On Fri, Jul 15, 2011 at 7:56 AM, David Peverley <pev@xxxxxxxxxxxxxxxxx> wrote: > Checking via GDB to see what the code looks like : > (gdb) disassemble kernel_entry > Dump of assembler code for function kernel_entry: > 0x80504590 <kernel_entry+0>: add %ah,0x8(%eax) > 0x80504593 <kernel_entry+3>: inc %eax In the future, you may want to use "disass/r kernel_entry" to see the hex instructions, and correlate that output to what you see in the file. It would also be helpful to use a cross gdb built for MIPS. But the "objdump -d" output is probably enough to go on. e.g. mipsel-percello-linux-gnu-objdump -d vmlinux | grep -A5 'kernel_entry.:' > I then convert this vmlinux file to a binary image : > $ mipsel-percello-linux-gnu-objcopy -O binary ./vmlinux ./vmlinux.bin.test > > and then hexdump to see what's at the kernel_entry offset : > $ xxd -g xxd -g 4 -s 0x4550 -l 0x80 ./vmlinux.bin.test > 0004550: 00600840 0010013c 1f002134 25400101 .`.@...<..!4%@.. > 0004560: 1f000839 00608840 c0000000 5080083c ...9.`.@....P..< > 0004570: bc450825 08000001 00000000 9c80083c .E.%...........< > 0004580: 00700825 000000ad a180093c dcb22925 .p.%.......<..)% > 0004590: 04000825 feff0915 000000ad 9c80013c ...%...........< > 00045a0: 647624ac 9c80013c 687625ac 9c80013c dv$....<hv%....< > 00045b0: 6c7626ac 9c80013c 707627ac 00208040 lv&....<pv'.. .@ > 00045c0: 96801c3c 00609c27 e01f1d24 21e8bc03 ...<.`.'...$!... > > The code for kernel_entry seems to be at an offset of 0x0004550. This > is -0x40 bytes below where I'd expected! Am I missing something? I can > confirm this by using my bootloader to load vmlinux.bin a > (load_address + 0x40) and it runs just fine. I think the key here is the load address of the first ELF section. >From the objcopy man page: "objcopy can be used to generate a raw binary file by using an output target of binary (e.g., use -O binary). When objcopy generates a raw binary file, it will essentially produce a memory dump of the contents of the input object file. All symbols and relocation information will be discarded. The memory dump will start at the load address of the lowest section copied into the output file." So, the kernel_entry address really does not matter in this context. Although obviously it's good to know whether the kernel_entry code is getting loaded into the right memory location. Posting the output from "objdump -x vmlinux" (or maybe even a URL from which to download your vmlinux file) may help shed light on what is happening. BTW: I enable CONFIG_BOOT_RAW on my images so that I don't have to worry about finding the ELF start address after converting vmlinux into a raw binary image. This lets me jump to a fixed start address, even if the address of kernel_entry moves around.
