* [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing
@ 2017-01-09 20:51 James Hogan
2017-01-09 20:51 ` James Hogan
` (10 more replies)
0 siblings, 11 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Note: My intention is to take this series via the MIPS KVM tree along
with the others for v4.11, with a topic branch containing the MIPS
architecture change, so an ack is welcome for patch 1 in particular.
This series first converts MIPS KVM to use page tables for its GPA ->
HPA mappings instead of a linear array. The linear array was only really
meant to be temporary, and isn't sparse so its wasteful of memory. It
also never handled resizing of the array for multiple or changed memory
regions, which a sparse page table pretty much handles automatically.
We then go on to implement the shadow flushing architecture callbacks to
allow the mappings (page tables and TLB entries) to be flushed in
response to memory region changes. This is fairly straightforward for
GPA which is shared between VCPUs as the kvm->mmu_lock can protect it,
but GVA page tables are specific to a VCPU so are accessed locklessly.
This would make it unsafe to directly modify any GVA page tables, so we
wire up the TLB flush VCPU request so that we can tell a possibly
running VCPU to flush its own GVA mappings.
Since MIPS KVM emulation code can also access GVA mappings directly, we
use the READING_SHADOW_PAGE_TABLES VCPU mode similar to how x86 does to
locklessly protect these accesses. This ensures that either the flush
will take place before the GVA access, or an IPI will be sent to confirm
receipt of the request which will be delayed until after the GVA access
is complete.
The patches are roughly grouped as follows:
Patch 1:
This is a MIPS architecture change needed for patch 9. As I mentioned
above I intend to combine this into a topic branch which can be merged
into both the MIPS architecture tree and the MIPS KVM tree.
Patch 2:
This singularly converts GPA to use page tables.
Patches 3-10:
These implement shadow flushing, first laying the ground work to allow
TLB flush requests to work and to protect direct GVA access from
asynchronous flush requests.
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 (10):
MIPS: Add return errors to protected cache ops
KVM: MIPS/MMU: Convert guest physical map to page table
KVM: MIPS: Update vcpu->mode and vcpu->cpu
KVM: MIPS/T&E: Handle TLB invalidation requests
KVM: MIPS/T&E: Reduce stale ASID checks
KVM: MIPS/T&E: Add lockless GVA access helpers
KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans
KVM: MIPS/MMU: Use lockless GVA helpers for get_inst()
KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation
KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot
arch/mips/include/asm/kvm_host.h | 35 ++-
arch/mips/include/asm/r4kcache.h | 55 +++--
arch/mips/kvm/dyntrans.c | 26 +-
arch/mips/kvm/emulate.c | 149 +++++--------
arch/mips/kvm/mips.c | 92 +++++---
arch/mips/kvm/mmu.c | 367 ++++++++++++++++++++++++++++----
arch/mips/kvm/tlb.c | 35 +---
arch/mips/kvm/trap_emul.c | 185 ++++++++++++----
8 files changed, 690 insertions(+), 254 deletions(-)
--
git-series 0.8.10
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 1/10] MIPS: Add return errors to protected cache ops James Hogan
` (9 subsequent siblings)
10 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Note: My intention is to take this series via the MIPS KVM tree along
with the others for v4.11, with a topic branch containing the MIPS
architecture change, so an ack is welcome for patch 1 in particular.
This series first converts MIPS KVM to use page tables for its GPA ->
HPA mappings instead of a linear array. The linear array was only really
meant to be temporary, and isn't sparse so its wasteful of memory. It
also never handled resizing of the array for multiple or changed memory
regions, which a sparse page table pretty much handles automatically.
We then go on to implement the shadow flushing architecture callbacks to
allow the mappings (page tables and TLB entries) to be flushed in
response to memory region changes. This is fairly straightforward for
GPA which is shared between VCPUs as the kvm->mmu_lock can protect it,
but GVA page tables are specific to a VCPU so are accessed locklessly.
This would make it unsafe to directly modify any GVA page tables, so we
wire up the TLB flush VCPU request so that we can tell a possibly
running VCPU to flush its own GVA mappings.
Since MIPS KVM emulation code can also access GVA mappings directly, we
use the READING_SHADOW_PAGE_TABLES VCPU mode similar to how x86 does to
locklessly protect these accesses. This ensures that either the flush
will take place before the GVA access, or an IPI will be sent to confirm
receipt of the request which will be delayed until after the GVA access
is complete.
The patches are roughly grouped as follows:
Patch 1:
This is a MIPS architecture change needed for patch 9. As I mentioned
above I intend to combine this into a topic branch which can be merged
into both the MIPS architecture tree and the MIPS KVM tree.
Patch 2:
This singularly converts GPA to use page tables.
Patches 3-10:
These implement shadow flushing, first laying the ground work to allow
TLB flush requests to work and to protect direct GVA access from
asynchronous flush requests.
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 (10):
MIPS: Add return errors to protected cache ops
KVM: MIPS/MMU: Convert guest physical map to page table
KVM: MIPS: Update vcpu->mode and vcpu->cpu
KVM: MIPS/T&E: Handle TLB invalidation requests
KVM: MIPS/T&E: Reduce stale ASID checks
KVM: MIPS/T&E: Add lockless GVA access helpers
KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans
KVM: MIPS/MMU: Use lockless GVA helpers for get_inst()
KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation
KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot
arch/mips/include/asm/kvm_host.h | 35 ++-
arch/mips/include/asm/r4kcache.h | 55 +++--
arch/mips/kvm/dyntrans.c | 26 +-
arch/mips/kvm/emulate.c | 149 +++++--------
arch/mips/kvm/mips.c | 92 +++++---
arch/mips/kvm/mmu.c | 367 ++++++++++++++++++++++++++++----
arch/mips/kvm/tlb.c | 35 +---
arch/mips/kvm/trap_emul.c | 185 ++++++++++++----
8 files changed, 690 insertions(+), 254 deletions(-)
--
git-series 0.8.10
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH 1/10] MIPS: Add return errors to protected cache ops
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
2017-01-09 20:51 ` James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-17 16:05 ` Ralf Baechle
2017-01-09 20:51 ` [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table James Hogan
` (8 subsequent siblings)
10 siblings, 2 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Ralf Baechle, Paolo Bonzini,
Radim Krčmář,
kvm
The protected cache ops contain no out of line fixup code to return an
error code in the event of a fault, with the cache op being skipped in
that case. For KVM however we'd like to detect this case as page
faulting will be disabled so it could happen during normal operation if
the GVA page tables were flushed, and need to be handled by the caller.
Add the out-of-line fixup code to load the error value -EFAULT into the
return variable, and adapt the protected cache line functions to pass
the error back to the caller.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
arch/mips/include/asm/r4kcache.h | 55 +++++++++++++++++++++------------
1 file changed, 35 insertions(+), 20 deletions(-)
diff --git a/arch/mips/include/asm/r4kcache.h b/arch/mips/include/asm/r4kcache.h
index b42b513007a2..7227c158cbf8 100644
--- a/arch/mips/include/asm/r4kcache.h
+++ b/arch/mips/include/asm/r4kcache.h
@@ -147,49 +147,64 @@ static inline void flush_scache_line(unsigned long addr)
}
#define protected_cache_op(op,addr) \
+({ \
+ int __err = 0; \
__asm__ __volatile__( \
" .set push \n" \
" .set noreorder \n" \
" .set "MIPS_ISA_ARCH_LEVEL" \n" \
- "1: cache %0, (%1) \n" \
+ "1: cache %1, (%2) \n" \
"2: .set pop \n" \
+ " .section .fixup,\"ax\" \n" \
+ "3: li %0, %3 \n" \
+ " j 2b \n" \
+ " .previous \n" \
" .section __ex_table,\"a\" \n" \
- " "STR(PTR)" 1b, 2b \n" \
+ " "STR(PTR)" 1b, 3b \n" \
" .previous" \
- : \
- : "i" (op), "r" (addr))
+ : "+r" (__err) \
+ : "i" (op), "r" (addr), "i" (-EFAULT)); \
+ __err; \
+})
+
#define protected_cachee_op(op,addr) \
+({ \
+ int __err = 0; \
__asm__ __volatile__( \
" .set push \n" \
" .set noreorder \n" \
" .set mips0 \n" \
" .set eva \n" \
- "1: cachee %0, (%1) \n" \
+ "1: cachee %1, (%2) \n" \
"2: .set pop \n" \
+ " .section .fixup,\"ax\" \n" \
+ "3: li %0, %3 \n" \
+ " j 2b \n" \
+ " .previous \n" \
" .section __ex_table,\"a\" \n" \
- " "STR(PTR)" 1b, 2b \n" \
+ " "STR(PTR)" 1b, 3b \n" \
" .previous" \
- : \
- : "i" (op), "r" (addr))
+ : "+r" (__err) \
+ : "i" (op), "r" (addr), "i" (-EFAULT)); \
+ __err; \
+})
/*
* The next two are for badland addresses like signal trampolines.
*/
-static inline void protected_flush_icache_line(unsigned long addr)
+static inline int protected_flush_icache_line(unsigned long addr)
{
switch (boot_cpu_type()) {
case CPU_LOONGSON2:
- protected_cache_op(Hit_Invalidate_I_Loongson2, addr);
- break;
+ return protected_cache_op(Hit_Invalidate_I_Loongson2, addr);
default:
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Invalidate_I, addr);
+ return protected_cachee_op(Hit_Invalidate_I, addr);
#else
- protected_cache_op(Hit_Invalidate_I, addr);
+ return protected_cache_op(Hit_Invalidate_I, addr);
#endif
- break;
}
}
@@ -199,21 +214,21 @@ static inline void protected_flush_icache_line(unsigned long addr)
* caches. We're talking about one cacheline unnecessarily getting invalidated
* here so the penalty isn't overly hard.
*/
-static inline void protected_writeback_dcache_line(unsigned long addr)
+static inline int protected_writeback_dcache_line(unsigned long addr)
{
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Writeback_Inv_D, addr);
+ return protected_cachee_op(Hit_Writeback_Inv_D, addr);
#else
- protected_cache_op(Hit_Writeback_Inv_D, addr);
+ return protected_cache_op(Hit_Writeback_Inv_D, addr);
#endif
}
-static inline void protected_writeback_scache_line(unsigned long addr)
+static inline int protected_writeback_scache_line(unsigned long addr)
{
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Writeback_Inv_SD, addr);
+ return protected_cachee_op(Hit_Writeback_Inv_SD, addr);
#else
- protected_cache_op(Hit_Writeback_Inv_SD, addr);
+ return protected_cache_op(Hit_Writeback_Inv_SD, addr);
#endif
}
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 1/10] MIPS: Add return errors to protected cache ops
2017-01-09 20:51 ` [PATCH 1/10] MIPS: Add return errors to protected cache ops James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-17 16:05 ` Ralf Baechle
1 sibling, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Ralf Baechle, Paolo Bonzini,
Radim Krčmář,
kvm
The protected cache ops contain no out of line fixup code to return an
error code in the event of a fault, with the cache op being skipped in
that case. For KVM however we'd like to detect this case as page
faulting will be disabled so it could happen during normal operation if
the GVA page tables were flushed, and need to be handled by the caller.
Add the out-of-line fixup code to load the error value -EFAULT into the
return variable, and adapt the protected cache line functions to pass
the error back to the caller.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
---
arch/mips/include/asm/r4kcache.h | 55 +++++++++++++++++++++------------
1 file changed, 35 insertions(+), 20 deletions(-)
diff --git a/arch/mips/include/asm/r4kcache.h b/arch/mips/include/asm/r4kcache.h
index b42b513007a2..7227c158cbf8 100644
--- a/arch/mips/include/asm/r4kcache.h
+++ b/arch/mips/include/asm/r4kcache.h
@@ -147,49 +147,64 @@ static inline void flush_scache_line(unsigned long addr)
}
#define protected_cache_op(op,addr) \
+({ \
+ int __err = 0; \
__asm__ __volatile__( \
" .set push \n" \
" .set noreorder \n" \
" .set "MIPS_ISA_ARCH_LEVEL" \n" \
- "1: cache %0, (%1) \n" \
+ "1: cache %1, (%2) \n" \
"2: .set pop \n" \
+ " .section .fixup,\"ax\" \n" \
+ "3: li %0, %3 \n" \
+ " j 2b \n" \
+ " .previous \n" \
" .section __ex_table,\"a\" \n" \
- " "STR(PTR)" 1b, 2b \n" \
+ " "STR(PTR)" 1b, 3b \n" \
" .previous" \
- : \
- : "i" (op), "r" (addr))
+ : "+r" (__err) \
+ : "i" (op), "r" (addr), "i" (-EFAULT)); \
+ __err; \
+})
+
#define protected_cachee_op(op,addr) \
+({ \
+ int __err = 0; \
__asm__ __volatile__( \
" .set push \n" \
" .set noreorder \n" \
" .set mips0 \n" \
" .set eva \n" \
- "1: cachee %0, (%1) \n" \
+ "1: cachee %1, (%2) \n" \
"2: .set pop \n" \
+ " .section .fixup,\"ax\" \n" \
+ "3: li %0, %3 \n" \
+ " j 2b \n" \
+ " .previous \n" \
" .section __ex_table,\"a\" \n" \
- " "STR(PTR)" 1b, 2b \n" \
+ " "STR(PTR)" 1b, 3b \n" \
" .previous" \
- : \
- : "i" (op), "r" (addr))
+ : "+r" (__err) \
+ : "i" (op), "r" (addr), "i" (-EFAULT)); \
+ __err; \
+})
/*
* The next two are for badland addresses like signal trampolines.
*/
-static inline void protected_flush_icache_line(unsigned long addr)
+static inline int protected_flush_icache_line(unsigned long addr)
{
switch (boot_cpu_type()) {
case CPU_LOONGSON2:
- protected_cache_op(Hit_Invalidate_I_Loongson2, addr);
- break;
+ return protected_cache_op(Hit_Invalidate_I_Loongson2, addr);
default:
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Invalidate_I, addr);
+ return protected_cachee_op(Hit_Invalidate_I, addr);
#else
- protected_cache_op(Hit_Invalidate_I, addr);
+ return protected_cache_op(Hit_Invalidate_I, addr);
#endif
- break;
}
}
@@ -199,21 +214,21 @@ static inline void protected_flush_icache_line(unsigned long addr)
* caches. We're talking about one cacheline unnecessarily getting invalidated
* here so the penalty isn't overly hard.
*/
-static inline void protected_writeback_dcache_line(unsigned long addr)
+static inline int protected_writeback_dcache_line(unsigned long addr)
{
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Writeback_Inv_D, addr);
+ return protected_cachee_op(Hit_Writeback_Inv_D, addr);
#else
- protected_cache_op(Hit_Writeback_Inv_D, addr);
+ return protected_cache_op(Hit_Writeback_Inv_D, addr);
#endif
}
-static inline void protected_writeback_scache_line(unsigned long addr)
+static inline int protected_writeback_scache_line(unsigned long addr)
{
#ifdef CONFIG_EVA
- protected_cachee_op(Hit_Writeback_Inv_SD, addr);
+ return protected_cachee_op(Hit_Writeback_Inv_SD, addr);
#else
- protected_cache_op(Hit_Writeback_Inv_SD, addr);
+ return protected_cache_op(Hit_Writeback_Inv_SD, addr);
#endif
}
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 1/10] MIPS: Add return errors to protected cache ops James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu James Hogan
` (7 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Current guest physical memory is mapped to host physical addresses using
a single linear array (guest_pmap of length guest_pmap_npages). This was
only really meant to be temporary, and isn't sparse, so its wasteful of
memory. A small amount of RAM at GPA 0 and a small boot exception vector
at GPA 0x1fc00000 cannot be represented without a full 128KiB guest_pmap
allocation (MIPS32 with 16KiB pages), which is one reason why QEMU
currently runs its boot code at the top of RAM instead of the usual boot
exception vector address.
Instead use the existing infrastructure for host virtual page table
management to allocate a page table for guest physical memory too. This
should be sufficient for now, assuming the size of physical memory
doesn't exceed the size of virtual memory. It may need extending in
future to handle XPA (eXtended Physical Addressing) in 32-bit guests, as
supported by VZ guests on P5600.
Some of this code is based loosely on Cavium's VZ KVM implementation.
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 | 7 +-
arch/mips/kvm/mips.c | 48 +----
arch/mips/kvm/mmu.c | 293 ++++++++++++++++++++++++++++----
3 files changed, 278 insertions(+), 70 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index ac9d41900597..b2d968ed2757 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -153,9 +153,8 @@ struct kvm_arch_memory_slot {
};
struct kvm_arch {
- /* Guest GVA->HPA page table */
- unsigned long *guest_pmap;
- unsigned long guest_pmap_npages;
+ /* Guest physical mm */
+ struct mm_struct gpa_mm;
};
#define N_MIPS_COPROC_REGS 32
@@ -633,6 +632,8 @@ enum kvm_mips_flush {
KMF_GPA = 0x2,
};
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);
+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,
bool user);
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 4dc6cf2b7c50..90c28fbf4829 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -22,6 +22,7 @@
#include <asm/page.h>
#include <asm/cacheflush.h>
#include <asm/mmu_context.h>
+#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <linux/kvm_host.h>
@@ -94,6 +95,11 @@ void kvm_arch_check_processor_compat(void *rtn)
int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
{
+ /* Allocate page table to map GPA -> RPA */
+ kvm->arch.gpa_mm.pgd = kvm_pgd_alloc();
+ if (!kvm->arch.gpa_mm.pgd)
+ return -ENOMEM;
+
return 0;
}
@@ -112,13 +118,6 @@ void kvm_mips_free_vcpus(struct kvm *kvm)
unsigned int i;
struct kvm_vcpu *vcpu;
- /* Put the pages we reserved for the guest pmap */
- for (i = 0; i < kvm->arch.guest_pmap_npages; i++) {
- if (kvm->arch.guest_pmap[i] != KVM_INVALID_PAGE)
- kvm_release_pfn_clean(kvm->arch.guest_pmap[i]);
- }
- kfree(kvm->arch.guest_pmap);
-
kvm_for_each_vcpu(i, vcpu, kvm) {
kvm_arch_vcpu_free(vcpu);
}
@@ -133,9 +132,17 @@ void kvm_mips_free_vcpus(struct kvm *kvm)
mutex_unlock(&kvm->lock);
}
+static void kvm_mips_free_gpa_pt(struct kvm *kvm)
+{
+ /* It should always be safe to remove after flushing the whole range */
+ WARN_ON(!kvm_mips_flush_gpa_pt(kvm, 0, ~0));
+ pgd_free(NULL, kvm->arch.gpa_mm.pgd);
+}
+
void kvm_arch_destroy_vm(struct kvm *kvm)
{
kvm_mips_free_vcpus(kvm);
+ kvm_mips_free_gpa_pt(kvm);
}
long kvm_arch_dev_ioctl(struct file *filp, unsigned int ioctl,
@@ -164,36 +171,9 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
const struct kvm_memory_slot *new,
enum kvm_mr_change change)
{
- unsigned long npages = 0;
- int i;
-
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);
-
- /* Setup Guest PMAP table */
- if (!kvm->arch.guest_pmap) {
- if (mem->slot == 0)
- npages = mem->memory_size >> PAGE_SHIFT;
-
- if (npages) {
- kvm->arch.guest_pmap_npages = npages;
- kvm->arch.guest_pmap =
- kzalloc(npages * sizeof(unsigned long), GFP_KERNEL);
-
- if (!kvm->arch.guest_pmap) {
- kvm_err("Failed to allocate guest PMAP\n");
- return;
- }
-
- kvm_debug("Allocated space for Guest PMAP Table (%ld pages) @ %p\n",
- npages, kvm->arch.guest_pmap);
-
- /* Now setup the page table */
- for (i = 0; i < npages; i++)
- kvm->arch.guest_pmap[i] = KVM_INVALID_PAGE;
- }
- }
}
static inline void dump_handler(const char *symbol, void *start, void *end)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 6379ac1bc7b9..09f5da706d9a 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -63,6 +63,63 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu)
}
/**
+ * kvm_pgd_init() - Initialise KVM GPA page directory.
+ * @page: Pointer to page directory (PGD) for KVM GPA.
+ *
+ * Initialise a KVM GPA page directory with pointers to the invalid table, i.e.
+ * representing no mappings. This is similar to pgd_init(), however it
+ * initialises all the page directory pointers, not just the ones corresponding
+ * to the userland address space (since it is for the guest physical address
+ * space rather than a virtual address space).
+ */
+static void kvm_pgd_init(void *page)
+{
+ unsigned long *p, *end;
+ unsigned long entry;
+
+#ifdef __PAGETABLE_PMD_FOLDED
+ entry = (unsigned long)invalid_pte_table;
+#else
+ entry = (unsigned long)invalid_pmd_table;
+#endif
+
+ p = (unsigned long *)page;
+ end = p + PTRS_PER_PGD;
+
+ do {
+ p[0] = entry;
+ p[1] = entry;
+ p[2] = entry;
+ p[3] = entry;
+ p[4] = entry;
+ p += 8;
+ p[-3] = entry;
+ p[-2] = entry;
+ p[-1] = entry;
+ } while (p != end);
+}
+
+/**
+ * kvm_pgd_alloc() - Allocate and initialise a KVM GPA page directory.
+ *
+ * Allocate a blank KVM GPA page directory (PGD) for representing guest physical
+ * to host physical page mappings.
+ *
+ * Returns: Pointer to new KVM GPA page directory.
+ * NULL on allocation failure.
+ */
+pgd_t *kvm_pgd_alloc(void)
+{
+ pgd_t *ret;
+
+ ret = (pgd_t *)__get_free_pages(GFP_KERNEL, PGD_ORDER);
+ if (ret)
+ kvm_pgd_init(ret);
+
+ return ret;
+}
+
+/**
* kvm_mips_walk_pgd() - Walk page table with optional allocation.
* @pgd: Page directory pointer.
* @addr: Address to index page table using.
@@ -112,15 +169,182 @@ static pte_t *kvm_mips_walk_pgd(pgd_t *pgd, struct kvm_mmu_memory_cache *cache,
return pte_offset(pmd, addr);
}
-static int kvm_mips_map_page(struct kvm *kvm, gfn_t gfn)
+/* Caller must hold kvm->mm_lock */
+static pte_t *kvm_mips_pte_for_gpa(struct kvm *kvm,
+ struct kvm_mmu_memory_cache *cache,
+ unsigned long addr)
{
- int srcu_idx, err = 0;
- kvm_pfn_t pfn;
+ return kvm_mips_walk_pgd(kvm->arch.gpa_mm.pgd, cache, addr);
+}
- if (kvm->arch.guest_pmap[gfn] != KVM_INVALID_PAGE)
- return 0;
+/*
+ * kvm_mips_flush_gpa_{pte,pmd,pud,pgd,pt}.
+ * Flush a range of guest physical address space from the VM's GPA page tables.
+ */
+
+static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ int i_min = __pte_offset(start_gpa);
+ int i_max = __pte_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PTE - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i) {
+ if (!pte_present(pte[i]))
+ continue;
+
+ kvm_release_pfn_clean(pte_pfn(pte[i]));
+ set_pte(pte + i, __pte(0));
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pmd(pmd_t *pmd, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pte_t *pte;
+ unsigned long end = ~0ul;
+ int i_min = __pmd_offset(start_gpa);
+ int i_max = __pmd_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PMD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pmd_present(pmd[i]))
+ continue;
+
+ pte = pte_offset(pmd + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pte(pte, start_gpa, end)) {
+ pmd_clear(pmd + i);
+ pte_free_kernel(NULL, pte);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pud(pud_t *pud, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pmd_t *pmd;
+ unsigned long end = ~0ul;
+ int i_min = __pud_offset(start_gpa);
+ int i_max = __pud_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PUD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pud_present(pud[i]))
+ continue;
+
+ pmd = pmd_offset(pud + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pmd(pmd, start_gpa, end)) {
+ pud_clear(pud + i);
+ pmd_free(NULL, pmd);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pgd(pgd_t *pgd, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pud_t *pud;
+ unsigned long end = ~0ul;
+ int i_min = pgd_index(start_gpa);
+ int i_max = pgd_index(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PGD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pgd_present(pgd[i]))
+ continue;
+
+ pud = pud_offset(pgd + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pud(pud, start_gpa, end)) {
+ pgd_clear(pgd + i);
+ pud_free(NULL, pud);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+/**
+ * kvm_mips_flush_gpa_pt() - Flush a range of guest physical addresses.
+ * @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.
+ *
+ * Flushes a range of GPA mappings from the GPA page tables.
+ *
+ * The caller must hold the @kvm->mmu_lock spinlock.
+ *
+ * Returns: Whether its safe to remove the top level page directory because
+ * all lower levels have been removed.
+ */
+bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
+{
+ return kvm_mips_flush_gpa_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.
+ * @gpa: Guest physical address of fault.
+ * @out_entry: New PTE for @gpa (written on success unless NULL).
+ * @out_buddy: New PTE for @gpa's buddy (written on success unless
+ * NULL).
+ *
+ * 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.
+ *
+ * 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
+ * execution.
+ * -EFAULT if there is no memory region at @gpa or a write was
+ * attempted to a read-only memory region. This is usually handled
+ * as an MMIO access.
+ */
+static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
+ pte_t *out_entry, pte_t *out_buddy)
+{
+ struct kvm *kvm = vcpu->kvm;
+ struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
+ gfn_t gfn = gpa >> PAGE_SHIFT;
+ int srcu_idx, err;
+ kvm_pfn_t pfn;
+ pte_t *ptep, entry, old_pte;
+ unsigned long prot_bits;
srcu_idx = srcu_read_lock(&kvm->srcu);
+
+ /* We need a minimum of cached pages ready for page table creation */
+ err = mmu_topup_memory_cache(memcache, KVM_MMU_CACHE_MIN_PAGES,
+ KVM_NR_MEM_OBJS);
+ if (err)
+ goto out;
+
pfn = gfn_to_pfn(kvm, gfn);
if (is_error_noslot_pfn(pfn)) {
@@ -129,7 +353,25 @@ static int kvm_mips_map_page(struct kvm *kvm, gfn_t gfn)
goto out;
}
- kvm->arch.guest_pmap[gfn] = pfn;
+ spin_lock(&kvm->mmu_lock);
+
+ ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
+
+ prot_bits = __READABLE | _PAGE_PRESENT | __WRITEABLE;
+ entry = pfn_pte(pfn, __pgprot(prot_bits));
+
+ 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)
+ *out_entry = *ptep;
+ if (out_buddy)
+ *out_buddy = *ptep_buddy(ptep);
+
+ spin_unlock(&kvm->mmu_lock);
out:
srcu_read_unlock(&kvm->srcu, srcu_idx);
return err;
@@ -318,11 +560,10 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
struct kvm_vcpu *vcpu)
{
- gfn_t gfn;
+ unsigned long gpa;
kvm_pfn_t pfn0, pfn1;
- unsigned long vaddr = 0;
- struct kvm *kvm = vcpu->kvm;
- pte_t *ptep_gva;
+ unsigned long vaddr;
+ pte_t pte_gpa[2], *ptep_gva;
if (KVM_GUEST_KSEGX(badvaddr) != KVM_GUEST_KSEG0) {
kvm_err("%s: Invalid BadVaddr: %#lx\n", __func__, badvaddr);
@@ -332,23 +573,17 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
/* Find host PFNs */
- gfn = (KVM_GUEST_CPHYSADDR(badvaddr) >> PAGE_SHIFT);
- if ((gfn | 1) >= kvm->arch.guest_pmap_npages) {
- kvm_err("%s: Invalid gfn: %#llx, BadVaddr: %#lx\n", __func__,
- gfn, badvaddr);
- kvm_mips_dump_host_tlbs();
- return -1;
- }
+ gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
vaddr = badvaddr & (PAGE_MASK << 1);
- if (kvm_mips_map_page(vcpu->kvm, gfn) < 0)
+ if (kvm_mips_map_page(vcpu, gpa, &pte_gpa[0], NULL) < 0)
return -1;
- if (kvm_mips_map_page(vcpu->kvm, gfn ^ 0x1) < 0)
+ if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, &pte_gpa[1], NULL) < 0)
return -1;
- pfn0 = kvm->arch.guest_pmap[gfn & ~0x1];
- pfn1 = kvm->arch.guest_pmap[gfn | 0x1];
+ pfn0 = pte_pfn(pte_gpa[0]);
+ pfn1 = pte_pfn(pte_gpa[1]);
/* Find GVA page table entry */
@@ -371,11 +606,9 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
struct kvm_mips_tlb *tlb,
unsigned long gva)
{
- struct kvm *kvm = vcpu->kvm;
kvm_pfn_t pfn;
- gfn_t gfn;
long tlb_lo = 0;
- pte_t *ptep_gva;
+ pte_t pte_gpa, *ptep_gva;
unsigned int idx;
bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
@@ -388,16 +621,10 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
tlb_lo = tlb->tlb_lo[idx];
/* Find host PFN */
- gfn = mips3_tlbpfn_to_paddr(tlb_lo) >> PAGE_SHIFT;
- if (gfn >= kvm->arch.guest_pmap_npages) {
- kvm_err("%s: Invalid gfn: %#llx, EHi: %#lx\n",
- __func__, gfn, tlb->tlb_hi);
- kvm_mips_dump_guest_tlbs(vcpu);
- return -1;
- }
- if (kvm_mips_map_page(kvm, gfn) < 0)
+ if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), &pte_gpa,
+ NULL) < 0)
return -1;
- pfn = kvm->arch.guest_pmap[gfn];
+ pfn = pte_pfn(pte_gpa);
/* Find GVA page table entry */
ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table
2017-01-09 20:51 ` [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Current guest physical memory is mapped to host physical addresses using
a single linear array (guest_pmap of length guest_pmap_npages). This was
only really meant to be temporary, and isn't sparse, so its wasteful of
memory. A small amount of RAM at GPA 0 and a small boot exception vector
at GPA 0x1fc00000 cannot be represented without a full 128KiB guest_pmap
allocation (MIPS32 with 16KiB pages), which is one reason why QEMU
currently runs its boot code at the top of RAM instead of the usual boot
exception vector address.
Instead use the existing infrastructure for host virtual page table
management to allocate a page table for guest physical memory too. This
should be sufficient for now, assuming the size of physical memory
doesn't exceed the size of virtual memory. It may need extending in
future to handle XPA (eXtended Physical Addressing) in 32-bit guests, as
supported by VZ guests on P5600.
Some of this code is based loosely on Cavium's VZ KVM implementation.
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 | 7 +-
arch/mips/kvm/mips.c | 48 +----
arch/mips/kvm/mmu.c | 293 ++++++++++++++++++++++++++++----
3 files changed, 278 insertions(+), 70 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index ac9d41900597..b2d968ed2757 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -153,9 +153,8 @@ struct kvm_arch_memory_slot {
};
struct kvm_arch {
- /* Guest GVA->HPA page table */
- unsigned long *guest_pmap;
- unsigned long guest_pmap_npages;
+ /* Guest physical mm */
+ struct mm_struct gpa_mm;
};
#define N_MIPS_COPROC_REGS 32
@@ -633,6 +632,8 @@ enum kvm_mips_flush {
KMF_GPA = 0x2,
};
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);
+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,
bool user);
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 4dc6cf2b7c50..90c28fbf4829 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -22,6 +22,7 @@
#include <asm/page.h>
#include <asm/cacheflush.h>
#include <asm/mmu_context.h>
+#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <linux/kvm_host.h>
@@ -94,6 +95,11 @@ void kvm_arch_check_processor_compat(void *rtn)
int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
{
+ /* Allocate page table to map GPA -> RPA */
+ kvm->arch.gpa_mm.pgd = kvm_pgd_alloc();
+ if (!kvm->arch.gpa_mm.pgd)
+ return -ENOMEM;
+
return 0;
}
@@ -112,13 +118,6 @@ void kvm_mips_free_vcpus(struct kvm *kvm)
unsigned int i;
struct kvm_vcpu *vcpu;
- /* Put the pages we reserved for the guest pmap */
- for (i = 0; i < kvm->arch.guest_pmap_npages; i++) {
- if (kvm->arch.guest_pmap[i] != KVM_INVALID_PAGE)
- kvm_release_pfn_clean(kvm->arch.guest_pmap[i]);
- }
- kfree(kvm->arch.guest_pmap);
-
kvm_for_each_vcpu(i, vcpu, kvm) {
kvm_arch_vcpu_free(vcpu);
}
@@ -133,9 +132,17 @@ void kvm_mips_free_vcpus(struct kvm *kvm)
mutex_unlock(&kvm->lock);
}
+static void kvm_mips_free_gpa_pt(struct kvm *kvm)
+{
+ /* It should always be safe to remove after flushing the whole range */
+ WARN_ON(!kvm_mips_flush_gpa_pt(kvm, 0, ~0));
+ pgd_free(NULL, kvm->arch.gpa_mm.pgd);
+}
+
void kvm_arch_destroy_vm(struct kvm *kvm)
{
kvm_mips_free_vcpus(kvm);
+ kvm_mips_free_gpa_pt(kvm);
}
long kvm_arch_dev_ioctl(struct file *filp, unsigned int ioctl,
@@ -164,36 +171,9 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
const struct kvm_memory_slot *new,
enum kvm_mr_change change)
{
- unsigned long npages = 0;
- int i;
-
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);
-
- /* Setup Guest PMAP table */
- if (!kvm->arch.guest_pmap) {
- if (mem->slot == 0)
- npages = mem->memory_size >> PAGE_SHIFT;
-
- if (npages) {
- kvm->arch.guest_pmap_npages = npages;
- kvm->arch.guest_pmap =
- kzalloc(npages * sizeof(unsigned long), GFP_KERNEL);
-
- if (!kvm->arch.guest_pmap) {
- kvm_err("Failed to allocate guest PMAP\n");
- return;
- }
-
- kvm_debug("Allocated space for Guest PMAP Table (%ld pages) @ %p\n",
- npages, kvm->arch.guest_pmap);
-
- /* Now setup the page table */
- for (i = 0; i < npages; i++)
- kvm->arch.guest_pmap[i] = KVM_INVALID_PAGE;
- }
- }
}
static inline void dump_handler(const char *symbol, void *start, void *end)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 6379ac1bc7b9..09f5da706d9a 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -63,6 +63,63 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu)
}
/**
+ * kvm_pgd_init() - Initialise KVM GPA page directory.
+ * @page: Pointer to page directory (PGD) for KVM GPA.
+ *
+ * Initialise a KVM GPA page directory with pointers to the invalid table, i.e.
+ * representing no mappings. This is similar to pgd_init(), however it
+ * initialises all the page directory pointers, not just the ones corresponding
+ * to the userland address space (since it is for the guest physical address
+ * space rather than a virtual address space).
+ */
+static void kvm_pgd_init(void *page)
+{
+ unsigned long *p, *end;
+ unsigned long entry;
+
+#ifdef __PAGETABLE_PMD_FOLDED
+ entry = (unsigned long)invalid_pte_table;
+#else
+ entry = (unsigned long)invalid_pmd_table;
+#endif
+
+ p = (unsigned long *)page;
+ end = p + PTRS_PER_PGD;
+
+ do {
+ p[0] = entry;
+ p[1] = entry;
+ p[2] = entry;
+ p[3] = entry;
+ p[4] = entry;
+ p += 8;
+ p[-3] = entry;
+ p[-2] = entry;
+ p[-1] = entry;
+ } while (p != end);
+}
+
+/**
+ * kvm_pgd_alloc() - Allocate and initialise a KVM GPA page directory.
+ *
+ * Allocate a blank KVM GPA page directory (PGD) for representing guest physical
+ * to host physical page mappings.
+ *
+ * Returns: Pointer to new KVM GPA page directory.
+ * NULL on allocation failure.
+ */
+pgd_t *kvm_pgd_alloc(void)
+{
+ pgd_t *ret;
+
+ ret = (pgd_t *)__get_free_pages(GFP_KERNEL, PGD_ORDER);
+ if (ret)
+ kvm_pgd_init(ret);
+
+ return ret;
+}
+
+/**
* kvm_mips_walk_pgd() - Walk page table with optional allocation.
* @pgd: Page directory pointer.
* @addr: Address to index page table using.
@@ -112,15 +169,182 @@ static pte_t *kvm_mips_walk_pgd(pgd_t *pgd, struct kvm_mmu_memory_cache *cache,
return pte_offset(pmd, addr);
}
-static int kvm_mips_map_page(struct kvm *kvm, gfn_t gfn)
+/* Caller must hold kvm->mm_lock */
+static pte_t *kvm_mips_pte_for_gpa(struct kvm *kvm,
+ struct kvm_mmu_memory_cache *cache,
+ unsigned long addr)
{
- int srcu_idx, err = 0;
- kvm_pfn_t pfn;
+ return kvm_mips_walk_pgd(kvm->arch.gpa_mm.pgd, cache, addr);
+}
- if (kvm->arch.guest_pmap[gfn] != KVM_INVALID_PAGE)
- return 0;
+/*
+ * kvm_mips_flush_gpa_{pte,pmd,pud,pgd,pt}.
+ * Flush a range of guest physical address space from the VM's GPA page tables.
+ */
+
+static bool kvm_mips_flush_gpa_pte(pte_t *pte, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ int i_min = __pte_offset(start_gpa);
+ int i_max = __pte_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PTE - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i) {
+ if (!pte_present(pte[i]))
+ continue;
+
+ kvm_release_pfn_clean(pte_pfn(pte[i]));
+ set_pte(pte + i, __pte(0));
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pmd(pmd_t *pmd, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pte_t *pte;
+ unsigned long end = ~0ul;
+ int i_min = __pmd_offset(start_gpa);
+ int i_max = __pmd_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PMD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pmd_present(pmd[i]))
+ continue;
+
+ pte = pte_offset(pmd + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pte(pte, start_gpa, end)) {
+ pmd_clear(pmd + i);
+ pte_free_kernel(NULL, pte);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pud(pud_t *pud, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pmd_t *pmd;
+ unsigned long end = ~0ul;
+ int i_min = __pud_offset(start_gpa);
+ int i_max = __pud_offset(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PUD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pud_present(pud[i]))
+ continue;
+
+ pmd = pmd_offset(pud + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pmd(pmd, start_gpa, end)) {
+ pud_clear(pud + i);
+ pmd_free(NULL, pmd);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+static bool kvm_mips_flush_gpa_pgd(pgd_t *pgd, unsigned long start_gpa,
+ unsigned long end_gpa)
+{
+ pud_t *pud;
+ unsigned long end = ~0ul;
+ int i_min = pgd_index(start_gpa);
+ int i_max = pgd_index(end_gpa);
+ bool safe_to_remove = (i_min == 0 && i_max == PTRS_PER_PGD - 1);
+ int i;
+
+ for (i = i_min; i <= i_max; ++i, start_gpa = 0) {
+ if (!pgd_present(pgd[i]))
+ continue;
+
+ pud = pud_offset(pgd + i, 0);
+ if (i == i_max)
+ end = end_gpa;
+
+ if (kvm_mips_flush_gpa_pud(pud, start_gpa, end)) {
+ pgd_clear(pgd + i);
+ pud_free(NULL, pud);
+ } else {
+ safe_to_remove = false;
+ }
+ }
+ return safe_to_remove;
+}
+
+/**
+ * kvm_mips_flush_gpa_pt() - Flush a range of guest physical addresses.
+ * @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.
+ *
+ * Flushes a range of GPA mappings from the GPA page tables.
+ *
+ * The caller must hold the @kvm->mmu_lock spinlock.
+ *
+ * Returns: Whether its safe to remove the top level page directory because
+ * all lower levels have been removed.
+ */
+bool kvm_mips_flush_gpa_pt(struct kvm *kvm, gfn_t start_gfn, gfn_t end_gfn)
+{
+ return kvm_mips_flush_gpa_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.
+ * @gpa: Guest physical address of fault.
+ * @out_entry: New PTE for @gpa (written on success unless NULL).
+ * @out_buddy: New PTE for @gpa's buddy (written on success unless
+ * NULL).
+ *
+ * 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.
+ *
+ * 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
+ * execution.
+ * -EFAULT if there is no memory region at @gpa or a write was
+ * attempted to a read-only memory region. This is usually handled
+ * as an MMIO access.
+ */
+static int kvm_mips_map_page(struct kvm_vcpu *vcpu, unsigned long gpa,
+ pte_t *out_entry, pte_t *out_buddy)
+{
+ struct kvm *kvm = vcpu->kvm;
+ struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
+ gfn_t gfn = gpa >> PAGE_SHIFT;
+ int srcu_idx, err;
+ kvm_pfn_t pfn;
+ pte_t *ptep, entry, old_pte;
+ unsigned long prot_bits;
srcu_idx = srcu_read_lock(&kvm->srcu);
+
+ /* We need a minimum of cached pages ready for page table creation */
+ err = mmu_topup_memory_cache(memcache, KVM_MMU_CACHE_MIN_PAGES,
+ KVM_NR_MEM_OBJS);
+ if (err)
+ goto out;
+
pfn = gfn_to_pfn(kvm, gfn);
if (is_error_noslot_pfn(pfn)) {
@@ -129,7 +353,25 @@ static int kvm_mips_map_page(struct kvm *kvm, gfn_t gfn)
goto out;
}
- kvm->arch.guest_pmap[gfn] = pfn;
+ spin_lock(&kvm->mmu_lock);
+
+ ptep = kvm_mips_pte_for_gpa(kvm, memcache, gpa);
+
+ prot_bits = __READABLE | _PAGE_PRESENT | __WRITEABLE;
+ entry = pfn_pte(pfn, __pgprot(prot_bits));
+
+ 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)
+ *out_entry = *ptep;
+ if (out_buddy)
+ *out_buddy = *ptep_buddy(ptep);
+
+ spin_unlock(&kvm->mmu_lock);
out:
srcu_read_unlock(&kvm->srcu, srcu_idx);
return err;
@@ -318,11 +560,10 @@ void kvm_mips_flush_gva_pt(pgd_t *pgd, enum kvm_mips_flush flags)
int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
struct kvm_vcpu *vcpu)
{
- gfn_t gfn;
+ unsigned long gpa;
kvm_pfn_t pfn0, pfn1;
- unsigned long vaddr = 0;
- struct kvm *kvm = vcpu->kvm;
- pte_t *ptep_gva;
+ unsigned long vaddr;
+ pte_t pte_gpa[2], *ptep_gva;
if (KVM_GUEST_KSEGX(badvaddr) != KVM_GUEST_KSEG0) {
kvm_err("%s: Invalid BadVaddr: %#lx\n", __func__, badvaddr);
@@ -332,23 +573,17 @@ int kvm_mips_handle_kseg0_tlb_fault(unsigned long badvaddr,
/* Find host PFNs */
- gfn = (KVM_GUEST_CPHYSADDR(badvaddr) >> PAGE_SHIFT);
- if ((gfn | 1) >= kvm->arch.guest_pmap_npages) {
- kvm_err("%s: Invalid gfn: %#llx, BadVaddr: %#lx\n", __func__,
- gfn, badvaddr);
- kvm_mips_dump_host_tlbs();
- return -1;
- }
+ gpa = KVM_GUEST_CPHYSADDR(badvaddr & (PAGE_MASK << 1));
vaddr = badvaddr & (PAGE_MASK << 1);
- if (kvm_mips_map_page(vcpu->kvm, gfn) < 0)
+ if (kvm_mips_map_page(vcpu, gpa, &pte_gpa[0], NULL) < 0)
return -1;
- if (kvm_mips_map_page(vcpu->kvm, gfn ^ 0x1) < 0)
+ if (kvm_mips_map_page(vcpu, gpa | PAGE_SIZE, &pte_gpa[1], NULL) < 0)
return -1;
- pfn0 = kvm->arch.guest_pmap[gfn & ~0x1];
- pfn1 = kvm->arch.guest_pmap[gfn | 0x1];
+ pfn0 = pte_pfn(pte_gpa[0]);
+ pfn1 = pte_pfn(pte_gpa[1]);
/* Find GVA page table entry */
@@ -371,11 +606,9 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
struct kvm_mips_tlb *tlb,
unsigned long gva)
{
- struct kvm *kvm = vcpu->kvm;
kvm_pfn_t pfn;
- gfn_t gfn;
long tlb_lo = 0;
- pte_t *ptep_gva;
+ pte_t pte_gpa, *ptep_gva;
unsigned int idx;
bool kernel = KVM_GUEST_KERNEL_MODE(vcpu);
@@ -388,16 +621,10 @@ int kvm_mips_handle_mapped_seg_tlb_fault(struct kvm_vcpu *vcpu,
tlb_lo = tlb->tlb_lo[idx];
/* Find host PFN */
- gfn = mips3_tlbpfn_to_paddr(tlb_lo) >> PAGE_SHIFT;
- if (gfn >= kvm->arch.guest_pmap_npages) {
- kvm_err("%s: Invalid gfn: %#llx, EHi: %#lx\n",
- __func__, gfn, tlb->tlb_hi);
- kvm_mips_dump_guest_tlbs(vcpu);
- return -1;
- }
- if (kvm_mips_map_page(kvm, gfn) < 0)
+ if (kvm_mips_map_page(vcpu, mips3_tlbpfn_to_paddr(tlb_lo), &pte_gpa,
+ NULL) < 0)
return -1;
- pfn = kvm->arch.guest_pmap[gfn];
+ pfn = pte_pfn(pte_gpa);
/* Find GVA page table entry */
ptep_gva = kvm_trap_emul_pte_for_gva(vcpu, gva);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (2 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests James Hogan
` (6 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Keep the vcpu->mode and vcpu->cpu variables up to date so that
kvm_make_all_cpus_request() has a chance of functioning correctly. This
will soon need to be used for kvm_flush_remote_tlbs().
We can easily update vcpu->cpu when the VCPU context is loaded or saved,
which will happen when accessing guest context and when the guest is
scheduled in and out.
We need to be a little careful with vcpu->mode though, as we will in
future be checking for outstanding VCPU requests, and this must be done
after the value of IN_GUEST_MODE in vcpu->mode is visible to other CPUs.
Otherwise the other CPU could fail to trigger an IPI to wait for
completion dispite the VCPU request not being seen.
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 | 18 ++++++++++++++++++
arch/mips/kvm/mmu.c | 2 ++
2 files changed, 20 insertions(+), 0 deletions(-)
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 90c28fbf4829..325e98367b30 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -365,6 +365,14 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
guest_enter_irqoff();
trace_kvm_enter(vcpu);
+ /*
+ * Make sure the read of VCPU requests in vcpu_run() callback is not
+ * reordered ahead of the write to vcpu->mode, or we could miss a TLB
+ * flush request while the requester sees the VCPU as outside of guest
+ * mode and not needing an IPI.
+ */
+ smp_store_mb(vcpu->mode, IN_GUEST_MODE);
+
r = kvm_mips_callbacks->vcpu_run(run, vcpu);
trace_kvm_out(vcpu);
@@ -1331,6 +1339,8 @@ int kvm_mips_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
u32 inst;
int ret = RESUME_GUEST;
+ vcpu->mode = OUTSIDE_GUEST_MODE;
+
/* re-enable HTW before enabling interrupts */
htw_start();
@@ -1486,6 +1496,14 @@ int kvm_mips_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
if (ret == RESUME_GUEST) {
trace_kvm_reenter(vcpu);
+ /*
+ * Make sure the read of VCPU requests in vcpu_reenter()
+ * callback is not reordered ahead of the write to vcpu->mode,
+ * or we could miss a TLB flush request while the requester sees
+ * the VCPU as outside of guest mode and not needing an IPI.
+ */
+ smp_store_mb(vcpu->mode, IN_GUEST_MODE);
+
kvm_mips_callbacks->vcpu_reenter(run, vcpu);
/*
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 09f5da706d9a..e41ee36dd626 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -696,6 +696,7 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
local_irq_save(flags);
+ vcpu->cpu = cpu;
if (vcpu->arch.last_sched_cpu != cpu) {
kvm_debug("[%d->%d]KVM VCPU[%d] switch\n",
vcpu->arch.last_sched_cpu, cpu, vcpu->vcpu_id);
@@ -723,6 +724,7 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
cpu = smp_processor_id();
vcpu->arch.last_sched_cpu = cpu;
+ vcpu->cpu = -1;
/* save guest state in registers */
kvm_mips_callbacks->vcpu_put(vcpu, cpu);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu
2017-01-09 20:51 ` [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Keep the vcpu->mode and vcpu->cpu variables up to date so that
kvm_make_all_cpus_request() has a chance of functioning correctly. This
will soon need to be used for kvm_flush_remote_tlbs().
We can easily update vcpu->cpu when the VCPU context is loaded or saved,
which will happen when accessing guest context and when the guest is
scheduled in and out.
We need to be a little careful with vcpu->mode though, as we will in
future be checking for outstanding VCPU requests, and this must be done
after the value of IN_GUEST_MODE in vcpu->mode is visible to other CPUs.
Otherwise the other CPU could fail to trigger an IPI to wait for
completion dispite the VCPU request not being seen.
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 | 18 ++++++++++++++++++
arch/mips/kvm/mmu.c | 2 ++
2 files changed, 20 insertions(+), 0 deletions(-)
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 90c28fbf4829..325e98367b30 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -365,6 +365,14 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
guest_enter_irqoff();
trace_kvm_enter(vcpu);
+ /*
+ * Make sure the read of VCPU requests in vcpu_run() callback is not
+ * reordered ahead of the write to vcpu->mode, or we could miss a TLB
+ * flush request while the requester sees the VCPU as outside of guest
+ * mode and not needing an IPI.
+ */
+ smp_store_mb(vcpu->mode, IN_GUEST_MODE);
+
r = kvm_mips_callbacks->vcpu_run(run, vcpu);
trace_kvm_out(vcpu);
@@ -1331,6 +1339,8 @@ int kvm_mips_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
u32 inst;
int ret = RESUME_GUEST;
+ vcpu->mode = OUTSIDE_GUEST_MODE;
+
/* re-enable HTW before enabling interrupts */
htw_start();
@@ -1486,6 +1496,14 @@ int kvm_mips_handle_exit(struct kvm_run *run, struct kvm_vcpu *vcpu)
if (ret == RESUME_GUEST) {
trace_kvm_reenter(vcpu);
+ /*
+ * Make sure the read of VCPU requests in vcpu_reenter()
+ * callback is not reordered ahead of the write to vcpu->mode,
+ * or we could miss a TLB flush request while the requester sees
+ * the VCPU as outside of guest mode and not needing an IPI.
+ */
+ smp_store_mb(vcpu->mode, IN_GUEST_MODE);
+
kvm_mips_callbacks->vcpu_reenter(run, vcpu);
/*
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 09f5da706d9a..e41ee36dd626 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -696,6 +696,7 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
local_irq_save(flags);
+ vcpu->cpu = cpu;
if (vcpu->arch.last_sched_cpu != cpu) {
kvm_debug("[%d->%d]KVM VCPU[%d] switch\n",
vcpu->arch.last_sched_cpu, cpu, vcpu->vcpu_id);
@@ -723,6 +724,7 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
cpu = smp_processor_id();
vcpu->arch.last_sched_cpu = cpu;
+ vcpu->cpu = -1;
/* save guest state in registers */
kvm_mips_callbacks->vcpu_put(vcpu, cpu);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (3 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks James Hogan
` (5 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Add handling of TLB invalidation requests before entering guest mode.
This will allow asynchonous invalidation of the VCPU mappings when
physical memory regions are altered. Should the CPU running the VCPU
already be in guest mode an IPI will be sent to trigger a guest exit.
The reload_asid path will be used in a future patch for when GVA is
about to be directly accessed by KVM.
In the process, the stale user ASID check in the re-entry path (for lazy
user GVA flushing) is generalised to check the ASID for the current
guest mode, in case a TLB invalidation request was handled. This has the
side effect of making the ASID checks on vcpu_load too conservative,
which will be addressed in a later 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/trap_emul.c | 71 +++++++++++++++++++++++++++++++++++-----
1 file changed, 63 insertions(+), 8 deletions(-)
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index a92772098294..c6a16267f084 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -775,31 +775,86 @@ static int kvm_trap_emul_vcpu_put(struct kvm_vcpu *vcpu, int cpu)
return 0;
}
+static void kvm_trap_emul_check_requests(struct kvm_vcpu *vcpu, int cpu,
+ bool reload_asid)
+{
+ struct mm_struct *kern_mm = &vcpu->arch.guest_kernel_mm;
+ struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
+ struct mm_struct *mm;
+ int i;
+
+ if (likely(!vcpu->requests))
+ return;
+
+ if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
+ /*
+ * Both kernel & user GVA mappings must be invalidated. The
+ * caller is just about to check whether the ASID is stale
+ * anyway so no need to reload it here.
+ */
+ kvm_mips_flush_gva_pt(kern_mm->pgd, KMF_GPA | KMF_KERN);
+ kvm_mips_flush_gva_pt(user_mm->pgd, KMF_GPA | KMF_USER);
+ for_each_possible_cpu(i) {
+ cpu_context(i, kern_mm) = 0;
+ cpu_context(i, user_mm) = 0;
+ }
+
+ /* Generate new ASID for current mode */
+ if (reload_asid) {
+ mm = KVM_GUEST_KERNEL_MODE(vcpu) ? kern_mm : user_mm;
+ get_new_mmu_context(mm, cpu);
+ htw_stop();
+ write_c0_entryhi(cpu_asid(cpu, mm));
+ TLBMISS_HANDLER_SETUP_PGD(mm->pgd);
+ htw_start();
+ }
+ }
+}
+
static void kvm_trap_emul_vcpu_reenter(struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
+ struct mm_struct *kern_mm = &vcpu->arch.guest_kernel_mm;
struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
+ struct mm_struct *mm;
struct mips_coproc *cop0 = vcpu->arch.cop0;
int i, cpu = smp_processor_id();
unsigned int gasid;
/*
- * Lazy host ASID regeneration / PT flush for guest user mode.
- * If the guest ASID has changed since the last guest usermode
- * execution, regenerate the host ASID so as to invalidate stale TLB
- * entries and flush GVA PT entries too.
+ * No need to reload ASID, IRQs are disabled already so there's no rush,
+ * and we'll check if we need to regenerate below anyway before
+ * re-entering the guest.
*/
- if (!KVM_GUEST_KERNEL_MODE(vcpu)) {
+ kvm_trap_emul_check_requests(vcpu, cpu, false);
+
+ if (KVM_GUEST_KERNEL_MODE(vcpu)) {
+ mm = kern_mm;
+ } else {
+ mm = user_mm;
+
+ /*
+ * Lazy host ASID regeneration / PT flush for guest user mode.
+ * If the guest ASID has changed since the last guest usermode
+ * execution, invalidate the stale TLB entries and flush GVA PT
+ * entries too.
+ */
gasid = kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID;
if (gasid != vcpu->arch.last_user_gasid) {
kvm_mips_flush_gva_pt(user_mm->pgd, KMF_USER);
- get_new_mmu_context(user_mm, cpu);
for_each_possible_cpu(i)
- if (i != cpu)
- cpu_context(i, user_mm) = 0;
+ cpu_context(i, user_mm) = 0;
vcpu->arch.last_user_gasid = gasid;
}
}
+
+ /*
+ * Check if ASID is stale. This may happen due to a TLB flush request or
+ * a lazy user MM invalidation.
+ */
+ if ((cpu_context(cpu, mm) ^ asid_cache(cpu)) &
+ asid_version_mask(cpu))
+ get_new_mmu_context(mm, cpu);
}
static int kvm_trap_emul_vcpu_run(struct kvm_run *run, struct kvm_vcpu *vcpu)
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests
2017-01-09 20:51 ` [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Add handling of TLB invalidation requests before entering guest mode.
This will allow asynchonous invalidation of the VCPU mappings when
physical memory regions are altered. Should the CPU running the VCPU
already be in guest mode an IPI will be sent to trigger a guest exit.
The reload_asid path will be used in a future patch for when GVA is
about to be directly accessed by KVM.
In the process, the stale user ASID check in the re-entry path (for lazy
user GVA flushing) is generalised to check the ASID for the current
guest mode, in case a TLB invalidation request was handled. This has the
side effect of making the ASID checks on vcpu_load too conservative,
which will be addressed in a later 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/trap_emul.c | 71 +++++++++++++++++++++++++++++++++++-----
1 file changed, 63 insertions(+), 8 deletions(-)
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index a92772098294..c6a16267f084 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -775,31 +775,86 @@ static int kvm_trap_emul_vcpu_put(struct kvm_vcpu *vcpu, int cpu)
return 0;
}
+static void kvm_trap_emul_check_requests(struct kvm_vcpu *vcpu, int cpu,
+ bool reload_asid)
+{
+ struct mm_struct *kern_mm = &vcpu->arch.guest_kernel_mm;
+ struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
+ struct mm_struct *mm;
+ int i;
+
+ if (likely(!vcpu->requests))
+ return;
+
+ if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
+ /*
+ * Both kernel & user GVA mappings must be invalidated. The
+ * caller is just about to check whether the ASID is stale
+ * anyway so no need to reload it here.
+ */
+ kvm_mips_flush_gva_pt(kern_mm->pgd, KMF_GPA | KMF_KERN);
+ kvm_mips_flush_gva_pt(user_mm->pgd, KMF_GPA | KMF_USER);
+ for_each_possible_cpu(i) {
+ cpu_context(i, kern_mm) = 0;
+ cpu_context(i, user_mm) = 0;
+ }
+
+ /* Generate new ASID for current mode */
+ if (reload_asid) {
+ mm = KVM_GUEST_KERNEL_MODE(vcpu) ? kern_mm : user_mm;
+ get_new_mmu_context(mm, cpu);
+ htw_stop();
+ write_c0_entryhi(cpu_asid(cpu, mm));
+ TLBMISS_HANDLER_SETUP_PGD(mm->pgd);
+ htw_start();
+ }
+ }
+}
+
static void kvm_trap_emul_vcpu_reenter(struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
+ struct mm_struct *kern_mm = &vcpu->arch.guest_kernel_mm;
struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
+ struct mm_struct *mm;
struct mips_coproc *cop0 = vcpu->arch.cop0;
int i, cpu = smp_processor_id();
unsigned int gasid;
/*
- * Lazy host ASID regeneration / PT flush for guest user mode.
- * If the guest ASID has changed since the last guest usermode
- * execution, regenerate the host ASID so as to invalidate stale TLB
- * entries and flush GVA PT entries too.
+ * No need to reload ASID, IRQs are disabled already so there's no rush,
+ * and we'll check if we need to regenerate below anyway before
+ * re-entering the guest.
*/
- if (!KVM_GUEST_KERNEL_MODE(vcpu)) {
+ kvm_trap_emul_check_requests(vcpu, cpu, false);
+
+ if (KVM_GUEST_KERNEL_MODE(vcpu)) {
+ mm = kern_mm;
+ } else {
+ mm = user_mm;
+
+ /*
+ * Lazy host ASID regeneration / PT flush for guest user mode.
+ * If the guest ASID has changed since the last guest usermode
+ * execution, invalidate the stale TLB entries and flush GVA PT
+ * entries too.
+ */
gasid = kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID;
if (gasid != vcpu->arch.last_user_gasid) {
kvm_mips_flush_gva_pt(user_mm->pgd, KMF_USER);
- get_new_mmu_context(user_mm, cpu);
for_each_possible_cpu(i)
- if (i != cpu)
- cpu_context(i, user_mm) = 0;
+ cpu_context(i, user_mm) = 0;
vcpu->arch.last_user_gasid = gasid;
}
}
+
+ /*
+ * Check if ASID is stale. This may happen due to a TLB flush request or
+ * a lazy user MM invalidation.
+ */
+ if ((cpu_context(cpu, mm) ^ asid_cache(cpu)) &
+ asid_version_mask(cpu))
+ get_new_mmu_context(mm, cpu);
}
static int kvm_trap_emul_vcpu_run(struct kvm_run *run, struct kvm_vcpu *vcpu)
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (4 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers James Hogan
` (4 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
The stale ASID checks taking place on VCPU load can be reduced:
- Now that we check for a stale ASID on guest re-entry, there is no need
to do so when loading the VCPU outside of guest context, since it will
happen before entering the guest. Note that a lot of KVM VCPU ioctls
will cause the VCPU to be loaded but guest context won't be entered.
- There is no need to check for a stale kernel_mm ASID when the guest is
in user mode and vice versa. In fact doing so can potentially be
problematic since the user_mm ASID regeneration may trigger a new ASID
cycle, which would cause the kern_mm ASID to become stale after it has
been checked for staleness.
Therefore only check the ASID for the mm corresponding to the current
guest mode, and only if we're already in guest context. We drop some of
the related kvm_debug() calls here too.
