2009/8/28 M. Mohan Kumar <mohan at in.ibm.com>: > > Wilbur, I am not sure about ppc32 arch. Check > kexec/purgatory/arch/ppc64/v2wrap.S which makes all secondary cpus to spin > in a loop at the physical address 0x60 > > Regards, > M. Mohan Kumar > I've read the code of kexec-elf-ppc64.c and v2wrap.S, but got quit confused . Hmmm... as matter of fact, I'm not quit familliar with the principle of purgatory , because I modified the standard kexec-tool that , I only passed in the kernel segment , without any ramdisk or other segments. My questions are as followed: q1: const unsigned char purgatory[] =... (kexec/purgatory.c) where are the elements in Array 'purgatory' coming from ? Is it the binary form of v2wrap.S ? (kexec_elf_ppc64.c) q2: elf_rel_build_load(info, &info->rhdr, (const char *)purgatory,\ purgatory_size, 0, max_addr, 1, 0); Does the 'purgatory' here stand for the array above or v2wrap.S ? or both? What is this code used for? q3: elf_rel_get_symbol(&info->rhdr, "purgatory_start", slave_code, sizeof(slave_code)); master_entry = slave_code[0]; Does master_entry euqal to the address of instruction 'b master' in v2wrap.S ? q4: memcpy(slave_code, info->segment[0].buf, sizeof(slave_code)); why coying 256 bytes of kernel data to slave_code , modify first 4 bytes to master_entry,then copy back to info->rhdr ? Why doing this? I saw comment saying : 'a copy of the first 0x100 bytes of this code(v2wrap.S) is copied to 0, and , The above 0x100 bytes at purgatory_start are replaced with the code from the kernel' After this, how the slave cpus jumps to 'slave' point in v2wrap.S , while master cpu jumps to 'purgatory_start' ? Thank you regards, wilbur
