On Sun, Jun 7, 2009 at 18:52, Robin Getz wrote: > On Sun 7 Jun 2009 18:20, Mike Frysinger pondered: >> On Sun, Jun 7, 2009 at 18:15, Robin Getz wrote: >> > On Sun 7 Jun 2009 17:24, [email protected] pondered: >> >> Revision >> >> 6605 >> > <http://blackfin.uclinux.org/gf/project/linux-kernel/scmsvn/?action=browse&path=/&view=rev&root=linux-kernel&revision=6605> >> >> Author >> >> vapier <http://blackfin.uclinux.org/gf/user/vapier/> >> >> Date >> >> 2009-06-07 16:24:10 -0500 (Sun, 07 Jun 2009) >> > >> >> #ifdef CONFIG_BUG >> >> + >> >> +#define BFIN_BUG_OPCODE 0xefcd >> > >> > So - I think hooking GENERIC_BUG up is good -- but before we willy-nilly >> > start >> > defining new opcodes - there needs to be a little more discussion about >> > exactly what it should be - and this discussion needs to include more than >> > just us... >> > >> > I'm also assuming that we might want to add support for this in the >> > assembler? >> > (eventually? or never?) >> >> there is no reason for the opcode to be seen outside of the kernel. >> the assembler shouldnt disassemble it. > > What happens when someone looks at it with objdump? or gdb (JTAG) or kgdb > sees it? > > I think "kernel bug/warn" or something is better than "invalid instruction" > (which is what would happen today)...
i think having it read ILLEGAL is fine. it's how every other arch works, and the instruction only gets executed when there is a problem -- BUG/WARN should not be triggered in the normal case. -mike _______________________________________________ Linux-kernel-commits mailing list [email protected] https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits