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 | 35 ++++++-----------------------------
1 file changed, 6 insertions(+), 29 deletions(-)
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index c6a16267f084..0cb76e1aac0e 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -714,35 +714,15 @@ static int kvm_trap_emul_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
struct mm_struct *mm;
- /* Allocate new kernel and user ASIDs if needed */
-
- if ((cpu_context(cpu, kern_mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu)) {
- get_new_mmu_context(kern_mm, cpu);
-
- kvm_debug("[%d]: cpu_context: %#lx\n", cpu,
- cpu_context(cpu, current->mm));
- kvm_debug("[%d]: Allocated new ASID for Guest Kernel: %#lx\n",
- cpu, cpu_context(cpu, kern_mm));
- }
-
- if ((cpu_context(cpu, user_mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu)) {
- get_new_mmu_context(user_mm, cpu);
-
- kvm_debug("[%d]: cpu_context: %#lx\n", cpu,
- cpu_context(cpu, current->mm));
- kvm_debug("[%d]: Allocated new ASID for Guest User: %#lx\n",
- cpu, cpu_context(cpu, user_mm));
- }
-
/*
- * Were we in guest context? If so then the pre-empted ASID is
- * no longer valid, we need to set it to what it should be based
- * on the mode of the Guest (Kernel/User)
+ * Were we in guest context? If so, restore the appropriate ASID based
+ * on the mode of the Guest (Kernel/User).
*/
if (current->flags & PF_VCPU) {
mm = KVM_GUEST_KERNEL_MODE(vcpu) ? kern_mm : user_mm;
+ if ((cpu_context(cpu, mm) ^ asid_cache(cpu)) &
+ asid_version_mask(cpu))
+ get_new_mmu_context(mm, cpu);
write_c0_entryhi(cpu_asid(cpu, mm));
TLBMISS_HANDLER_SETUP_PGD(mm->pgd);
cpumask_clear_cpu(cpu, mm_cpumask(current->active_mm));
@@ -760,11 +740,8 @@ static int kvm_trap_emul_vcpu_put(struct kvm_vcpu *vcpu, int cpu)
if (current->flags & PF_VCPU) {
/* Restore normal Linux process memory map */
if (((cpu_context(cpu, current->mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu))) {
- kvm_debug("%s: Dropping MMU Context: %#lx\n", __func__,
- cpu_context(cpu, current->mm));
+ asid_version_mask(cpu)))
get_new_mmu_context(current->mm, cpu);
- }
write_c0_entryhi(cpu_asid(cpu, current->mm));
TLBMISS_HANDLER_SETUP_PGD(current->mm->pgd);
cpumask_set_cpu(cpu, mm_cpumask(current->mm));
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks
2017-01-09 20:51 ` [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
The stale ASID checks taking place on VCPU load can be reduced:
- Now that we check for a stale ASID on guest re-entry, there is no need
to do so when loading the VCPU outside of guest context, since it will
happen before entering the guest. Note that a lot of KVM VCPU ioctls
will cause the VCPU to be loaded but guest context won't be entered.
- There is no need to check for a stale kernel_mm ASID when the guest is
in user mode and vice versa. In fact doing so can potentially be
problematic since the user_mm ASID regeneration may trigger a new ASID
cycle, which would cause the kern_mm ASID to become stale after it has
been checked for staleness.
Therefore only check the ASID for the mm corresponding to the current
guest mode, and only if we're already in guest context. We drop some of
the related kvm_debug() calls here too.
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 | 35 ++++++-----------------------------
1 file changed, 6 insertions(+), 29 deletions(-)
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index c6a16267f084..0cb76e1aac0e 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -714,35 +714,15 @@ static int kvm_trap_emul_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
struct mm_struct *user_mm = &vcpu->arch.guest_user_mm;
struct mm_struct *mm;
- /* Allocate new kernel and user ASIDs if needed */
-
- if ((cpu_context(cpu, kern_mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu)) {
- get_new_mmu_context(kern_mm, cpu);
-
- kvm_debug("[%d]: cpu_context: %#lx\n", cpu,
- cpu_context(cpu, current->mm));
- kvm_debug("[%d]: Allocated new ASID for Guest Kernel: %#lx\n",
- cpu, cpu_context(cpu, kern_mm));
- }
-
- if ((cpu_context(cpu, user_mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu)) {
- get_new_mmu_context(user_mm, cpu);
-
- kvm_debug("[%d]: cpu_context: %#lx\n", cpu,
- cpu_context(cpu, current->mm));
- kvm_debug("[%d]: Allocated new ASID for Guest User: %#lx\n",
- cpu, cpu_context(cpu, user_mm));
- }
-
/*
- * Were we in guest context? If so then the pre-empted ASID is
- * no longer valid, we need to set it to what it should be based
- * on the mode of the Guest (Kernel/User)
+ * Were we in guest context? If so, restore the appropriate ASID based
+ * on the mode of the Guest (Kernel/User).
*/
if (current->flags & PF_VCPU) {
mm = KVM_GUEST_KERNEL_MODE(vcpu) ? kern_mm : user_mm;
+ if ((cpu_context(cpu, mm) ^ asid_cache(cpu)) &
+ asid_version_mask(cpu))
+ get_new_mmu_context(mm, cpu);
write_c0_entryhi(cpu_asid(cpu, mm));
TLBMISS_HANDLER_SETUP_PGD(mm->pgd);
cpumask_clear_cpu(cpu, mm_cpumask(current->active_mm));
@@ -760,11 +740,8 @@ static int kvm_trap_emul_vcpu_put(struct kvm_vcpu *vcpu, int cpu)
if (current->flags & PF_VCPU) {
/* Restore normal Linux process memory map */
if (((cpu_context(cpu, current->mm) ^ asid_cache(cpu)) &
- asid_version_mask(cpu))) {
- kvm_debug("%s: Dropping MMU Context: %#lx\n", __func__,
- cpu_context(cpu, current->mm));
+ asid_version_mask(cpu)))
get_new_mmu_context(current->mm, cpu);
- }
write_c0_entryhi(cpu_asid(cpu, current->mm));
TLBMISS_HANDLER_SETUP_PGD(current->mm->pgd);
cpumask_set_cpu(cpu, mm_cpumask(current->mm));
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (5 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans James Hogan
` (3 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Add helpers to allow for lockless direct access to the GVA space, by
changing the VCPU mode to READING_SHADOW_PAGE_TABLES for the duration of
the access. This allows asynchronous TLB flush requests in future
patches to safely trigger either a TLB flush before the direct GVA space
access, or a delay until the in-progress lockless direct access is
complete.
The kvm_trap_emul_gva_lockless_begin() and
kvm_trap_emul_gva_lockless_end() helpers take care of guarding the
direct GVA accesses, and kvm_trap_emul_gva_fault() tries to handle a
uaccess fault resulting from a flush having taken place.
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 | 15 ++++++++-
arch/mips/kvm/mmu.c | 51 ++++++++++++++++++++++++++-
arch/mips/kvm/trap_emul.c | 65 +++++++++++++++++++++++++++++++++-
3 files changed, 131 insertions(+), 0 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index b2d968ed2757..491689fcbaac 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -243,6 +243,7 @@ enum emulation_result {
#define TLB_ASID(x) ((x).tlb_hi & KVM_ENTRYHI_ASID)
#define TLB_LO_IDX(x, va) (((va) >> PAGE_SHIFT) & 1)
#define TLB_IS_VALID(x, va) ((x).tlb_lo[TLB_LO_IDX(x, va)] & ENTRYLO_V)
+#define TLB_IS_DIRTY(x, va) ((x).tlb_lo[TLB_LO_IDX(x, va)] & ENTRYLO_D)
#define TLB_HI_VPN2_HIT(x, y) ((TLB_VPN2(x) & ~(x).tlb_mask) == \
((y) & VPN2_MASK & ~(x).tlb_mask))
#define TLB_HI_ASID_HIT(x, y) (TLB_IS_GLOBAL(x) || \
@@ -637,6 +638,20 @@ 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,
bool user);
+void kvm_trap_emul_gva_lockless_begin(struct kvm_vcpu *vcpu);
+void kvm_trap_emul_gva_lockless_end(struct kvm_vcpu *vcpu);
+
+enum kvm_mips_fault_result {
+ KVM_MIPS_MAPPED = 0,
+ KVM_MIPS_GVA,
+ KVM_MIPS_GPA,
+ KVM_MIPS_TLB,
+ KVM_MIPS_TLBINV,
+ KVM_MIPS_TLBMOD,
+};
+enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
+ unsigned long gva,
+ bool write);
/* Emulation */
int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index e41ee36dd626..32c317de6c0a 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -732,6 +732,57 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
local_irq_restore(flags);
}
+/**
+ * kvm_trap_emul_gva_fault() - Safely attempt to handle a GVA access fault.
+ * @vcpu: Virtual CPU.
+ * @gva: Guest virtual address to be accessed.
+ * @write: True if write attempted (must be dirtied and made writable).
+ *
+ * Safely attempt to handle a GVA fault, mapping GVA pages if necessary, and
+ * dirtying the page if @write so that guest instructions can be modified.
+ *
+ * Returns: KVM_MIPS_MAPPED on success.
+ * KVM_MIPS_GVA if bad guest virtual address.
+ * KVM_MIPS_GPA if bad guest physical address.
+ * KVM_MIPS_TLB if guest TLB not present.
+ * KVM_MIPS_TLBINV if guest TLB present but not valid.
+ * KVM_MIPS_TLBMOD if guest TLB read only.
+ */
+enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
+ unsigned long gva,
+ bool write)
+{
+ struct mips_coproc *cop0 = vcpu->arch.cop0;
+ struct kvm_mips_tlb *tlb;
+ int index;
+
+ if (KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG0) {
+ if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu) < 0)
+ return KVM_MIPS_GPA;
+ } else if ((KVM_GUEST_KSEGX(gva) < KVM_GUEST_KSEG0) ||
+ KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG23) {
+ /* Address should be in the guest TLB */
+ index = kvm_mips_guest_tlb_lookup(vcpu, (gva & VPN2_MASK) |
+ (kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID));
+ if (index < 0)
+ return KVM_MIPS_TLB;
+ tlb = &vcpu->arch.guest_tlb[index];
+
+ /* Entry should be valid, and dirty for writes */
+ if (!TLB_IS_VALID(*tlb, gva))
+ return KVM_MIPS_TLBINV;
+ if (write && !TLB_IS_DIRTY(*tlb, gva))
+ return KVM_MIPS_TLBMOD;
+
+ if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva))
+ return KVM_MIPS_GPA;
+ } else {
+ return KVM_MIPS_GVA;
+ }
+
+ return KVM_MIPS_MAPPED;
+}
+
int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out)
{
int err;
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 0cb76e1aac0e..2cc637ed9e95 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -788,6 +788,71 @@ static void kvm_trap_emul_check_requests(struct kvm_vcpu *vcpu, int cpu,
}
}
+/**
+ * kvm_trap_emul_gva_lockless_begin() - Begin lockless access to GVA space.
+ * @vcpu: VCPU pointer.
+ *
+ * Call before a GVA space access outside of guest mode, to ensure that
+ * asynchronous TLB flush requests are handled or delayed until completion of
+ * the GVA access (as indicated by a matching kvm_trap_emul_gva_lockless_end()).
+ *
+ * Should be called with IRQs already enabled.
+ */
+void kvm_trap_emul_gva_lockless_begin(struct kvm_vcpu *vcpu)
+{
+ /* We re-enable IRQs in kvm_trap_emul_gva_lockless_end() */
+ WARN_ON_ONCE(irqs_disabled());
+
+ /*
+ * The caller is about to access the GVA space, so we set the mode to
+ * force TLB flush requests to send an IPI, and also disable IRQs to
+ * delay IPI handling until kvm_trap_emul_gva_lockless_end().
+ */
+ local_irq_disable();
+
+ /*
+ * Make sure the read of VCPU requests is not reordered ahead of the
+ * write to vcpu->mode, or we could miss a TLB flush request while
+ * the requester sees the VCPU as outside of guest mode and not needing
+ * an IPI.
+ */
+ smp_store_mb(vcpu->mode, READING_SHADOW_PAGE_TABLES);
+
+ /*
+ * If a TLB flush has been requested (potentially while
+ * OUTSIDE_GUEST_MODE and assumed immediately effective), perform it
+ * before accessing the GVA space, and be sure to reload the ASID if
+ * necessary as it'll be immediately used.
+ *
+ * TLB flush requests after this check will trigger an IPI due to the
+ * mode change above, which will be delayed due to IRQs disabled.
+ */
+ kvm_trap_emul_check_requests(vcpu, smp_processor_id(), true);
+}
+
+/**
+ * kvm_trap_emul_gva_lockless_end() - End lockless access to GVA space.
+ * @vcpu: VCPU pointer.
+ *
+ * Called after a GVA space access outside of guest mode. Should have a matching
+ * call to kvm_trap_emul_gva_lockless_begin().
+ */
+void kvm_trap_emul_gva_lockless_end(struct kvm_vcpu *vcpu)
+{
+ /*
+ * Make sure the write to vcpu->mode is not reordered in front of GVA
+ * accesses, or a TLB flush requester may not think it necessary to send
+ * an IPI.
+ */
+ smp_store_release(&vcpu->mode, OUTSIDE_GUEST_MODE);
+
+ /*
+ * Now that the access to GVA space is complete, its safe for pending
+ * TLB flush request IPIs to be handled (which indicates completion).
+ */
+ local_irq_enable();
+}
+
static void kvm_trap_emul_vcpu_reenter(struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers
2017-01-09 20:51 ` [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Add helpers to allow for lockless direct access to the GVA space, by
changing the VCPU mode to READING_SHADOW_PAGE_TABLES for the duration of
the access. This allows asynchronous TLB flush requests in future
patches to safely trigger either a TLB flush before the direct GVA space
access, or a delay until the in-progress lockless direct access is
complete.
The kvm_trap_emul_gva_lockless_begin() and
kvm_trap_emul_gva_lockless_end() helpers take care of guarding the
direct GVA accesses, and kvm_trap_emul_gva_fault() tries to handle a
uaccess fault resulting from a flush having taken place.
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 | 15 ++++++++-
arch/mips/kvm/mmu.c | 51 ++++++++++++++++++++++++++-
arch/mips/kvm/trap_emul.c | 65 +++++++++++++++++++++++++++++++++-
3 files changed, 131 insertions(+), 0 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index b2d968ed2757..491689fcbaac 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -243,6 +243,7 @@ enum emulation_result {
#define TLB_ASID(x) ((x).tlb_hi & KVM_ENTRYHI_ASID)
#define TLB_LO_IDX(x, va) (((va) >> PAGE_SHIFT) & 1)
#define TLB_IS_VALID(x, va) ((x).tlb_lo[TLB_LO_IDX(x, va)] & ENTRYLO_V)
+#define TLB_IS_DIRTY(x, va) ((x).tlb_lo[TLB_LO_IDX(x, va)] & ENTRYLO_D)
#define TLB_HI_VPN2_HIT(x, y) ((TLB_VPN2(x) & ~(x).tlb_mask) == \
((y) & VPN2_MASK & ~(x).tlb_mask))
#define TLB_HI_ASID_HIT(x, y) (TLB_IS_GLOBAL(x) || \
@@ -637,6 +638,20 @@ 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,
bool user);
+void kvm_trap_emul_gva_lockless_begin(struct kvm_vcpu *vcpu);
+void kvm_trap_emul_gva_lockless_end(struct kvm_vcpu *vcpu);
+
+enum kvm_mips_fault_result {
+ KVM_MIPS_MAPPED = 0,
+ KVM_MIPS_GVA,
+ KVM_MIPS_GPA,
+ KVM_MIPS_TLB,
+ KVM_MIPS_TLBINV,
+ KVM_MIPS_TLBMOD,
+};
+enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
+ unsigned long gva,
+ bool write);
/* Emulation */
int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out);
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index e41ee36dd626..32c317de6c0a 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -732,6 +732,57 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
local_irq_restore(flags);
}
+/**
+ * kvm_trap_emul_gva_fault() - Safely attempt to handle a GVA access fault.
