On Sun, Dec 23, 2012 at 10:00:26AM -0800, Yinghai Lu wrote: > On Sun, Dec 23, 2012 at 6:33 AM, H. Peter Anvin <h...@zytor.com> wrote: > > Explanation please? > > you have following change in the patch > > /* Finally jump to run C code and to be on real kernel address > * Since we are running on identity-mapped space we have to jump > * to the full 64bit address, this is only possible as indirect > * jump. In addition we need to ensure %cs is set so we make this > - * a far return. > + * a far jump. > */ > - movq initial_code(%rip),%rax > pushq $0 # fake return address to stop unwinder > - pushq $__KERNEL_CS # set correct cs > - pushq %rax # target address in negative space > - lretq > + /* gas 2.22 is buggy and mis-assembles ljmpq */ > + rex64 ljmp *initial_code(%rip) > > #ifdef CONFIG_HOTPLUG_CPU > /* > > remove that change, AMD systems works again.
Right, the original code did a RET FAR by popping CS and rIP from the stack. And we did prepare the stack properly before that so it worked. Now, the ljmp translates to a JMP FAR: ffffffff8100016e: 48 ff 2d ab a5 7f 00 rex.W ljmpq *0x7fa5ab(%rip) # ffffffff817fa720 <initial_code> ffffffff81000175: 66 66 2e 0f 1f 84 00 data32 nopw %cs:0x0(%rax,%rax,1) ffffffff8100017c: 00 00 00 00 and in 64-bit mode it has for an operand a 16-bit selector followed by a 32-bit offset. Now, Intel SDM says also this: REX.W + FF /5 JMP m16:64 A Valid N.E. Jump far, absolute indirect, address given in m16:64. And I don't think AMD supports a 64-bit offset. At least I don't see it in the APM where it has only: JMP FAR mem16:16 FF /5 Far jump indirect, with the target specified by a far pointer in memory. JMP FAR mem16:32 FF /5 Far jump indirect, with the target specified by a far pointer in memory. This is at least what I can see at a quick scan. I could ask around if AMD actually supports that FF /5 with a REX.W prefix and it is not only a documentation omission. HTH. -- Regards/Gruss, Boris. Sent from a fat crate under my desk. Formatting is fine. -- -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/