Re: [PATCH 09/11] KVM: MMU: simplify mmu_need_write_protect

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 12/16/2015 04:05 PM, Kai Huang wrote:


On 12/15/2015 05:25 PM, Xiao Guangrong wrote:


On 12/15/2015 04:43 PM, Kai Huang wrote:


On 12/01/2015 02:26 AM, Xiao Guangrong wrote:
Now, all non-leaf shadow page are page tracked, if gfn is not tracked
there is no non-leaf shadow page of gfn is existed, we can directly
make the shadow page of gfn to unsync

Signed-off-by: Xiao Guangrong <guangrong.xiao@xxxxxxxxxxxxxxx>
---
  arch/x86/kvm/mmu.c | 26 ++++++++------------------
  1 file changed, 8 insertions(+), 18 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 5a2ca73..f89e77f 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -2461,41 +2461,31 @@ static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page
*sp)
      kvm_mmu_mark_parents_unsync(sp);
  }
-static void kvm_unsync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
+static bool kvm_unsync_pages(struct kvm_vcpu *vcpu, gfn_t gfn,
+                 bool can_unsync)
  {
      struct kvm_mmu_page *s;
      for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
+        if (!can_unsync)
+            return true;
How about moving this right before for_each_gfn_indirect_valid_sp? As can_unsync is passed as
parameter, so there's no point checking it several times.


We can not do this. What we are doing here is checking if we have shadow page mapping
for 'gfn':
a) if no, it can be writable.
I think in this case you should also check whether the GFN is being write protection tracked. Ex, if
the spte never exists when you add the GFN to write protection tracking, and in this case I think
mmu_need_write_protect should also report true. Right?

We have already checked it:

static bool mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
                                   bool can_unsync)
{
        if (kvm_page_track_check_mode(vcpu, gfn, KVM_PAGE_TRACK_WRITE))
                return true;

        return kvm_unsync_pages(vcpu, gfn, can_unsync);
}



b) if yes, check 'can_unsync' to see if these shadow pages can make to be 'unsync'.

Your suggestion can break the point a).

A further thinking is can we move it to mmu_need_write_protect? Passing can_unsync as parameter to
kvm_unsync_pages sounds a little bit odd.

+
          if (s->unsync)
              continue;
          WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
How about large page mapping? Such as if guest uses 2M mapping and its shadow is indirect, does
above WARN_ON still meet? As you removed the PT level check in mmu_need_write_protect.

The lager mapping are on the non-leaf shadow pages which can be figured out by
kvm_page_track_check_mode() before we call this function.
Actually I am not quite understanding how large page mapping is implemented. I see in
kvm_mmu_get_page, when sp is allocated, it is large page mapping disabled, but I think we do support
large shadow mapping, right? I mean theoretically if guest uses  2M mapping and shadow mapping can
certainly use 2M mapping as well, and the 2M shadow mapping can also be 'unsynced' (as a leaf
mapping table). But in your series I see if we write protect some  GFN, the shadow large page
mapping is always disabled.

Am I wrong?

If the large page contains the page which is used as page table, kvm does not map large page for
it, the reason is we track the 4k page instead of the whole large page to reduce write emulation.

--
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



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux