My mistake, the last set of listed registers is in this case a list of registers implicitly clobbered by the inline assembly. If I understand correctly this pmap_copy_page() function is incompatible with preservation of the frame pointer. Perhaps falling back on a non-optimized pmap_copy_page() when DDB is enabled is the right course of action excepting an alternative optimized implementation which preserves r31.
Warm Regards, Adam -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Scislowicz, Adam Sent: Thursday, September 06, 2012 5:12 PM To: [email protected] Subject: Use of r31 on the powerpc/booke:mpc85xx architecture In sys/arch/powerpc/booke/booke_pmap.c:pmap_copy_page() The r31 register is used. When compiling for DDB support with "-fno-omit-frame-pointer" the compiler complains about the use of register 31. I am new to this architecture but it seems that register is used as the frame pointer. If you modify the inline assembly to use r23-r30 instead of r24-r31 it seems to solve the problem, but I wanted to ask if this is the right way to change this particular code as I am new to this architecture and don't want to introduce any unpredictable side effects when making the change locally. It also seems that DDB support may not be functioning yet on the powerpc/booke architecture. Is this the known state of things with this port? I am using a P2020RDB-like system and using the P2020RDB config file which references the board code in mpc85xx. While the system otherwise works to some extent: I can boot and use user-space tools when not using DDB. However when I compile with DDB support I get a trap fairly early in the kernel initialization process and then my watchdog reboots the system from out of under me. Warm Regards, Adam http://www.Taglocity.com Tags: netbsd-upstream
