On my S3C2410 (arm920t) system, OpenOCD always prints erroneous (actually totally bogus) CP15 values. Independent of the cp15 register I try to read, it always gives mi different constant values that have no relation to the real register values.
If on a fully-running Linux system (with working userspace) OpenPCD tells me that bit 0 in cp15 r1 (mmu) is switched off, than that’s impossible.
Changes written to a CP15 register via OpenOCD are visible when reading them back with OpenOCD. But it must be some random memory location somewhere, but not CP15 on the arm920 core.
any idea on how to hunt this one down?
Oh, and please let me know whether there’s some bugzilla/trac/… system for reporting bugs like this (and the other one). Also, is there any mailinglist ? Web forums are hard to read when offline/travelling/…
I’ve checked the arm920t CP15 support on a CSB337 (AT91RM9200) and couldn’t reproduce your problems.
To debug these problems, you could recompile the code with DEBUG_JTAG_IO enabled in src/jtag/jtag.h, and DEBUG_INSTRUCTION_EXECUTION enabled in src/target/arm920t.c and src/target/arm9tdmi.c.
This should provide a lot of additional debug information to help track the problem down.
The berlios project site at http://developer.berlios.de/projects/openocd/ provides a bugtracker, a patch submission system, web access to the SVN repository and other facilities. I’ve just created a mailing list for development, but I’d like to keep support requests in one place (i.e. this forum). So far, most users seem to come from a Forum world, who don’t use mailinglists (which I’d prefer) very often. For example the LPC2000 group actually discussed moving away from yahoo in favor of a forum solution…
Regards,
Dominic