[PATCH 0/13] KVM: MIPS: Dirty logging, SYNC_MMU & READONLY_MEM

[PATCH 0/13] KVM: MIPS: Dirty logging, SYNC_MMU & READONLY_MEM

[James Hogan]

Note: My intention is to take this series via the MIPS KVM tree. This
series is based on my other series posted this cycle:
[0/10] KVM: MIPS: Implement GVA page tables and shadow flushing
[0/3]  KVM: MIPS: Use CP0_BadInstr[P] for emulation
[0/30] KVM: MIPS: Implement GVA page tables

This series adds proper dirty page logging, KVM_CAP_SYNC_MMU, and
KVM_CAP_READONLY_MEM support to MIPS KVM.

The existing dirty page logging support (integral to live migration) was
basically incomplete as pages were never marked read only or recorded as
being dirty on a TLB modified exception. This series incrementally adds
proper support, using the dirty bit in the GPA and GVA page tables to
trigger TLB modified exceptions.

Support is added for KVM_CAP_SYNC_MMU, which adds MMU notifiers so that
KVM can react to asynchronous (and synchronous) host virtual MM changes.
This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning, as well as paving the way for KVM_CAP_READONLY_MEM.

Finally support is added for read only memory regions
(KVM_CAP_READONLY_MEM), which can be supported fairly minimally once
dirty page logging and KVM_CAP_SYNC_MMU are in place. This allows memory
regions to be marked read only to the guest so that reads work but
writes trigger MMIO.

The patches are roughly grouped as follows:

Patches 1-2:
  Preliminary changes for readonly regions / dirty logging.

Patches 3-5:
  Add handling of read only regions (or areas with no region) as MMIO,
  and also call fault handling for dirty logging.

Patches 6-11:
  Incrementally add proper dirty logging support, first making ranges of
  GPA clean when log read or region is made readonly, then adding the
  fast path fault handling for dirty logging, and finally transferring
  GPA protection bits (including dirty) across to GVA page tables so
  they take effect.

Patch 12:
  Add KVM_CAP_SYNC_MMU support.

Patch 13:
  Enable KVM_CAP_READONLY_MEM. This is after KVM_CAP_SYNC_MMU support as
  it needs to use gfn_to_pfn_prot() to know whether the GFN is read
  only, which opens up the possibility of getting copy-on-write enabled
  pages, which aren't safe without KVM_CAP_SYNC_MMU since the guest
  mappings wouldn't get updated after a copy-on-write took place.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org

James Hogan (13):
  KVM: MIPS/T&E: Ignore user writes to CP0_Config7
  KVM: MIPS: Pass type of fault down to kvm_mips_map_page()
  KVM: MIPS/T&E: Abstract bad access handling
  KVM: MIPS/T&E: Treat unhandled guest KSeg0 as MMIO
  KVM: MIPS/T&E: Handle read only GPA in TLB mod
  KVM: MIPS/MMU: Add GPA PT mkclean helper
  KVM: MIPS/MMU: Use generic dirty log & protect helper
  KVM: MIPS: Clean & flush on dirty page logging enable
  KVM: MIPS/MMU: Handle dirty logging on GPA faults
  KVM: MIPS/MMU: Pass GPA PTE bits to KSeg0 GVA PTEs
  KVM: MIPS/MMU: Pass GPA PTE bits to mapped GVA PTEs
  KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU
  KVM: MIPS: Claim KVM_CAP_READONLY_MEM support

 arch/mips/include/asm/kvm_host.h |  28 +-
 arch/mips/include/uapi/asm/kvm.h |   2 +-
 arch/mips/kvm/Kconfig            |   2 +-
 arch/mips/kvm/emulate.c          |  38 +--
 arch/mips/kvm/mips.c             |  68 ++--
 arch/mips/kvm/mmu.c              | 546 ++++++++++++++++++++++++++++----
 arch/mips/kvm/trap_emul.c        | 203 +++++++-----
 7 files changed, 693 insertions(+), 194 deletions(-)

-- 
git-series 0.8.10

[PATCH 0/13] KVM: MIPS: Dirty logging, SYNC_MMU & READONLY_MEM

[James Hogan]

Note: My intention is to take this series via the MIPS KVM tree. This
series is based on my other series posted this cycle:
[0/10] KVM: MIPS: Implement GVA page tables and shadow flushing
[0/3]  KVM: MIPS: Use CP0_BadInstr[P] for emulation
[0/30] KVM: MIPS: Implement GVA page tables

This series adds proper dirty page logging, KVM_CAP_SYNC_MMU, and
KVM_CAP_READONLY_MEM support to MIPS KVM.

The existing dirty page logging support (integral to live migration) was
basically incomplete as pages were never marked read only or recorded as
being dirty on a TLB modified exception. This series incrementally adds
proper support, using the dirty bit in the GPA and GVA page tables to
trigger TLB modified exceptions.

Support is added for KVM_CAP_SYNC_MMU, which adds MMU notifiers so that
KVM can react to asynchronous (and synchronous) host virtual MM changes.
This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning, as well as paving the way for KVM_CAP_READONLY_MEM.

Finally support is added for read only memory regions
(KVM_CAP_READONLY_MEM), which can be supported fairly minimally once
dirty page logging and KVM_CAP_SYNC_MMU are in place. This allows memory
regions to be marked read only to the guest so that reads work but
writes trigger MMIO.

The patches are roughly grouped as follows:

Patches 1-2:
  Preliminary changes for readonly regions / dirty logging.

Patches 3-5:
  Add handling of read only regions (or areas with no region) as MMIO,
  and also call fault handling for dirty logging.

Patches 6-11:
  Incrementally add proper dirty logging support, first making ranges of
  GPA clean when log read or region is made readonly, then adding the
  fast path fault handling for dirty logging, and finally transferring
  GPA protection bits (including dirty) across to GVA page tables so
  they take effect.

Patch 12:
  Add KVM_CAP_SYNC_MMU support.

Patch 13:
  Enable KVM_CAP_READONLY_MEM. This is after KVM_CAP_SYNC_MMU support as
  it needs to use gfn_to_pfn_prot() to know whether the GFN is read
  only, which opens up the possibility of getting copy-on-write enabled
  pages, which aren't safe without KVM_CAP_SYNC_MMU since the guest
  mappings wouldn't get updated after a copy-on-write took place.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org

James Hogan (13):
  KVM: MIPS/T&E: Ignore user writes to CP0_Config7
  KVM: MIPS: Pass type of fault down to kvm_mips_map_page()
  KVM: MIPS/T&E: Abstract bad access handling
  KVM: MIPS/T&E: Treat unhandled guest KSeg0 as MMIO
  KVM: MIPS/T&E: Handle read only GPA in TLB mod
  KVM: MIPS/MMU: Add GPA PT mkclean helper
  KVM: MIPS/MMU: Use generic dirty log & protect helper
  KVM: MIPS: Clean & flush on dirty page logging enable
  KVM: MIPS/MMU: Handle dirty logging on GPA faults
  KVM: MIPS/MMU: Pass GPA PTE bits to KSeg0 GVA PTEs
  KVM: MIPS/MMU: Pass GPA PTE bits to mapped GVA PTEs
  KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU
  KVM: MIPS: Claim KVM_CAP_READONLY_MEM support

 arch/mips/include/asm/kvm_host.h |  28 +-
 arch/mips/include/uapi/asm/kvm.h |   2 +-
 arch/mips/kvm/Kconfig            |   2 +-
 arch/mips/kvm/emulate.c          |  38 +--
 arch/mips/kvm/mips.c             |  68 ++--
 arch/mips/kvm/mmu.c              | 546 ++++++++++++++++++++++++++++----
 arch/mips/kvm/trap_emul.c        | 203 +++++++-----
 7 files changed, 693 insertions(+), 194 deletions(-)

-- 
git-series 0.8.10

[PATCH 1/13] KVM: MIPS/T&E: Ignore user writes to CP0_Config7

[James Hogan]

Ignore userland writes to CP0_Config7 rather than reporting an error,
since we do allow reads of this register and it is claimed to exist in
the ioctl API.

This allows userland to blindly save and restore KVM registers without
having to special case certain registers as not being writable, for
example during live migration once dirty page logging is fixed.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/trap_emul.c | 3 +++
 1 file changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index a9a8c8e450e4..970ab216b355 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -705,6 +705,9 @@ static int kvm_trap_emul_set_one_reg(struct kvm_vcpu *vcpu,
 			kvm_write_c0_guest_config5(cop0, v);
 		}
 		break;
+	case KVM_REG_MIPS_CP0_CONFIG7:
+		/* writes ignored */
+		break;
 	case KVM_REG_MIPS_COUNT_CTL:
 		ret = kvm_mips_set_count_ctl(vcpu, v);
 		break;
-- 
git-series 0.8.10

[PATCH 1/13] KVM: MIPS/T&E: Ignore user writes to CP0_Config7

[James Hogan]

Ignore userland writes to CP0_Config7 rather than reporting an error,
since we do allow reads of this register and it is claimed to exist in
the ioctl API.

This allows userland to blindly save and restore KVM registers without
having to special case certain registers as not being writable, for
example during live migration once dirty page logging is fixed.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/trap_emul.c | 3 +++
 1 file changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index a9a8c8e450e4..970ab216b355 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -705,6 +705,9 @@ static int kvm_trap_emul_set_one_reg(struct kvm_vcpu *vcpu,
 			kvm_write_c0_guest_config5(cop0, v);
 		}
 		break;
+	case KVM_REG_MIPS_CP0_CONFIG7:
+		/* writes ignored */
+		break;
 	case KVM_REG_MIPS_COUNT_CTL:
 		ret = kvm_mips_set_count_ctl(vcpu, v);
 		break;
-- 
git-series 0.8.10

[PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

kvm_mips_map_page() will need to know whether the fault was due to a
read or a write in order to support dirty page tracking,
KVM_CAP_SYNC_MMU, and read only memory regions, so get that information
passed down to it via new bool write_fault arguments to various
functions.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  9 ++++++---
 arch/mips/kvm/emulate.c          |  7 ++++---
 arch/mips/kvm/mmu.c              | 21 +++++++++++++--------
 arch/mips/kvm/trap_emul.c        |  4 ++--
 4 files changed, 25 insertions(+), 16 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index c24c1c23196b..70c2dd353468 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -597,19 +597,22 @@ u32 kvm_get_user_asid(struct kvm_vcpu *vcpu);
 u32 kvm_get_commpage_asid (struct kvm_vcpu *vcpu);
 
 extern int kvm_mips_handle_kseg0_tlb_fault(unsigned long badbaddr,
-					   struct kvm_vcpu *vcpu);
+					   struct kvm_vcpu *vcpu,
+					   bool write_fault);
 
 extern int kvm_mips_handle_commpage_tlb_fault(unsigned long badvaddr,
 					      struct kvm_vcpu *vcpu);
 
 extern int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 						struct kvm_mips_tlb *tlb,
-						unsigned long gva);
+						unsigned long gva,
+						bool write_fault);
 
 extern enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 						     u32 *opc,
 						     struct kvm_run *run,
-						     struct kvm_vcpu *vcpu);
+						     struct kvm_vcpu *vcpu,
+						     bool write_fault);
 
 extern enum emulation_result kvm_mips_handle_tlbmod(u32 cause,
 						    u32 *opc,
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index 72eb307a61a7..a47f8af9193e 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -2705,7 +2705,8 @@ enum emulation_result kvm_mips_check_privilege(u32 cause,
 enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 					      u32 *opc,
 					      struct kvm_run *run,
-					      struct kvm_vcpu *vcpu)
+					      struct kvm_vcpu *vcpu,
+					      bool write_fault)
 {
 	enum emulation_result er = EMULATE_DONE;
 	u32 exccode = (cause >> CAUSEB_EXCCODE) & 0x1f;
@@ -2761,8 +2762,8 @@ enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 			 * OK we have a Guest TLB entry, now inject it into the
 			 * shadow host TLB
 			 */
-			if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb,
-								 va)) {
+			if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, va,
+								 write_fault)) {
 				kvm_err("%s: handling mapped seg tlb fault for %lx, index: %u, vcpu: %p, ASID: %#lx\n",
 					__func__, va, index, vcpu,
 					read_c0_entryhi());
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index b3da473e1569..1af65f2e6bb7 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -308,6 +308,7 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
+ * @write_fault:	Whether the fault was due to a write.
  * @out_entry:		New PTE for @gpa (written on success unless NULL).
  * @out_buddy:		New PTE for @gpa's buddy (written on success unless
  *			NULL).
@@ -327,6 +328,7 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
  *		as an MMIO access.
  */
 static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
+			     bool write_fault,
 			     pte_t *out_entry, pte_t *out_buddy)
 {
 	struct kvm *kvm = vcpu->kvm;
@@ -558,7 +560,8 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
 
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
-				    struct kvm_vcpu *vcpu)
+				    struct kvm_vcpu *vcpu,
+				    bool write_fault)
 {
 	unsigned long gpa;
 	kvm_pfn_t pfn0, pfn1;
@@ -576,10 +579,11 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 	gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
 	vaddr = badvaddr & (PAGE_MASK << 1);
 
-	if (kvm_mips_map_page(vcpu, gpa, &pte_gpa[0], NULL) < 0)
+	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[0], NULL) < 0)
 		return -1;
 
-	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, &pte_gpa[1], NULL) < 0)
+	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, write_fault, &pte_gpa[1],
+			      NULL) < 0)
 		return -1;
 
 	pfn0 = pte_pfn(pte_gpa[0]);
@@ -604,7 +608,8 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 
 int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 					 struct kvm_mips_tlb *tlb,
-					 unsigned long gva)
+					 unsigned long gva,
+					 bool write_fault)
 {
 	kvm_pfn_t pfn;
 	long tlb_lo = 0;
@@ -621,8 +626,8 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 		tlb_lo = tlb->tlb_lo[idx];
 
 	/* Find host PFN */
-	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), &pte_gpa,
-			      NULL) < 0)
+	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), write_fault,
+			      &pte_gpa, NULL) < 0)
 		return -1;
 	pfn = pte_pfn(pte_gpa);
 
@@ -757,7 +762,7 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 	int index;
 
 	if (KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG0) {
-		if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu) < 0)
+		if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu, write) < 0)
 			return KVM_MIPS_GPA;
 	} else if ((KVM_GUEST_KSEGX(gva) < KVM_GUEST_KSEG0) ||
 		   KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG23) {
@@ -774,7 +779,7 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 		if (write && !TLB_IS_DIRTY(*tlb, gva))
 			return KVM_MIPS_TLBMOD;
 
-		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva))
+		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva, write))
 			return KVM_MIPS_GPA;
 	} else {
 		return KVM_MIPS_GVA;
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 970ab216b355..33888f1a89f4 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -159,7 +159,7 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 *     into the shadow host TLB
 		 */
 
-		er = kvm_mips_handle_tlbmiss(cause, opc, run, vcpu);
+		er = kvm_mips_handle_tlbmiss(cause, opc, run, vcpu, store);
 		if (er == EMULATE_DONE)
 			ret = RESUME_GUEST;
 		else {
@@ -172,7 +172,7 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 * not expect to ever get them
 		 */
 		if (kvm_mips_handle_kseg0_tlb_fault
-		    (vcpu->arch.host_cp0_badvaddr, vcpu) < 0) {
+		    (vcpu->arch.host_cp0_badvaddr, vcpu, store) < 0) {
 			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 			ret = RESUME_HOST;
 		}
-- 
git-series 0.8.10

[PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

kvm_mips_map_page() will need to know whether the fault was due to a
read or a write in order to support dirty page tracking,
KVM_CAP_SYNC_MMU, and read only memory regions, so get that information
passed down to it via new bool write_fault arguments to various
functions.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  9 ++++++---
 arch/mips/kvm/emulate.c          |  7 ++++---
 arch/mips/kvm/mmu.c              | 21 +++++++++++++--------
 arch/mips/kvm/trap_emul.c        |  4 ++--
 4 files changed, 25 insertions(+), 16 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index c24c1c23196b..70c2dd353468 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -597,19 +597,22 @@ u32 kvm_get_user_asid(struct kvm_vcpu *vcpu);
 u32 kvm_get_commpage_asid (struct kvm_vcpu *vcpu);
 
 extern int kvm_mips_handle_kseg0_tlb_fault(unsigned long badbaddr,
-					   struct kvm_vcpu *vcpu);
+					   struct kvm_vcpu *vcpu,
+					   bool write_fault);
 
 extern int kvm_mips_handle_commpage_tlb_fault(unsigned long badvaddr,
 					      struct kvm_vcpu *vcpu);
 
 extern int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 						struct kvm_mips_tlb *tlb,
-						unsigned long gva);
+						unsigned long gva,
+						bool write_fault);
 
 extern enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 						     u32 *opc,
 						     struct kvm_run *run,
-						     struct kvm_vcpu *vcpu);
+						     struct kvm_vcpu *vcpu,
+						     bool write_fault);
 
 extern enum emulation_result kvm_mips_handle_tlbmod(u32 cause,
 						    u32 *opc,
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index 72eb307a61a7..a47f8af9193e 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -2705,7 +2705,8 @@ enum emulation_result kvm_mips_check_privilege(u32 cause,
 enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 					      u32 *opc,
 					      struct kvm_run *run,
-					      struct kvm_vcpu *vcpu)
+					      struct kvm_vcpu *vcpu,
+					      bool write_fault)
 {
 	enum emulation_result er = EMULATE_DONE;
 	u32 exccode = (cause >> CAUSEB_EXCCODE) & 0x1f;
@@ -2761,8 +2762,8 @@ enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 			 * OK we have a Guest TLB entry, now inject it into the
 			 * shadow host TLB
 			 */
-			if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb,
-								 va)) {
+			if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, va,
+								 write_fault)) {
 				kvm_err("%s: handling mapped seg tlb fault for %lx, index: %u, vcpu: %p, ASID: %#lx\n",
 					__func__, va, index, vcpu,
 					read_c0_entryhi());
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index b3da473e1569..1af65f2e6bb7 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -308,6 +308,7 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
+ * @write_fault:	Whether the fault was due to a write.
  * @out_entry:		New PTE for @gpa (written on success unless NULL).
  * @out_buddy:		New PTE for @gpa's buddy (written on success unless
  *			NULL).
@@ -327,6 +328,7 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
  *		as an MMIO access.
  */
 static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
+			     bool write_fault,
 			     pte_t *out_entry, pte_t *out_buddy)
 {
 	struct kvm *kvm = vcpu->kvm;
@@ -558,7 +560,8 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
 
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
-				    struct kvm_vcpu *vcpu)
+				    struct kvm_vcpu *vcpu,
+				    bool write_fault)
 {
 	unsigned long gpa;
 	kvm_pfn_t pfn0, pfn1;
@@ -576,10 +579,11 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 	gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
 	vaddr = badvaddr & (PAGE_MASK << 1);
 
-	if (kvm_mips_map_page(vcpu, gpa, &pte_gpa[0], NULL) < 0)
+	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[0], NULL) < 0)
 		return -1;
 
-	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, &pte_gpa[1], NULL) < 0)
+	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, write_fault, &pte_gpa[1],
+			      NULL) < 0)
 		return -1;
 
 	pfn0 = pte_pfn(pte_gpa[0]);
@@ -604,7 +608,8 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 
 int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 					 struct kvm_mips_tlb *tlb,
-					 unsigned long gva)
+					 unsigned long gva,
+					 bool write_fault)
 {
 	kvm_pfn_t pfn;
 	long tlb_lo = 0;
@@ -621,8 +626,8 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 		tlb_lo = tlb->tlb_lo[idx];
 
 	/* Find host PFN */
-	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), &pte_gpa,
-			      NULL) < 0)
+	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), write_fault,
+			      &pte_gpa, NULL) < 0)
 		return -1;
 	pfn = pte_pfn(pte_gpa);
 
@@ -757,7 +762,7 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 	int index;
 
 	if (KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG0) {
-		if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu) < 0)
+		if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu, write) < 0)
 			return KVM_MIPS_GPA;
 	} else if ((KVM_GUEST_KSEGX(gva) < KVM_GUEST_KSEG0) ||
 		   KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG23) {
@@ -774,7 +779,7 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 		if (write && !TLB_IS_DIRTY(*tlb, gva))
 			return KVM_MIPS_TLBMOD;
 
-		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva))
+		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva, write))
 			return KVM_MIPS_GPA;
 	} else {
 		return KVM_MIPS_GVA;
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 970ab216b355..33888f1a89f4 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -159,7 +159,7 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 *     into the shadow host TLB
 		 */
 
-		er = kvm_mips_handle_tlbmiss(cause, opc, run, vcpu);
+		er = kvm_mips_handle_tlbmiss(cause, opc, run, vcpu, store);
 		if (er == EMULATE_DONE)
 			ret = RESUME_GUEST;
 		else {
@@ -172,7 +172,7 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 * not expect to ever get them
 		 */
 		if (kvm_mips_handle_kseg0_tlb_fault
-		    (vcpu->arch.host_cp0_badvaddr, vcpu) < 0) {
+		    (vcpu->arch.host_cp0_badvaddr, vcpu, store) < 0) {
 			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 			ret = RESUME_HOST;
 		}
-- 
git-series 0.8.10

[PATCH 3/13] KVM: MIPS/T&E: Abstract bad access handling

[James Hogan]

Abstract the handling of bad guest loads and stores which may need to
trigger an MMIO, so that the same code can be used in a later patch for
guest KSeg0 addresses (TLB exception handling) as well as for host KSeg1
addresses (existing address error exception and TLB exception handling).