+ * @vcpu: Virtual CPU.
+ * @gva: Guest virtual address to be accessed.
+ * @write: True if write attempted (must be dirtied and made writable).
+ *
+ * Safely attempt to handle a GVA fault, mapping GVA pages if necessary, and
+ * dirtying the page if @write so that guest instructions can be modified.
+ *
+ * Returns: KVM_MIPS_MAPPED on success.
+ * KVM_MIPS_GVA if bad guest virtual address.
+ * KVM_MIPS_GPA if bad guest physical address.
+ * KVM_MIPS_TLB if guest TLB not present.
+ * KVM_MIPS_TLBINV if guest TLB present but not valid.
+ * KVM_MIPS_TLBMOD if guest TLB read only.
+ */
+enum kvm_mips_fault_result kvm_trap_emul_gva_fault(struct kvm_vcpu *vcpu,
+ unsigned long gva,
+ bool write)
+{
+ struct mips_coproc *cop0 = vcpu->arch.cop0;
+ struct kvm_mips_tlb *tlb;
+ int index;
+
+ if (KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG0) {
+ if (kvm_mips_handle_kseg0_tlb_fault(gva, vcpu) < 0)
+ return KVM_MIPS_GPA;
+ } else if ((KVM_GUEST_KSEGX(gva) < KVM_GUEST_KSEG0) ||
+ KVM_GUEST_KSEGX(gva) == KVM_GUEST_KSEG23) {
+ /* Address should be in the guest TLB */
+ index = kvm_mips_guest_tlb_lookup(vcpu, (gva & VPN2_MASK) |
+ (kvm_read_c0_guest_entryhi(cop0) & KVM_ENTRYHI_ASID));
+ if (index < 0)
+ return KVM_MIPS_TLB;
+ tlb = &vcpu->arch.guest_tlb[index];
+
+ /* Entry should be valid, and dirty for writes */
+ if (!TLB_IS_VALID(*tlb, gva))
+ return KVM_MIPS_TLBINV;
+ if (write && !TLB_IS_DIRTY(*tlb, gva))
+ return KVM_MIPS_TLBMOD;
+
+ if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb, gva))
+ return KVM_MIPS_GPA;
+ } else {
+ return KVM_MIPS_GVA;
+ }
+
+ return KVM_MIPS_MAPPED;
+}
+
int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out)
{
int err;
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 0cb76e1aac0e..2cc637ed9e95 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -788,6 +788,71 @@ static void kvm_trap_emul_check_requests(struct kvm_vcpu *vcpu, int cpu,
}
}
+/**
+ * kvm_trap_emul_gva_lockless_begin() - Begin lockless access to GVA space.
+ * @vcpu: VCPU pointer.
+ *
+ * Call before a GVA space access outside of guest mode, to ensure that
+ * asynchronous TLB flush requests are handled or delayed until completion of
+ * the GVA access (as indicated by a matching kvm_trap_emul_gva_lockless_end()).
+ *
+ * Should be called with IRQs already enabled.
+ */
+void kvm_trap_emul_gva_lockless_begin(struct kvm_vcpu *vcpu)
+{
+ /* We re-enable IRQs in kvm_trap_emul_gva_lockless_end() */
+ WARN_ON_ONCE(irqs_disabled());
+
+ /*
+ * The caller is about to access the GVA space, so we set the mode to
+ * force TLB flush requests to send an IPI, and also disable IRQs to
+ * delay IPI handling until kvm_trap_emul_gva_lockless_end().
+ */
+ local_irq_disable();
+
+ /*
+ * Make sure the read of VCPU requests is not reordered ahead of the
+ * write to vcpu->mode, or we could miss a TLB flush request while
+ * the requester sees the VCPU as outside of guest mode and not needing
+ * an IPI.
+ */
+ smp_store_mb(vcpu->mode, READING_SHADOW_PAGE_TABLES);
+
+ /*
+ * If a TLB flush has been requested (potentially while
+ * OUTSIDE_GUEST_MODE and assumed immediately effective), perform it
+ * before accessing the GVA space, and be sure to reload the ASID if
+ * necessary as it'll be immediately used.
+ *
+ * TLB flush requests after this check will trigger an IPI due to the
+ * mode change above, which will be delayed due to IRQs disabled.
+ */
+ kvm_trap_emul_check_requests(vcpu, smp_processor_id(), true);
+}
+
+/**
+ * kvm_trap_emul_gva_lockless_end() - End lockless access to GVA space.
+ * @vcpu: VCPU pointer.
+ *
+ * Called after a GVA space access outside of guest mode. Should have a matching
+ * call to kvm_trap_emul_gva_lockless_begin().
+ */
+void kvm_trap_emul_gva_lockless_end(struct kvm_vcpu *vcpu)
+{
+ /*
+ * Make sure the write to vcpu->mode is not reordered in front of GVA
+ * accesses, or a TLB flush requester may not think it necessary to send
+ * an IPI.
+ */
+ smp_store_release(&vcpu->mode, OUTSIDE_GUEST_MODE);
+
+ /*
+ * Now that the access to GVA space is complete, its safe for pending
+ * TLB flush request IPIs to be handled (which indicates completion).
+ */
+ local_irq_enable();
+}
+
static void kvm_trap_emul_vcpu_reenter(struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (6 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers James Hogan
@ 2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:52 ` [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst() James Hogan
` (2 subsequent siblings)
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the dynamic translation of
guest instructions. This will allow it to handle asynchronous TLB
flushes when they are implemented.
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/dyntrans.c | 26 ++++++++++++++++++++++++--
1 file changed, 24 insertions(+), 2 deletions(-)
diff --git a/arch/mips/kvm/dyntrans.c b/arch/mips/kvm/dyntrans.c
index 60ebf5862d2b..f8e772564d74 100644
--- a/arch/mips/kvm/dyntrans.c
+++ b/arch/mips/kvm/dyntrans.c
@@ -33,10 +33,32 @@ static int kvm_mips_trans_replace(struct kvm_vcpu *vcpu, u32 *opc,
unsigned long vaddr = (unsigned long)opc;
int err;
+retry:
+ /* The GVA page table is still active so use the Linux TLB handlers */
+ kvm_trap_emul_gva_lockless_begin(vcpu);
err = put_user(replace.word, opc);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
if (unlikely(err)) {
- kvm_err("%s: Invalid address: %p\n", __func__, opc);
- return err;
+ /*
+ * We write protect clean pages in GVA page table so normal
+ * Linux TLB mod handler doesn't silently dirty the page.
+ * Its also possible we raced with a GVA invalidation.
+ * Try to force the page to become dirty.
+ */
+ err = kvm_trap_emul_gva_fault(vcpu, vaddr, true);
+ if (unlikely(err)) {
+ kvm_info("%s: Address unwriteable: %p\n",
+ __func__, opc);
+ return -EFAULT;
+ }
+
+ /*
+ * Try again. This will likely trigger a TLB refill, which will
+ * fetch the new dirty entry from the GVA page table, which
+ * should then succeed.
+ */
+ goto retry;
}
__local_flush_icache_user_range(vaddr, vaddr + 4);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans
2017-01-09 20:51 ` [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans James Hogan
@ 2017-01-09 20:51 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:51 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the dynamic translation of
guest instructions. This will allow it to handle asynchronous TLB
flushes when they are implemented.
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/dyntrans.c | 26 ++++++++++++++++++++++++--
1 file changed, 24 insertions(+), 2 deletions(-)
diff --git a/arch/mips/kvm/dyntrans.c b/arch/mips/kvm/dyntrans.c
index 60ebf5862d2b..f8e772564d74 100644
--- a/arch/mips/kvm/dyntrans.c
+++ b/arch/mips/kvm/dyntrans.c
@@ -33,10 +33,32 @@ static int kvm_mips_trans_replace(struct kvm_vcpu *vcpu, u32 *opc,
unsigned long vaddr = (unsigned long)opc;
int err;
+retry:
+ /* The GVA page table is still active so use the Linux TLB handlers */
+ kvm_trap_emul_gva_lockless_begin(vcpu);
err = put_user(replace.word, opc);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
if (unlikely(err)) {
- kvm_err("%s: Invalid address: %p\n", __func__, opc);
- return err;
+ /*
+ * We write protect clean pages in GVA page table so normal
+ * Linux TLB mod handler doesn't silently dirty the page.
+ * Its also possible we raced with a GVA invalidation.
+ * Try to force the page to become dirty.
+ */
+ err = kvm_trap_emul_gva_fault(vcpu, vaddr, true);
+ if (unlikely(err)) {
+ kvm_info("%s: Address unwriteable: %p\n",
+ __func__, opc);
+ return -EFAULT;
+ }
+
+ /*
+ * Try again. This will likely trigger a TLB refill, which will
+ * fetch the new dirty entry from the GVA page table, which
+ * should then succeed.
+ */
+ goto retry;
}
__local_flush_icache_user_range(vaddr, vaddr + 4);
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst()
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (7 preceding siblings ...)
2017-01-09 20:51 ` [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans James Hogan
@ 2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation James Hogan
2017-01-09 20:52 ` [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot James Hogan
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the reading of guest
instructions for emulation. This will allow it to handle asynchronous
TLB flushes when they are implemented.
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 | 21 ++++++++++++++++++---
1 file changed, 18 insertions(+), 3 deletions(-)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 32c317de6c0a..b3da473e1569 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -787,11 +787,26 @@ int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out)
{
int err;
+retry:
+ kvm_trap_emul_gva_lockless_begin(vcpu);
err = get_user(*out, opc);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
if (unlikely(err)) {
- kvm_err("%s: illegal address: %p\n", __func__, opc);
- return -EFAULT;
- }
+ /*
+ * Try to handle the fault, maybe we just raced with a GVA
+ * invalidation.
+ */
+ err = kvm_trap_emul_gva_fault(vcpu, (unsigned long)opc,
+ false);
+ if (unlikely(err)) {
+ kvm_err("%s: illegal address: %p\n",
+ __func__, opc);
+ return -EFAULT;
+ }
+ /* Hopefully it'll work now */
+ goto retry;
+ }
return 0;
}
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst()
2017-01-09 20:52 ` [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst() James Hogan
@ 2017-01-09 20:52 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the reading of guest
instructions for emulation. This will allow it to handle asynchronous
TLB flushes when they are implemented.
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 | 21 ++++++++++++++++++---
1 file changed, 18 insertions(+), 3 deletions(-)
diff --git a/arch/mips/kvm/mmu.c b/arch/mips/kvm/mmu.c
index 32c317de6c0a..b3da473e1569 100644
--- a/arch/mips/kvm/mmu.c
+++ b/arch/mips/kvm/mmu.c
@@ -787,11 +787,26 @@ int kvm_get_inst(u32 *opc, struct kvm_vcpu *vcpu, u32 *out)
{
int err;
+retry:
+ kvm_trap_emul_gva_lockless_begin(vcpu);
err = get_user(*out, opc);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
if (unlikely(err)) {
- kvm_err("%s: illegal address: %p\n", __func__, opc);
- return -EFAULT;
- }
+ /*
+ * Try to handle the fault, maybe we just raced with a GVA
+ * invalidation.
+ */
+ err = kvm_trap_emul_gva_fault(vcpu, (unsigned long)opc,
+ false);
+ if (unlikely(err)) {
+ kvm_err("%s: illegal address: %p\n",
+ __func__, opc);
+ return -EFAULT;
+ }
+ /* Hopefully it'll work now */
+ goto retry;
+ }
return 0;
}
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (8 preceding siblings ...)
2017-01-09 20:52 ` [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst() James Hogan
@ 2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot James Hogan
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the reading of guest
instructions for emulation. This will allow it to handle asynchronous
TLB flushes when they are implemented.
This is a little more complicated than the other two cases (get_inst()
and dynamic translation) due to the need to emulate the appropriate
guest TLB exception when the address isn't present or isn't valid in the
guest TLB.
Since there are several protected cache ops that may need to be
performed safely, this is abstracted by kvm_mips_guest_cache_op() which
is passed a protected cache op function pointer and takes care of the
lockless operation and fault handling / retry if the op should fail,
taking advantage of the new errors which the protected cache ops can now
return. This allows the existing advance fault handling which relied on
host TLB lookups to be removed, along with the now unused
kvm_mips_host_tlb_lookup(),
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 | 2 +-
arch/mips/kvm/emulate.c | 149 ++++++++++++++------------------
arch/mips/kvm/tlb.c | 35 +--------
3 files changed, 67 insertions(+), 119 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 491689fcbaac..de20a6da5a4d 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -226,6 +226,7 @@ enum emulation_result {
EMULATE_FAIL, /* can't emulate this instruction */
EMULATE_WAIT, /* WAIT instruction */
EMULATE_PRIV_FAIL,
+ EMULATE_EXCEPT, /* A guest exception has been generated */
};
#define mips3_paddr_to_tlbpfn(x) \
@@ -614,7 +615,6 @@ extern int kvm_mips_host_tlb_inv(struct kvm_vcpu *vcpu, unsigned long entryhi,
extern int kvm_mips_guest_tlb_lookup(struct kvm_vcpu *vcpu,
unsigned long entryhi);
-extern int kvm_mips_host_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long vaddr);
/* MMU handling */
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index b295a4a1496f..72eb307a61a7 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -1697,12 +1697,57 @@ enum emulation_result kvm_mips_emulate_load(union mips_instruction inst,
return er;
}
+static enum emulation_result kvm_mips_guest_cache_op(int (*fn)(unsigned long),
+ unsigned long curr_pc,
+ unsigned long addr,
+ struct kvm_run *run,
+ struct kvm_vcpu *vcpu,
+ u32 cause)
+{
+ int err;
+
+ for (;;) {
+ /* Carefully attempt the cache operation */
+ kvm_trap_emul_gva_lockless_begin(vcpu);
+ err = fn(addr);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
+ if (likely(!err))
+ return EMULATE_DONE;
+
+ /*
+ * Try to handle the fault and retry, maybe we just raced with a
+ * GVA invalidation.
