KVM: arm64: Add support for FEAT_TLBIRANGE
v7 cover KVM: arm64: Add support for FEAT_TLBIRANGE
In certain code paths, KVM/ARM currently invalidates the entire VM's page-tables instead of just invalidating a necessary range. For example, when collapsing a table PTE to a block PTE, instead of iterating over each PTE and flushing them, KVM uses 'vmalls12e1is' TLBI operation to flush all the entries. This is inefficient since the guest would have to refill the TLBs again, even for the addresses that aren't covered by the table entry. The performance impact would scale poorly if many addresses in the VM is going through this remapping.
For architectures that implement FEAT_TLBIRANGE, KVM can replace such inefficient paths by performing the invalidations only on the range of addresses that are in scope. This series tries to achieve the same in the areas of stage-2 map, unmap and write-protecting the pages.
As suggested by Oliver in the original v5 of the series [1], I'm posting the series by including v2 of David Matlack's 'KVM: Add a common API for range-based TLB invalidation' series [2].
Patches 1-5 includes David M.'s patches 1, 2, 6, and 7 from [2].
Patch-6 refactors the core arm64's __flush_tlb_range() to be used by other entities.
Patch-7,8 adds a range-based TLBI mechanism for KVM (VHE and nVHE).
Patch-9 implements the kvm_arch_flush_remote_tlbs_range() for arm64.
Patch-10 aims to flush only the memslot that undergoes a write-protect, instead of the entire VM.
Patch-11 operates on stage2_try_break_pte() to use the range based TLBI instructions when collapsing a table entry. The map path is the immediate consumer of this when KVM remaps a table entry into a block.
Patch-12 modifies the stage-2 unmap path in which, if the system supports FEAT_TLBIRANGE, the TLB invalidations are skipped during the page-table. walk. Instead it's done in one go after the entire walk is finished.
The series is based off of upstream v6.5-rc1.
The performance evaluation was done on a hardware that supports FEAT_TLBIRANGE, on a VHE configuration, using a modified kvm_page_table_test.
The modified version updates the guest code in the ADJUST_MAPPINGS case to not only access this page but also to access up to 512 pages backwards for every new page it iterates through. This is done to test the effect of TLBI misses after KVM has handled a fault.
The series captures the impact in the map and unmap paths as described above.
$ kvm_page_table_test -m 2 -v 128 -s anonymous_hugetlb_2mb -b $i
+--------+------------------------------+------------------------------+ | mem_sz | ADJUST_MAPPINGS (s) | Unmap VM (s) | | (GB) | Baseline | Baseline + series | Baseline | Baseline + series | +--------+----------|-------------------+------------------------------+ | 1 | 3.33 | 3.22 | 0.009 | 0.005 | | 2 | 7.39 | 7.32 | 0.012 | 0.006 | | 4 | 13.49 | 10.50 | 0.017 | 0.008 | | 8 | 21.60 | 21.50 | 0.027 | 0.011 | | 16 | 57.02 | 43.63 | 0.046 | 0.018 | | 32 | 95.92 | 83.26 | 0.087 | 0.030 | | 64 | 199.57 | 165.14 | 0.146 | 0.055 | | 128 | 423.65 | 349.37 | 0.280 | 0.100 | +--------+----------+-------------------+----------+-------------------+
$ kvm_page_table_test -m 2 -b 128G -s anonymous_hugetlb_2mb -v $i
+--------+------------------------------+ | vCPUs | ADJUST_MAPPINGS (s) | | | Baseline | Baseline + series | +--------+----------|-------------------+ | 1 | 111.44 | 114.63 | | 2 | 102.88 | 74.64 | | 4 | 134.83 | 98.78 | | 8 | 98.81 | 95.01 | | 16 | 127.41 | 99.05 | | 32 | 105.35 | 91.75 | | 64 | 201.13 | 163.63 | | 128 | 423.65 | 349.37 | +--------+----------+-------------------+
For the ADJUST_MAPPINGS cases, which maps back the 4K table entries to 2M hugepages, the series sees an average improvement of ~15%. For unmapping 2M hugepages, we see a gain of 2x to 3x.
$ kvm_page_table_test -m 2 -b $i
+--------+------------------------------+ | mem_sz | Unmap VM (s) | | (GB) | Baseline | Baseline + series | +--------+------------------------------+ | 1 | 0.54 | 0.13 | | 2 | 1.07 | 0.25 | | 4 | 2.10 | 0.47 | | 8 | 4.19 | 0.92 | | 16 | 8.35 | 1.92 | | 32 | 16.66 | 3.61 | | 64 | 32.36 | 7.62 | | 128 | 64.65 | 14.39 | +--------+----------+-------------------+
The series sees an average gain of 4x when the guest backed by PAGE_SIZE (4K) pages.
Other testing:
- Booted on x86_64 and ran KVM selftests.
- Build tested for MIPS and RISCV architectures against defconfig.
v6:
https://lore.kernel.org/all/20230715005405.3689586-1-rananta at google.com/ Thank you, Philippe and Shaoqin for the reviews and suggestions
- Split the patch-2/11 to separate the removal of CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL and arm64 to switch to using kvm_arch_flush_remote_tlbs(). (Philippe)
- Align the 'pages' argument with 'kvm' in patch-3/11. (Shaoqin)
- Call __tlb_switch_to_guest() before __flush_tlb_range_op() in the VHE's implementation of __kvm_tlb_flush_vmid_range(). (Shaoqin)
v5 (RESEND):
https://lore.kernel.org/all/20230621175002.2832640-1-rananta at google.com/ Thanks, Gavin for the suggestions:
- Adjusted the comment on patch-2 to align with the code.
- Fixed checkpatch.pl warning on patch-5.
v5:
https://lore.kernel.org/all/20230606192858.3600174-1-rananta at google.com/ Thank you, Marc and Oliver for the comments
- Introduced a helper, kvm_tlb_flush_vmid_range(), to handle the decision of using range-based TLBI instructions or invalidating the entire VMID, rather than depending on __kvm_tlb_flush_vmid_range() for it.
- kvm_tlb_flush_vmid_range() splits the range-based invalidations if the requested range exceeds MAX_TLBI_RANGE_PAGES.
- All the users in need of invalidating the TLB upon a range now depends on kvm_tlb_flush_vmid_range() rather than directly on __kvm_tlb_flush_vmid_range().
- stage2_unmap_defer_tlb_flush() introduces a WARN_ON() to track if there's any change in TLBIRANGE or FWB support during the unmap process as the features are based on alternative patching and the TLBI operations solely depend
on this check.
- Corrected an incorrect hunk being present on v4's patch-3.
- Updated the patches changelog and code comments as per the suggestions.
v4:
https://lore.kernel.org/all/20230519005231.3027912-1-rananta at google.com/ Thanks again, Oliver for all the comments
- Updated the __kvm_tlb_flush_vmid_range() implementation for nVHE to adjust with the modfied __tlb_switch_to_guest() that accepts a new 'bool nsh' arg.
- Renamed stage2_put_pte() to stage2_unmap_put_pte() and removed the 'skip_flush' argument.
- Defined stage2_unmap_defer_tlb_flush() to check if the PTE flushes can be deferred during the unmap table walk. It's being called from stage2_unmap_put_pte() and kvm_pgtable_stage2_unmap().
- Got rid of the 'struct stage2_unmap_data'.
v3:
https://lore.kernel.org/all/20230414172922.812640-1-rananta at google.com/ Thanks, Oliver for all the suggestions.
- The core flush API (__kvm_tlb_flush_vmid_range()) now checks if the system support FEAT_TLBIRANGE or not, thus elimiating the redundancy in the upper layers.
- If FEAT_TLBIRANGE is not supported, the implementation falls back to invalidating all the TLB entries with the VMID, instead of doing an iterative flush for the range.
- The kvm_arch_flush_remote_tlbs_range() doesn't return -EOPNOTSUPP if the system doesn't implement FEAT_TLBIRANGE. It depends on __kvm_tlb_flush_vmid_range() to do take care of the decisions and return 0 regardless of the underlying feature support.
- __kvm_tlb_flush_vmid_range() doesn't take 'level' as input to calculate the 'stride'. Instead, it always assumes PAGE_SIZE.
- Fast unmap path is eliminated. Instead, the existing unmap walker is modified to skip the TLBIs during the walk, and do it all at once after the walk, using the range-based instructions.
v2:
https://lore.kernel.org/all/20230206172340.2639971-1-rananta at google.com/
- Rebased the series on top of David Matlack's series for common TLB invalidation API[1].
- Implement kvm_arch_flush_remote_tlbs_range() for arm64, by extending the support introduced by [1].