We now use kvm_mips_emulate_store() and kvm_mips_emulate_load() directly
rather than the more generic kvm_mips_emulate_inst(), as there is no
need to expose emulation of any other instructions.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/trap_emul.c | 119 +++++++++++++++++++++++----------------
 1 file changed, 72 insertions(+), 47 deletions(-)

diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 33888f1a89f4..79c30a0baad1 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -85,6 +85,75 @@ static int kvm_trap_emul_handle_cop_unusable(struct kvm_vcpu *vcpu)
 	return ret;
 }
 
+static int kvm_mips_bad_load(u32 cause, u32 *opc, struct kvm_run *run,
+			     struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er;
+	union mips_instruction inst;
+	int err;
+
+	/* A code fetch fault doesn't count as an MMIO */
+	if (kvm_is_ifetch_fault(&vcpu->arch)) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Fetch the instruction. */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Emulate the load */
+	er = kvm_mips_emulate_load(inst, cause, run, vcpu);
+	if (er == EMULATE_FAIL) {
+		kvm_err("Emulate load from MMIO space failed\n");
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+	} else {
+		run->exit_reason = KVM_EXIT_MMIO;
+	}
+	return RESUME_HOST;
+}
+
+static int kvm_mips_bad_store(u32 cause, u32 *opc, struct kvm_run *run,
+			      struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er;
+	union mips_instruction inst;
+	int err;
+
+	/* Fetch the instruction. */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Emulate the store */
+	er = kvm_mips_emulate_store(inst, cause, run, vcpu);
+	if (er == EMULATE_FAIL) {
+		kvm_err("Emulate store to MMIO space failed\n");
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+	} else {
+		run->exit_reason = KVM_EXIT_MMIO;
+	}
+	return RESUME_HOST;
+}
+
+static int kvm_mips_bad_access(u32 cause, u32 *opc, struct kvm_run *run,
+			       struct kvm_vcpu *vcpu, bool store)
+{
+	if (store)
+		return kvm_mips_bad_store(cause, opc, run, vcpu);
+	else
+		return kvm_mips_bad_load(cause, opc, run, vcpu);
+}
+
 static int kvm_trap_emul_handle_tlb_mod(struct kvm_vcpu *vcpu)
 {
 	struct kvm_run *run = vcpu->run;
@@ -178,28 +247,11 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		}
 	} else if (KVM_GUEST_KERNEL_MODE(vcpu)
 		   && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
-		/* A code fetch fault doesn't count as an MMIO */
-		if (!store && kvm_is_ifetch_fault(&vcpu->arch)) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			return RESUME_HOST;
-		}
-
 		/*
 		 * With EVA we may get a TLB exception instead of an address
 		 * error when the guest performs MMIO to KSeg1 addresses.
 		 */
-		kvm_debug("Emulate %s MMIO space\n",
-			  store ? "Store to" : "Load from");
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate %s MMIO space failed\n",
-				store ? "Store to" : "Load from");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_access(cause, opc, run, vcpu, store);
 	} else {
 		kvm_err("Illegal TLB %s fault address , cause %#x, PC: %p, BadVaddr: %#lx\n",
 			store ? "ST" : "LD", cause, opc, badvaddr);
@@ -227,21 +279,11 @@ static int kvm_trap_emul_handle_addr_err_st(struct kvm_vcpu *vcpu)
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
 	int ret = RESUME_GUEST;
 
 	if (KVM_GUEST_KERNEL_MODE(vcpu)
 	    && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
-		kvm_debug("Emulate Store to MMIO space\n");
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate Store to MMIO space failed\n");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_store(cause, opc, run, vcpu);
 	} else {
 		kvm_err("Address Error (STORE): cause %#x, PC: %p, BadVaddr: %#lx\n",
 			cause, opc, badvaddr);
@@ -257,32 +299,15 @@ static int kvm_trap_emul_handle_addr_err_ld(struct kvm_vcpu *vcpu)
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
 	int ret = RESUME_GUEST;
 
 	if (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1) {
-		/* A code fetch fault doesn't count as an MMIO */
-		if (kvm_is_ifetch_fault(&vcpu->arch)) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			return RESUME_HOST;
-		}
-
-		kvm_debug("Emulate Load from MMIO space @ %#lx\n", badvaddr);
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate Load from MMIO space failed\n");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_load(cause, opc, run, vcpu);
 	} else {
 		kvm_err("Address Error (LOAD): cause %#x, PC: %p, BadVaddr: %#lx\n",
 			cause, opc, badvaddr);
 		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 		ret = RESUME_HOST;
-		er = EMULATE_FAIL;
 	}
 	return ret;
 }
-- 
git-series 0.8.10

[PATCH 3/13] KVM: MIPS/T&E: Abstract bad access handling

[James Hogan]

Abstract the handling of bad guest loads and stores which may need to
trigger an MMIO, so that the same code can be used in a later patch for
guest KSeg0 addresses (TLB exception handling) as well as for host KSeg1
addresses (existing address error exception and TLB exception handling).

We now use kvm_mips_emulate_store() and kvm_mips_emulate_load() directly
rather than the more generic kvm_mips_emulate_inst(), as there is no
need to expose emulation of any other instructions.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/trap_emul.c | 119 +++++++++++++++++++++++----------------
 1 file changed, 72 insertions(+), 47 deletions(-)

diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 33888f1a89f4..79c30a0baad1 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -85,6 +85,75 @@ static int kvm_trap_emul_handle_cop_unusable(struct kvm_vcpu *vcpu)
 	return ret;
 }
 
+static int kvm_mips_bad_load(u32 cause, u32 *opc, struct kvm_run *run,
+			     struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er;
+	union mips_instruction inst;
+	int err;
+
+	/* A code fetch fault doesn't count as an MMIO */
+	if (kvm_is_ifetch_fault(&vcpu->arch)) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Fetch the instruction. */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Emulate the load */
+	er = kvm_mips_emulate_load(inst, cause, run, vcpu);
+	if (er == EMULATE_FAIL) {
+		kvm_err("Emulate load from MMIO space failed\n");
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+	} else {
+		run->exit_reason = KVM_EXIT_MMIO;
+	}
+	return RESUME_HOST;
+}
+
+static int kvm_mips_bad_store(u32 cause, u32 *opc, struct kvm_run *run,
+			      struct kvm_vcpu *vcpu)
+{
+	enum emulation_result er;
+	union mips_instruction inst;
+	int err;
+
+	/* Fetch the instruction. */
+	if (cause & CAUSEF_BD)
+		opc += 1;
+	err = kvm_get_badinstr(opc, vcpu, &inst.word);
+	if (err) {
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+		return RESUME_HOST;
+	}
+
+	/* Emulate the store */
+	er = kvm_mips_emulate_store(inst, cause, run, vcpu);
+	if (er == EMULATE_FAIL) {
+		kvm_err("Emulate store to MMIO space failed\n");
+		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+	} else {
+		run->exit_reason = KVM_EXIT_MMIO;
+	}
+	return RESUME_HOST;
+}
+
+static int kvm_mips_bad_access(u32 cause, u32 *opc, struct kvm_run *run,
+			       struct kvm_vcpu *vcpu, bool store)
+{
+	if (store)
+		return kvm_mips_bad_store(cause, opc, run, vcpu);
+	else
+		return kvm_mips_bad_load(cause, opc, run, vcpu);
+}
+
 static int kvm_trap_emul_handle_tlb_mod(struct kvm_vcpu *vcpu)
 {
 	struct kvm_run *run = vcpu->run;
@@ -178,28 +247,11 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		}
 	} else if (KVM_GUEST_KERNEL_MODE(vcpu)
 		   && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
-		/* A code fetch fault doesn't count as an MMIO */
-		if (!store && kvm_is_ifetch_fault(&vcpu->arch)) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			return RESUME_HOST;
-		}
-
 		/*
 		 * With EVA we may get a TLB exception instead of an address
 		 * error when the guest performs MMIO to KSeg1 addresses.
 		 */
-		kvm_debug("Emulate %s MMIO space\n",
-			  store ? "Store to" : "Load from");
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate %s MMIO space failed\n",
-				store ? "Store to" : "Load from");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_access(cause, opc, run, vcpu, store);
 	} else {
 		kvm_err("Illegal TLB %s fault address , cause %#x, PC: %p, BadVaddr: %#lx\n",
 			store ? "ST" : "LD", cause, opc, badvaddr);
@@ -227,21 +279,11 @@ static int kvm_trap_emul_handle_addr_err_st(struct kvm_vcpu *vcpu)
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
 	int ret = RESUME_GUEST;
 
 	if (KVM_GUEST_KERNEL_MODE(vcpu)
 	    && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
-		kvm_debug("Emulate Store to MMIO space\n");
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate Store to MMIO space failed\n");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_store(cause, opc, run, vcpu);
 	} else {
 		kvm_err("Address Error (STORE): cause %#x, PC: %p, BadVaddr: %#lx\n",
 			cause, opc, badvaddr);
@@ -257,32 +299,15 @@ static int kvm_trap_emul_handle_addr_err_ld(struct kvm_vcpu *vcpu)
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
 	int ret = RESUME_GUEST;
 
 	if (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1) {
-		/* A code fetch fault doesn't count as an MMIO */
-		if (kvm_is_ifetch_fault(&vcpu->arch)) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			return RESUME_HOST;
-		}
-
-		kvm_debug("Emulate Load from MMIO space @ %#lx\n", badvaddr);
-		er = kvm_mips_emulate_inst(cause, opc, run, vcpu);
-		if (er == EMULATE_FAIL) {
-			kvm_err("Emulate Load from MMIO space failed\n");
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		} else {
-			run->exit_reason = KVM_EXIT_MMIO;
-			ret = RESUME_HOST;
-		}
+		ret = kvm_mips_bad_load(cause, opc, run, vcpu);
 	} else {
 		kvm_err("Address Error (LOAD): cause %#x, PC: %p, BadVaddr: %#lx\n",
 			cause, opc, badvaddr);
 		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
 		ret = RESUME_HOST;
-		er = EMULATE_FAIL;
 	}
 	return ret;
 }
-- 
git-series 0.8.10

[PATCH 4/13] KVM: MIPS/T&E: Treat unhandled guest KSeg0 as MMIO

[James Hogan]

Treat unhandled accesses to guest KSeg0 as MMIO, rather than only host
KSeg0 addresses. This will allow read only memory regions (such as the
Malta boot flash as emulated by QEMU) to have writes (before reads)
treated as MMIO, and unallocated physical addresses to have all accesses
treated as MMIO.

The MMIO emulation uses the gva_to_gpa callback, so this is also updated
for trap & emulate to handle guest KSeg0 addresses.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c       |  1 -
 arch/mips/kvm/trap_emul.c | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 1af65f2e6bb7..934bcc3732da 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -350,7 +350,6 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	pfn = gfn_to_pfn(kvm, gfn);
 
 	if (is_error_noslot_pfn(pfn)) {
-		kvm_err("Couldn't get pfn for gfn %#llx!\n", gfn);
 		err = -EFAULT;
 		goto out;
 	}
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 79c30a0baad1..236390db6219 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -23,9 +23,12 @@ static gpa_t kvm_trap_emul_gva_to_gpa_cb(gva_t gva)
 {
 	gpa_t gpa;
 	gva_t kseg = KSEGX(gva);
+	gva_t gkseg = KVM_GUEST_KSEGX(gva);
 
 	if ((kseg == CKSEG0) || (kseg == CKSEG1))
 		gpa = CPHYSADDR(gva);
+	else if (gkseg == KVM_GUEST_KSEG0)
+		gpa = KVM_GUEST_CPHYSADDR(gva);
 	else {
 		kvm_err("%s: cannot find GPA for GVA: %#lx\n", __func__, gva);
 		kvm_mips_dump_host_tlbs();
@@ -240,11 +243,8 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 * All KSEG0 faults are handled by KVM, as the guest kernel does
 		 * not expect to ever get them
 		 */
-		if (kvm_mips_handle_kseg0_tlb_fault
-		    (vcpu->arch.host_cp0_badvaddr, vcpu, store) < 0) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		}
+		if (kvm_mips_handle_kseg0_tlb_fault(badvaddr, vcpu, store) < 0)
+			ret = kvm_mips_bad_access(cause, opc, run, vcpu, store);
 	} else if (KVM_GUEST_KERNEL_MODE(vcpu)
 		   && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
 		/*
-- 
git-series 0.8.10

[PATCH 4/13] KVM: MIPS/T&E: Treat unhandled guest KSeg0 as MMIO

[James Hogan]

Treat unhandled accesses to guest KSeg0 as MMIO, rather than only host
KSeg0 addresses. This will allow read only memory regions (such as the
Malta boot flash as emulated by QEMU) to have writes (before reads)
treated as MMIO, and unallocated physical addresses to have all accesses
treated as MMIO.

The MMIO emulation uses the gva_to_gpa callback, so this is also updated
for trap & emulate to handle guest KSeg0 addresses.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c       |  1 -
 arch/mips/kvm/trap_emul.c | 10 +++++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 1af65f2e6bb7..934bcc3732da 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -350,7 +350,6 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	pfn = gfn_to_pfn(kvm, gfn);
 
 	if (is_error_noslot_pfn(pfn)) {
-		kvm_err("Couldn't get pfn for gfn %#llx!\n", gfn);
 		err = -EFAULT;
 		goto out;
 	}
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 79c30a0baad1..236390db6219 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -23,9 +23,12 @@ static gpa_t kvm_trap_emul_gva_to_gpa_cb(gva_t gva)
 {
 	gpa_t gpa;
 	gva_t kseg = KSEGX(gva);
+	gva_t gkseg = KVM_GUEST_KSEGX(gva);
 
 	if ((kseg == CKSEG0) || (kseg == CKSEG1))
 		gpa = CPHYSADDR(gva);
+	else if (gkseg == KVM_GUEST_KSEG0)
+		gpa = KVM_GUEST_CPHYSADDR(gva);
 	else {
 		kvm_err("%s: cannot find GPA for GVA: %#lx\n", __func__, gva);
 		kvm_mips_dump_host_tlbs();
@@ -240,11 +243,8 @@ static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
 		 * All KSEG0 faults are handled by KVM, as the guest kernel does
 		 * not expect to ever get them
 		 */
-		if (kvm_mips_handle_kseg0_tlb_fault
-		    (vcpu->arch.host_cp0_badvaddr, vcpu, store) < 0) {
-			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
-		}
+		if (kvm_mips_handle_kseg0_tlb_fault(badvaddr, vcpu, store) < 0)
+			ret = kvm_mips_bad_access(cause, opc, run, vcpu, store);
 	} else if (KVM_GUEST_KERNEL_MODE(vcpu)
 		   && (KSEGX(badvaddr) == CKSEG0 || KSEGX(badvaddr) == CKSEG1)) {
 		/*
-- 
git-series 0.8.10

[PATCH 5/13] KVM: MIPS/T&E: Handle read only GPA in TLB mod

[James Hogan]

Rewrite TLB modified exception handling to handle read only GPA memory
regions, instead of unconditionally passing the exception to the guest.

If the guest TLB is not the cause of the exception we call into the
normal TLB fault handling depending on the memory segment, which will
soon attempt to remap the physical page to be writable (handling dirty
page tracking or copy on write in the process).

Failing that we fall back to treating it as MMIO, due to a read only
memory region. Once the capability is enabled, this will allow read only
memory regions (such as the Malta boot flash as emulated by QEMU) to
have writes treated as MMIO, while still allowing reads to run
untrapped.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  5 +--
 arch/mips/kvm/emulate.c          | 31 +---------------
 arch/mips/kvm/trap_emul.c        | 69 ++++++++++++++++++++-------------
 3 files changed, 43 insertions(+), 62 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 70c2dd353468..da401a75a204 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -614,11 +614,6 @@ extern enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 						     struct kvm_vcpu *vcpu,
 						     bool write_fault);
 
-extern enum emulation_result kvm_mips_handle_tlbmod(u32 cause,
-						    u32 *opc,
-						    struct kvm_run *run,
-						    struct kvm_vcpu *vcpu);
-
 extern void kvm_mips_dump_host_tlbs(void);
 extern void kvm_mips_dump_guest_tlbs(struct kvm_vcpu *vcpu);
 extern int kvm_mips_host_tlb_inv(struct kvm_vcpu *vcpu, unsigned long entryhi,
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index a47f8af9193e..eaa2fa091808 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -2125,37 +2125,6 @@ enum emulation_result kvm_mips_emulate_tlbinv_st(u32 cause,
 	return EMULATE_DONE;
 }
 
-/* TLBMOD: store into address matching TLB with Dirty bit off */
-enum emulation_result kvm_mips_handle_tlbmod(u32 cause, u32 *opc,
-					     struct kvm_run *run,
-					     struct kvm_vcpu *vcpu)
-{
-	enum emulation_result er = EMULATE_DONE;
-#ifdef DEBUG
-	struct mips_coproc *cop0 = vcpu->arch.cop0;
-	unsigned long entryhi = (vcpu->arch.host_cp0_badvaddr & VPN2_MASK) |
-			(kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID);
-	bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
-	int index;
-
-	/* If address not in the guest TLB, then we are in trouble */
-	index = kvm_mips_guest_tlb_lookup(vcpu, entryhi);
-	if (index < 0) {
-		/* XXXKYMA Invalidate and retry */
-		kvm_mips_host_tlb_inv(vcpu, vcpu->arch.host_cp0_badvaddr,
-				      !kernel, kernel);
-		kvm_err("%s: host got TLBMOD for %#lx but entry not present in Guest TLB\n",
-		     __func__, entryhi);
-		kvm_mips_dump_guest_tlbs(vcpu);
-		kvm_mips_dump_host_tlbs();
-		return EMULATE_FAIL;
-	}
-#endif
-
-	er = kvm_mips_emulate_tlbmod(cause, opc, run, vcpu);
-	return er;
-}
-
 enum emulation_result kvm_mips_emulate_tlbmod(u32 cause,
 					      u32 *opc,
 					      struct kvm_run *run,
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 236390db6219..da5acd0ac005 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -159,46 +159,63 @@ static int kvm_mips_bad_access(u32 cause, u32 *opc, struct kvm_run *run,
 
 static int kvm_trap_emul_handle_tlb_mod(struct kvm_vcpu *vcpu)
 {
+	struct mips_coproc *cop0 = vcpu->arch.cop0;
 	struct kvm_run *run = vcpu->run;
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
-	int ret = RESUME_GUEST;
+	struct kvm_mips_tlb *tlb;
+	unsigned long entryhi;
+	int index;
 
 	if (KVM_GUEST_KSEGX(badvaddr) < KVM_GUEST_KSEG0
 	    || KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG23) {
-		kvm_debug("USER/KSEG23 ADDR TLB MOD fault: cause %#x, PC: %p, BadVaddr: %#lx\n",
-			  cause, opc, badvaddr);
-		er = kvm_mips_handle_tlbmod(cause, opc, run, vcpu);
+		/*
+		 * First find the mapping in the guest TLB. If the failure to
+		 * write was due to the guest TLB, it should be up to the guest
+		 * to handle it.
+		 */
+		entryhi = (badvaddr & VPN2_MASK) |
+			  (kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID);
+		index = kvm_mips_guest_tlb_lookup(vcpu, entryhi);
 
-		if (er == EMULATE_DONE)
-			ret = RESUME_GUEST;
-		else {
+		/*
+		 * These should never happen.
+		 * They would indicate stale host TLB entries.
+		 */
+		if (unlikely(index < 0)) {
 			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
+			return RESUME_HOST;
 		}
-	} else if (KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG0) {
+		tlb = vcpu->arch.guest_tlb + index;
+		if (unlikely(!TLB_IS_VALID(*tlb, badvaddr))) {
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+			return RESUME_HOST;
+		}
+
 		/*
-		 * XXXKYMA: The guest kernel does not expect to get this fault
-		 * when we are not using HIGHMEM. Need to address this in a
-		 * HIGHMEM kernel
+		 * Guest entry not dirty? That would explain the TLB modified
+		 * exception. Relay that on to the guest so it can handle it.
 		 */
-		kvm_err("TLB MOD fault not handled, cause %#x, PC: %p, BadVaddr: %#lx\n",
-			cause, opc, badvaddr);
-		kvm_mips_dump_host_tlbs();
-		kvm_arch_vcpu_dump_regs(vcpu);
-		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-		ret = RESUME_HOST;
+		if (!TLB_IS_DIRTY(*tlb, badvaddr)) {
+			kvm_mips_emulate_tlbmod(cause, opc, run, vcpu);
+			return RESUME_GUEST;
+		}
+
+		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, badvaddr,
+							 true))
+			/* Not writable, needs handling as MMIO */
+			return kvm_mips_bad_store(cause, opc, run, vcpu);
+		return RESUME_GUEST;
+	} else if (KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG0) {
+		if (kvm_mips_handle_kseg0_tlb_fault(badvaddr, vcpu, true) < 0)
+			/* Not writable, needs handling as MMIO */
+			return kvm_mips_bad_store(cause, opc, run, vcpu);
+		return RESUME_GUEST;
 	} else {
-		kvm_err("Illegal TLB Mod fault address , cause %#x, PC: %p, BadVaddr: %#lx\n",
-			cause, opc, badvaddr);
-		kvm_mips_dump_host_tlbs();
-		kvm_arch_vcpu_dump_regs(vcpu);
-		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-		ret = RESUME_HOST;
+		/* host kernel addresses are all handled as MMIO */
+		return kvm_mips_bad_store(cause, opc, run, vcpu);
 	}
-	return ret;
 }
 
 static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
-- 
git-series 0.8.10

[PATCH 5/13] KVM: MIPS/T&E: Handle read only GPA in TLB mod

[James Hogan]

Rewrite TLB modified exception handling to handle read only GPA memory
regions, instead of unconditionally passing the exception to the guest.

If the guest TLB is not the cause of the exception we call into the
normal TLB fault handling depending on the memory segment, which will
soon attempt to remap the physical page to be writable (handling dirty
page tracking or copy on write in the process).

Failing that we fall back to treating it as MMIO, due to a read only
memory region. Once the capability is enabled, this will allow read only
memory regions (such as the Malta boot flash as emulated by QEMU) to
have writes treated as MMIO, while still allowing reads to run
untrapped.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  5 +--
 arch/mips/kvm/emulate.c          | 31 +---------------
 arch/mips/kvm/trap_emul.c        | 69 ++++++++++++++++++++-------------
 3 files changed, 43 insertions(+), 62 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 70c2dd353468..da401a75a204 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -614,11 +614,6 @@ extern enum emulation_result kvm_mips_handle_tlbmiss(u32 cause,
 						     struct kvm_vcpu *vcpu,
 						     bool write_fault);
 
-extern enum emulation_result kvm_mips_handle_tlbmod(u32 cause,
-						    u32 *opc,
-						    struct kvm_run *run,
-						    struct kvm_vcpu *vcpu);
-
 extern void kvm_mips_dump_host_tlbs(void);
 extern void kvm_mips_dump_guest_tlbs(struct kvm_vcpu *vcpu);
 extern int kvm_mips_host_tlb_inv(struct kvm_vcpu *vcpu, unsigned long entryhi,
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index a47f8af9193e..eaa2fa091808 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -2125,37 +2125,6 @@ enum emulation_result kvm_mips_emulate_tlbinv_st(u32 cause,
 	return EMULATE_DONE;
 }
 
-/* TLBMOD: store into address matching TLB with Dirty bit off */
-enum emulation_result kvm_mips_handle_tlbmod(u32 cause, u32 *opc,
-					     struct kvm_run *run,
-					     struct kvm_vcpu *vcpu)
-{
-	enum emulation_result er = EMULATE_DONE;
-#ifdef DEBUG
-	struct mips_coproc *cop0 = vcpu->arch.cop0;
-	unsigned long entryhi = (vcpu->arch.host_cp0_badvaddr & VPN2_MASK) |
-			(kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID);
-	bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
-	int index;
-
-	/* If address not in the guest TLB, then we are in trouble */
-	index = kvm_mips_guest_tlb_lookup(vcpu, entryhi);
-	if (index < 0) {
-		/* XXXKYMA Invalidate and retry */
-		kvm_mips_host_tlb_inv(vcpu, vcpu->arch.host_cp0_badvaddr,
-				      !kernel, kernel);
-		kvm_err("%s: host got TLBMOD for %#lx but entry not present in Guest TLB\n",
-		     __func__, entryhi);
-		kvm_mips_dump_guest_tlbs(vcpu);
-		kvm_mips_dump_host_tlbs();
-		return EMULATE_FAIL;
-	}
-#endif
-
-	er = kvm_mips_emulate_tlbmod(cause, opc, run, vcpu);
-	return er;
-}
-
 enum emulation_result kvm_mips_emulate_tlbmod(u32 cause,
 					      u32 *opc,
 					      struct kvm_run *run,
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 236390db6219..da5acd0ac005 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -159,46 +159,63 @@ static int kvm_mips_bad_access(u32 cause, u32 *opc, struct kvm_run *run,
 
 static int kvm_trap_emul_handle_tlb_mod(struct kvm_vcpu *vcpu)
 {
+	struct mips_coproc *cop0 = vcpu->arch.cop0;
 	struct kvm_run *run = vcpu->run;
 	u32 __user *opc = (u32 __user *) vcpu->arch.pc;
 	unsigned long badvaddr = vcpu->arch.host_cp0_badvaddr;
 	u32 cause = vcpu->arch.host_cp0_cause;
-	enum emulation_result er = EMULATE_DONE;
-	int ret = RESUME_GUEST;
+	struct kvm_mips_tlb *tlb;
+	unsigned long entryhi;
+	int index;
 
 	if (KVM_GUEST_KSEGX(badvaddr) < KVM_GUEST_KSEG0
 	    || KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG23) {
-		kvm_debug("USER/KSEG23 ADDR TLB MOD fault: cause %#x, PC: %p, BadVaddr: %#lx\n",
-			  cause, opc, badvaddr);
-		er = kvm_mips_handle_tlbmod(cause, opc, run, vcpu);
+		/*
+		 * First find the mapping in the guest TLB. If the failure to
+		 * write was due to the guest TLB, it should be up to the guest
+		 * to handle it.
+		 */
+		entryhi = (badvaddr & VPN2_MASK) |
+			  (kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID);
+		index = kvm_mips_guest_tlb_lookup(vcpu, entryhi);
 
-		if (er == EMULATE_DONE)
-			ret = RESUME_GUEST;
-		else {
+		/*
+		 * These should never happen.
+		 * They would indicate stale host TLB entries.
+		 */
+		if (unlikely(index < 0)) {
 			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-			ret = RESUME_HOST;
+			return RESUME_HOST;
 		}
-	} else if (KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG0) {
+		tlb = vcpu->arch.guest_tlb + index;
+		if (unlikely(!TLB_IS_VALID(*tlb, badvaddr))) {
+			run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
+			return RESUME_HOST;
+		}
+
 		/*
-		 * XXXKYMA: The guest kernel does not expect to get this fault
-		 * when we are not using HIGHMEM. Need to address this in a
-		 * HIGHMEM kernel
+		 * Guest entry not dirty? That would explain the TLB modified
+		 * exception. Relay that on to the guest so it can handle it.
 		 */
-		kvm_err("TLB MOD fault not handled, cause %#x, PC: %p, BadVaddr: %#lx\n",
-			cause, opc, badvaddr);
-		kvm_mips_dump_host_tlbs();
-		kvm_arch_vcpu_dump_regs(vcpu);
-		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-		ret = RESUME_HOST;
+		if (!TLB_IS_DIRTY(*tlb, badvaddr)) {
+			kvm_mips_emulate_tlbmod(cause, opc, run, vcpu);
+			return RESUME_GUEST;
+		}
+
+		if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, badvaddr,
+							 true))
+			/* Not writable, needs handling as MMIO */
+			return kvm_mips_bad_store(cause, opc, run, vcpu);
+		return RESUME_GUEST;
+	} else if (KVM_GUEST_KSEGX(badvaddr) == KVM_GUEST_KSEG0) {
+		if (kvm_mips_handle_kseg0_tlb_fault(badvaddr, vcpu, true) < 0)
+			/* Not writable, needs handling as MMIO */
+			return kvm_mips_bad_store(cause, opc, run, vcpu);
+		return RESUME_GUEST;
 	} else {
-		kvm_err("Illegal TLB Mod fault address , cause %#x, PC: %p, BadVaddr: %#lx\n",
-			cause, opc, badvaddr);
-		kvm_mips_dump_host_tlbs();
-		kvm_arch_vcpu_dump_regs(vcpu);
-		run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
-		ret = RESUME_HOST;
+		/* host kernel addresses are all handled as MMIO */
+		return kvm_mips_bad_store(cause, opc, run, vcpu);
 	}
-	return ret;
 }
 
 static int kvm_trap_emul_handle_tlb_miss(struct kvm_vcpu *vcpu, bool store)
-- 
git-series 0.8.10

[PATCH 6/13] KVM: MIPS/MMU: Add GPA PT mkclean helper

[James Hogan]

Add a helper function to make a range of guest physical address (GPA)
mappings in the GPA page table clean so that writes can be caught. This
will be used in a few places to manage dirty page logging.

Note that until the dirty bit is transferred from GPA page table entries
to GVA page table entries in an upcoming patch this won't trigger a TLB
modified exception on write.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |   1 +-
 arch/mips/kvm/mmu.c              | 124 ++++++++++++++++++++++++++++++++-
 2 files changed, 125 insertions(+), 0 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index da401a75a204..8f12385ccebd 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -640,6 +640,7 @@ enum kvm_mips_flush {
 };
 void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags);
 bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn);
+int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn);
 pgd_t *kvm_pgd_alloc(void);
 void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);
 void kvm_trap_emul_invalidate_gva(struct kvm_vcpu *vcpu, unsigned long addr,
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 934bcc3732da..892fd0ede718 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -304,6 +304,130 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
 				      end_gfn << PAGE_SHIFT);
 }
 
+#define BUILD_PTE_RANGE_OP(name, op)					\
+static int kvm_mips_##name##_pte(pte_t *pte, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	int i_min = __pte_offset(start);				\
+	int i_max = __pte_offset(end);					\
+	int i;								\
+	pte_t old, new;							\
+									\
+	for (i = i_min; i <= i_max; ++i) {				\
+		if (!pte_present(pte[i]))				\
+			continue;					\
+									\
+		old = pte[i];						\
+		new = op(old);						\
+		if (pte_val(new) == pte_val(old))			\
+			continue;					\
+		set_pte(pte + i, new);					\
+		ret = 1;						\
+	}								\
+	return ret;							\
+}									\
+									\
+/* returns true if anything was done */					\
+static int kvm_mips_##name##_pmd(pmd_t *pmd, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pte_t *pte;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = __pmd_offset(start);				\
+	int i_max = __pmd_offset(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pmd_present(pmd[i]))				\
+			continue;					\
+									\
+		pte = pte_offset(pmd + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pte(pte, start, cur_end);	\
+	}								\
+	return ret;							\
+}									\
+									\
+static int kvm_mips_##name##_pud(pud_t *pud, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pmd_t *pmd;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = __pud_offset(start);				\
+	int i_max = __pud_offset(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pud_present(pud[i]))				\
+			continue;					\
+									\
+		pmd = pmd_offset(pud + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pmd(pmd, start, cur_end);	\
+	}								\
+	return ret;							\
+}									\
+									\
+static int kvm_mips_##name##_pgd(pgd_t *pgd, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pud_t *pud;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = pgd_index(start);					\
+	int i_max = pgd_index(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pgd_present(pgd[i]))				\
+			continue;					\
+									\
+		pud = pud_offset(pgd + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pud(pud, start, cur_end);	\
+	}								\
+	return ret;							\
+}
+
+/*
+ * kvm_mips_mkclean_gpa_pt.
+ * Mark a range of guest physical address space clean (writes fault) in the VM's
+ * GPA page table to allow dirty page tracking.
+ */
+
+BUILD_PTE_RANGE_OP(mkclean, pte_mkclean)
+
+/**
+ * kvm_mips_mkclean_gpa_pt() - Make a range of guest physical addresses clean.
+ * @kvm:	KVM pointer.
+ * @start_gfn:	Guest frame number of first page in GPA range to flush.
+ * @end_gfn:	Guest frame number of last page in GPA range to flush.
+ *
+ * Make a range of GPA mappings clean so that guest writes will fault and
+ * trigger dirty page logging.
+ *
+ * The caller must hold the @kvm->mmu_lock spinlock.
+ *
+ * Returns:	Whether any GPA mappings were modified, which would require
+ *		derived mappings (GVA page tables & TLB enties) to be
+ *		invalidated.
+ */
+int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
+{
+	return kvm_mips_mkclean_pgd(kvm->arch.gpa_mm.pgd,
+				    start_gfn << PAGE_SHIFT,
+				    end_gfn << PAGE_SHIFT);
+}
+
 /**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
-- 
git-series 0.8.10

[PATCH 6/13] KVM: MIPS/MMU: Add GPA PT mkclean helper

[James Hogan]

Add a helper function to make a range of guest physical address (GPA)
mappings in the GPA page table clean so that writes can be caught. This
will be used in a few places to manage dirty page logging.

Note that until the dirty bit is transferred from GPA page table entries
to GVA page table entries in an upcoming patch this won't trigger a TLB
modified exception on write.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |   1 +-
 arch/mips/kvm/mmu.c              | 124 ++++++++++++++++++++++++++++++++-
 2 files changed, 125 insertions(+), 0 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index da401a75a204..8f12385ccebd 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -640,6 +640,7 @@ enum kvm_mips_flush {
 };
 void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags);
 bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn);
+int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn);
 pgd_t *kvm_pgd_alloc(void);
 void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);
 void kvm_trap_emul_invalidate_gva(struct kvm_vcpu *vcpu, unsigned long addr,
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 934bcc3732da..892fd0ede718 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -304,6 +304,130 @@ bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
 				      end_gfn << PAGE_SHIFT);
 }
 
+#define BUILD_PTE_RANGE_OP(name, op)					\
+static int kvm_mips_##name##_pte(pte_t *pte, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	int i_min = __pte_offset(start);				\
+	int i_max = __pte_offset(end);					\
+	int i;								\
+	pte_t old, new;							\
+									\
+	for (i = i_min; i <= i_max; ++i) {				\
+		if (!pte_present(pte[i]))				\
+			continue;					\
+									\
+		old = pte[i];						\
+		new = op(old);						\
+		if (pte_val(new) == pte_val(old))			\
+			continue;					\
+		set_pte(pte + i, new);					\
+		ret = 1;						\
+	}								\
+	return ret;							\
+}									\
+									\
+/* returns true if anything was done */					\
+static int kvm_mips_##name##_pmd(pmd_t *pmd, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pte_t *pte;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = __pmd_offset(start);				\
+	int i_max = __pmd_offset(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pmd_present(pmd[i]))				\
+			continue;					\
+									\
+		pte = pte_offset(pmd + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pte(pte, start, cur_end);	\
+	}								\
+	return ret;							\
+}									\
+									\
+static int kvm_mips_##name##_pud(pud_t *pud, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pmd_t *pmd;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = __pud_offset(start);				\
+	int i_max = __pud_offset(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pud_present(pud[i]))				\
+			continue;					\
+									\
+		pmd = pmd_offset(pud + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pmd(pmd, start, cur_end);	\
+	}								\
+	return ret;							\
+}									\
+									\
+static int kvm_mips_##name##_pgd(pgd_t *pgd, unsigned long start,	\
+				 unsigned long end)			\
+{									\
+	int ret = 0;							\
+	pud_t *pud;							\
+	unsigned long cur_end = ~0ul;					\
+	int i_min = pgd_index(start);					\
+	int i_max = pgd_index(end);					\
+	int i;								\
+									\
+	for (i = i_min; i <= i_max; ++i, start = 0) {			\
+		if (!pgd_present(pgd[i]))				\
+			continue;					\
+									\
+		pud = pud_offset(pgd + i, 0);				\
+		if (i == i_max)						\
+			cur_end = end;					\
+									\
+		ret |= kvm_mips_##name##_pud(pud, start, cur_end);	\
+	}								\
+	return ret;							\
+}
+
+/*
+ * kvm_mips_mkclean_gpa_pt.
+ * Mark a range of guest physical address space clean (writes fault) in the VM's
+ * GPA page table to allow dirty page tracking.
+ */
+
+BUILD_PTE_RANGE_OP(mkclean, pte_mkclean)
+
+/**
+ * kvm_mips_mkclean_gpa_pt() - Make a range of guest physical addresses clean.
+ * @kvm:	KVM pointer.
+ * @start_gfn:	Guest frame number of first page in GPA range to flush.
+ * @end_gfn:	Guest frame number of last page in GPA range to flush.
+ *
+ * Make a range of GPA mappings clean so that guest writes will fault and
+ * trigger dirty page logging.
+ *
+ * The caller must hold the @kvm->mmu_lock spinlock.
+ *
+ * Returns:	Whether any GPA mappings were modified, which would require
+ *		derived mappings (GVA page tables & TLB enties) to be
+ *		invalidated.
+ */
+int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
+{
+	return kvm_mips_mkclean_pgd(kvm->arch.gpa_mm.pgd,
+				    start_gfn << PAGE_SHIFT,
+				    end_gfn << PAGE_SHIFT);
+}
+
 /**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
-- 
git-series 0.8.10

[PATCH 7/13] KVM: MIPS/MMU: Use generic dirty log & protect helper

[James Hogan]

MIPS hasn't up to this point properly supported dirty page logging, as
pages in slots with dirty logging enabled aren't made clean, and tlbmod
exceptions from writes to clean pages have been assumed to be due to
guest TLB protection and unconditionally passed to the guest.

Use the generic dirty logging helper kvm_get_dirty_log_protect() to
properly implement kvm_vm_ioctl_get_dirty_log(), similar to how ARM
does. This uses xchg to clear the dirty bits when reading them, rather
than wiping them out afterwards with a memset, which would potentially
wipe recently set bits that weren't caught by kvm_get_dirty_log(). It
also makes the pages clean again using the
kvm_arch_mmu_enable_log_dirty_pt_masked() architecture callback so that
further writes after the shadow memslot is flushed will trigger tlbmod
exceptions and dirty handling.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/Kconfig |  1 +
 arch/mips/kvm/mips.c  | 42 +++++++++++++++++++++++-------------------
 arch/mips/kvm/mmu.c   | 22 ++++++++++++++++++++++
 3 files changed, 46 insertions(+), 19 deletions(-)

diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 7c56d6b124d1..85c4593b634a 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -20,6 +20,7 @@ config KVM
 	select EXPORT_UASM
 	select PREEMPT_NOTIFIERS
 	select ANON_INODES
+	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
 	select SRCU
 	---help---
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index b0d9fe2176c0..bee7f5eabd4f 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1091,42 +1091,46 @@ long kvm_arch_vcpu_ioctl(struct file *filp, unsigned int ioctl,
 	return r;
 }
 
-/* Get (and clear) the dirty memory log for a memory slot. */
+/**
+ * kvm_vm_ioctl_get_dirty_log - get and clear the log of dirty pages in a slot
+ * @kvm: kvm instance
+ * @log: slot id and address to which we copy the log
+ *
+ * Steps 1-4 below provide general overview of dirty page logging. See
+ * kvm_get_dirty_log_protect() function description for additional details.
+ *
+ * We call kvm_get_dirty_log_protect() to handle steps 1-3, upon return we
+ * always flush the TLB (step 4) even if previous step failed  and the dirty
+ * bitmap may be corrupt. Regardless of previous outcome the KVM logging API
+ * does not preclude user space subsequent dirty log read. Flushing TLB ensures
+ * writes will be marked dirty for next log read.
+ *
+ *   1. Take a snapshot of the bit and clear it if needed.
+ *   2. Write protect the corresponding page.
+ *   3. Copy the snapshot to the userspace.
+ *   4. Flush TLB's if needed.
+ */
 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
 {
 	struct kvm_memslots *slots;
 	struct kvm_memory_slot *memslot;
-	unsigned long ga, ga_end;
-	int is_dirty = 0;
+	bool is_dirty = false;
 	int r;
-	unsigned long n;
 
 	mutex_lock(&kvm->slots_lock);
 
-	r = kvm_get_dirty_log(kvm, log, &is_dirty);
-	if (r)
-		goto out;
+	r = kvm_get_dirty_log_protect(kvm, log, &is_dirty);
 
-	/* If nothing is dirty, don't bother messing with page tables. */
 	if (is_dirty) {
 		slots = kvm_memslots(kvm);
 		memslot = id_to_memslot(slots, log->slot);
 
-		ga = memslot->base_gfn << PAGE_SHIFT;
-		ga_end = ga + (memslot->npages << PAGE_SHIFT);
-
-		kvm_info("%s: dirty, ga: %#lx, ga_end %#lx\n", __func__, ga,
-			 ga_end);
-
-		n = kvm_dirty_bitmap_bytes(memslot);
-		memset(memslot->dirty_bitmap, 0, n);
+		/* Let implementation handle TLB/GVA invalidation */
+		kvm_mips_callbacks->flush_shadow_memslot(kvm, memslot);
 	}
 
-	r = 0;
-out:
 	mutex_unlock(&kvm->slots_lock);
 	return r;
-
 }
 
 long kvm_arch_vm_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 892fd0ede718..63a6d542ecb3 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -429,6 +429,28 @@ int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
 }
 
 /**
+ * kvm_arch_mmu_enable_log_dirty_pt_masked() - write protect dirty pages
+ * @kvm:	The KVM pointer
+ * @slot:	The memory slot associated with mask
+ * @gfn_offset:	The gfn offset in memory slot
+ * @mask:	The mask of dirty pages at offset 'gfn_offset' in this memory
+ *		slot to be write protected
+ *
+ * Walks bits set in mask write protects the associated pte's. Caller must
+ * acquire @kvm->mmu_lock.
+ */
+void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
+		struct kvm_memory_slot *slot,
+		gfn_t gfn_offset, unsigned long mask)
+{
+	gfn_t base_gfn = slot->base_gfn + gfn_offset;
+	gfn_t start = base_gfn +  __ffs(mask);
+	gfn_t end = base_gfn + __fls(mask);
+
+	kvm_mips_mkclean_gpa_pt(kvm, start, end);
+}
+
+/**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
-- 
git-series 0.8.10

[PATCH 7/13] KVM: MIPS/MMU: Use generic dirty log & protect helper

[James Hogan]

MIPS hasn't up to this point properly supported dirty page logging, as
pages in slots with dirty logging enabled aren't made clean, and tlbmod
exceptions from writes to clean pages have been assumed to be due to
guest TLB protection and unconditionally passed to the guest.

Use the generic dirty logging helper kvm_get_dirty_log_protect() to
properly implement kvm_vm_ioctl_get_dirty_log(), similar to how ARM
does. This uses xchg to clear the dirty bits when reading them, rather
than wiping them out afterwards with a memset, which would potentially
wipe recently set bits that weren't caught by kvm_get_dirty_log(). It
also makes the pages clean again using the
kvm_arch_mmu_enable_log_dirty_pt_masked() architecture callback so that
further writes after the shadow memslot is flushed will trigger tlbmod
exceptions and dirty handling.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/Kconfig |  1 +
 arch/mips/kvm/mips.c  | 42 +++++++++++++++++++++++-------------------
 arch/mips/kvm/mmu.c   | 22 ++++++++++++++++++++++
 3 files changed, 46 insertions(+), 19 deletions(-)

diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 7c56d6b124d1..85c4593b634a 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -20,6 +20,7 @@ config KVM
 	select EXPORT_UASM
 	select PREEMPT_NOTIFIERS
 	select ANON_INODES
+	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
 	select SRCU
 	---help---
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index b0d9fe2176c0..bee7f5eabd4f 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1091,42 +1091,46 @@ long kvm_arch_vcpu_ioctl(struct file *filp, unsigned int ioctl,
 	return r;
 }
 
-/* Get (and clear) the dirty memory log for a memory slot. */
+/**
+ * kvm_vm_ioctl_get_dirty_log - get and clear the log of dirty pages in a slot
+ * @kvm: kvm instance
+ * @log: slot id and address to which we copy the log
+ *
+ * Steps 1-4 below provide general overview of dirty page logging. See
+ * kvm_get_dirty_log_protect() function description for additional details.
+ *
+ * We call kvm_get_dirty_log_protect() to handle steps 1-3, upon return we
+ * always flush the TLB (step 4) even if previous step failed  and the dirty
+ * bitmap may be corrupt. Regardless of previous outcome the KVM logging API
+ * does not preclude user space subsequent dirty log read. Flushing TLB ensures
+ * writes will be marked dirty for next log read.
+ *
+ *   1. Take a snapshot of the bit and clear it if needed.
+ *   2. Write protect the corresponding page.
+ *   3. Copy the snapshot to the userspace.
+ *   4. Flush TLB's if needed.
+ */
 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log)
 {
 	struct kvm_memslots *slots;
 	struct kvm_memory_slot *memslot;
-	unsigned long ga, ga_end;
-	int is_dirty = 0;
+	bool is_dirty = false;
 	int r;
-	unsigned long n;
 
 	mutex_lock(&kvm->slots_lock);
 
-	r = kvm_get_dirty_log(kvm, log, &is_dirty);
-	if (r)
-		goto out;
+	r = kvm_get_dirty_log_protect(kvm, log, &is_dirty);
 
-	/* If nothing is dirty, don't bother messing with page tables. */
 	if (is_dirty) {
 		slots = kvm_memslots(kvm);
 		memslot = id_to_memslot(slots, log->slot);
 
-		ga = memslot->base_gfn << PAGE_SHIFT;
-		ga_end = ga + (memslot->npages << PAGE_SHIFT);
-
-		kvm_info("%s: dirty, ga: %#lx, ga_end %#lx\n", __func__, ga,
-			 ga_end);
-
-		n = kvm_dirty_bitmap_bytes(memslot);
-		memset(memslot->dirty_bitmap, 0, n);
+		/* Let implementation handle TLB/GVA invalidation */
+		kvm_mips_callbacks->flush_shadow_memslot(kvm, memslot);
 	}
 
-	r = 0;
-out:
 	mutex_unlock(&kvm->slots_lock);
 	return r;
-
 }
 
 long kvm_arch_vm_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 892fd0ede718..63a6d542ecb3 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -429,6 +429,28 @@ int kvm_mips_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
 }
 
 /**
+ * kvm_arch_mmu_enable_log_dirty_pt_masked() - write protect dirty pages
+ * @kvm:	The KVM pointer
+ * @slot:	The memory slot associated with mask
+ * @gfn_offset:	The gfn offset in memory slot
+ * @mask:	The mask of dirty pages at offset 'gfn_offset' in this memory
+ *		slot to be write protected
+ *
+ * Walks bits set in mask write protects the associated pte's. Caller must
+ * acquire @kvm->mmu_lock.
+ */
+void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
+		struct kvm_memory_slot *slot,
+		gfn_t gfn_offset, unsigned long mask)
+{
+	gfn_t base_gfn = slot->base_gfn + gfn_offset;
+	gfn_t start = base_gfn +  __ffs(mask);
+	gfn_t end = base_gfn + __fls(mask);
+
+	kvm_mips_mkclean_gpa_pt(kvm, start, end);
+}
+
+/**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
-- 
git-series 0.8.10

[PATCH 8/13] KVM: MIPS: Clean & flush on dirty page logging enable

[James Hogan]

When an existing memory region has dirty page logging enabled, make the
entire slot clean (read only) so that writes will immediately start
logging dirty pages (once the dirty bit is transferred from GPA to GVA
page tables in an upcoming patch).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mips.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+), 0 deletions(-)

diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index bee7f5eabd4f..5b7e9a053f77 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -197,9 +197,33 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
 				   const struct kvm_memory_slot *new,
 				   enum kvm_mr_change change)
 {
+	int needs_flush;
+
 	kvm_debug("%s: kvm: %p slot: %d, GPA: %llx, size: %llx, QVA: %llx\n",
 		  __func__, kvm, mem->slot, mem->guest_phys_addr,
 		  mem->memory_size, mem->userspace_addr);
+
+	/*
+	 * If dirty page logging is enabled, write protect all pages in the slot
+	 * ready for dirty logging.
+	 *
+	 * There is no need to do this in any of the following cases:
+	 * CREATE:	No dirty mappings will already exist.
+	 * MOVE/DELETE:	The old mappings will already have been cleaned up by
+	 *		kvm_arch_flush_shadow_memslot()
+	 */
+	if (change == KVM_MR_FLAGS_ONLY &&
+	    (!(old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
+	     new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+		spin_lock(&kvm->mmu_lock);
+		/* Write protect GPA page table entries */
+		needs_flush = kvm_mips_mkclean_gpa_pt(kvm, new->base_gfn,
+					new->base_gfn + new->npages - 1);
+		/* Let implementation do the rest */
+		if (needs_flush)
+			kvm_mips_callbacks->flush_shadow_memslot(kvm, new);
+		spin_unlock(&kvm->mmu_lock);
+	}
 }
 
 static inline void dump_handler(const char *symbol, void *start, void *end)
-- 
git-series 0.8.10

[PATCH 8/13] KVM: MIPS: Clean & flush on dirty page logging enable

[James Hogan]

When an existing memory region has dirty page logging enabled, make the
entire slot clean (read only) so that writes will immediately start
logging dirty pages (once the dirty bit is transferred from GPA to GVA
page tables in an upcoming patch).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mips.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+), 0 deletions(-)

diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index bee7f5eabd4f..5b7e9a053f77 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -197,9 +197,33 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
 				   const struct kvm_memory_slot *new,
 				   enum kvm_mr_change change)
 {
+	int needs_flush;
+
 	kvm_debug("%s: kvm: %p slot: %d, GPA: %llx, size: %llx, QVA: %llx\n",
 		  __func__, kvm, mem->slot, mem->guest_phys_addr,
 		  mem->memory_size, mem->userspace_addr);
+
+	/*
+	 * If dirty page logging is enabled, write protect all pages in the slot
+	 * ready for dirty logging.
+	 *
+	 * There is no need to do this in any of the following cases:
+	 * CREATE:	No dirty mappings will already exist.
+	 * MOVE/DELETE:	The old mappings will already have been cleaned up by
+	 *		kvm_arch_flush_shadow_memslot()
+	 */
+	if (change == KVM_MR_FLAGS_ONLY &&
+	    (!(old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
+	     new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+		spin_lock(&kvm->mmu_lock);
+		/* Write protect GPA page table entries */
+		needs_flush = kvm_mips_mkclean_gpa_pt(kvm, new->base_gfn,
+					new->base_gfn + new->npages - 1);
+		/* Let implementation do the rest */
+		if (needs_flush)
+			kvm_mips_callbacks->flush_shadow_memslot(kvm, new);
+		spin_unlock(&kvm->mmu_lock);
+	}
 }
 
 static inline void dump_handler(const char *symbol, void *start, void *end)
-- 
git-series 0.8.10

[PATCH 9/13] KVM: MIPS/MMU: Handle dirty logging on GPA faults

[James Hogan]

Update kvm_mips_map_page() to handle logging of dirty guest physical
pages. Upcoming patches will propagate the dirty bit to the GVA page
tables.

A fast path is added for handling protection bits that can be resolved
without calling into KVM, currently just dirtying of clean pages being
written to.

The slow path marks the GPA page table entry writable only on writes,
and at the same time marks the page dirty in the dirty page logging
bitmask.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 74 +++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 70 insertions(+), 4 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 63a6d542ecb3..7962eea4ebc3 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -451,6 +451,58 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 }
 
 /**
+ * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
+ * @vcpu:		VCPU pointer.
+ * @gpa:		Guest physical address of fault.
+ * @write_fault:	Whether the fault was due to a write.
+ * @out_entry:		New PTE for @gpa (written on success unless NULL).
+ * @out_buddy:		New PTE for @gpa's buddy (written on success unless
+ *			NULL).
+ *
+ * Perform fast path GPA fault handling, doing all that can be done without
+ * calling into KVM. This handles dirtying of clean pages (for dirty page
+ * logging).
+ *
+ * Returns:	0 on success, in which case we can update derived mappings and
+ *		resume guest execution.
+ *		-EFAULT on failure due to absent GPA mapping or write to
+ *		read-only page, in which case KVM must be consulted.
+ */
+static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
+				   bool write_fault,
+				   pte_t *out_entry, pte_t *out_buddy)
+{
+	struct kvm *kvm = vcpu->kvm;
+	gfn_t gfn = gpa >> PAGE_SHIFT;
+	pte_t *ptep;
+	int ret = 0;
+
+	spin_lock(&kvm->mmu_lock);
+
+	/* Fast path - just check GPA page table for an existing entry */
+	ptep = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	if (!ptep || !pte_present(*ptep)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (write_fault && !pte_dirty(*ptep)) {
+		/* Track dirtying of pages */
+		set_pte(ptep, pte_mkdirty(*ptep));
+		mark_page_dirty(kvm, gfn);
+	}
+
+	if (out_entry)
+		*out_entry = *ptep;
+	if (out_buddy)
+		*out_buddy = *ptep_buddy(ptep);
+
+out:
+	spin_unlock(&kvm->mmu_lock);
+	return ret;
+}
+
+/**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
@@ -462,9 +514,9 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of asking KVM for the corresponding PFN, and creating a
- * mapping in the GPA page tables. Derived mappings (GVA page tables and TLBs)
- * must be handled by the caller.
+ * This takes care of marking pages dirty (dirty page tracking), asking KVM for
+ * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
+ * mappings (GVA page tables and TLBs) must be handled by the caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -485,7 +537,12 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	pte_t *ptep, entry, old_pte;
 	unsigned long prot_bits;
 
+	/* Try the fast path to handle clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
+	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
+				      out_buddy);
+	if (!err)
+		goto out;
 
 	/* We need a minimum of cached pages ready for page table creation */
 	err = mmu_topup_memory_cache(memcache, KVM_MMU_CACHE_MIN_PAGES,
@@ -493,6 +550,7 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
+	/* Slow path - ask KVM core whether we can access this GPA */
 	pfn = gfn_to_pfn(kvm, gfn);
 
 	if (is_error_noslot_pfn(pfn)) {
@@ -502,11 +560,19 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 	spin_lock(&kvm->mmu_lock);
 
+	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
-	prot_bits = __READABLE | _PAGE_PRESENT | __WRITEABLE;
+	/* Set up the PTE */
+	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
+		_page_cachable_default;
+	if (write_fault) {
+		prot_bits |= __WRITEABLE;
+		mark_page_dirty(kvm, gfn);
+	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
+	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
 	if (pte_present(old_pte))
-- 
git-series 0.8.10

[PATCH 9/13] KVM: MIPS/MMU: Handle dirty logging on GPA faults

[James Hogan]

Update kvm_mips_map_page() to handle logging of dirty guest physical
pages. Upcoming patches will propagate the dirty bit to the GVA page
tables.

A fast path is added for handling protection bits that can be resolved
without calling into KVM, currently just dirtying of clean pages being
written to.

The slow path marks the GPA page table entry writable only on writes,
and at the same time marks the page dirty in the dirty page logging
bitmask.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 74 +++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 70 insertions(+), 4 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 63a6d542ecb3..7962eea4ebc3 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -451,6 +451,58 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 }
 
 /**
+ * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
+ * @vcpu:		VCPU pointer.
+ * @gpa:		Guest physical address of fault.
+ * @write_fault:	Whether the fault was due to a write.
+ * @out_entry:		New PTE for @gpa (written on success unless NULL).
+ * @out_buddy:		New PTE for @gpa's buddy (written on success unless
+ *			NULL).
+ *
+ * Perform fast path GPA fault handling, doing all that can be done without
+ * calling into KVM. This handles dirtying of clean pages (for dirty page
+ * logging).
+ *
+ * Returns:	0 on success, in which case we can update derived mappings and
+ *		resume guest execution.
+ *		-EFAULT on failure due to absent GPA mapping or write to
+ *		read-only page, in which case KVM must be consulted.
+ */
+static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
+				   bool write_fault,
+				   pte_t *out_entry, pte_t *out_buddy)
+{
+	struct kvm *kvm = vcpu->kvm;
+	gfn_t gfn = gpa >> PAGE_SHIFT;
+	pte_t *ptep;
+	int ret = 0;
+
+	spin_lock(&kvm->mmu_lock);
+
+	/* Fast path - just check GPA page table for an existing entry */
+	ptep = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	if (!ptep || !pte_present(*ptep)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (write_fault && !pte_dirty(*ptep)) {
+		/* Track dirtying of pages */
+		set_pte(ptep, pte_mkdirty(*ptep));
+		mark_page_dirty(kvm, gfn);
+	}
+
+	if (out_entry)
+		*out_entry = *ptep;
+	if (out_buddy)
+		*out_buddy = *ptep_buddy(ptep);
+
+out:
+	spin_unlock(&kvm->mmu_lock);
+	return ret;
+}
+
+/**
  * kvm_mips_map_page() - Map a guest physical page.
  * @vcpu:		VCPU pointer.
  * @gpa:		Guest physical address of fault.
@@ -462,9 +514,9 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of asking KVM for the corresponding PFN, and creating a
- * mapping in the GPA page tables. Derived mappings (GVA page tables and TLBs)
- * must be handled by the caller.
+ * This takes care of marking pages dirty (dirty page tracking), asking KVM for
+ * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
+ * mappings (GVA page tables and TLBs) must be handled by the caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -485,7 +537,12 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	pte_t *ptep, entry, old_pte;
 	unsigned long prot_bits;
 
+	/* Try the fast path to handle clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
+	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
+				      out_buddy);
+	if (!err)
+		goto out;
 
 	/* We need a minimum of cached pages ready for page table creation */
 	err = mmu_topup_memory_cache(memcache, KVM_MMU_CACHE_MIN_PAGES,
@@ -493,6 +550,7 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
+	/* Slow path - ask KVM core whether we can access this GPA */
 	pfn = gfn_to_pfn(kvm, gfn);
 
 	if (is_error_noslot_pfn(pfn)) {
@@ -502,11 +560,19 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 	spin_lock(&kvm->mmu_lock);
 
+	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
-	prot_bits = __READABLE | _PAGE_PRESENT | __WRITEABLE;
+	/* Set up the PTE */
+	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
+		_page_cachable_default;
+	if (write_fault) {
+		prot_bits |= __WRITEABLE;
+		mark_page_dirty(kvm, gfn);
+	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
+	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
 	if (pte_present(old_pte))
-- 
git-series 0.8.10

[PATCH 10/13] KVM: MIPS/MMU: Pass GPA PTE bits to KSeg0 GVA PTEs

[James Hogan]

Propagate the GPA PTE protection bits on to the GVA PTEs on a KSeg0
fault (except _PAGE_WRITE), rather than always overriding the
protection. This allows dirty page tracking to work in KSeg0 as a clear
dirty bit in the GPA PTE will propagate to the GVA PTEs.

This makes it simpler to use a single kvm_mips_map_page() to obtain both
the main GPA PTE and its buddy (which may be invalid), which also allows
memory regions to be fully accessible when they don't start and end on a
2*PAGE_SIZE boundary.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 49 ++++++++++++++++++++++++----------------------
 1 file changed, 26 insertions(+), 23 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 7962eea4ebc3..9cc941864aa8 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -769,15 +769,27 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
 	}
 }
 
+static pte_t kvm_mips_gpa_pte_to_gva_unmapped(pte_t pte)
+{
+	/*
+	 * Don't leak writeable but clean entries from GPA page tables. We don't
+	 * want the normal Linux tlbmod handler to handle dirtying when KVM
+	 * accesses guest memory.
+	 */
+	if (!pte_dirty(pte))
+		pte = pte_wrprotect(pte);
+
+	return pte;
+}
+
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 				    struct kvm_vcpu *vcpu,
 				    bool write_fault)
 {
 	unsigned long gpa;
-	kvm_pfn_t pfn0, pfn1;
-	unsigned long vaddr;
 	pte_t pte_gpa[2], *ptep_gva;
+	int idx;
 
 	if (KVM_GUEST_KSEGX(badvaddr) != KVM_GUEST_KSEG0) {
 		kvm_err("%s: Invalid BadVaddr: %#lx\n", __func__, badvaddr);
@@ -785,35 +797,26 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 		return -1;
 	}
 
-	/* Find host PFNs */
-
-	gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
-	vaddr = badvaddr & (PAGE_MASK << 1);
-
-	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[0], NULL) < 0)
-		return -1;
-
-	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, write_fault, &pte_gpa[1],
-			      NULL) < 0)
+	/* Get the GPA page table entry */
+	gpa = KVM_GUEST_CPHYSADDR(badvaddr);
+	idx = (badvaddr >> PAGE_SHIFT) & 1;
+	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[idx],
+			      &pte_gpa[!idx]) < 0)
 		return -1;
 
-	pfn0 = pte_pfn(pte_gpa[0]);
-	pfn1 = pte_pfn(pte_gpa[1]);
-
-	/* Find GVA page table entry */
-
-	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, vaddr);
+	/* Get the GVA page table entry */
+	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, badvaddr & ~PAGE_SIZE);
 	if (!ptep_gva) {
-		kvm_err("No ptep for gva %lx\n", vaddr);
+		kvm_err("No ptep for gva %lx\n", badvaddr);
 		return -1;
 	}
 
-	/* Write host PFNs into GVA page table */
-	ptep_gva[0] = pte_mkyoung(pte_mkdirty(pfn_pte(pfn0, PAGE_SHARED)));
-	ptep_gva[1] = pte_mkyoung(pte_mkdirty(pfn_pte(pfn1, PAGE_SHARED)));
+	/* Copy a pair of entries from GPA page table to GVA page table */
+	ptep_gva[0] = kvm_mips_gpa_pte_to_gva_unmapped(pte_gpa[0]);
+	ptep_gva[1] = kvm_mips_gpa_pte_to_gva_unmapped(pte_gpa[1]);
 
 	/* Invalidate this entry in the TLB, guest kernel ASID only */
-	kvm_mips_host_tlb_inv(vcpu, vaddr, false, true);
+	kvm_mips_host_tlb_inv(vcpu, badvaddr, false, true);
 	return 0;
 }
 
-- 
git-series 0.8.10

[PATCH 10/13] KVM: MIPS/MMU: Pass GPA PTE bits to KSeg0 GVA PTEs

[James Hogan]

Propagate the GPA PTE protection bits on to the GVA PTEs on a KSeg0
fault (except _PAGE_WRITE), rather than always overriding the
protection. This allows dirty page tracking to work in KSeg0 as a clear
dirty bit in the GPA PTE will propagate to the GVA PTEs.

This makes it simpler to use a single kvm_mips_map_page() to obtain both
the main GPA PTE and its buddy (which may be invalid), which also allows
memory regions to be fully accessible when they don't start and end on a
2*PAGE_SIZE boundary.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 49 ++++++++++++++++++++++++----------------------
 1 file changed, 26 insertions(+), 23 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 7962eea4ebc3..9cc941864aa8 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -769,15 +769,27 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
 	}
 }
 
+static pte_t kvm_mips_gpa_pte_to_gva_unmapped(pte_t pte)
+{
+	/*
+	 * Don't leak writeable but clean entries from GPA page tables. We don't
+	 * want the normal Linux tlbmod handler to handle dirtying when KVM
+	 * accesses guest memory.
+	 */
+	if (!pte_dirty(pte))
+		pte = pte_wrprotect(pte);
+
+	return pte;
+}
+
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 				    struct kvm_vcpu *vcpu,
 				    bool write_fault)
 {
 	unsigned long gpa;
-	kvm_pfn_t pfn0, pfn1;
-	unsigned long vaddr;
 	pte_t pte_gpa[2], *ptep_gva;
+	int idx;
 
 	if (KVM_GUEST_KSEGX(badvaddr) != KVM_GUEST_KSEG0) {
 		kvm_err("%s: Invalid BadVaddr: %#lx\n", __func__, badvaddr);
@@ -785,35 +797,26 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 		return -1;
 	}
 
-	/* Find host PFNs */
-
-	gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
-	vaddr = badvaddr & (PAGE_MASK << 1);
-
-	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[0], NULL) < 0)
-		return -1;
-
-	if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, write_fault, &pte_gpa[1],
-			      NULL) < 0)
+	/* Get the GPA page table entry */
+	gpa = KVM_GUEST_CPHYSADDR(badvaddr);
+	idx = (badvaddr >> PAGE_SHIFT) & 1;
+	if (kvm_mips_map_page(vcpu, gpa, write_fault, &pte_gpa[idx],
+			      &pte_gpa[!idx]) < 0)
 		return -1;
 
-	pfn0 = pte_pfn(pte_gpa[0]);
-	pfn1 = pte_pfn(pte_gpa[1]);
-
-	/* Find GVA page table entry */
-
-	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, vaddr);
+	/* Get the GVA page table entry */
+	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, badvaddr & ~PAGE_SIZE);
 	if (!ptep_gva) {
-		kvm_err("No ptep for gva %lx\n", vaddr);
+		kvm_err("No ptep for gva %lx\n", badvaddr);
 		return -1;
 	}
 
-	/* Write host PFNs into GVA page table */
-	ptep_gva[0] = pte_mkyoung(pte_mkdirty(pfn_pte(pfn0, PAGE_SHARED)));
-	ptep_gva[1] = pte_mkyoung(pte_mkdirty(pfn_pte(pfn1, PAGE_SHARED)));
+	/* Copy a pair of entries from GPA page table to GVA page table */
+	ptep_gva[0] = kvm_mips_gpa_pte_to_gva_unmapped(pte_gpa[0]);
+	ptep_gva[1] = kvm_mips_gpa_pte_to_gva_unmapped(pte_gpa[1]);
 
 	/* Invalidate this entry in the TLB, guest kernel ASID only */
-	kvm_mips_host_tlb_inv(vcpu, vaddr, false, true);
+	kvm_mips_host_tlb_inv(vcpu, badvaddr, false, true);
 	return 0;
 }
 
-- 
git-series 0.8.10

[PATCH 11/13] KVM: MIPS/MMU: Pass GPA PTE bits to mapped GVA PTEs

[James Hogan]

Propagate the GPA PTE protection bits on to the GVA PTEs on a mapped
fault (except _PAGE_WRITE, and filtered by the guest TLB entry), rather
than always overriding the protection. This allows dirty page tracking
to work in mapped guest segments as a clear dirty bit in the GPA PTE
will propagate to the GVA PTEs even when the guest TLB has the dirty bit
set.

Since the filtering of protection bits is now abstracted, if the buddy
GVA PTE is also valid, we obtain the corresponding GPA PTE using a
simple non-allocating walk and load that into the GVA PTE similarly
(which may itself be invalid).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 56 ++++++++++++++++++++++++++++++----------------
 1 file changed, 37 insertions(+), 19 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 9cc941864aa8..8a01bbd276fc 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -782,6 +782,15 @@ static pte_t kvm_mips_gpa_pte_to_gva_unmapped(pte_t pte)
 	return pte;
 }
 
+static pte_t kvm_mips_gpa_pte_to_gva_mapped(pte_t pte, long entrylo)
+{
+	/* Guest EntryLo overrides host EntryLo */
+	if (!(entrylo & ENTRYLO_D))
+		pte = pte_mkclean(pte);
+
+	return kvm_mips_gpa_pte_to_gva_unmapped(pte);
+}
+
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 				    struct kvm_vcpu *vcpu,
@@ -825,39 +834,48 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 					 unsigned long gva,
 					 bool write_fault)
 {
-	kvm_pfn_t pfn;
-	long tlb_lo = 0;
-	pte_t pte_gpa, *ptep_gva;
-	unsigned int idx;
+	struct kvm *kvm = vcpu->kvm;
+	long tlb_lo[2];
+	pte_t pte_gpa[2], *ptep_buddy, *ptep_gva;
+	unsigned int idx = TLB_LO_IDX(*tlb, gva);
 	bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
 
+	tlb_lo[0] = tlb->tlb_lo[0];
+	tlb_lo[1] = tlb->tlb_lo[1];
+
 	/*
 	 * The commpage address must not be mapped to anything else if the guest
 	 * TLB contains entries nearby, or commpage accesses will break.
 	 */
-	idx = TLB_LO_IDX(*tlb, gva);
-	if ((gva ^ KVM_GUEST_COMMPAGE_ADDR) & VPN2_MASK & PAGE_MASK)
-		tlb_lo = tlb->tlb_lo[idx];
+	if (!((gva ^ KVM_GUEST_COMMPAGE_ADDR) & VPN2_MASK & (PAGE_MASK << 1)))
+		tlb_lo[TLB_LO_IDX(*tlb, KVM_GUEST_COMMPAGE_ADDR)] = 0;
 
-	/* Find host PFN */
-	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), write_fault,
-			      &pte_gpa, NULL) < 0)
+	/* Get the GPA page table entry */
+	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo[idx]),
+			      write_fault, &pte_gpa[idx], NULL) < 0)
 		return -1;
-	pfn = pte_pfn(pte_gpa);
 
-	/* Find GVA page table entry */
-	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva);
+	/* And its GVA buddy's GPA page table entry if it also exists */
+	pte_gpa[!idx] = pfn_pte(0, __pgprot(0));
+	if (tlb_lo[!idx] & ENTRYLO_V) {
+		spin_lock(&kvm->mmu_lock);
+		ptep_buddy = kvm_mips_pte_for_gpa(kvm, NULL,
+					mips3_tlbpfn_to_paddr(tlb_lo[!idx]));
+		if (ptep_buddy)
+			pte_gpa[!idx] = *ptep_buddy;
+		spin_unlock(&kvm->mmu_lock);
+	}
+
+	/* Get the GVA page table entry pair */
+	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva & ~PAGE_SIZE);
 	if (!ptep_gva) {
 		kvm_err("No ptep for gva %lx\n", gva);
 		return -1;
 	}
 
-	/* Write PFN into GVA page table, taking attributes from Guest TLB */
-	*ptep_gva = pfn_pte(pfn, (!(tlb_lo & ENTRYLO_V)) ? __pgprot(0) :
-				 (tlb_lo & ENTRYLO_D) ? PAGE_SHARED :
-				 PAGE_READONLY);
-	if (pte_present(*ptep_gva))
-		*ptep_gva = pte_mkyoung(pte_mkdirty(*ptep_gva));
+	/* Copy a pair of entries from GPA page table to GVA page table */
+	ptep_gva[0] = kvm_mips_gpa_pte_to_gva_mapped(pte_gpa[0], tlb_lo[0]);
+	ptep_gva[1] = kvm_mips_gpa_pte_to_gva_mapped(pte_gpa[1], tlb_lo[1]);
 
 	/* Invalidate this entry in the TLB, current guest mode ASID only */
 	kvm_mips_host_tlb_inv(vcpu, gva, !kernel, kernel);
-- 
git-series 0.8.10

[PATCH 11/13] KVM: MIPS/MMU: Pass GPA PTE bits to mapped GVA PTEs

[James Hogan]

Propagate the GPA PTE protection bits on to the GVA PTEs on a mapped
fault (except _PAGE_WRITE, and filtered by the guest TLB entry), rather
than always overriding the protection. This allows dirty page tracking
to work in mapped guest segments as a clear dirty bit in the GPA PTE
will propagate to the GVA PTEs even when the guest TLB has the dirty bit
set.

Since the filtering of protection bits is now abstracted, if the buddy
GVA PTE is also valid, we obtain the corresponding GPA PTE using a
simple non-allocating walk and load that into the GVA PTE similarly
(which may itself be invalid).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/kvm/mmu.c | 56 ++++++++++++++++++++++++++++++----------------
 1 file changed, 37 insertions(+), 19 deletions(-)

diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 9cc941864aa8..8a01bbd276fc 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -782,6 +782,15 @@ static pte_t kvm_mips_gpa_pte_to_gva_unmapped(pte_t pte)
 	return pte;
 }
 
+static pte_t kvm_mips_gpa_pte_to_gva_mapped(pte_t pte, long entrylo)
+{
+	/* Guest EntryLo overrides host EntryLo */
+	if (!(entrylo & ENTRYLO_D))
+		pte = pte_mkclean(pte);
+
+	return kvm_mips_gpa_pte_to_gva_unmapped(pte);
+}
+
 /* XXXKYMA: Must be called with interrupts disabled */
 int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
 				    struct kvm_vcpu *vcpu,
@@ -825,39 +834,48 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
 					 unsigned long gva,
 					 bool write_fault)
 {
-	kvm_pfn_t pfn;
-	long tlb_lo = 0;
-	pte_t pte_gpa, *ptep_gva;
-	unsigned int idx;
+	struct kvm *kvm = vcpu->kvm;
+	long tlb_lo[2];
+	pte_t pte_gpa[2], *ptep_buddy, *ptep_gva;
+	unsigned int idx = TLB_LO_IDX(*tlb, gva);
 	bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
 
+	tlb_lo[0] = tlb->tlb_lo[0];
+	tlb_lo[1] = tlb->tlb_lo[1];
+
 	/*
 	 * The commpage address must not be mapped to anything else if the guest
 	 * TLB contains entries nearby, or commpage accesses will break.
 	 */
-	idx = TLB_LO_IDX(*tlb, gva);
-	if ((gva ^ KVM_GUEST_COMMPAGE_ADDR) & VPN2_MASK & PAGE_MASK)
-		tlb_lo = tlb->tlb_lo[idx];
+	if (!((gva ^ KVM_GUEST_COMMPAGE_ADDR) & VPN2_MASK & (PAGE_MASK << 1)))
+		tlb_lo[TLB_LO_IDX(*tlb, KVM_GUEST_COMMPAGE_ADDR)] = 0;
 
-	/* Find host PFN */
-	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), write_fault,
-			      &pte_gpa, NULL) < 0)
+	/* Get the GPA page table entry */
+	if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo[idx]),
+			      write_fault, &pte_gpa[idx], NULL) < 0)
 		return -1;
-	pfn = pte_pfn(pte_gpa);
 
-	/* Find GVA page table entry */
-	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva);
+	/* And its GVA buddy's GPA page table entry if it also exists */
+	pte_gpa[!idx] = pfn_pte(0, __pgprot(0));
+	if (tlb_lo[!idx] & ENTRYLO_V) {
+		spin_lock(&kvm->mmu_lock);
+		ptep_buddy = kvm_mips_pte_for_gpa(kvm, NULL,
+					mips3_tlbpfn_to_paddr(tlb_lo[!idx]));
+		if (ptep_buddy)
+			pte_gpa[!idx] = *ptep_buddy;
+		spin_unlock(&kvm->mmu_lock);
+	}
+
+	/* Get the GVA page table entry pair */
+	ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva & ~PAGE_SIZE);
 	if (!ptep_gva) {
 		kvm_err("No ptep for gva %lx\n", gva);
 		return -1;
 	}
 
-	/* Write PFN into GVA page table, taking attributes from Guest TLB */
-	*ptep_gva = pfn_pte(pfn, (!(tlb_lo & ENTRYLO_V)) ? __pgprot(0) :
-				 (tlb_lo & ENTRYLO_D) ? PAGE_SHARED :
-				 PAGE_READONLY);
-	if (pte_present(*ptep_gva))
-		*ptep_gva = pte_mkyoung(pte_mkdirty(*ptep_gva));
+	/* Copy a pair of entries from GPA page table to GVA page table */
+	ptep_gva[0] = kvm_mips_gpa_pte_to_gva_mapped(pte_gpa[0], tlb_lo[0]);
+	ptep_gva[1] = kvm_mips_gpa_pte_to_gva_mapped(pte_gpa[1], tlb_lo[1]);
 
 	/* Invalidate this entry in the TLB, current guest mode ASID only */
 	kvm_mips_host_tlb_inv(vcpu, gva, !kernel, kernel);
-- 
git-series 0.8.10

[PATCH 12/13] KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU

[James Hogan]

Implement the SYNC_MMU capability for KVM MIPS, allowing changes in the
underlying user host virtual address (HVA) mappings to be promptly
reflected in the corresponding guest physical address (GPA) mappings.

This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning.

There are two main aspects of this change, described below.

The KVM MMU notifier architecture callbacks are implemented so we can be
notified of changes in the HVA mappings. These arrange for the guest
physical address (GPA) page tables to be modified and possibly for
derived mappings (GVA page tables and TLBs) to be flushed.

 - kvm_unmap_hva[_range]() - These deal with HVA mappings being removed,
   for example before a copy-on-write takes place, which requires the
   corresponding GPA page table mappings to be removed too.

 - kvm_set_spte_hva() - These update a GPA page table entry to match the
   new HVA entry, but must be careful to respect KVM specific
   configuration such as not dirtying a clean guest page which is dirty
   to the host, and write protecting writable pages in read only
   memslots (which will soon be supported).

 - kvm[_test]_age_hva() - These update GPA page table entries to be old
   (invalid) so that access can be tracked, making them young again.

The GPA page fault handling (kvm_mips_map_page) is updated to use
gfn_to_pfn_prot() (which may provide read-only pages), to handle
asynchronous page table invalidation from MMU notifier callbacks, and to
handle more cases in the fast path.

 - mmu_notifier_seq is used to detect asynchronous page table
   invalidations while we're holding a pfn from gfn_to_pfn_prot()
   outside of kvm->mmu_lock, retrying if invalidations have taken place,
   e.g. a COW or a KSM page merge.

 - The fast path (_kvm_mips_map_page_fast) now handles marking old pages
   as young / accessed, and disallowing dirtying of clean pages that
   aren't actually writable (e.g. shared pages that should COW, and
   read-only memory regions when they are enabled in a future patch).

 - Due to the use of MMU notifications we no longer need to keep the
   page references after we've updated the GPA page tables.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  13 ++-
 arch/mips/kvm/Kconfig            |   1 +-
 arch/mips/kvm/mips.c             |   1 +-
 arch/mips/kvm/mmu.c              | 235 +++++++++++++++++++++++++++++---
 4 files changed, 233 insertions(+), 17 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 8f12385ccebd..31e7f3797f52 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -660,6 +660,19 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 						   unsigned long gva,
 						   bool write);
 
+#define KVM_ARCH_WANT_MMU_NOTIFIER
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
+int kvm_unmap_hva_range(struct kvm *kvm,
+			unsigned long start, unsigned long end);
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
+
+static inline void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
+							 unsigned long address)
+{
+}
+
 /* Emulation */
 int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
 enum emulation_result update_pc(struct kvm_vcpu *vcpu, u32 cause);
diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 85c4593b634a..65067327db12 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -22,6 +22,7 @@ config KVM
 	select ANON_INODES
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
+	select MMU_NOTIFIER
 	select SRCU
 	---help---
 	  Support for hosting Guest kernels.
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 5b7e9a053f77..425e4139742c 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 8a01bbd276fc..001a3e6ce2c4 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -194,7 +194,6 @@ static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
 		if (!pte_present(pte[i]))
 			continue;
 
-		kvm_release_pfn_clean(pte_pfn(pte[i]));
 		set_pte(pte + i, __pte(0));
 	}
 	return safe_to_remove;
@@ -450,6 +449,155 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 	kvm_mips_mkclean_gpa_pt(kvm, start, end);
 }
 
+/*
+ * kvm_mips_mkold_gpa_pt.
+ * Mark a range of guest physical address space old (all accesses fault) in the
+ * VM's GPA page table to allow detection of commonly used pages.
+ */
+
+BUILD_PTE_RANGE_OP(mkold, pte_mkold)
+
+static int kvm_mips_mkold_gpa_pt(struct kvm *kvm, gfn_t start_gfn,
+				 gfn_t end_gfn)
+{
+	return kvm_mips_mkold_pgd(kvm->arch.gpa_mm.pgd,
+				  start_gfn << PAGE_SHIFT,
+				  end_gfn << PAGE_SHIFT);
+}
+
+static int handle_hva_to_gpa(struct kvm *kvm,
+			     unsigned long start,
+			     unsigned long end,
+			     int (*handler)(struct kvm *kvm, gfn_t gfn,
+					    gpa_t gfn_end,
+					    struct kvm_memory_slot *memslot,
+					    void *data),
+			     void *data)
+{
+	struct kvm_memslots *slots;
+	struct kvm_memory_slot *memslot;
+	int ret = 0;
+
+	slots = kvm_memslots(kvm);
+
+	/* we only care about the pages that the guest sees */
+	kvm_for_each_memslot(memslot, slots) {
+		unsigned long hva_start, hva_end;
+		gfn_t gfn, gfn_end;
+
+		hva_start = max(start, memslot->userspace_addr);
+		hva_end = min(end, memslot->userspace_addr +
+					(memslot->npages << PAGE_SHIFT));
+		if (hva_start >= hva_end)
+			continue;
+
+		/*
+		 * {gfn(page) | page intersects with [hva_start, hva_end)} =
+		 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
+		 */
+		gfn = hva_to_gfn_memslot(hva_start, memslot);
+		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
+
+		ret |= handler(kvm, gfn, gfn_end, memslot, data);
+	}
+
+	return ret;
+}
+
+
+static int kvm_unmap_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				 struct kvm_memory_slot *memslot, void *data)
+{
+	kvm_mips_flush_gpa_pt(kvm, gfn, gfn_end);
+	return 1;
+}
+
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
+{
+	unsigned long end = hva + PAGE_SIZE;
+
+	handle_hva_to_gpa(kvm, hva, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	handle_hva_to_gpa(kvm, start, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+static int kvm_set_spte_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t hva_pte = *(pte_t *)data;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	pte_t old_pte;
+
+	if (!gpa_pte)
+		return 0;
+
+	/* Mapping may need adjusting depending on memslot flags */
+	old_pte = *gpa_pte;
+	if (memslot->flags & KVM_MEM_LOG_DIRTY_PAGES && !pte_dirty(old_pte))
+		hva_pte = pte_mkclean(hva_pte);
+	else if (memslot->flags & KVM_MEM_READONLY)
+		hva_pte = pte_wrprotect(hva_pte);
+
+	set_pte(gpa_pte, hva_pte);
+
+	/* Replacing an absent or old page doesn't need flushes */
+	if (!pte_present(old_pte) || !pte_young(old_pte))
+		return 0;
+
+	/* Pages swapped, aged, moved, or cleaned require flushes */
+	return !pte_present(hva_pte) ||
+	       !pte_young(hva_pte) ||
+	       pte_pfn(old_pte) != pte_pfn(hva_pte) ||
+	       (pte_dirty(old_pte) && !pte_dirty(hva_pte));
+}
+
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
+{
+	unsigned long end = hva + PAGE_SIZE;
+	int ret;
+
+	ret = handle_hva_to_gpa(kvm, hva, end, &kvm_set_spte_handler, &pte);
+	if (ret)
+		kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+static int kvm_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+			       struct kvm_memory_slot *memslot, void *data)
+{
+	return kvm_mips_mkold_gpa_pt(kvm, gfn, gfn_end);
+}
+
+static int kvm_test_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				    struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+
+	if (!gpa_pte)
+		return 0;
+	return pte_young(*gpa_pte);
+}
+
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	return handle_hva_to_gpa(kvm, start, end, kvm_age_hva_handler, NULL);
+}
+
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
+{
+	return handle_hva_to_gpa(kvm, hva, hva, kvm_test_age_hva_handler, NULL);
+}
+
 /**
  * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
  * @vcpu:		VCPU pointer.
@@ -460,8 +608,8 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  *			NULL).
  *
  * Perform fast path GPA fault handling, doing all that can be done without
- * calling into KVM. This handles dirtying of clean pages (for dirty page
- * logging).
+ * calling into KVM. This handles marking old pages young (for idle page
+ * tracking), and dirtying of clean pages (for dirty page logging).
  *
  * Returns:	0 on success, in which case we can update derived mappings and
  *		resume guest execution.
@@ -475,6 +623,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 	struct kvm *kvm = vcpu->kvm;
 	gfn_t gfn = gpa >> PAGE_SHIFT;
 	pte_t *ptep;
+	kvm_pfn_t pfn;
+	bool pfn_valid = false;
 	int ret = 0;
 
 	spin_lock(&kvm->mmu_lock);
@@ -486,10 +636,24 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 		goto out;
 	}
 
+	/* Track access to pages marked old */
+	if (!pte_young(*ptep)) {
+		set_pte(ptep, pte_mkyoung(*ptep));
+		pfn = pte_pfn(*ptep);
+		pfn_valid = true;
+		/* call kvm_set_pfn_accessed() after unlock */
+	}
 	if (write_fault && !pte_dirty(*ptep)) {
-		/* Track dirtying of pages */
+		if (!pte_write(*ptep)) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		/* Track dirtying of writeable pages */
 		set_pte(ptep, pte_mkdirty(*ptep));
+		pfn = pte_pfn(*ptep);
 		mark_page_dirty(kvm, gfn);
+		kvm_set_pfn_dirty(pfn);
 	}
 
 	if (out_entry)
@@ -499,6 +663,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 out:
 	spin_unlock(&kvm->mmu_lock);
+	if (pfn_valid)
+		kvm_set_pfn_accessed(pfn);
 	return ret;
 }
 
@@ -514,9 +680,10 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of marking pages dirty (dirty page tracking), asking KVM for
- * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
- * mappings (GVA page tables and TLBs) must be handled by the caller.
+ * This takes care of marking pages young or dirty (idle/dirty page tracking),
+ * asking KVM for the corresponding PFN, and creating a mapping in the GPA page
+ * tables. Derived mappings (GVA page tables and TLBs) must be handled by the
+ * caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -535,9 +702,11 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	int srcu_idx, err;
 	kvm_pfn_t pfn;
 	pte_t *ptep, entry, old_pte;
+	bool writeable;
 	unsigned long prot_bits;
+	unsigned long mmu_seq;
 
-	/* Try the fast path to handle clean pages */
+	/* Try the fast path to handle old / clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
 	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
 				      out_buddy);
@@ -550,33 +719,63 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
-	/* Slow path - ask KVM core whether we can access this GPA */
-	pfn = gfn_to_pfn(kvm, gfn);
+retry:
+	/*
+	 * Used to check for invalidations in progress, of the pfn that is
+	 * returned by pfn_to_pfn_prot below.
+	 */
+	mmu_seq = kvm->mmu_notifier_seq;
+	/*
+	 * Ensure the read of mmu_notifier_seq isn't reordered with PTE reads in
+	 * gfn_to_pfn_prot() (which calls get_user_pages()), so that we don't
+	 * risk the page we get a reference to getting unmapped before we have a
+	 * chance to grab the mmu_lock without mmu_notifier_retry() noticing.
+	 *
+	 * This smp_rmb() pairs with the effective smp_wmb() of the combination
+	 * of the pte_unmap_unlock() after the PTE is zapped, and the
+	 * spin_lock() in kvm_mmu_notifier_invalidate_<page|range_end>() before
+	 * mmu_notifier_seq is incremented.
+	 */
+	smp_rmb();
 
+	/* Slow path - ask KVM core whether we can access this GPA */
+	pfn = gfn_to_pfn_prot(kvm, gfn, write_fault, &writeable);
 	if (is_error_noslot_pfn(pfn)) {
 		err = -EFAULT;
 		goto out;
 	}
 
 	spin_lock(&kvm->mmu_lock);
+	/* Check if an invalidation has taken place since we got pfn */
+	if (mmu_notifier_retry(kvm, mmu_seq)) {
+		/*
+		 * This can happen when mappings are changed asynchronously, but
+		 * also synchronously if a COW is triggered by
+		 * gfn_to_pfn_prot().
+		 */
+		spin_unlock(&kvm->mmu_lock);
+		kvm_release_pfn_clean(pfn);
+		goto retry;
+	}
 
 	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
 	/* Set up the PTE */
-	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
-		_page_cachable_default;
-	if (write_fault) {
-		prot_bits |= __WRITEABLE;
-		mark_page_dirty(kvm, gfn);
+	prot_bits = _PAGE_PRESENT | __READABLE | _page_cachable_default;
+	if (writeable) {
+		prot_bits |= _PAGE_WRITE;
+		if (write_fault) {
+			prot_bits |= __WRITEABLE;
+			mark_page_dirty(kvm, gfn);
+			kvm_set_pfn_dirty(pfn);
+		}
 	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
 	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
-	if (pte_present(old_pte))
-		kvm_release_pfn_clean(pte_pfn(old_pte));
 
 	err = 0;
 	if (out_entry)
@@ -585,6 +784,8 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 		*out_buddy = *ptep_buddy(ptep);
 
 	spin_unlock(&kvm->mmu_lock);
+	kvm_release_pfn_clean(pfn);
+	kvm_set_pfn_accessed(pfn);
 out:
 	srcu_read_unlock(&kvm->srcu, srcu_idx);
 	return err;
-- 
git-series 0.8.10

[PATCH 12/13] KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU

[James Hogan]

Implement the SYNC_MMU capability for KVM MIPS, allowing changes in the
underlying user host virtual address (HVA) mappings to be promptly
reflected in the corresponding guest physical address (GPA) mappings.

This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning.

There are two main aspects of this change, described below.

The KVM MMU notifier architecture callbacks are implemented so we can be
notified of changes in the HVA mappings. These arrange for the guest
physical address (GPA) page tables to be modified and possibly for
derived mappings (GVA page tables and TLBs) to be flushed.

 - kvm_unmap_hva[_range]() - These deal with HVA mappings being removed,
   for example before a copy-on-write takes place, which requires the
   corresponding GPA page table mappings to be removed too.

 - kvm_set_spte_hva() - These update a GPA page table entry to match the
   new HVA entry, but must be careful to respect KVM specific
   configuration such as not dirtying a clean guest page which is dirty
   to the host, and write protecting writable pages in read only
   memslots (which will soon be supported).

 - kvm[_test]_age_hva() - These update GPA page table entries to be old
   (invalid) so that access can be tracked, making them young again.

The GPA page fault handling (kvm_mips_map_page) is updated to use
gfn_to_pfn_prot() (which may provide read-only pages), to handle
asynchronous page table invalidation from MMU notifier callbacks, and to
handle more cases in the fast path.

 - mmu_notifier_seq is used to detect asynchronous page table
   invalidations while we're holding a pfn from gfn_to_pfn_prot()
   outside of kvm->mmu_lock, retrying if invalidations have taken place,
   e.g. a COW or a KSM page merge.

 - The fast path (_kvm_mips_map_page_fast) now handles marking old pages
   as young / accessed, and disallowing dirtying of clean pages that
   aren't actually writable (e.g. shared pages that should COW, and
   read-only memory regions when they are enabled in a future patch).

 - Due to the use of MMU notifications we no longer need to keep the
   page references after we've updated the GPA page tables.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/asm/kvm_host.h |  13 ++-
 arch/mips/kvm/Kconfig            |   1 +-
 arch/mips/kvm/mips.c             |   1 +-
 arch/mips/kvm/mmu.c              | 235 +++++++++++++++++++++++++++++---
 4 files changed, 233 insertions(+), 17 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 8f12385ccebd..31e7f3797f52 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -660,6 +660,19 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 						   unsigned long gva,
 						   bool write);
 
+#define KVM_ARCH_WANT_MMU_NOTIFIER
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
+int kvm_unmap_hva_range(struct kvm *kvm,
+			unsigned long start, unsigned long end);
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
+
+static inline void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
+							 unsigned long address)
+{
+}
+
 /* Emulation */
 int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
 enum emulation_result update_pc(struct kvm_vcpu *vcpu, u32 cause);
diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 85c4593b634a..65067327db12 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -22,6 +22,7 @@ config KVM
 	select ANON_INODES
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
+	select MMU_NOTIFIER
 	select SRCU
 	---help---
 	  Support for hosting Guest kernels.
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 5b7e9a053f77..425e4139742c 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 8a01bbd276fc..001a3e6ce2c4 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -194,7 +194,6 @@ static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
 		if (!pte_present(pte[i]))
 			continue;
 
-		kvm_release_pfn_clean(pte_pfn(pte[i]));
 		set_pte(pte + i, __pte(0));
 	}
 	return safe_to_remove;
@@ -450,6 +449,155 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 	kvm_mips_mkclean_gpa_pt(kvm, start, end);
 }
 
+/*
+ * kvm_mips_mkold_gpa_pt.
+ * Mark a range of guest physical address space old (all accesses fault) in the
+ * VM's GPA page table to allow detection of commonly used pages.
+ */
+
+BUILD_PTE_RANGE_OP(mkold, pte_mkold)
+
+static int kvm_mips_mkold_gpa_pt(struct kvm *kvm, gfn_t start_gfn,
+				 gfn_t end_gfn)
+{
+	return kvm_mips_mkold_pgd(kvm->arch.gpa_mm.pgd,
+				  start_gfn << PAGE_SHIFT,
+				  end_gfn << PAGE_SHIFT);
+}
+
+static int handle_hva_to_gpa(struct kvm *kvm,
+			     unsigned long start,
+			     unsigned long end,
+			     int (*handler)(struct kvm *kvm, gfn_t gfn,
+					    gpa_t gfn_end,
+					    struct kvm_memory_slot *memslot,
+					    void *data),
+			     void *data)
+{
+	struct kvm_memslots *slots;
+	struct kvm_memory_slot *memslot;
+	int ret = 0;
+
+	slots = kvm_memslots(kvm);
+
+	/* we only care about the pages that the guest sees */
+	kvm_for_each_memslot(memslot, slots) {
+		unsigned long hva_start, hva_end;
+		gfn_t gfn, gfn_end;
+
+		hva_start = max(start, memslot->userspace_addr);
+		hva_end = min(end, memslot->userspace_addr +
+					(memslot->npages << PAGE_SHIFT));
+		if (hva_start >= hva_end)
+			continue;
+
+		/*
+		 * {gfn(page) | page intersects with [hva_start, hva_end)} =
+		 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
+		 */
+		gfn = hva_to_gfn_memslot(hva_start, memslot);
+		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
+
+		ret |= handler(kvm, gfn, gfn_end, memslot, data);
+	}
+
+	return ret;
+}
+
+
+static int kvm_unmap_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				 struct kvm_memory_slot *memslot, void *data)
+{
+	kvm_mips_flush_gpa_pt(kvm, gfn, gfn_end);
+	return 1;
+}
+
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
+{
+	unsigned long end = hva + PAGE_SIZE;
+
+	handle_hva_to_gpa(kvm, hva, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	handle_hva_to_gpa(kvm, start, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+static int kvm_set_spte_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t hva_pte = *(pte_t *)data;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	pte_t old_pte;
+
+	if (!gpa_pte)
+		return 0;
+
+	/* Mapping may need adjusting depending on memslot flags */
+	old_pte = *gpa_pte;
+	if (memslot->flags & KVM_MEM_LOG_DIRTY_PAGES && !pte_dirty(old_pte))
+		hva_pte = pte_mkclean(hva_pte);
+	else if (memslot->flags & KVM_MEM_READONLY)
+		hva_pte = pte_wrprotect(hva_pte);
+
+	set_pte(gpa_pte, hva_pte);
+
+	/* Replacing an absent or old page doesn't need flushes */
+	if (!pte_present(old_pte) || !pte_young(old_pte))
+		return 0;
+
+	/* Pages swapped, aged, moved, or cleaned require flushes */
+	return !pte_present(hva_pte) ||
+	       !pte_young(hva_pte) ||
+	       pte_pfn(old_pte) != pte_pfn(hva_pte) ||
+	       (pte_dirty(old_pte) && !pte_dirty(hva_pte));
+}
+
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
+{
+	unsigned long end = hva + PAGE_SIZE;
+	int ret;
+
+	ret = handle_hva_to_gpa(kvm, hva, end, &kvm_set_spte_handler, &pte);
+	if (ret)
+		kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+static int kvm_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+			       struct kvm_memory_slot *memslot, void *data)
+{
+	return kvm_mips_mkold_gpa_pt(kvm, gfn, gfn_end);
+}
+
+static int kvm_test_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				    struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+
+	if (!gpa_pte)
+		return 0;
+	return pte_young(*gpa_pte);
+}
+
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	return handle_hva_to_gpa(kvm, start, end, kvm_age_hva_handler, NULL);
+}
+
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
+{
+	return handle_hva_to_gpa(kvm, hva, hva, kvm_test_age_hva_handler, NULL);
+}
+
 /**
  * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
  * @vcpu:		VCPU pointer.
@@ -460,8 +608,8 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  *			NULL).
  *
  * Perform fast path GPA fault handling, doing all that can be done without
- * calling into KVM. This handles dirtying of clean pages (for dirty page
- * logging).
+ * calling into KVM. This handles marking old pages young (for idle page
+ * tracking), and dirtying of clean pages (for dirty page logging).
  *
  * Returns:	0 on success, in which case we can update derived mappings and
  *		resume guest execution.
@@ -475,6 +623,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 	struct kvm *kvm = vcpu->kvm;
 	gfn_t gfn = gpa >> PAGE_SHIFT;
 	pte_t *ptep;
+	kvm_pfn_t pfn;
+	bool pfn_valid = false;
 	int ret = 0;
 
 	spin_lock(&kvm->mmu_lock);
@@ -486,10 +636,24 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 		goto out;
 	}
 
+	/* Track access to pages marked old */
+	if (!pte_young(*ptep)) {
+		set_pte(ptep, pte_mkyoung(*ptep));
+		pfn = pte_pfn(*ptep);
+		pfn_valid = true;
+		/* call kvm_set_pfn_accessed() after unlock */
+	}
 	if (write_fault && !pte_dirty(*ptep)) {
-		/* Track dirtying of pages */
+		if (!pte_write(*ptep)) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		/* Track dirtying of writeable pages */
 		set_pte(ptep, pte_mkdirty(*ptep));
+		pfn = pte_pfn(*ptep);
 		mark_page_dirty(kvm, gfn);
+		kvm_set_pfn_dirty(pfn);
 	}
 
 	if (out_entry)
@@ -499,6 +663,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 out:
 	spin_unlock(&kvm->mmu_lock);
+	if (pfn_valid)
+		kvm_set_pfn_accessed(pfn);
 	return ret;
 }
 
@@ -514,9 +680,10 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of marking pages dirty (dirty page tracking), asking KVM for
- * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
- * mappings (GVA page tables and TLBs) must be handled by the caller.
+ * This takes care of marking pages young or dirty (idle/dirty page tracking),
+ * asking KVM for the corresponding PFN, and creating a mapping in the GPA page
+ * tables. Derived mappings (GVA page tables and TLBs) must be handled by the
+ * caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -535,9 +702,11 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	int srcu_idx, err;
 	kvm_pfn_t pfn;
 	pte_t *ptep, entry, old_pte;
+	bool writeable;
 	unsigned long prot_bits;
+	unsigned long mmu_seq;
 
-	/* Try the fast path to handle clean pages */
+	/* Try the fast path to handle old / clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
 	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
 				      out_buddy);
@@ -550,33 +719,63 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
-	/* Slow path - ask KVM core whether we can access this GPA */
-	pfn = gfn_to_pfn(kvm, gfn);
+retry:
+	/*
+	 * Used to check for invalidations in progress, of the pfn that is
+	 * returned by pfn_to_pfn_prot below.
+	 */
+	mmu_seq = kvm->mmu_notifier_seq;
+	/*
+	 * Ensure the read of mmu_notifier_seq isn't reordered with PTE reads in
+	 * gfn_to_pfn_prot() (which calls get_user_pages()), so that we don't
+	 * risk the page we get a reference to getting unmapped before we have a
+	 * chance to grab the mmu_lock without mmu_notifier_retry() noticing.
+	 *
+	 * This smp_rmb() pairs with the effective smp_wmb() of the combination
+	 * of the pte_unmap_unlock() after the PTE is zapped, and the
+	 * spin_lock() in kvm_mmu_notifier_invalidate_<page|range_end>() before
+	 * mmu_notifier_seq is incremented.
+	 */
+	smp_rmb();
 
+	/* Slow path - ask KVM core whether we can access this GPA */
+	pfn = gfn_to_pfn_prot(kvm, gfn, write_fault, &writeable);
 	if (is_error_noslot_pfn(pfn)) {
 		err = -EFAULT;
 		goto out;
 	}
 
 	spin_lock(&kvm->mmu_lock);
+	/* Check if an invalidation has taken place since we got pfn */
+	if (mmu_notifier_retry(kvm, mmu_seq)) {
+		/*
+		 * This can happen when mappings are changed asynchronously, but
+		 * also synchronously if a COW is triggered by
+		 * gfn_to_pfn_prot().
+		 */
+		spin_unlock(&kvm->mmu_lock);
+		kvm_release_pfn_clean(pfn);
+		goto retry;
+	}
 
 	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
 	/* Set up the PTE */
-	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
-		_page_cachable_default;
-	if (write_fault) {
-		prot_bits |= __WRITEABLE;
-		mark_page_dirty(kvm, gfn);
+	prot_bits = _PAGE_PRESENT | __READABLE | _page_cachable_default;
+	if (writeable) {
+		prot_bits |= _PAGE_WRITE;
+		if (write_fault) {
+			prot_bits |= __WRITEABLE;
+			mark_page_dirty(kvm, gfn);
+			kvm_set_pfn_dirty(pfn);
+		}
 	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
 	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
-	if (pte_present(old_pte))
-		kvm_release_pfn_clean(pte_pfn(old_pte));
 
 	err = 0;
 	if (out_entry)
@@ -585,6 +784,8 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 		*out_buddy = *ptep_buddy(ptep);
 
 	spin_unlock(&kvm->mmu_lock);
+	kvm_release_pfn_clean(pfn);
+	kvm_set_pfn_accessed(pfn);
 out:
 	srcu_read_unlock(&kvm->srcu, srcu_idx);
 	return err;
-- 
git-series 0.8.10

[PATCH v2 12/13] KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU

[James Hogan]

Implement the SYNC_MMU capability for KVM MIPS, allowing changes in the
underlying user host virtual address (HVA) mappings to be promptly
reflected in the corresponding guest physical address (GPA) mappings.

This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning.

There are two main aspects of this change, described below.

The KVM MMU notifier architecture callbacks are implemented so we can be
notified of changes in the HVA mappings. These arrange for the guest
physical address (GPA) page tables to be modified and possibly for
derived mappings (GVA page tables and TLBs) to be flushed.

 - kvm_unmap_hva[_range]() - These deal with HVA mappings being removed,
   for example before a copy-on-write takes place, which requires the
   corresponding GPA page table mappings to be removed too.

 - kvm_set_spte_hva() - These update a GPA page table entry to match the
   new HVA entry, but must be careful to respect KVM specific
   configuration such as not dirtying a clean guest page which is dirty
   to the host, and write protecting writable pages in read only
   memslots (which will soon be supported).

 - kvm[_test]_age_hva() - These update GPA page table entries to be old
   (invalid) so that access can be tracked, making them young again.

The GPA page fault handling (kvm_mips_map_page) is updated to use
gfn_to_pfn_prot() (which may provide read-only pages), to handle
asynchronous page table invalidation from MMU notifier callbacks, and to
handle more cases in the fast path.

 - mmu_notifier_seq is used to detect asynchronous page table
   invalidations while we're holding a pfn from gfn_to_pfn_prot()
   outside of kvm->mmu_lock, retrying if invalidations have taken place,
   e.g. a COW or a KSM page merge.

 - The fast path (_kvm_mips_map_page_fast) now handles marking old pages
   as young / accessed, and disallowing dirtying of clean pages that
   aren't actually writable (e.g. shared pages that should COW, and
   read-only memory regions when they are enabled in a future patch).