+ */
+ switch (kvm_trap_emul_gva_fault(vcpu, addr, false)) {
+ case KVM_MIPS_GVA:
+ case KVM_MIPS_GPA:
+ /* bad virtual or physical address */
+ return EMULATE_FAIL;
+ case KVM_MIPS_TLB:
+ /* no matching guest TLB */
+ vcpu->arch.host_cp0_badvaddr = addr;
+ vcpu->arch.pc = curr_pc;
+ kvm_mips_emulate_tlbmiss_ld(cause, NULL, run, vcpu);
+ return EMULATE_EXCEPT;
+ case KVM_MIPS_TLBINV:
+ /* invalid matching guest TLB */
+ vcpu->arch.host_cp0_badvaddr = addr;
+ vcpu->arch.pc = curr_pc;
+ kvm_mips_emulate_tlbinv_ld(cause, NULL, run, vcpu);
+ return EMULATE_EXCEPT;
+ default:
+ break;
+ };
+ }
+}
+
+
enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
u32 *opc, u32 cause,
struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
- struct mips_coproc *cop0 = vcpu->arch.cop0;
enum emulation_result er = EMULATE_DONE;
u32 cache, op_inst, op, base;
s16 offset;
@@ -1759,81 +1804,16 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
goto done;
}
- preempt_disable();
- if (KVM_GUEST_KSEGX(va) == KVM_GUEST_KSEG0) {
- if (kvm_mips_host_tlb_lookup(vcpu, va) < 0 &&
- kvm_mips_handle_kseg0_tlb_fault(va, vcpu)) {
- kvm_err("%s: handling mapped kseg0 tlb fault for %lx, vcpu: %p, ASID: %#lx\n",
- __func__, va, vcpu, read_c0_entryhi());
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
- }
- } else if ((KVM_GUEST_KSEGX(va) < KVM_GUEST_KSEG0) ||
- KVM_GUEST_KSEGX(va) == KVM_GUEST_KSEG23) {
- int index;
-
- /* If an entry already exists then skip */
- if (kvm_mips_host_tlb_lookup(vcpu, va) >= 0)
- goto skip_fault;
-
- /*
- * If address not in the guest TLB, then give the guest a fault,
- * the resulting handler will do the right thing
- */
- index = kvm_mips_guest_tlb_lookup(vcpu, (va & VPN2_MASK) |
- (kvm_read_c0_guest_entryhi
- (cop0) & KVM_ENTRYHI_ASID));
-
- if (index < 0) {
- vcpu->arch.host_cp0_badvaddr = va;
- vcpu->arch.pc = curr_pc;
- er = kvm_mips_emulate_tlbmiss_ld(cause, NULL, run,
- vcpu);
- preempt_enable();
- goto dont_update_pc;
- } else {
- struct kvm_mips_tlb *tlb = &vcpu->arch.guest_tlb[index];
- /*
- * Check if the entry is valid, if not then setup a TLB
- * invalid exception to the guest
- */
- if (!TLB_IS_VALID(*tlb, va)) {
- vcpu->arch.host_cp0_badvaddr = va;
- vcpu->arch.pc = curr_pc;
- er = kvm_mips_emulate_tlbinv_ld(cause, NULL,
- run, vcpu);
- preempt_enable();
- goto dont_update_pc;
- }
- /*
- * We fault an entry from the guest tlb to the
- * shadow host TLB
- */
- if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb,
- va)) {
- kvm_err("%s: handling mapped seg tlb fault for %lx, index: %u, vcpu: %p, ASID: %#lx\n",
- __func__, va, index, vcpu,
- read_c0_entryhi());
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
- }
- }
- } else {
- kvm_err("INVALID CACHE INDEX/ADDRESS (cache: %#x, op: %#x, base[%d]: %#lx, offset: %#x\n",
- cache, op, base, arch->gprs[base], offset);
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
-
- }
-
-skip_fault:
/* XXXKYMA: Only a subset of cache ops are supported, used by Linux */
if (op_inst == Hit_Writeback_Inv_D || op_inst == Hit_Invalidate_D) {
- protected_writeback_dcache_line(va);
-
+ /*
+ * Perform the dcache part of icache synchronisation on the
+ * guest's behalf.
+ */
+ er = kvm_mips_guest_cache_op(protected_writeback_dcache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
#ifdef CONFIG_KVM_MIPS_DYN_TRANS
/*
* Replace the CACHE instruction, with a SYNCI, not the same,
@@ -1842,8 +1822,15 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
kvm_mips_trans_cache_va(inst, opc, vcpu);
#endif
} else if (op_inst == Hit_Invalidate_I) {
- protected_writeback_dcache_line(va);
- protected_flush_icache_line(va);
+ /* Perform the icache synchronisation on the guest's behalf */
+ er = kvm_mips_guest_cache_op(protected_writeback_dcache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
+ er = kvm_mips_guest_cache_op(protected_flush_icache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
#ifdef CONFIG_KVM_MIPS_DYN_TRANS
/* Replace the CACHE instruction, with a SYNCI */
@@ -1855,17 +1842,13 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
er = EMULATE_FAIL;
}
- preempt_enable();
done:
/* Rollback PC only if emulation was unsuccessful */
if (er == EMULATE_FAIL)
vcpu->arch.pc = curr_pc;
-
-dont_update_pc:
- /*
- * This is for exceptions whose emulation updates the PC, so do not
- * overwrite the PC under any circumstances
- */
+ /* Guest exception needs guest to resume */
+ if (er == EMULATE_EXCEPT)
+ er = EMULATE_DONE;
return er;
}
diff --git a/arch/mips/kvm/tlb.c b/arch/mips/kvm/tlb.c
index 4b0528361f56..085f8f48233e 100644
--- a/arch/mips/kvm/tlb.c
+++ b/arch/mips/kvm/tlb.c
@@ -117,41 +117,6 @@ int kvm_mips_guest_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long entryhi)
}
EXPORT_SYMBOL_GPL(kvm_mips_guest_tlb_lookup);
-int kvm_mips_host_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long vaddr)
-{
- unsigned long old_entryhi, flags;
- int idx;
-
- local_irq_save(flags);
-
- old_entryhi = read_c0_entryhi();
-
- if (KVM_GUEST_KERNEL_MODE(vcpu))
- write_c0_entryhi((vaddr & VPN2_MASK) |
- kvm_mips_get_kernel_asid(vcpu));
- else {
- write_c0_entryhi((vaddr & VPN2_MASK) |
- kvm_mips_get_user_asid(vcpu));
- }
-
- mtc0_tlbw_hazard();
-
- tlb_probe();
- tlb_probe_hazard();
- idx = read_c0_index();
-
- /* Restore old ASID */
- write_c0_entryhi(old_entryhi);
- mtc0_tlbw_hazard();
-
- local_irq_restore(flags);
-
- kvm_debug("Host TLB lookup, %#lx, idx: %2d\n", vaddr, idx);
-
- return idx;
-}
-EXPORT_SYMBOL_GPL(kvm_mips_host_tlb_lookup);
-
static int _kvm_mips_host_tlb_inv(unsigned long entryhi)
{
int idx;
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation
2017-01-09 20:52 ` [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation James Hogan
@ 2017-01-09 20:52 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Use the lockless GVA helpers to implement the reading of guest
instructions for emulation. This will allow it to handle asynchronous
TLB flushes when they are implemented.
This is a little more complicated than the other two cases (get_inst()
and dynamic translation) due to the need to emulate the appropriate
guest TLB exception when the address isn't present or isn't valid in the
guest TLB.
Since there are several protected cache ops that may need to be
performed safely, this is abstracted by kvm_mips_guest_cache_op() which
is passed a protected cache op function pointer and takes care of the
lockless operation and fault handling / retry if the op should fail,
taking advantage of the new errors which the protected cache ops can now
return. This allows the existing advance fault handling which relied on
host TLB lookups to be removed, along with the now unused
kvm_mips_host_tlb_lookup(),
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 | 2 +-
arch/mips/kvm/emulate.c | 149 ++++++++++++++------------------
arch/mips/kvm/tlb.c | 35 +--------
3 files changed, 67 insertions(+), 119 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index 491689fcbaac..de20a6da5a4d 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -226,6 +226,7 @@ enum emulation_result {
EMULATE_FAIL, /* can't emulate this instruction */
EMULATE_WAIT, /* WAIT instruction */
EMULATE_PRIV_FAIL,
+ EMULATE_EXCEPT, /* A guest exception has been generated */
};
#define mips3_paddr_to_tlbpfn(x) \
@@ -614,7 +615,6 @@ extern int kvm_mips_host_tlb_inv(struct kvm_vcpu *vcpu, unsigned long entryhi,
extern int kvm_mips_guest_tlb_lookup(struct kvm_vcpu *vcpu,
unsigned long entryhi);
-extern int kvm_mips_host_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long vaddr);
/* MMU handling */
diff --git a/arch/mips/kvm/emulate.c b/arch/mips/kvm/emulate.c
index b295a4a1496f..72eb307a61a7 100644
--- a/arch/mips/kvm/emulate.c
+++ b/arch/mips/kvm/emulate.c
@@ -1697,12 +1697,57 @@ enum emulation_result kvm_mips_emulate_load(union mips_instruction inst,
return er;
}
+static enum emulation_result kvm_mips_guest_cache_op(int (*fn)(unsigned long),
+ unsigned long curr_pc,
+ unsigned long addr,
+ struct kvm_run *run,
+ struct kvm_vcpu *vcpu,
+ u32 cause)
+{
+ int err;
+
+ for (;;) {
+ /* Carefully attempt the cache operation */
+ kvm_trap_emul_gva_lockless_begin(vcpu);
+ err = fn(addr);
+ kvm_trap_emul_gva_lockless_end(vcpu);
+
+ if (likely(!err))
+ return EMULATE_DONE;
+
+ /*
+ * Try to handle the fault and retry, maybe we just raced with a
+ * GVA invalidation.
+ */
+ switch (kvm_trap_emul_gva_fault(vcpu, addr, false)) {
+ case KVM_MIPS_GVA:
+ case KVM_MIPS_GPA:
+ /* bad virtual or physical address */
+ return EMULATE_FAIL;
+ case KVM_MIPS_TLB:
+ /* no matching guest TLB */
+ vcpu->arch.host_cp0_badvaddr = addr;
+ vcpu->arch.pc = curr_pc;
+ kvm_mips_emulate_tlbmiss_ld(cause, NULL, run, vcpu);
+ return EMULATE_EXCEPT;
+ case KVM_MIPS_TLBINV:
+ /* invalid matching guest TLB */
+ vcpu->arch.host_cp0_badvaddr = addr;
+ vcpu->arch.pc = curr_pc;
+ kvm_mips_emulate_tlbinv_ld(cause, NULL, run, vcpu);
+ return EMULATE_EXCEPT;
+ default:
+ break;
+ };
+ }
+}
+
+
enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
u32 *opc, u32 cause,
struct kvm_run *run,
struct kvm_vcpu *vcpu)
{
- struct mips_coproc *cop0 = vcpu->arch.cop0;
enum emulation_result er = EMULATE_DONE;
u32 cache, op_inst, op, base;
s16 offset;
@@ -1759,81 +1804,16 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
goto done;
}
- preempt_disable();
- if (KVM_GUEST_KSEGX(va) == KVM_GUEST_KSEG0) {
- if (kvm_mips_host_tlb_lookup(vcpu, va) < 0 &&
- kvm_mips_handle_kseg0_tlb_fault(va, vcpu)) {
- kvm_err("%s: handling mapped kseg0 tlb fault for %lx, vcpu: %p, ASID: %#lx\n",
- __func__, va, vcpu, read_c0_entryhi());
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
- }
- } else if ((KVM_GUEST_KSEGX(va) < KVM_GUEST_KSEG0) ||
- KVM_GUEST_KSEGX(va) == KVM_GUEST_KSEG23) {
- int index;
-
- /* If an entry already exists then skip */
- if (kvm_mips_host_tlb_lookup(vcpu, va) >= 0)
- goto skip_fault;
-
- /*
- * If address not in the guest TLB, then give the guest a fault,
- * the resulting handler will do the right thing
- */
- index = kvm_mips_guest_tlb_lookup(vcpu, (va & VPN2_MASK) |
- (kvm_read_c0_guest_entryhi
- (cop0) & KVM_ENTRYHI_ASID));
-
- if (index < 0) {
- vcpu->arch.host_cp0_badvaddr = va;
- vcpu->arch.pc = curr_pc;
- er = kvm_mips_emulate_tlbmiss_ld(cause, NULL, run,
- vcpu);
- preempt_enable();
- goto dont_update_pc;
- } else {
- struct kvm_mips_tlb *tlb = &vcpu->arch.guest_tlb[index];
- /*
- * Check if the entry is valid, if not then setup a TLB
- * invalid exception to the guest
- */
- if (!TLB_IS_VALID(*tlb, va)) {
- vcpu->arch.host_cp0_badvaddr = va;
- vcpu->arch.pc = curr_pc;
- er = kvm_mips_emulate_tlbinv_ld(cause, NULL,
- run, vcpu);
- preempt_enable();
- goto dont_update_pc;
- }
- /*
- * We fault an entry from the guest tlb to the
- * shadow host TLB
- */
- if (kvm_mips_handle_mapped_seg_tlb_fault(vcpu, tlb,
- va)) {
- kvm_err("%s: handling mapped seg tlb fault for %lx, index: %u, vcpu: %p, ASID: %#lx\n",
- __func__, va, index, vcpu,
- read_c0_entryhi());
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
- }
- }
- } else {
- kvm_err("INVALID CACHE INDEX/ADDRESS (cache: %#x, op: %#x, base[%d]: %#lx, offset: %#x\n",
- cache, op, base, arch->gprs[base], offset);
- er = EMULATE_FAIL;
- preempt_enable();
- goto done;
-
- }
-
-skip_fault:
/* XXXKYMA: Only a subset of cache ops are supported, used by Linux */
if (op_inst == Hit_Writeback_Inv_D || op_inst == Hit_Invalidate_D) {
- protected_writeback_dcache_line(va);
-
+ /*
+ * Perform the dcache part of icache synchronisation on the
+ * guest's behalf.
+ */
+ er = kvm_mips_guest_cache_op(protected_writeback_dcache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
#ifdef CONFIG_KVM_MIPS_DYN_TRANS
/*
* Replace the CACHE instruction, with a SYNCI, not the same,
@@ -1842,8 +1822,15 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
kvm_mips_trans_cache_va(inst, opc, vcpu);
#endif
} else if (op_inst == Hit_Invalidate_I) {
- protected_writeback_dcache_line(va);
- protected_flush_icache_line(va);
+ /* Perform the icache synchronisation on the guest's behalf */
+ er = kvm_mips_guest_cache_op(protected_writeback_dcache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
+ er = kvm_mips_guest_cache_op(protected_flush_icache_line,
+ curr_pc, va, run, vcpu, cause);
+ if (er != EMULATE_DONE)
+ goto done;
#ifdef CONFIG_KVM_MIPS_DYN_TRANS
/* Replace the CACHE instruction, with a SYNCI */
@@ -1855,17 +1842,13 @@ enum emulation_result kvm_mips_emulate_cache(union mips_instruction inst,
er = EMULATE_FAIL;
}
- preempt_enable();
done:
/* Rollback PC only if emulation was unsuccessful */
if (er == EMULATE_FAIL)
vcpu->arch.pc = curr_pc;
-
-dont_update_pc:
- /*
- * This is for exceptions whose emulation updates the PC, so do not
- * overwrite the PC under any circumstances
- */
+ /* Guest exception needs guest to resume */
+ if (er == EMULATE_EXCEPT)
+ er = EMULATE_DONE;
return er;
}
diff --git a/arch/mips/kvm/tlb.c b/arch/mips/kvm/tlb.c
index 4b0528361f56..085f8f48233e 100644
--- a/arch/mips/kvm/tlb.c
+++ b/arch/mips/kvm/tlb.c
@@ -117,41 +117,6 @@ int kvm_mips_guest_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long entryhi)
}
EXPORT_SYMBOL_GPL(kvm_mips_guest_tlb_lookup);
-int kvm_mips_host_tlb_lookup(struct kvm_vcpu *vcpu, unsigned long vaddr)
-{
- unsigned long old_entryhi, flags;
- int idx;
-
- local_irq_save(flags);
-
- old_entryhi = read_c0_entryhi();
-
- if (KVM_GUEST_KERNEL_MODE(vcpu))
- write_c0_entryhi((vaddr & VPN2_MASK) |
- kvm_mips_get_kernel_asid(vcpu));
- else {
- write_c0_entryhi((vaddr & VPN2_MASK) |
- kvm_mips_get_user_asid(vcpu));
- }
-
- mtc0_tlbw_hazard();
-
- tlb_probe();
- tlb_probe_hazard();
- idx = read_c0_index();
-
- /* Restore old ASID */
- write_c0_entryhi(old_entryhi);
- mtc0_tlbw_hazard();
-
- local_irq_restore(flags);
-
- kvm_debug("Host TLB lookup, %#lx, idx: %2d\n", vaddr, idx);
-
- return idx;
-}
-EXPORT_SYMBOL_GPL(kvm_mips_host_tlb_lookup);
-
static int _kvm_mips_host_tlb_inv(unsigned long entryhi)
{
int idx;
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
` (9 preceding siblings ...)