- Use kvm_flush_remote_tlbs_memslot() introduced by [1] to flush only the current memslot after write-protect.
- Modified the __kvm_tlb_flush_range() macro to accepts 'level' as an argument to calculate the 'stride' instead of just using PAGE_SIZE.
- Split the patch that introduces the range-based TLBI to KVM and the implementation of IPA-based invalidation into its own patches.
- Dropped the patch that tries to optimize the mmu notifiers paths.
- Rename the function kvm_table_pte_flush() to kvm_pgtable_stage2_flush_range(), and accept the range of addresses to flush. [Oliver]
- Drop the 'tlb_level' argument for stage2_try_break_pte() and directly pass '0' as 'tlb_level' to kvm_pgtable_stage2_flush_range(). [Oliver]
v1:
https://lore.kernel.org/all/20230109215347.3119271-1-rananta at google.com/
Thank you.
Raghavendra
[1]: https://lore.kernel.org/all/ZIrONR6cSegiK1e2 at linux.dev/
[2]:
https://lore.kernel.org/linux-arm-kernel/20230126184025.2294823-1-dmatlack at google.com/
David Matlack (3):
KVM: Rename kvm_arch_flush_remote_tlb() to
kvm_arch_flush_remote_tlbs()
KVM: Allow range-based TLB invalidation from common code
KVM: Move kvm_arch_flush_remote_tlbs_memslot() to common code
Raghavendra Rao Ananta (9):
KVM: arm64: Use kvm_arch_flush_remote_tlbs()
KVM: Remove CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
KVM: arm64: Implement __kvm_tlb_flush_vmid_range()
KVM: arm64: Define kvm_tlb_flush_vmid_range()
KVM: arm64: Implement kvm_arch_flush_remote_tlbs_range()
KVM: arm64: Flush only the memslot after write-protect
KVM: arm64: Invalidate the table entries upon a range
KVM: arm64: Use TLBI range-based intructions for unmap
arch/arm64/include/asm/kvm_asm.h | 3 + arch/arm64/include/asm/kvm_host.h | 6 ++ arch/arm64/include/asm/kvm_pgtable.h | 10 +++ arch/arm64/include/asm/tlbflush.h | 109 ++++++++++++++------------- arch/arm64/kvm/Kconfig | 1 - arch/arm64/kvm/arm.c | 6 -- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++ arch/arm64/kvm/hyp/pgtable.c | 90 +++++++++++++++++++--- arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++ arch/arm64/kvm/mmu.c | 15 +++- arch/mips/include/asm/kvm_host.h | 4 +- arch/mips/kvm/mips.c | 12 +-- arch/riscv/kvm/mmu.c | 6 -- arch/x86/include/asm/kvm_host.h | 7 +- arch/x86/kvm/mmu/mmu.c | 25 ++---- arch/x86/kvm/mmu/mmu_internal.h | 3 - arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 20 +++-- virt/kvm/Kconfig | 3 - virt/kvm/kvm_main.c | 35 +++++++-- 21 files changed, 294 insertions(+), 131 deletions(-)
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
PATCH v7 1/12 KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
From: David Matlack <[email protected]>
Rename kvm_arch_flush_remote_tlb() and the associated macro __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB to kvm_arch_flush_remote_tlbs() and __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS respectively.
Making the name plural matches kvm_flush_remote_tlbs() and makes it more clear that this function can affect more than one remote TLB.
No functional change intended.
