On 01/29/2013 03:48 PM, Gleb Natapov wrote: > On Tue, Jan 29, 2013 at 03:37:23PM +0800, Xiao Guangrong wrote: >> On 01/29/2013 02:50 PM, Gleb Natapov wrote: >>> On Tue, Jan 29, 2013 at 11:06:43AM +0800, Xiao Guangrong wrote: >>>> On 01/28/2013 06:59 PM, Gleb Natapov wrote: >>>>> On Fri, Jan 25, 2013 at 11:28:40AM +0800, Xiao Guangrong wrote: >>>>>> On 01/25/2013 09:17 AM, Takuya Yoshikawa wrote: >>>>>>> On Thu, 24 Jan 2013 15:03:57 -0700 >>>>>>> Alex Williamson <alex.williamson@xxxxxxxxxx> wrote: >>>>>>> >>>>>>>> A couple patches to make KVM IOMMU support honor read-only mappings. >>>>>>>> This causes an un-map, re-map when the read-only flag changes and >>>>>>>> makes use of it when setting IOMMU attributes. Thanks, >>>>>>> >>>>>>> Looks good to me. >>>>>>> >>>>>>> I think I can naturally update my patch after this gets merged. >>>>>>> >>>>>> >>>>>> Please wait. >>>>>> >>>>>> The commit c972f3b1 changed the write-protect behaviour - it does >>>>>> wirte-protection only when dirty flag is set. >>>>>> [ I did not see this commit when we discussed the problem before. ] >>>>>> >>>>>> Further more, i notice that write-protect is not enough, when do sync >>>>>> shadow page: >>>>>> >>>>>> FNAME(sync_page): >>>>>> >>>>>> host_writable = sp->spt[i] & SPTE_HOST_WRITEABLE; >>>>>> >>>>>> set_spte(vcpu, &sp->spt[i], pte_access, >>>>>> PT_PAGE_TABLE_LEVEL, gfn, >>>>>> spte_to_pfn(sp->spt[i]), true, false, >>>>>> host_writable); >>>>>> >>>>>> It sets spte based on the old value that means the readonly flag check >>>>>> is missed. We need to call kvm_arch_flush_shadow_all under this case. >>>>> Why not just disallow changing memory region KVM_MEM_READONLY flag >>>>> without deleting the region? >>>> >>>> It will introduce some restriction when VM-sharing-mem is being implemented, >>>> but we need to do some optimization for it, at least, properly write-protect >>>> readonly pages (fix sync_page()) instead of zap_all_page. >>>> >>> What is VM-sharing-mem? >> >> Sharing memory between different guests. >> > That much I can figure out for the name itself. My question is how this > sharing will work? Why KVM_MEM_READONLY is needed for it? Why ability to > change KVM_MEM_READONLY flag without destroying memory region will be > important. What's wrong with nahanni, shared memory device we have today? I'm not clear now or maybe my memory is wrong, i need to find the origin discussion... > >>> >>>> So, i guess we can do the simple fix first. >>>> >>> By simple fix you mean calling kvm_arch_flush_shadow_all() on READONLY >>> flag change? >> >> Simply disallow READONLY flag changing. > Ok, can somebody craft a patch? Takuya, will you? ;) -- 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