On 24/03/2020 12:29, Jan Beulich wrote: > Note that SDM revision 070 doesn't specify exception behavior for > ModRM.mod == 0b11; assuming #UD here.
Didn't I confirm this behaviour for you last time around? > @@ -10075,6 +10079,14 @@ x86_emulate( > : "0" ((uint32_t)src.val), "rm" (_regs.edx) ); > break; > > + case X86EMUL_OPC(0x0f38, 0xf9): /* movdiri mem,r */ > + vcpu_must_have(movdiri); > + generate_exception_if(dst.type != OP_MEM, EXC_UD); > + /* Ignore the non-temporal behavior for now. */ > + dst.val = src.val; > + sfence = true; Looking again at the SDM, I'm not entirely sure this is good enough. Even on top of WB/WP mappings, it needs to have WC properties, knock aliasing lines out of the cache, and ending up as a bus transaction. Also, I'm not convinced the current chunking algorithm for qemu which repeatedly subdivides down to 1, is compatible with the misaligned behaviour described, guaranteeing a split of two. ~Andrew