--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)
arch/mips/include/asm/kvm_host.h +2 -2
@@ -896,7 +896,7 @@ 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) {} -#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);#endif /* __MIPS_KVM_HOST_H__ */
arch/mips/kvm/mips.c +1 -1
@@ -981,7 +981,7 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) } -int kvm_arch_flush_remote_tlb (struct kvm *kvm) +int kvm_arch_flush_remote_tlbs (struct kvm *kvm) { kvm_mips_callbacks->prepare_flush_shadow (kvm);return 1 ;
arch/x86/include/asm/kvm_host.h +2 -2
@@ -1794,8 +1794,8 @@ static inline struct kvm *kvm_arch_alloc_vm(void)#define __KVM_HAVE_ARCH_VM_FREEvoid kvm_arch_free_vm (struct kvm *kvm); -#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -static inline int kvm_arch_flush_remote_tlb (struct kvm *kvm) +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +static inline int kvm_arch_flush_remote_tlbs (struct kvm *kvm) {if (kvm_x86_ops.flush_remote_tlbs && !static_call (kvm_x86_flush_remote_tlbs)(kvm))
include/linux/kvm_host.h +2 -2
@@ -1479,8 +1479,8 @@ static inline void kvm_arch_free_vm(struct kvm *kvm) }#endif -#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -static inline int kvm_arch_flush_remote_tlb (struct kvm *kvm) +#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +static inline int kvm_arch_flush_remote_tlbs (struct kvm *kvm) {return -ENOTSUPP ; }
virt/kvm/kvm_main.c +1 -1
@@ -361,7 +361,7 @@ void kvm_flush_remote_tlbs(struct kvm *kvm)* kvm_make_all_cpus_request() reads vcpu->mode. We reuse that * barrier here. */ -if (!kvm_arch_flush_remote_tlb (kvm) +if (!kvm_arch_flush_remote_tlbs (kvm) ||kvm_make_all_cpus_request (kvm,KVM_REQ_TLB_FLUSH )) ++kvm->stat.generic.remote_tlb_flush; }
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Marc Zyngier
maintainerRe: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 30 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Rename
kvm_arch_flush_remote_tlb()and the associated macro__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBtokvm_arch_flush_remote_tlbs()and__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSrespectively.Making the name plural matches
kvm_flush_remote_tlbs()and makes it more clear that this function can affect more than one remote TLB.No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Philippe Mathieu-Daud? <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)arch/mips/include/asm/kvm_host.h +0
-0
arch/mips/include/asm/kvm_host.h static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} · line 900
Quoted 7 lines
arch/mips/include/asm/kvm_host.h +2
-2@@ -896,7 +896,7 @@ 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) {} -#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);
How about making this prototype global? I don't see a point in having it per-architecture, specially as you are adding arm64 to that mix in the following patch.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 36 lines · On Thu, Jul 27, 2023 at 3:24?AM Marc Zyngier <[email protected]> wrote:
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Rename
kvm_arch_flush_remote_tlb()and the associated macro__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBtokvm_arch_flush_remote_tlbs()and__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSrespectively.Making the name plural matches
kvm_flush_remote_tlbs()and makes it more clear that this function can affect more than one remote TLB.No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Philippe Mathieu-Daud? <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)arch/mips/include/asm/kvm_host.h +0
-0@@ -896,7 +896,7 @@ 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) {}
Quoted 9 lines
arch/mips/include/asm/kvm_host.h static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} · line 900
Quoted 4 lines
arch/mips/include/asm/kvm_host.h +2
-2-#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);How about making this prototype global? I don't see a point in having it per-architecture, specially as you are adding arm64 to that mix in the following patch.
We can make it global, but I'm not sure what was the intention of the original author. My guess is that he was following the same style that we have for some of the other kvm_arch_*() functions (kvm_arch_free_vm() for example)?
- Raghavendra
Quoted 4 lines
M.--
Without deviation from the norm, progress is not possible.
Sean Christopherson
maintainerRe: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 46 lines · Raghavendra Rao Ananta
Quoted 36 lines · On Thu, Jul 27, 2023 at 3:24?AM Marc Zyngier <[email protected]> wrote:
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Rename
kvm_arch_flush_remote_tlb()and the associated macro__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBtokvm_arch_flush_remote_tlbs()and__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSrespectively.Making the name plural matches
kvm_flush_remote_tlbs()and makes it more clear that this function can affect more than one remote TLB.No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Philippe Mathieu-Daud? <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)arch/mips/include/asm/kvm_host.h +0
-0@@ -896,7 +896,7 @@ 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) {}Quoted 9 lines
arch/mips/include/asm/kvm_host.h static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} · line 900
Quoted 4 lines
arch/mips/include/asm/kvm_host.h +2
-2-#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);How about making this prototype global? I don't see a point in having it per-architecture, specially as you are adding arm64 to that mix in the following patch.
Quoted 4 lines
We can make it global, but I'm not sure what was the intention of the original author. My guess is that he was following the same style that we have for some of the other kvm_arch_*() functions (
kvm_arch_free_vm()for example)?
Heh, KVM has a *lot* of code that was written with questionable style. I agree with Marc, I can't think of a single reason not to have the definition in common code. Declaring the function doesn't preclude a "static inline" implementation, and we could even keep the prototype under an #ifdef, e.g.
include/linux/kvm_host.h +2 -0
@@ -1484,6 +1484,8 @@ static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) {return -ENOTSUPP ; } +#else +int kvm_arch_flush_remote_tlb (struct kvm *kvm);#endif #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
Re: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 64 lines · On Mon, Jul 31, 2023 at 2:42?PM Sean Christopherson <[email protected]> wrote:
Quoted 46 lines · Raghavendra Rao Ananta
Quoted 36 lines · On Thu, Jul 27, 2023 at 3:24?AM Marc Zyngier <[email protected]> wrote:
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Rename
kvm_arch_flush_remote_tlb()and the associated macro__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBtokvm_arch_flush_remote_tlbs()and__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSrespectively.Making the name plural matches
kvm_flush_remote_tlbs()and makes it more clear that this function can affect more than one remote TLB.No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Philippe Mathieu-Daud? <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)arch/mips/include/asm/kvm_host.h +0
-0@@ -896,7 +896,7 @@ 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) {}Quoted 9 lines
arch/mips/include/asm/kvm_host.h static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} · line 900
Quoted 4 lines
arch/mips/include/asm/kvm_host.h +2
-2-#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);How about making this prototype global? I don't see a point in having it per-architecture, specially as you are adding arm64 to that mix in the following patch.
Quoted 4 lines
We can make it global, but I'm not sure what was the intention of the original author. My guess is that he was following the same style that we have for some of the other kvm_arch_*() functions (
kvm_arch_free_vm()for example)?Heh, KVM has a *lot* of code that was written with questionable style. I agree with Marc, I can't think of a single reason not to have the definition in common code. Declaring the function doesn't preclude a "static inline" implementation, and we could even keep the prototype under an #ifdef, e.g.
include/linux/kvm_host.h +0
-0
include/linux/kvm_host.h static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) · line 1491
Quoted 8 lines
include/linux/kvm_host.h +2
-0@@ -1484,6 +1484,8 @@ static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) {return -ENOTSUPP ; } +#else +int kvm_arch_flush_remote_tlb (struct kvm *kvm);#endif #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
Thanks for the suggestions; I can go with a common declaration. Along with that, do we want to keep defining __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS in the arch code that supports it or convert it into a CONFIG_?
- Raghavendra
Marc Zyngier
maintainerRe: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 72 lines · Raghavendra Rao Ananta
Quoted 70 lines · On Mon, Jul 31, 2023 at 2:42?PM Sean Christopherson <[email protected]> wrote:
Quoted 46 lines · Raghavendra Rao Ananta
Quoted 36 lines · On Thu, Jul 27, 2023 at 3:24?AM Marc Zyngier <[email protected]> wrote:
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Rename
kvm_arch_flush_remote_tlb()and the associated macro__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBtokvm_arch_flush_remote_tlbs()and__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSrespectively.Making the name plural matches
kvm_flush_remote_tlbs()and makes it more clear that this function can affect more than one remote TLB.No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Philippe Mathieu-Daud? <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/mips/include/asm/kvm_host.h | 4 ++-- arch/mips/kvm/mips.c | 2 +- arch/x86/include/asm/kvm_host.h | 4 ++-- include/linux/kvm_host.h | 4 ++-- virt/kvm/kvm_main.c | 2 +- 5 files changed, 8 insertions(+), 8 deletions(-)arch/mips/include/asm/kvm_host.h +0
-0@@ -896,7 +896,7 @@ 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) {}Quoted 9 lines
arch/mips/include/asm/kvm_host.h static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {} · line 900
Quoted 4 lines
arch/mips/include/asm/kvm_host.h +2
-2-#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB -int kvm_arch_flush_remote_tlb (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);How about making this prototype global? I don't see a point in having it per-architecture, specially as you are adding arm64 to that mix in the following patch.
Quoted 4 lines
We can make it global, but I'm not sure what was the intention of the original author. My guess is that he was following the same style that we have for some of the other kvm_arch_*() functions (
kvm_arch_free_vm()for example)?Heh, KVM has a *lot* of code that was written with questionable style. I agree with Marc, I can't think of a single reason not to have the definition in common code. Declaring the function doesn't preclude a "static inline" implementation, and we could even keep the prototype under an #ifdef, e.g.
include/linux/kvm_host.h +2
-0@@ -1484,6 +1484,8 @@ static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) {return -ENOTSUPP ; } +#else +int kvm_arch_flush_remote_tlb (struct kvm *kvm);#endif
Quoted 6 lines
include/linux/kvm_host.h static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) · line 1491
Quoted 2 lines
include/linux/kvm_host.h +0
-0#ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMAThanks for the suggestions; I can go with a common declaration. Along with that, do we want to keep defining
__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSin the arch code that supports it or convert it into aCONFIG_?
This isn't something that a user can select, more something that is an architectural decision. Maybe in a later patch if there is a consensus around that, but probably not as part of this series.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Sean Christopherson
maintainerRe: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Quoted 10 lines · Marc Zyngier
Quoted 4 lines · Raghavendra Rao Ananta
Thanks for the suggestions; I can go with a common declaration. Along with that, do we want to keep defining
__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSin the arch code that supports it or convert it into aCONFIG_?This isn't something that a user can select, more something that is an architectural decision. Maybe in a later patch if there is a consensus around that, but probably not as part of this series.
+1. I agree it's annoying that KVM uses a mix of Kconfigs and manual #defines for the various "KVM_HAVE" knobs, but we have so many of both that one-off conversions without a real need don't make much sense.
Re: [PATCH v7 01/12] KVM: Rename kvm_arch_flush_remote_tlb() to kvm_arch_flush_remote_tlbs()
Okay, so just the #define in the respective arch header with a global declaration. I'll consider this in v8.
Thanks,
Raghavendra
Quoted 16 lines · On Wed, Aug 2, 2023 at 9:10?AM Sean Christopherson <[email protected]> wrote:
Quoted 10 lines · Marc Zyngier
Quoted 4 lines · Raghavendra Rao Ananta
Thanks for the suggestions; I can go with a common declaration. Along with that, do we want to keep defining
__KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSin the arch code that supports it or convert it into aCONFIG_?This isn't something that a user can select, more something that is an architectural decision. Maybe in a later patch if there is a consensus around that, but probably not as part of this series.
+1. I agree it's annoying that KVM uses a mix of Kconfigs and manual #defines for the various "
KVM_HAVE" knobs, but we have so many of both that one-off conversions without a real need don't make much sense.
PATCH v7 2/12 KVM: arm64: Use kvm_arch_flush_remote_tlbs()
In reply to Raghavendra Rao AnantaStop depending on CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL and opt to standardize on kvm_arch_flush_remote_tlbs() since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.
This adds an extra function call to the ARM64 kvm_flush_remote_tlbs() path, but that is a small cost in comparison to flushing remote TLBs.
In addition, instead of just incrementing remote_tlb_flush_requests stat, the generic interface would also increment the remote_tlb_flush stat.
--- arch/arm64/include/asm/kvm_host.h | 3 +++ arch/arm64/kvm/Kconfig | 1 - arch/arm64/kvm/mmu.c | 6 +++--- 3 files changed, 6 insertions(+), 4 deletions(-)
arch/arm64/include/asm/kvm_host.h +3 -0
@@ -1111,6 +1111,9 @@ int __init kvm_set_ipa_limit(void);#define __KVM_HAVE_ARCH_VM_ALLOCstruct kvm *kvm_arch_alloc_vm (void ); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm); +static inline bool kvm_vm_is_protected (struct kvm *kvm) {return false;
arch/arm64/kvm/Kconfig +0 -1
@@ -25,7 +25,6 @@ menuconfig KVM select MMU_NOTIFIER select PREEMPT_NOTIFIERS select HAVE_KVM_CPU_RELAX_INTERCEPT - select HAVE_KVM_ARCH_TLB_FLUSH_ALL select KVM_MMIO select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_XFER_TO_GUEST_WORK
arch/arm64/kvm/mmu.c +3 -3
@@ -161,15 +161,15 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) }/** -* kvm_flush_remote_tlbs() - flush all VM TLB entries for v7/8 +* kvm_arch_flush_remote_tlbs() - flush all VM TLB entries for v7/8 * @kvm: pointer to kvm structure. * * Interface to HYP function to flush all VM TLB entries */ -void kvm_flush_remote_tlbs (struct kvm *kvm) +int kvm_arch_flush_remote_tlbs (struct kvm *kvm) { - ++kvm->stat.generic.remote_tlb_flush_requests;kvm_call_hyp (__kvm_tlb_flush_vmid, &kvm->arch.mmu); +return 0 ; }static bool kvm_is_device_pfn (unsigned long pfn)
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Re: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 5 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.
Quoted 8 lines
This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.In addition, instead of just incrementing
remote_tlb_flush_requestsstat, the generic interface would also increment theremote_tlb_flushstat.Signed-off-by Raghavendra Rao Ananta <[email protected]>
Quoted 55 lines
--- arch/arm64/include/asm/kvm_host.h | 3 +++ arch/arm64/kvm/Kconfig | 1 - arch/arm64/kvm/mmu.c | 6 +++--- 3 files changed, 6 insertions(+), 4 deletions(-)arch/arm64/include/asm/kvm_host.h +3
-0@@ -1111,6 +1111,9 @@ int __init kvm_set_ipa_limit(void);#define __KVM_HAVE_ARCH_VM_ALLOCstruct kvm *kvm_arch_alloc_vm (void ); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm); +static inline bool kvm_vm_is_protected (struct kvm *kvm) {return false;arch/arm64/kvm/Kconfig +0
-1@@ -25,7 +25,6 @@ menuconfig KVM select MMU_NOTIFIER select PREEMPT_NOTIFIERS select HAVE_KVM_CPU_RELAX_INTERCEPT - select HAVE_KVM_ARCH_TLB_FLUSH_ALL select KVM_MMIO select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_XFER_TO_GUEST_WORKarch/arm64/kvm/mmu.c +3
-3@@ -161,15 +161,15 @@ static bool memslot_is_logging(struct kvm_memory_slot *memslot) }/** -* kvm_flush_remote_tlbs() - flush all VM TLB entries for v7/8 +* kvm_arch_flush_remote_tlbs() - flush all VM TLB entries for v7/8 * @kvm: pointer to kvm structure. * * Interface to HYP function to flush all VM TLB entries */ -void kvm_flush_remote_tlbs (struct kvm *kvm) +int kvm_arch_flush_remote_tlbs (struct kvm *kvm) { - ++kvm->stat.generic.remote_tlb_flush_requests;kvm_call_hyp (__kvm_tlb_flush_vmid, &kvm->arch.mmu); +return 0 ; }static bool kvm_is_device_pfn (unsigned long pfn)
Signature · 1 line
Shaoqin
Marc Zyngier
maintainerRe: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
In reply to Raghavendra Rao AnantaQuoted 8 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.
Well, there is no such thing as a "remote TLB" anyway. We either have a non-shareable or inner-shareable invalidation. The notion of remote would imply that we track who potentially has a TLB, which we obviously don't.
More x86 nonsense...
Quoted 17 lines
In addition, instead of just incrementing
remote_tlb_flush_requestsstat, the generic interface would also increment theremote_tlb_flushstat.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 3 +++ arch/arm64/kvm/Kconfig | 1 - arch/arm64/kvm/mmu.c | 6 +++--- 3 files changed, 6 insertions(+), 4 deletions(-)arch/arm64/include/asm/kvm_host.h +0
-0
arch/arm64/include/asm/kvm_host.h int __init kvm_set_ipa_limit(void); · line 1115
Quoted 5 lines
arch/arm64/include/asm/kvm_host.h +2
-0@@ -1111,6 +1111,9 @@ int __init kvm_set_ipa_limit(void);#define __KVM_HAVE_ARCH_VM_ALLOCstruct kvm *kvm_arch_alloc_vm (void ); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS +int kvm_arch_flush_remote_tlbs (struct kvm *kvm);
See my earlier comment about making this prototype global.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Sean Christopherson
maintainerRe: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 8 lines · Marc Zyngier
Quoted 6 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.
Quoted 7 lines
Quoted 2 lines
This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.Well, there is no such thing as a "remote TLB" anyway. We either have a non-shareable or inner-shareable invalidation. The notion of remote would imply that we track who potentially has a TLB, which we obviously don't.
Maybe kvm_arch_flush_vm_tlbs()? The "remote" part is misleading even on x86 when running on Hyper-V, as the flush may be done via a single hypercall and by kicking "remote" vCPUs.
Marc Zyngier
maintainerRe: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 13 lines · Sean Christopherson
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.
Quoted 8 lines
Quoted 4 lines
Well, there is no such thing as a "remote TLB" anyway. We either have a non-shareable or inner-shareable invalidation. The notion of remote would imply that we track who potentially has a TLB, which we obviously don't.