 - Due to the use of MMU notifications we no longer need to keep the
   page references after we've updated the GPA page tables.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
Changes in v2:
- Silence bogus GCC 4.9 warning by initialising pfn to 0:

  arch/mips/kvm/mmu.c In function ‘kvm_mips_map_page’:
  arch/mips/kvm/mmu.c +667 :3: error: ‘pfn’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
     kvm_set_pfn_accessed(pfn);
     ^
  arch/mips/kvm/mmu.c +626 :12: note: ‘pfn’ was declared here
    kvm_pfn_t pfn;
              ^
---
 arch/mips/include/asm/kvm_host.h |  13 ++-
 arch/mips/kvm/Kconfig            |   1 +-
 arch/mips/kvm/mips.c             |   1 +-
 arch/mips/kvm/mmu.c              | 235 +++++++++++++++++++++++++++++---
 4 files changed, 233 insertions(+), 17 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index a7394940119c..718dfffa17d5 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -663,6 +663,19 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 						   unsigned long gva,
 						   bool write);
 
+#define KVM_ARCH_WANT_MMU_NOTIFIER
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
+int kvm_unmap_hva_range(struct kvm *kvm,
+			unsigned long start, unsigned long end);
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
+
+static inline void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
+							 unsigned long address)
+{
+}
+
 /* Emulation */
 int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
 enum emulation_result update_pc(struct kvm_vcpu *vcpu, u32 cause);
diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 85c4593b634a..65067327db12 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -22,6 +22,7 @@ config KVM
 	select ANON_INODES
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
+	select MMU_NOTIFIER
 	select SRCU
 	---help---
 	  Support for hosting Guest kernels.
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 5b7e9a053f77..425e4139742c 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 8a01bbd276fc..cb0faade311e 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -194,7 +194,6 @@ static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
 		if (!pte_present(pte[i]))
 			continue;
 
-		kvm_release_pfn_clean(pte_pfn(pte[i]));
 		set_pte(pte + i, __pte(0));
 	}
 	return safe_to_remove;
@@ -450,6 +449,155 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 	kvm_mips_mkclean_gpa_pt(kvm, start, end);
 }
 
+/*
+ * kvm_mips_mkold_gpa_pt.
+ * Mark a range of guest physical address space old (all accesses fault) in the
+ * VM's GPA page table to allow detection of commonly used pages.
+ */
+
+BUILD_PTE_RANGE_OP(mkold, pte_mkold)
+
+static int kvm_mips_mkold_gpa_pt(struct kvm *kvm, gfn_t start_gfn,
+				 gfn_t end_gfn)
+{
+	return kvm_mips_mkold_pgd(kvm->arch.gpa_mm.pgd,
+				  start_gfn << PAGE_SHIFT,
+				  end_gfn << PAGE_SHIFT);
+}
+
+static int handle_hva_to_gpa(struct kvm *kvm,
+			     unsigned long start,
+			     unsigned long end,
+			     int (*handler)(struct kvm *kvm, gfn_t gfn,
+					    gpa_t gfn_end,
+					    struct kvm_memory_slot *memslot,
+					    void *data),
+			     void *data)
+{
+	struct kvm_memslots *slots;
+	struct kvm_memory_slot *memslot;
+	int ret = 0;
+
+	slots = kvm_memslots(kvm);
+
+	/* we only care about the pages that the guest sees */
+	kvm_for_each_memslot(memslot, slots) {
+		unsigned long hva_start, hva_end;
+		gfn_t gfn, gfn_end;
+
+		hva_start = max(start, memslot->userspace_addr);
+		hva_end = min(end, memslot->userspace_addr +
+					(memslot->npages << PAGE_SHIFT));
+		if (hva_start >= hva_end)
+			continue;
+
+		/*
+		 * {gfn(page) | page intersects with [hva_start, hva_end)} =
+		 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
+		 */
+		gfn = hva_to_gfn_memslot(hva_start, memslot);
+		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
+
+		ret |= handler(kvm, gfn, gfn_end, memslot, data);
+	}
+
+	return ret;
+}
+
+
+static int kvm_unmap_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				 struct kvm_memory_slot *memslot, void *data)
+{
+	kvm_mips_flush_gpa_pt(kvm, gfn, gfn_end);
+	return 1;
+}
+
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
+{
+	unsigned long end = hva + PAGE_SIZE;
+
+	handle_hva_to_gpa(kvm, hva, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	handle_hva_to_gpa(kvm, start, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+static int kvm_set_spte_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t hva_pte = *(pte_t *)data;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	pte_t old_pte;
+
+	if (!gpa_pte)
+		return 0;
+
+	/* Mapping may need adjusting depending on memslot flags */
+	old_pte = *gpa_pte;
+	if (memslot->flags & KVM_MEM_LOG_DIRTY_PAGES && !pte_dirty(old_pte))
+		hva_pte = pte_mkclean(hva_pte);
+	else if (memslot->flags & KVM_MEM_READONLY)
+		hva_pte = pte_wrprotect(hva_pte);
+
+	set_pte(gpa_pte, hva_pte);
+
+	/* Replacing an absent or old page doesn't need flushes */
+	if (!pte_present(old_pte) || !pte_young(old_pte))
+		return 0;
+
+	/* Pages swapped, aged, moved, or cleaned require flushes */
+	return !pte_present(hva_pte) ||
+	       !pte_young(hva_pte) ||
+	       pte_pfn(old_pte) != pte_pfn(hva_pte) ||
+	       (pte_dirty(old_pte) && !pte_dirty(hva_pte));
+}
+
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
+{
+	unsigned long end = hva + PAGE_SIZE;
+	int ret;
+
+	ret = handle_hva_to_gpa(kvm, hva, end, &kvm_set_spte_handler, &pte);
+	if (ret)
+		kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+static int kvm_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+			       struct kvm_memory_slot *memslot, void *data)
+{
+	return kvm_mips_mkold_gpa_pt(kvm, gfn, gfn_end);
+}
+
+static int kvm_test_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				    struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+
+	if (!gpa_pte)
+		return 0;
+	return pte_young(*gpa_pte);
+}
+
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	return handle_hva_to_gpa(kvm, start, end, kvm_age_hva_handler, NULL);
+}
+
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
+{
+	return handle_hva_to_gpa(kvm, hva, hva, kvm_test_age_hva_handler, NULL);
+}
+
 /**
  * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
  * @vcpu:		VCPU pointer.
@@ -460,8 +608,8 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  *			NULL).
  *
  * Perform fast path GPA fault handling, doing all that can be done without
- * calling into KVM. This handles dirtying of clean pages (for dirty page
- * logging).
+ * calling into KVM. This handles marking old pages young (for idle page
+ * tracking), and dirtying of clean pages (for dirty page logging).
  *
  * Returns:	0 on success, in which case we can update derived mappings and
  *		resume guest execution.
@@ -475,6 +623,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 	struct kvm *kvm = vcpu->kvm;
 	gfn_t gfn = gpa >> PAGE_SHIFT;
 	pte_t *ptep;
+	kvm_pfn_t pfn = 0;	/* silence bogus GCC warning */
+	bool pfn_valid = false;
 	int ret = 0;
 
 	spin_lock(&kvm->mmu_lock);
@@ -486,10 +636,24 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 		goto out;
 	}
 
+	/* Track access to pages marked old */
+	if (!pte_young(*ptep)) {
+		set_pte(ptep, pte_mkyoung(*ptep));
+		pfn = pte_pfn(*ptep);
+		pfn_valid = true;
+		/* call kvm_set_pfn_accessed() after unlock */
+	}
 	if (write_fault && !pte_dirty(*ptep)) {
-		/* Track dirtying of pages */
+		if (!pte_write(*ptep)) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		/* Track dirtying of writeable pages */
 		set_pte(ptep, pte_mkdirty(*ptep));
+		pfn = pte_pfn(*ptep);
 		mark_page_dirty(kvm, gfn);
+		kvm_set_pfn_dirty(pfn);
 	}
 
 	if (out_entry)
@@ -499,6 +663,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 out:
 	spin_unlock(&kvm->mmu_lock);
+	if (pfn_valid)
+		kvm_set_pfn_accessed(pfn);
 	return ret;
 }
 
@@ -514,9 +680,10 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of marking pages dirty (dirty page tracking), asking KVM for
- * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
- * mappings (GVA page tables and TLBs) must be handled by the caller.
+ * This takes care of marking pages young or dirty (idle/dirty page tracking),
+ * asking KVM for the corresponding PFN, and creating a mapping in the GPA page
+ * tables. Derived mappings (GVA page tables and TLBs) must be handled by the
+ * caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -535,9 +702,11 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	int srcu_idx, err;
 	kvm_pfn_t pfn;
 	pte_t *ptep, entry, old_pte;
+	bool writeable;
 	unsigned long prot_bits;
+	unsigned long mmu_seq;
 
-	/* Try the fast path to handle clean pages */
+	/* Try the fast path to handle old / clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
 	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
 				      out_buddy);
@@ -550,33 +719,63 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
-	/* Slow path - ask KVM core whether we can access this GPA */
-	pfn = gfn_to_pfn(kvm, gfn);
+retry:
+	/*
+	 * Used to check for invalidations in progress, of the pfn that is
+	 * returned by pfn_to_pfn_prot below.
+	 */
+	mmu_seq = kvm->mmu_notifier_seq;
+	/*
+	 * Ensure the read of mmu_notifier_seq isn't reordered with PTE reads in
+	 * gfn_to_pfn_prot() (which calls get_user_pages()), so that we don't
+	 * risk the page we get a reference to getting unmapped before we have a
+	 * chance to grab the mmu_lock without mmu_notifier_retry() noticing.
+	 *
+	 * This smp_rmb() pairs with the effective smp_wmb() of the combination
+	 * of the pte_unmap_unlock() after the PTE is zapped, and the
+	 * spin_lock() in kvm_mmu_notifier_invalidate_<page|range_end>() before
+	 * mmu_notifier_seq is incremented.
+	 */
+	smp_rmb();
 
+	/* Slow path - ask KVM core whether we can access this GPA */
+	pfn = gfn_to_pfn_prot(kvm, gfn, write_fault, &writeable);
 	if (is_error_noslot_pfn(pfn)) {
 		err = -EFAULT;
 		goto out;
 	}
 
 	spin_lock(&kvm->mmu_lock);
+	/* Check if an invalidation has taken place since we got pfn */
+	if (mmu_notifier_retry(kvm, mmu_seq)) {
+		/*
+		 * This can happen when mappings are changed asynchronously, but
+		 * also synchronously if a COW is triggered by
+		 * gfn_to_pfn_prot().
+		 */
+		spin_unlock(&kvm->mmu_lock);
+		kvm_release_pfn_clean(pfn);
+		goto retry;
+	}
 
 	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
 	/* Set up the PTE */
-	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
-		_page_cachable_default;
-	if (write_fault) {
-		prot_bits |= __WRITEABLE;
-		mark_page_dirty(kvm, gfn);
+	prot_bits = _PAGE_PRESENT | __READABLE | _page_cachable_default;
+	if (writeable) {
+		prot_bits |= _PAGE_WRITE;
+		if (write_fault) {
+			prot_bits |= __WRITEABLE;
+			mark_page_dirty(kvm, gfn);
+			kvm_set_pfn_dirty(pfn);
+		}
 	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
 	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
-	if (pte_present(old_pte))
-		kvm_release_pfn_clean(pte_pfn(old_pte));
 
 	err = 0;
 	if (out_entry)
@@ -585,6 +784,8 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 		*out_buddy = *ptep_buddy(ptep);
 
 	spin_unlock(&kvm->mmu_lock);
+	kvm_release_pfn_clean(pfn);
+	kvm_set_pfn_accessed(pfn);
 out:
 	srcu_read_unlock(&kvm->srcu, srcu_idx);
 	return err;
-- 
git-series 0.8.10

[PATCH v2 12/13] KVM: MIPS/MMU: Implement KVM_CAP_SYNC_MMU

[James Hogan]

Implement the SYNC_MMU capability for KVM MIPS, allowing changes in the
underlying user host virtual address (HVA) mappings to be promptly
reflected in the corresponding guest physical address (GPA) mappings.

This allows for several features to work with guest RAM which require
mappings to be altered or protected, such as copy-on-write, KSM (Kernel
Samepage Merging), idle page tracking, memory swapping, and guest memory
ballooning.

There are two main aspects of this change, described below.

The KVM MMU notifier architecture callbacks are implemented so we can be
notified of changes in the HVA mappings. These arrange for the guest
physical address (GPA) page tables to be modified and possibly for
derived mappings (GVA page tables and TLBs) to be flushed.

 - kvm_unmap_hva[_range]() - These deal with HVA mappings being removed,
   for example before a copy-on-write takes place, which requires the
   corresponding GPA page table mappings to be removed too.

 - kvm_set_spte_hva() - These update a GPA page table entry to match the
   new HVA entry, but must be careful to respect KVM specific
   configuration such as not dirtying a clean guest page which is dirty
   to the host, and write protecting writable pages in read only
   memslots (which will soon be supported).

 - kvm[_test]_age_hva() - These update GPA page table entries to be old
   (invalid) so that access can be tracked, making them young again.

The GPA page fault handling (kvm_mips_map_page) is updated to use
gfn_to_pfn_prot() (which may provide read-only pages), to handle
asynchronous page table invalidation from MMU notifier callbacks, and to
handle more cases in the fast path.

 - mmu_notifier_seq is used to detect asynchronous page table
   invalidations while we're holding a pfn from gfn_to_pfn_prot()
   outside of kvm->mmu_lock, retrying if invalidations have taken place,
   e.g. a COW or a KSM page merge.

 - The fast path (_kvm_mips_map_page_fast) now handles marking old pages
   as young / accessed, and disallowing dirtying of clean pages that
   aren't actually writable (e.g. shared pages that should COW, and
   read-only memory regions when they are enabled in a future patch).

 - Due to the use of MMU notifications we no longer need to keep the
   page references after we've updated the GPA page tables.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
Changes in v2:
- Silence bogus GCC 4.9 warning by initialising pfn to 0:

  arch/mips/kvm/mmu.c In function ‘kvm_mips_map_page’:
  arch/mips/kvm/mmu.c +667 :3: error: ‘pfn’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
     kvm_set_pfn_accessed(pfn);
     ^
  arch/mips/kvm/mmu.c +626 :12: note: ‘pfn’ was declared here
    kvm_pfn_t pfn;
              ^
---
 arch/mips/include/asm/kvm_host.h |  13 ++-
 arch/mips/kvm/Kconfig            |   1 +-
 arch/mips/kvm/mips.c             |   1 +-
 arch/mips/kvm/mmu.c              | 235 +++++++++++++++++++++++++++++---
 4 files changed, 233 insertions(+), 17 deletions(-)

diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index a7394940119c..718dfffa17d5 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -663,6 +663,19 @@ enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
 						   unsigned long gva,
 						   bool write);
 
+#define KVM_ARCH_WANT_MMU_NOTIFIER
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
+int kvm_unmap_hva_range(struct kvm *kvm,
+			unsigned long start, unsigned long end);
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
+
+static inline void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
+							 unsigned long address)
+{
+}
+
 /* Emulation */
 int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
 enum emulation_result update_pc(struct kvm_vcpu *vcpu, u32 cause);
diff --git a/arch/mips/kvm/Kconfig b/arch/mips/kvm/Kconfig
index 85c4593b634a..65067327db12 100644
--- a/arch/mips/kvm/Kconfig
+++ b/arch/mips/kvm/Kconfig
@@ -22,6 +22,7 @@ config KVM
 	select ANON_INODES
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_MMIO
+	select MMU_NOTIFIER
 	select SRCU
 	---help---
 	  Support for hosting Guest kernels.
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 5b7e9a053f77..425e4139742c 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 8a01bbd276fc..cb0faade311e 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -194,7 +194,6 @@ static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
 		if (!pte_present(pte[i]))
 			continue;
 
-		kvm_release_pfn_clean(pte_pfn(pte[i]));
 		set_pte(pte + i, __pte(0));
 	}
 	return safe_to_remove;
@@ -450,6 +449,155 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 	kvm_mips_mkclean_gpa_pt(kvm, start, end);
 }
 
+/*
+ * kvm_mips_mkold_gpa_pt.
+ * Mark a range of guest physical address space old (all accesses fault) in the
+ * VM's GPA page table to allow detection of commonly used pages.
+ */
+
+BUILD_PTE_RANGE_OP(mkold, pte_mkold)
+
+static int kvm_mips_mkold_gpa_pt(struct kvm *kvm, gfn_t start_gfn,
+				 gfn_t end_gfn)
+{
+	return kvm_mips_mkold_pgd(kvm->arch.gpa_mm.pgd,
+				  start_gfn << PAGE_SHIFT,
+				  end_gfn << PAGE_SHIFT);
+}
+
+static int handle_hva_to_gpa(struct kvm *kvm,
+			     unsigned long start,
+			     unsigned long end,
+			     int (*handler)(struct kvm *kvm, gfn_t gfn,
+					    gpa_t gfn_end,
+					    struct kvm_memory_slot *memslot,
+					    void *data),
+			     void *data)
+{
+	struct kvm_memslots *slots;
+	struct kvm_memory_slot *memslot;
+	int ret = 0;
+
+	slots = kvm_memslots(kvm);
+
+	/* we only care about the pages that the guest sees */
+	kvm_for_each_memslot(memslot, slots) {
+		unsigned long hva_start, hva_end;
+		gfn_t gfn, gfn_end;
+
+		hva_start = max(start, memslot->userspace_addr);
+		hva_end = min(end, memslot->userspace_addr +
+					(memslot->npages << PAGE_SHIFT));
+		if (hva_start >= hva_end)
+			continue;
+
+		/*
+		 * {gfn(page) | page intersects with [hva_start, hva_end)} =
+		 * {gfn_start, gfn_start+1, ..., gfn_end-1}.
+		 */
+		gfn = hva_to_gfn_memslot(hva_start, memslot);
+		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
+
+		ret |= handler(kvm, gfn, gfn_end, memslot, data);
+	}
+
+	return ret;
+}
+
+
+static int kvm_unmap_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				 struct kvm_memory_slot *memslot, void *data)
+{
+	kvm_mips_flush_gpa_pt(kvm, gfn, gfn_end);
+	return 1;
+}
+
+int kvm_unmap_hva(struct kvm *kvm, unsigned long hva)
+{
+	unsigned long end = hva + PAGE_SIZE;
+
+	handle_hva_to_gpa(kvm, hva, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	handle_hva_to_gpa(kvm, start, end, &kvm_unmap_hva_handler, NULL);
+
+	kvm_mips_callbacks->flush_shadow_all(kvm);
+	return 0;
+}
+
+static int kvm_set_spte_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t hva_pte = *(pte_t *)data;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+	pte_t old_pte;
+
+	if (!gpa_pte)
+		return 0;
+
+	/* Mapping may need adjusting depending on memslot flags */
+	old_pte = *gpa_pte;
+	if (memslot->flags & KVM_MEM_LOG_DIRTY_PAGES && !pte_dirty(old_pte))
+		hva_pte = pte_mkclean(hva_pte);
+	else if (memslot->flags & KVM_MEM_READONLY)
+		hva_pte = pte_wrprotect(hva_pte);
+
+	set_pte(gpa_pte, hva_pte);
+
+	/* Replacing an absent or old page doesn't need flushes */
+	if (!pte_present(old_pte) || !pte_young(old_pte))
+		return 0;
+
+	/* Pages swapped, aged, moved, or cleaned require flushes */
+	return !pte_present(hva_pte) ||
+	       !pte_young(hva_pte) ||
+	       pte_pfn(old_pte) != pte_pfn(hva_pte) ||
+	       (pte_dirty(old_pte) && !pte_dirty(hva_pte));
+}
+
+void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte)
+{
+	unsigned long end = hva + PAGE_SIZE;
+	int ret;
+
+	ret = handle_hva_to_gpa(kvm, hva, end, &kvm_set_spte_handler, &pte);
+	if (ret)
+		kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+static int kvm_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+			       struct kvm_memory_slot *memslot, void *data)
+{
+	return kvm_mips_mkold_gpa_pt(kvm, gfn, gfn_end);
+}
+
+static int kvm_test_age_hva_handler(struct kvm *kvm, gfn_t gfn, gfn_t gfn_end,
+				    struct kvm_memory_slot *memslot, void *data)
+{
+	gpa_t gpa = gfn << PAGE_SHIFT;
+	pte_t *gpa_pte = kvm_mips_pte_for_gpa(kvm, NULL, gpa);
+
+	if (!gpa_pte)
+		return 0;
+	return pte_young(*gpa_pte);
+}
+
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
+{
+	return handle_hva_to_gpa(kvm, start, end, kvm_age_hva_handler, NULL);
+}
+
+int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
+{
+	return handle_hva_to_gpa(kvm, hva, hva, kvm_test_age_hva_handler, NULL);
+}
+
 /**
  * _kvm_mips_map_page_fast() - Fast path GPA fault handler.
  * @vcpu:		VCPU pointer.
@@ -460,8 +608,8 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
  *			NULL).
  *
  * Perform fast path GPA fault handling, doing all that can be done without
- * calling into KVM. This handles dirtying of clean pages (for dirty page
- * logging).
+ * calling into KVM. This handles marking old pages young (for idle page
+ * tracking), and dirtying of clean pages (for dirty page logging).
  *
  * Returns:	0 on success, in which case we can update derived mappings and
  *		resume guest execution.
@@ -475,6 +623,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 	struct kvm *kvm = vcpu->kvm;
 	gfn_t gfn = gpa >> PAGE_SHIFT;
 	pte_t *ptep;
+	kvm_pfn_t pfn = 0;	/* silence bogus GCC warning */
+	bool pfn_valid = false;
 	int ret = 0;
 
 	spin_lock(&kvm->mmu_lock);
@@ -486,10 +636,24 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 		goto out;
 	}
 
+	/* Track access to pages marked old */
+	if (!pte_young(*ptep)) {
+		set_pte(ptep, pte_mkyoung(*ptep));
+		pfn = pte_pfn(*ptep);
+		pfn_valid = true;
+		/* call kvm_set_pfn_accessed() after unlock */
+	}
 	if (write_fault && !pte_dirty(*ptep)) {
-		/* Track dirtying of pages */
+		if (!pte_write(*ptep)) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		/* Track dirtying of writeable pages */
 		set_pte(ptep, pte_mkdirty(*ptep));
+		pfn = pte_pfn(*ptep);
 		mark_page_dirty(kvm, gfn);
+		kvm_set_pfn_dirty(pfn);
 	}
 
 	if (out_entry)
@@ -499,6 +663,8 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
 
 out:
 	spin_unlock(&kvm->mmu_lock);
+	if (pfn_valid)
+		kvm_set_pfn_accessed(pfn);
 	return ret;
 }
 
