https://sourceware.org/bugzilla/show_bug.cgi?id=17712
--- Comment #2 from Michael Rolle <m at rolle dot name> --- Actually, a jump to an absolute 64-bit address using a 32-bit displacement isn't possible anyway, no matter what the linker or loader tries to do. So I think this instruction should be illegal in all cases. The purpose of the PCREL32 relocation is to provide an RIP-relative 32-bit offset to a target that will be in the same segment of the linked image file, and the resulting offset must fit into a signed 32-bit value. The loader can put the entire image segment at any memory virtual address it desires, and this does not affect any relative offsets within that segment. Addresses to other image segments (in the same image, or in a dynamic library) can be relocated by the loader with relocation records in the image. These, however, are full 64 bits. In order to jump/call to an absolute address, the program has to either (1) get the address in a register (a 64-bit 'MOV reg, offset' could be used) and do a jump/call to the register, or (2) have the address in memory somewhere and do an indirect jump/call to this memory location. In either case, the opcode will be FF /4 (jump) or FF /2 (call), rather than the E8 or E9 that gas is currently assembling. -- You are receiving this mail because: You are on the CC list for the bug. _______________________________________________ bug-binutils mailing list bug-binutils@gnu.org https://lists.gnu.org/mailman/listinfo/bug-binutils
