At 01:17 PM 2/8/2002 -0600, Navin Boppuri wrote: >Hello again, > >I have PPCBOOT 1.1.2 running on my hardware with an MPC855T processor >running at 64MHZ (both core and bus). I have SDRAM interfaced on this >hardware and I think the burst accesses are working fine since PPCBOOT is >executing well with burst accesses turned on. I have a 2.4.2 Linux kernel >programmed in flash and I am trying to boot this kernel on the hardware. > >I have a BDI hooked up to this hardware. > >The problem is when I start booting the kernel, PPCBOOT calculates the >checksum and says it is OK. It then uncompresses the kernel and even that >is successful. The processor then goes into the debug state and halts at >instruction 0xc000fa0c with a Software Emulation exception. When I do a >disassemble of the Linux kernel, I see that this instruction is the start >of the "printk" routine. No matter how many times I do this, I break at >exactly the same instruction.
This is almost invariably a very strong clue that you are trying to execute 0xFFxxxxxx as an instruction. It usually happens because you jumped to the wrong location or you have a memory fetch problem. In your case, everybody on the list is betting on option #2. I don't have a PPC board handy for disassembling a 0xFFFFFFFF opcode, but from the "PowerPC 603e RISC Microprocessor User's Manual" Appendix A, page A-15 it looks like 0xFFFFFFFF is a fnmaddx instruction (floating point multiply and add). From the MPC860 User's Manual: 7.1.2.11 Floating-Point Assist Exception The foating-point assist exception is not generated by the MPC860. Attempting to execute a foating-point causes an instruction implementation-speci?c software emulation exception. >I tried defining CONFIG_CLOCKS_IN_MHZ but that did not change anything. I >also tried running the processor at 64Mhz/32Mhz (core/memory bus) and >32Mhz/32Mhz but it did not make a difference. > >I was hoping that the kernel would come up atleast with burst accesses >turned off, but I see the same exact behaviour even then. > >I am not sure what to look for now. > >Thank you for all the help and putting up with my weird queries. > >Navin Boppuri. > > ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