@@ -514,9 +680,10 @@ static int _kvm_mips_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
  * Handle GPA faults by creating a new GPA mapping (or updating an existing
  * one).
  *
- * This takes care of marking pages dirty (dirty page tracking), asking KVM for
- * the corresponding PFN, and creating a mapping in the GPA page tables. Derived
- * mappings (GVA page tables and TLBs) must be handled by the caller.
+ * This takes care of marking pages young or dirty (idle/dirty page tracking),
+ * asking KVM for the corresponding PFN, and creating a mapping in the GPA page
+ * tables. Derived mappings (GVA page tables and TLBs) must be handled by the
+ * caller.
  *
  * Returns:	0 on success, in which case the caller may use the @out_entry
  *		and @out_buddy PTEs to update derived mappings and resume guest
@@ -535,9 +702,11 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	int srcu_idx, err;
 	kvm_pfn_t pfn;
 	pte_t *ptep, entry, old_pte;
+	bool writeable;
 	unsigned long prot_bits;
+	unsigned long mmu_seq;
 
-	/* Try the fast path to handle clean pages */
+	/* Try the fast path to handle old / clean pages */
 	srcu_idx = srcu_read_lock(&kvm->srcu);
 	err = _kvm_mips_map_page_fast(vcpu, gpa, write_fault, out_entry,
 				      out_buddy);
@@ -550,33 +719,63 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 	if (err)
 		goto out;
 
-	/* Slow path - ask KVM core whether we can access this GPA */
-	pfn = gfn_to_pfn(kvm, gfn);
+retry:
+	/*
+	 * Used to check for invalidations in progress, of the pfn that is
+	 * returned by pfn_to_pfn_prot below.
+	 */
+	mmu_seq = kvm->mmu_notifier_seq;
+	/*
+	 * Ensure the read of mmu_notifier_seq isn't reordered with PTE reads in
+	 * gfn_to_pfn_prot() (which calls get_user_pages()), so that we don't
+	 * risk the page we get a reference to getting unmapped before we have a
+	 * chance to grab the mmu_lock without mmu_notifier_retry() noticing.
+	 *
+	 * This smp_rmb() pairs with the effective smp_wmb() of the combination
+	 * of the pte_unmap_unlock() after the PTE is zapped, and the
+	 * spin_lock() in kvm_mmu_notifier_invalidate_<page|range_end>() before
+	 * mmu_notifier_seq is incremented.
+	 */
+	smp_rmb();
 
+	/* Slow path - ask KVM core whether we can access this GPA */
+	pfn = gfn_to_pfn_prot(kvm, gfn, write_fault, &writeable);
 	if (is_error_noslot_pfn(pfn)) {
 		err = -EFAULT;
 		goto out;
 	}
 
 	spin_lock(&kvm->mmu_lock);
+	/* Check if an invalidation has taken place since we got pfn */
+	if (mmu_notifier_retry(kvm, mmu_seq)) {
+		/*
+		 * This can happen when mappings are changed asynchronously, but
+		 * also synchronously if a COW is triggered by
+		 * gfn_to_pfn_prot().
+		 */
+		spin_unlock(&kvm->mmu_lock);
+		kvm_release_pfn_clean(pfn);
+		goto retry;
+	}
 
 	/* Ensure page tables are allocated */
 	ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
 
 	/* Set up the PTE */
-	prot_bits = __READABLE | _PAGE_PRESENT | _PAGE_WRITE |
-		_page_cachable_default;
-	if (write_fault) {
-		prot_bits |= __WRITEABLE;
-		mark_page_dirty(kvm, gfn);
+	prot_bits = _PAGE_PRESENT | __READABLE | _page_cachable_default;
+	if (writeable) {
+		prot_bits |= _PAGE_WRITE;
+		if (write_fault) {
+			prot_bits |= __WRITEABLE;
+			mark_page_dirty(kvm, gfn);
+			kvm_set_pfn_dirty(pfn);
+		}
 	}
 	entry = pfn_pte(pfn, __pgprot(prot_bits));
 
 	/* Write the PTE */
 	old_pte = *ptep;
 	set_pte(ptep, entry);
-	if (pte_present(old_pte))
-		kvm_release_pfn_clean(pte_pfn(old_pte));
 
 	err = 0;
 	if (out_entry)
@@ -585,6 +784,8 @@ static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
 		*out_buddy = *ptep_buddy(ptep);
 
 	spin_unlock(&kvm->mmu_lock);
+	kvm_release_pfn_clean(pfn);
+	kvm_set_pfn_accessed(pfn);
 out:
 	srcu_read_unlock(&kvm->srcu, srcu_idx);
 	return err;
-- 
git-series 0.8.10

[PATCH 13/13] KVM: MIPS: Claim KVM_CAP_READONLY_MEM support

[James Hogan]

Now that load/store faults due to read only memory regions are treated
as MMIO accesses it is safe to claim support for read only memory
regions (KVM_CAP_READONLY_MEM).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/uapi/asm/kvm.h | 2 ++
 arch/mips/kvm/mips.c             | 1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/mips/include/uapi/asm/kvm.h b/arch/mips/include/uapi/asm/kvm.h
index 6985eb59b085..a8a0199bf760 100644
--- a/arch/mips/include/uapi/asm/kvm.h
+++ b/arch/mips/include/uapi/asm/kvm.h
@@ -19,6 +19,8 @@
  * Some parts derived from the x86 version of this file.
  */
 
+#define __KVM_HAVE_READONLY_MEM
+
 /*
  * for KVM_GET_REGS and KVM_SET_REGS
  *
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 425e4139742c..cf501a5a0dfe 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_READONLY_MEM:
 	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
-- 
git-series 0.8.10

[PATCH 13/13] KVM: MIPS: Claim KVM_CAP_READONLY_MEM support

[James Hogan]

Now that load/store faults due to read only memory regions are treated
as MMIO accesses it is safe to claim support for read only memory
regions (KVM_CAP_READONLY_MEM).

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
 arch/mips/include/uapi/asm/kvm.h | 2 ++
 arch/mips/kvm/mips.c             | 1 +
 2 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/mips/include/uapi/asm/kvm.h b/arch/mips/include/uapi/asm/kvm.h
index 6985eb59b085..a8a0199bf760 100644
--- a/arch/mips/include/uapi/asm/kvm.h
+++ b/arch/mips/include/uapi/asm/kvm.h
@@ -19,6 +19,8 @@
  * Some parts derived from the x86 version of this file.
  */
 
+#define __KVM_HAVE_READONLY_MEM
+
 /*
  * for KVM_GET_REGS and KVM_SET_REGS
  *
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 425e4139742c..cf501a5a0dfe 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -1222,6 +1222,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	switch (ext) {
 	case KVM_CAP_ONE_REG:
 	case KVM_CAP_ENABLE_CAP:
+	case KVM_CAP_READONLY_MEM:
 	case KVM_CAP_SYNC_MMU:
 		r = 1;
 		break;
-- 
git-series 0.8.10

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 3196 bytes --]

Hi,

On Tue, Jan 17, 2017 at 07:06:39PM +0800, jiang.biao2@zte.com.cn wrote:
> ＞ kvm_mips_map_page() will need to know whether the fault was due to a
> ＞ read or a write in order to support dirty page tracking,
> ＞ KVM_CAP_SYNC_MMU, and read only memory regions, so get that information
> ＞ passed down to it via new bool write_fault arguments to various
> ＞ functions.
> 
> Hi, James,
> 
> 
> Maybe it's not good idea to add an argument and pass it down to various functions,  it will ,
> 
> make the the interface more complicated,  have more possibolity, and more difficult to test.
> 

kvm_mips_map_page() needs to know is whether it is being triggered in
response to a write, but it is abstracted and independent from what host
exception triggered it. It is only really concerned with the GPA space.

> 
> 
> 
> As far as I can see,  what you need is to let kvm_mips_map_page() know the *reason*.
> 
> would it be better to let kvm_mips_map_page() get *exccode* from the CP0 directly, or 
> 
> 
> provide any interface to get the fault reason?

I presume you mean from the saved host cause register in the VCPU
context (since intervening exceptions/interrupts will clobber the actual
CP0 Cause register).

It directly needs to know whether it can get away with a read-only
mapping, and although it directly depends on a GVA segment, it doesn't
necessarily relate to a memory access made by the guest.

kvm_mips_map_page() is called via:

 - kvm_mips_handle_kseg0_tlb_fault()
   for faults in guest KSeg0

 - kvm_mips_handle_mapped_seg_tlb_fault()
   for faults in guest TLB mapped segments

From these functions:

 - kvm_trap_emul_handle_tlb_mod() (write_fault = true)
   in response to a write to a read-only page (exccode = MOD)

 - kvm_trap_emul_handle_tlb_miss() (write_fault = true or false)
   in response to a read or write when TLB mapping absent or invalid
   (exccode = TLBL/TLBS)

 - via the kvm_trap_emul_gva_fault() helper when KVM needs to directly
   access GVA space:

   - kvm_get_inst() (write_fault = false)
     when reading an instruction from guest RAM (when BadInstr/BadInstrP
     registers unavailable) which needs to be emulated, i.e. reserved
     instruction (exccode = RI), cop0 unusuable (exccode = CPU), MMIO
     load/store (TLBL/TLBS/MOD/ADEL/ADES), branch delay slot handling
     for MMIO load/store, or debug output for an unhandled exccode
     (exccode = ?)

   - kvm_mips_trans_replace() (write_fault = true)
     to write a replacement instruction into guest memory (mainly
     exccode = CPU)

   - kvm_mips_guest_cache_op() (write_fault = false)
     cache instruction emulation (exccode = CPU)

So there is a many:many mapping from exccode to write_fault for these
exccodes:

 - CPU (CoProcessor Unusable)
   could be reading instruction or servicing a CACHE instruction
   (write_fault = false) or replacing an instruction (write_fault =
   true).

 - MOD, TLBS, ADES
   could be the write itself (write_fault = true), or a read of the
   instruction triggering the exception or the prior branch instruction
   (write_fault = false).

Cheers
James

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 3196 bytes --]

Hi,

On Tue, Jan 17, 2017 at 07:06:39PM +0800, jiang.biao2@zte.com.cn wrote:
> ＞ kvm_mips_map_page() will need to know whether the fault was due to a
> ＞ read or a write in order to support dirty page tracking,
> ＞ KVM_CAP_SYNC_MMU, and read only memory regions, so get that information
> ＞ passed down to it via new bool write_fault arguments to various
> ＞ functions.
> 
> Hi, James,
> 
> 
> Maybe it's not good idea to add an argument and pass it down to various functions,  it will ,
> 
> make the the interface more complicated,  have more possibolity, and more difficult to test.
> 

kvm_mips_map_page() needs to know is whether it is being triggered in
response to a write, but it is abstracted and independent from what host
exception triggered it. It is only really concerned with the GPA space.

> 
> 
> 
> As far as I can see,  what you need is to let kvm_mips_map_page() know the *reason*.
> 
> would it be better to let kvm_mips_map_page() get *exccode* from the CP0 directly, or 
> 
> 
> provide any interface to get the fault reason?

I presume you mean from the saved host cause register in the VCPU
context (since intervening exceptions/interrupts will clobber the actual
CP0 Cause register).

It directly needs to know whether it can get away with a read-only
mapping, and although it directly depends on a GVA segment, it doesn't
necessarily relate to a memory access made by the guest.

kvm_mips_map_page() is called via:

 - kvm_mips_handle_kseg0_tlb_fault()
   for faults in guest KSeg0

 - kvm_mips_handle_mapped_seg_tlb_fault()
   for faults in guest TLB mapped segments

From these functions:

 - kvm_trap_emul_handle_tlb_mod() (write_fault = true)
   in response to a write to a read-only page (exccode = MOD)

 - kvm_trap_emul_handle_tlb_miss() (write_fault = true or false)
   in response to a read or write when TLB mapping absent or invalid
   (exccode = TLBL/TLBS)

 - via the kvm_trap_emul_gva_fault() helper when KVM needs to directly
   access GVA space:

   - kvm_get_inst() (write_fault = false)
     when reading an instruction from guest RAM (when BadInstr/BadInstrP
     registers unavailable) which needs to be emulated, i.e. reserved
     instruction (exccode = RI), cop0 unusuable (exccode = CPU), MMIO
     load/store (TLBL/TLBS/MOD/ADEL/ADES), branch delay slot handling
     for MMIO load/store, or debug output for an unhandled exccode
     (exccode = ?)

   - kvm_mips_trans_replace() (write_fault = true)
     to write a replacement instruction into guest memory (mainly
     exccode = CPU)

   - kvm_mips_guest_cache_op() (write_fault = false)
     cache instruction emulation (exccode = CPU)

So there is a many:many mapping from exccode to write_fault for these
exccodes:

 - CPU (CoProcessor Unusable)
   could be reading instruction or servicing a CACHE instruction
   (write_fault = false) or replacing an instruction (write_fault =
   true).

 - MOD, TLBS, ADES
   could be the write itself (write_fault = true), or a read of the
   instruction triggering the exception or the prior branch instruction
   (write_fault = false).

Cheers
James

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 2511 bytes --]

Hi,

On Wed, Jan 18, 2017 at 04:18:46PM +0800, jiang.biao2@zte.com.cn wrote:
> Hi,
> 
> ＞ I presume you mean from the saved host cause register in the VCPU
> ＞ context (since intervening exceptions/interrupts will clobber the actual
> ＞ CP0 Cause register).
> ＞ 
> ＞ It directly needs to know whether it can get away with a read-only
> ＞ mapping, and although it directly depends on a GVA segment, it doesn't
> ＞ necessarily relate to a memory access made by the guest.
> ＞ 
> ＞ kvm_mips_map_page() is called via:
> ＞ 
> ＞ - kvm_mips_handle_kseg0_tlb_fault()
> ＞   for faults in guest KSeg0
> ＞ 
> ＞  - kvm_mips_handle_mapped_seg_tlb_fault()
> ＞    for faults in guest TLB mapped segments
> ＞ 
> ＞ From these functions:
> ＞ 
> ＞  - kvm_trap_emul_handle_tlb_mod() (write_fault = true)
> ＞   in response to a write to a read-only page (exccode = MOD)
> ＞ 
> ＞ - kvm_trap_emul_handle_tlb_miss() (write_fault = true or false)
> ＞   in response to a read or write when TLB mapping absent or invalid
> ＞   (exccode = TLBL/TLBS)
> ＞
> ＞ 
> ＞ So there is a many:many mapping from exccode to write_fault for these
> ＞ exccodes:
> ＞ 
> ＞  - CPU (CoProcessor Unusable)
> ＞    could be reading instruction or servicing a CACHE instruction
> ＞    (write_fault = false) or replacing an instruction (write_fault =
> ＞    true).
> 
> ＞  - MOD, TLBS, ADES
> ＞   could be the write itself (write_fault = true), or a read of the
> ＞   instruction triggering the exception or the prior branch instruction
> ＞   (write_fault = false).
> ＞ 
> Thanks for the detail, it is more complicted than I thought.
> 
> But there may be still bad smell from the long parameters, espacially from 
> 
> 
> bool type ones.  

Whats wrong with bool parameters?

It needs a GPA mapping created, either for a read or a write depending
on the caller. bool would seem ideally suited for just such a situation,
and in fact its exactly what the KVM GPA fault code path does to pass
whether the page needs to be writable:

kvm_mips_map_page() -> gfn_to_pfn_prot() -> __gfn_to_pfn_memslot() ->
hva_to_pfn() -> hva_to_pfn_slow().

so all this really does is extend that pattern up the other way as
necessary to be able to provide that information to gfn_to_pfn_prot().

Cheers
James

> 
> 
> Maybe there is better way to handle that, but I can not figure it out right now 
> 
> 
> because of the complexity.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 2511 bytes --]

Hi,

On Wed, Jan 18, 2017 at 04:18:46PM +0800, jiang.biao2@zte.com.cn wrote:
> Hi,
> 
> ＞ I presume you mean from the saved host cause register in the VCPU
> ＞ context (since intervening exceptions/interrupts will clobber the actual
> ＞ CP0 Cause register).
> ＞ 
> ＞ It directly needs to know whether it can get away with a read-only
> ＞ mapping, and although it directly depends on a GVA segment, it doesn't
> ＞ necessarily relate to a memory access made by the guest.
> ＞ 
> ＞ kvm_mips_map_page() is called via:
> ＞ 
> ＞ - kvm_mips_handle_kseg0_tlb_fault()
> ＞   for faults in guest KSeg0
> ＞ 
> ＞  - kvm_mips_handle_mapped_seg_tlb_fault()
> ＞    for faults in guest TLB mapped segments
> ＞ 
> ＞ From these functions:
> ＞ 
> ＞  - kvm_trap_emul_handle_tlb_mod() (write_fault = true)
> ＞   in response to a write to a read-only page (exccode = MOD)
> ＞ 
> ＞ - kvm_trap_emul_handle_tlb_miss() (write_fault = true or false)
> ＞   in response to a read or write when TLB mapping absent or invalid
> ＞   (exccode = TLBL/TLBS)
> ＞
> ＞ 
> ＞ So there is a many:many mapping from exccode to write_fault for these
> ＞ exccodes:
> ＞ 
> ＞  - CPU (CoProcessor Unusable)
> ＞    could be reading instruction or servicing a CACHE instruction
> ＞    (write_fault = false) or replacing an instruction (write_fault =
> ＞    true).
> 
> ＞  - MOD, TLBS, ADES
> ＞   could be the write itself (write_fault = true), or a read of the
> ＞   instruction triggering the exception or the prior branch instruction
> ＞   (write_fault = false).
> ＞ 
> Thanks for the detail, it is more complicted than I thought.
> 
> But there may be still bad smell from the long parameters, espacially from 
> 
> 
> bool type ones.  

Whats wrong with bool parameters?

It needs a GPA mapping created, either for a read or a write depending
on the caller. bool would seem ideally suited for just such a situation,
and in fact its exactly what the KVM GPA fault code path does to pass
whether the page needs to be writable:

kvm_mips_map_page() -> gfn_to_pfn_prot() -> __gfn_to_pfn_memslot() ->
hva_to_pfn() -> hva_to_pfn_slow().

so all this really does is extend that pattern up the other way as
necessary to be able to provide that information to gfn_to_pfn_prot().

Cheers
James

> 
> 
> Maybe there is better way to handle that, but I can not figure it out right now 
> 
> 
> because of the complexity.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 1162 bytes --]

Hi,

On Thu, Jan 19, 2017 at 04:29:51PM +0800, jiang.biao2@zte.com.cn wrote:
> Hi, James
> 
> 
> 
> 
> 
> 
> 
> ＞ Whats wrong with bool parameters?
> ＞ 
> ＞ It needs a GPA mapping created, either for a read or a write depending
> ＞ on the caller. bool would seem ideally suited for just such a situation,
> ＞ and in fact its exactly what the KVM GPA fault code path does to pass
> ＞ whether the page needs to be writable:
> ＞ 
> ＞ kvm_mips_map_page() -＞ gfn_to_pfn_prot() -＞ __gfn_to_pfn_memslot() -＞
> ＞ hva_to_pfn() -＞ hva_to_pfn_slow().
> ＞ 
> ＞ so all this really does is extend that pattern up the other way as
> ＞ necessary to be able to provide that information to gfn_to_pfn_prot().
> Bool parameter may make the code less readable.  :-)
> 
> 
> The way used is indeed consistent with the exist pattern, but the tramp data 
> 
> 
> passed around and long parameters list maybe code smell(not for sure for 
> 
> 
> the kernel :-) ),  which may be improved by some means.
> 
> No offense,  just personal opinion. :-)

No offense taken :-)

Thanks again for reviewing,

Cheers
James

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH 2/13] KVM: MIPS: Pass type of fault down to kvm_mips_map_page()

[James Hogan]

[-- Attachment #1: Type: text/plain, Size: 1162 bytes --]

Hi,

On Thu, Jan 19, 2017 at 04:29:51PM +0800, jiang.biao2@zte.com.cn wrote:
> Hi, James
> 
> 
> 
> 
> 
> 
> 
> ＞ Whats wrong with bool parameters?
> ＞ 
> ＞ It needs a GPA mapping created, either for a read or a write depending
> ＞ on the caller. bool would seem ideally suited for just such a situation,
> ＞ and in fact its exactly what the KVM GPA fault code path does to pass
> ＞ whether the page needs to be writable:
> ＞ 
> ＞ kvm_mips_map_page() -＞ gfn_to_pfn_prot() -＞ __gfn_to_pfn_memslot() -＞
> ＞ hva_to_pfn() -＞ hva_to_pfn_slow().
> ＞ 
> ＞ so all this really does is extend that pattern up the other way as
> ＞ necessary to be able to provide that information to gfn_to_pfn_prot().
> Bool parameter may make the code less readable.  :-)
> 
> 
> The way used is indeed consistent with the exist pattern, but the tramp data 
> 
> 
> passed around and long parameters list maybe code smell(not for sure for 
> 
> 
> the kernel :-) ),  which may be improved by some means.
> 
> No offense,  just personal opinion. :-)

No offense taken :-)

Thanks again for reviewing,

Cheers
James

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]