KVM: MMU: consider host cache mode in MMIO page check
authorHaozhong Zhang <haozhong.zhang@intel.com>
Wed, 20 Dec 2017 07:29:29 +0000 (15:29 +0800)
committerRadim Krčmář <rkrcmar@redhat.com>
Tue, 16 Jan 2018 15:40:09 +0000 (16:40 +0100)
Some reserved pages, such as those from NVDIMM DAX devices, are not
for MMIO, and can be mapped with cached memory type for better
performance. However, the above check misconceives those pages as
MMIO.  Because KVM maps MMIO pages with UC memory type, the
performance of guest accesses to those pages would be harmed.
Therefore, we check the host memory type in addition and only treat
UC/UC-/WC pages as MMIO.

Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
Reported-by: Cuevas Escareno, Ivan D <ivan.d.cuevas.escareno@intel.com>
Reported-by: Kumar, Karthik <karthik.kumar@intel.com>
Reviewed-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
arch/x86/kvm/mmu.c

index ff1e9ee259cf7cd0e9151d21b8292bd45545e380..1f1da400fcdeff04345736826787a5fea9b75ff4 100644 (file)
@@ -42,6 +42,7 @@
 #include <linux/kern_levels.h>
 
 #include <asm/page.h>
+#include <asm/pat.h>
 #include <asm/cmpxchg.h>
 #include <asm/io.h>
 #include <asm/vmx.h>
@@ -2708,7 +2709,18 @@ static bool mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
 static bool kvm_is_mmio_pfn(kvm_pfn_t pfn)
 {
        if (pfn_valid(pfn))
-               return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn));
+               return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn)) &&
+                       /*
+                        * Some reserved pages, such as those from NVDIMM
+                        * DAX devices, are not for MMIO, and can be mapped
+                        * with cached memory type for better performance.
+                        * However, the above check misconceives those pages
+                        * as MMIO, and results in KVM mapping them with UC
+                        * memory type, which would hurt the performance.
+                        * Therefore, we check the host memory type in addition
+                        * and only treat UC/UC-/WC pages as MMIO.
+                        */
+                       (!pat_enabled() || pat_pfn_immune_to_uc_mtrr(pfn));
 
        return true;
 }