On Tue, Apr 13, 2010 at 10:35:53AM +0300, Avi Kivity wrote: > On 04/13/2010 10:26 AM, Gleb Natapov wrote: > >On Tue, Apr 13, 2010 at 10:24:40AM +0300, Avi Kivity wrote: > >>On 04/13/2010 10:21 AM, Gleb Natapov wrote: > >>>May be I am missing something here, but it seams we can call > >>>kvm_mmu_pte_write() directly from emulator_cmpxchg_emulated() > >>>instead of passing mmu_only down to emulator_write_emulated_onepage() > >>>and call it there. > >>> > >>> > >>>@@ -3460,7 +3444,9 @@ static int emulator_cmpxchg_emulated(unsigned long addr, > >>> if (!exchanged) > >>> return X86EMUL_CMPXCHG_FAILED; > >>> > >>>- return __emulator_write_emulated(addr, new, bytes, vcpu, true); > >>>+ kvm_mmu_pte_write(vcpu, gpa, new, bytes, 1); > >>>+ > >>>+ return X86EMUL_CONTINUE; > >>> > >>The written range might cross a page boundary, which > >>kvm_mmu_pte_write() is not prepared to handle. > >> > >Don't we emulate exchange as write in this case? > > > > if (((gpa + bytes - 1)& PAGE_MASK) != (gpa& PAGE_MASK)) > > goto emul_write; > > We do, but that's unrelated. We still have to invalidate potential > ptes on both pages > The code path executed in case of cmpxchg crossing page boundary is not touched by the patch as far as I can see. In this case emulator_write_emulated() is executed with mmu_only false and kvm_mmu_pte_write() is called from emulator_write_phys(). -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