2017-01-09 20:52 ` [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation James Hogan
@ 2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` James Hogan
10 siblings, 1 reply; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Implement the kvm_arch_flush_shadow_all() and
kvm_arch_flush_shadow_memslot() KVM functions for MIPS to allow guest
physical mappings to be safely changed.
The general MIPS KVM code takes care of flushing of GPA page table
entries. kvm_arch_flush_shadow_all() flushes the whole GPA page table,
and is always called on the cleanup path so there is no need to acquire
the kvm->mmu_lock. kvm_arch_flush_shadow_memslot() flushes only the
range of mappings in the GPA page table corresponding to the slot being
flushed, and happens when memory regions are moved or deleted.
MIPS KVM implementation callbacks are added for handling the
implementation specific flushing of mappings derived from the GPA page
tables. These are implemented for trap_emul.c using
kvm_flush_remote_tlbs() which should now be functional, and will flush
the per-VCPU GVA page tables and ASIDS synchronously (before next
entering guest mode or directly accessing GVA space).
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 | 11 ++++++++---
arch/mips/kvm/mips.c | 26 ++++++++++++++++++++++++++
arch/mips/kvm/trap_emul.c | 14 ++++++++++++++
3 files changed, 48 insertions(+), 3 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index de20a6da5a4d..c24c1c23196b 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -531,6 +531,14 @@ struct kvm_mips_callbacks {
int (*vcpu_init)(struct kvm_vcpu *vcpu);
void (*vcpu_uninit)(struct kvm_vcpu *vcpu);
int (*vcpu_setup)(struct kvm_vcpu *vcpu);
+ void (*flush_shadow_all)(struct kvm *kvm);
+ /*
+ * Must take care of flushing any cached GPA PTEs (e.g. guest entries in
+ * VZ root TLB, or T&E GVA page tables and corresponding root TLB
+ * mappings).
+ */
+ void (*flush_shadow_memslot)(struct kvm *kvm,
+ const struct kvm_memory_slot *slot);
gpa_t (*gva_to_gpa)(gva_t gva);
void (*queue_timer_int)(struct kvm_vcpu *vcpu);
void (*dequeue_timer_int)(struct kvm_vcpu *vcpu);
@@ -824,9 +832,6 @@ static inline void kvm_arch_sync_events(struct kvm *kvm) {}
static inline void kvm_arch_free_memslot(struct kvm *kvm,
struct kvm_memory_slot *free, struct kvm_memory_slot *dont) {}
static inline void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots) {}
-static inline void kvm_arch_flush_shadow_all(struct kvm *kvm) {}
-static inline void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
- struct kvm_memory_slot *slot) {}
static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 325e98367b30..b0d9fe2176c0 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -157,6 +157,32 @@ int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
return 0;
}
+void kvm_arch_flush_shadow_all(struct kvm *kvm)
+{
+ /* Flush whole GPA */
+ kvm_mips_flush_gpa_pt(kvm, 0, ~0);
+
+ /* Let implementation do the rest */
+ kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
+ struct kvm_memory_slot *slot)
+{
+ /*
+ * The slot has been made invalid (ready for moving or deletion), so we
+ * need to ensure that it can no longer be accessed by any guest VCPUs.
+ */
+
+ spin_lock(&kvm->mmu_lock);
+ /* Flush slot from GPA */
+ kvm_mips_flush_gpa_pt(kvm, slot->base_gfn,
+ slot->base_gfn + slot->npages - 1);
+ /* Let implementation do the rest */
+ kvm_mips_callbacks->flush_shadow_memslot(kvm, slot);
+ spin_unlock(&kvm->mmu_lock);
+}
+
int kvm_arch_prepare_memory_region(struct kvm *kvm,
struct kvm_memory_slot *memslot,
const struct kvm_userspace_memory_region *mem,
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 2cc637ed9e95..a9a8c8e450e4 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -586,6 +586,18 @@ static int kvm_trap_emul_vcpu_setup(struct kvm_vcpu *vcpu)
return 0;
}
+static void kvm_trap_emul_flush_shadow_all(struct kvm *kvm)
+{
+ /* Flush GVA page tables and invalidate GVA ASIDs on all VCPUs */
+ kvm_flush_remote_tlbs(kvm);
+}
+
+static void kvm_trap_emul_flush_shadow_memslot(struct kvm *kvm,
+ const struct kvm_memory_slot *slot)
+{
+ kvm_trap_emul_flush_shadow_all(kvm);
+}
+
static unsigned long kvm_trap_emul_num_regs(struct kvm_vcpu *vcpu)
{
return 0;
@@ -967,6 +979,8 @@ static struct kvm_mips_callbacks kvm_trap_emul_callbacks = {
.vcpu_init = kvm_trap_emul_vcpu_init,
.vcpu_uninit = kvm_trap_emul_vcpu_uninit,
.vcpu_setup = kvm_trap_emul_vcpu_setup,
+ .flush_shadow_all = kvm_trap_emul_flush_shadow_all,
+ .flush_shadow_memslot = kvm_trap_emul_flush_shadow_memslot,
.gva_to_gpa = kvm_trap_emul_gva_to_gpa_cb,
.queue_timer_int = kvm_mips_queue_timer_int_cb,
.dequeue_timer_int = kvm_mips_dequeue_timer_int_cb,
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot
2017-01-09 20:52 ` [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot James Hogan
@ 2017-01-09 20:52 ` James Hogan
0 siblings, 0 replies; 23+ messages in thread
From: James Hogan @ 2017-01-09 20:52 UTC (permalink / raw)
To: linux-mips
Cc: James Hogan, Paolo Bonzini, Radim Krčmář,
Ralf Baechle, kvm
Implement the kvm_arch_flush_shadow_all() and
kvm_arch_flush_shadow_memslot() KVM functions for MIPS to allow guest
physical mappings to be safely changed.
The general MIPS KVM code takes care of flushing of GPA page table
entries. kvm_arch_flush_shadow_all() flushes the whole GPA page table,
and is always called on the cleanup path so there is no need to acquire
the kvm->mmu_lock. kvm_arch_flush_shadow_memslot() flushes only the
range of mappings in the GPA page table corresponding to the slot being
flushed, and happens when memory regions are moved or deleted.
MIPS KVM implementation callbacks are added for handling the
implementation specific flushing of mappings derived from the GPA page
tables. These are implemented for trap_emul.c using
kvm_flush_remote_tlbs() which should now be functional, and will flush
the per-VCPU GVA page tables and ASIDS synchronously (before next
entering guest mode or directly accessing GVA space).
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 | 11 ++++++++---
arch/mips/kvm/mips.c | 26 ++++++++++++++++++++++++++
arch/mips/kvm/trap_emul.c | 14 ++++++++++++++
3 files changed, 48 insertions(+), 3 deletions(-)
diff --git a/arch/mips/include/asm/kvm_host.h b/arch/mips/include/asm/kvm_host.h
index de20a6da5a4d..c24c1c23196b 100644
--- a/arch/mips/include/asm/kvm_host.h
+++ b/arch/mips/include/asm/kvm_host.h
@@ -531,6 +531,14 @@ struct kvm_mips_callbacks {
int (*vcpu_init)(struct kvm_vcpu *vcpu);
void (*vcpu_uninit)(struct kvm_vcpu *vcpu);
int (*vcpu_setup)(struct kvm_vcpu *vcpu);
+ void (*flush_shadow_all)(struct kvm *kvm);
+ /*
+ * Must take care of flushing any cached GPA PTEs (e.g. guest entries in
+ * VZ root TLB, or T&E GVA page tables and corresponding root TLB
+ * mappings).
+ */
+ void (*flush_shadow_memslot)(struct kvm *kvm,
+ const struct kvm_memory_slot *slot);
gpa_t (*gva_to_gpa)(gva_t gva);
void (*queue_timer_int)(struct kvm_vcpu *vcpu);
void (*dequeue_timer_int)(struct kvm_vcpu *vcpu);
@@ -824,9 +832,6 @@ static inline void kvm_arch_sync_events(struct kvm *kvm) {}
static inline void kvm_arch_free_memslot(struct kvm *kvm,
struct kvm_memory_slot *free, struct kvm_memory_slot *dont) {}
static inline void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots) {}
-static inline void kvm_arch_flush_shadow_all(struct kvm *kvm) {}
-static inline void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
- struct kvm_memory_slot *slot) {}
static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {}
static inline void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) {}
diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 325e98367b30..b0d9fe2176c0 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -157,6 +157,32 @@ int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
return 0;
}
+void kvm_arch_flush_shadow_all(struct kvm *kvm)
+{
+ /* Flush whole GPA */
+ kvm_mips_flush_gpa_pt(kvm, 0, ~0);
+
+ /* Let implementation do the rest */
+ kvm_mips_callbacks->flush_shadow_all(kvm);
+}
+
+void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
+ struct kvm_memory_slot *slot)
+{
+ /*
+ * The slot has been made invalid (ready for moving or deletion), so we
+ * need to ensure that it can no longer be accessed by any guest VCPUs.
+ */
+
+ spin_lock(&kvm->mmu_lock);
+ /* Flush slot from GPA */
+ kvm_mips_flush_gpa_pt(kvm, slot->base_gfn,
+ slot->base_gfn + slot->npages - 1);
+ /* Let implementation do the rest */
+ kvm_mips_callbacks->flush_shadow_memslot(kvm, slot);
+ spin_unlock(&kvm->mmu_lock);
+}
+
int kvm_arch_prepare_memory_region(struct kvm *kvm,
struct kvm_memory_slot *memslot,
const struct kvm_userspace_memory_region *mem,
diff --git a/arch/mips/kvm/trap_emul.c b/arch/mips/kvm/trap_emul.c
index 2cc637ed9e95..a9a8c8e450e4 100644
--- a/arch/mips/kvm/trap_emul.c
+++ b/arch/mips/kvm/trap_emul.c
@@ -586,6 +586,18 @@ static int kvm_trap_emul_vcpu_setup(struct kvm_vcpu *vcpu)
return 0;
}
+static void kvm_trap_emul_flush_shadow_all(struct kvm *kvm)
+{
+ /* Flush GVA page tables and invalidate GVA ASIDs on all VCPUs */
+ kvm_flush_remote_tlbs(kvm);
+}
+
+static void kvm_trap_emul_flush_shadow_memslot(struct kvm *kvm,
+ const struct kvm_memory_slot *slot)
+{
+ kvm_trap_emul_flush_shadow_all(kvm);
+}
+
static unsigned long kvm_trap_emul_num_regs(struct kvm_vcpu *vcpu)
{
return 0;
@@ -967,6 +979,8 @@ static struct kvm_mips_callbacks kvm_trap_emul_callbacks = {
.vcpu_init = kvm_trap_emul_vcpu_init,
.vcpu_uninit = kvm_trap_emul_vcpu_uninit,
.vcpu_setup = kvm_trap_emul_vcpu_setup,
+ .flush_shadow_all = kvm_trap_emul_flush_shadow_all,
+ .flush_shadow_memslot = kvm_trap_emul_flush_shadow_memslot,
.gva_to_gpa = kvm_trap_emul_gva_to_gpa_cb,
.queue_timer_int = kvm_mips_queue_timer_int_cb,
.dequeue_timer_int = kvm_mips_dequeue_timer_int_cb,
--
git-series 0.8.10
^ permalink raw reply related [flat|nested] 23+ messages in thread
* Re: [PATCH 1/10] MIPS: Add return errors to protected cache ops
2017-01-09 20:51 ` [PATCH 1/10] MIPS: Add return errors to protected cache ops James Hogan
2017-01-09 20:51 ` James Hogan
@ 2017-01-17 16:05 ` Ralf Baechle
1 sibling, 0 replies; 23+ messages in thread
From: Ralf Baechle @ 2017-01-17 16:05 UTC (permalink / raw)
To: James Hogan; +Cc: linux-mips, Paolo Bonzini, Radim Krčmář, kvm
On Mon, Jan 09, 2017 at 08:51:53PM +0000, James Hogan wrote:
> The protected cache ops contain no out of line fixup code to return an
> error code in the event of a fault, with the cache op being skipped in
> that case. For KVM however we'd like to detect this case as page
> faulting will be disabled so it could happen during normal operation if
> the GVA page tables were flushed, and need to be handled by the caller.
>
> Add the out-of-line fixup code to load the error value -EFAULT into the
> return variable, and adapt the protected cache line functions to pass
> the error back to the caller.
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Ralf
^ permalink raw reply [flat|nested] 23+ messages in thread
end of thread, other threads:[~2017-01-17 16:05 UTC | newest]
Thread overview: 23+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-01-09 20:51 [PATCH 0/10] KVM: MIPS: Implement GPA page tables and shadow flushing James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 1/10] MIPS: Add return errors to protected cache ops James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-17 16:05 ` Ralf Baechle
2017-01-09 20:51 ` [PATCH 2/10] KVM: MIPS/MMU: Convert guest physical map to page table James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 3/10] KVM: MIPS: Update vcpu->mode and vcpu->cpu James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 4/10] KVM: MIPS/T&E: Handle TLB invalidation requests James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 5/10] KVM: MIPS/T&E: Reduce stale ASID checks James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 6/10] KVM: MIPS/T&E: Add lockless GVA access helpers James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:51 ` [PATCH 7/10] KVM: MIPS/T&E: Use lockless GVA helpers for dyntrans James Hogan
2017-01-09 20:51 ` James Hogan
2017-01-09 20:52 ` [PATCH 8/10] KVM: MIPS/MMU: Use lockless GVA helpers for get_inst() James Hogan
2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` [PATCH 9/10] KVM: MIPS/Emulate: Use lockless GVA helpers for cache emulation James Hogan
2017-01-09 20:52 ` James Hogan
2017-01-09 20:52 ` [PATCH 10/10] KVM: MIPS: Implement kvm_arch_flush_shadow_all/memslot James Hogan
2017-01-09 20:52 ` James Hogan
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).