent that this would eventually obviate the need for avarice completely, but we'd then need some other tool to support software downloading through the JTAG ICE. Finally, the already done STK500v2 support, together with the somewhat similar protocols between both, made me start to work on avrdude. The result is now what I'd call alpha to beta quality. ``It works for me.'', basically. ;-) See below, the OCDEN fuse should not be written that frequently as it is now, and few other minor things need a bit of polishing. I wish I knew why the MCU doesn't start again after signing off, but I'm going to track that with Atmel (I hope). During the course of that work, I got quite a bit of contacts with Atmel Norway, as the JTAG ICE mkII protocol document appears to be, well, somewhat inaccurate in many respects. They've already updated the document in the web once, but there are still more errors in it, and quite a number of things remain unexplained (like the need to have the OCDEN fuse programmed when you want to access MTYPE_EEPROM or MTYPE_SPM memory). That doesn't mean I wouldn't consider avarice at all, but avrdude was the first thing to do it. I've also got a patch from Bernd Walter still in the pipeline that increases the write blocksize for ser_posix.c, in order to speed up the communication by lowering the syscall overhead. That's going to be committed later on. Please folks, give all that a try, both those of you who've got access to an STK500v2 and/or to a JTAG ICE mkII, as well as everyone else to see I didn't break anything. Please do also look at the doc changes to see if this all makes sense to you. ------------ Original CVS commit message: ----------------- Mega-commit to bring in both, the STK500v2 support from Erik Walthinsen, as well as JTAG ICE mkII support (by me). Erik's submission has been cleaned up a little bit, mostly to add his name and the current year to the copyright of the new file, remove trailing white space before importing the files, and fix the minor syntax errors in his avrdude.conf.in additions (missing semicolons). The JTAG ICE mkII support should be considered alpha to beta quality at this point. Few things are still to be done, like defering the hfuse (OCDEN) tweaks until they are really required. Also, for reasons not yet known, the target MCU doesn't start to run after signing off from the ICE, it needs a power-cycle first (at least on my STK500). Note that for the JTAG ICE, I did change a few things in the internal API. Notably I made the serial receive timeout configurable by the backends via an exported variable (done in both the Posix and the Win32 implementation), and I made the serial_recv() function return a -1 instead of bailing out with exit(1) upon encountering a receive timeout (currently only done in the Posix implementation). Both measures together allow me to receive a datastreem from the ICE at 115 kbps on a somewhat lossy PCI multi-UART card that occasionally drops a character. The JTAG ICE mkII protocol has enough of safety layers to allow recovering from these events, but the previous code wasn't prepared for any kind of recovery. The Win32 change for this still has to be done, and the traditional drivers need to be converted to exit(1) upon encountering a timeout (as they're now getting a -1 returned they didn't see before in that case). -- cheers, J"org .-.-. --... ...-- -.. . DL8DTL http://www.sax.de/~joerg/ NIC: JW11-RIPE Never trust an operating system you don't have sources for. ;-) _______________________________________________ avrdude-dev mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/avrdude-dev