Maybe
kvm_arch_flush_vm_tlbs()? The "remote" part is misleading even on x86 when running on Hyper-V, as the flush may be done via a single hypercall and by kicking "remote" vCPUs.
Yup, this would be much better.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Sure, I'll change it to kvm_arch_flush_vm_tlbs() in v8.
Thanks,
Raghavendra
Quoted 33 lines · On Wed, Aug 2, 2023 at 8:55?AM Marc Zyngier <[email protected]> wrote:
Quoted 13 lines · Sean Christopherson
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.Quoted 8 lines
Quoted 4 lines
Well, there is no such thing as a "remote TLB" anyway. We either have a non-shareable or inner-shareable invalidation. The notion of remote would imply that we track who potentially has a TLB, which we obviously don't.
Maybe
kvm_arch_flush_vm_tlbs()? The "remote" part is misleading even on x86 when running on Hyper-V, as the flush may be done via a single hypercall and by kicking "remote" vCPUs.Yup, this would be much better.
Thanks,
M.--
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 3 lines · On Wed, Aug 2, 2023 at 4:28?PM Raghavendra Rao Ananta <[email protected]>…
Sure, I'll change it to
kvm_arch_flush_vm_tlbs()in v8.
While working on the renaming, I realized that since this function is called from kvm_main.c's kvm_flush_remote_tlbs(). Do we want to rename this and the other kvm_flush_*() functions that the series introduces to match their kvm_arch_flush_*() counterparts? (spiraling more into this, we also have the 'remote_tlb_flush_requests' and 'remote_tlb_flush' stats)
Thank you.
Raghavendra
Quoted 37 lines
Thanks,
RaghavendraQuoted 33 lines · On Wed, Aug 2, 2023 at 8:55?AM Marc Zyngier <[email protected]> wrote:
Quoted 13 lines · Sean Christopherson
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · Raghavendra Rao Ananta
Stop depending on
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLand opt to standardize onkvm_arch_flush_remote_tlbs()since it avoids duplicating the generic TLB stats across architectures that implement their own remote TLB flush.This adds an extra function call to the ARM64
kvm_flush_remote_tlbs()path, but that is a small cost in comparison to flushing remote TLBs.Quoted 8 lines
Quoted 4 lines
Well, there is no such thing as a "remote TLB" anyway. We either have a non-shareable or inner-shareable invalidation. The notion of remote would imply that we track who potentially has a TLB, which we obviously don't.
Maybe
kvm_arch_flush_vm_tlbs()? The "remote" part is misleading even on x86 when running on Hyper-V, as the flush may be done via a single hypercall and by kicking "remote" vCPUs.Yup, this would be much better.
Thanks,
M.--
Without deviation from the norm, progress is not possible.
Sean Christopherson
maintainerRe: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 9 lines · Raghavendra Rao Ananta
Quoted 3 lines · On Wed, Aug 2, 2023 at 4:28?PM Raghavendra Rao Ananta <[email protected]>…
Sure, I'll change it to
kvm_arch_flush_vm_tlbs()in v8.While working on the renaming, I realized that since this function is called from kvm_main.c's
kvm_flush_remote_tlbs(). Do we want to rename this and the other kvm_flush_*() functions that the series introduces to match their kvm_arch_flush_*() counterparts?
Hmm, if we're going to rename one arch hook, then yes, I think it makes sense to rename all the common APIs and arch hooks to match.
However, x86 is rife with the "remote_tlbs" nomenclature, and renaming the common APIs will just push the inconsistencies into x86. While I 100% agree that the current naming is flawed, I am not willing to end up with x86 being partially converted.
I think I'm ok renaming all of x86's many hooks? But I'd definitely want input from more x86 folks, and the size and scope of this series would explode. Unless Marc objects and/or has a better idea, the least awful option is probably to ignore the poor "remote_tlbs" naming and tackle it in a separate series.
Sorry for not noticiing this earlier, I didn't realize just how much x86 uses remote_tlbs.
Quoted 2 lines
(spiraling more into this, we also have the '
remote_tlb_flush_requests' and 'remote_tlb_flush' stats)
Regardless of what we decide for the APIs, definitely leave the stats alone. The names are ABI. We could preserve the names and changes the struct fields, but that would be a net negative IMO.
Re: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 20 lines · On Tue, Aug 8, 2023 at 8:07?AM Sean Christopherson <[email protected]> wrote:
Quoted 9 lines · Raghavendra Rao Ananta
Quoted 3 lines · On Wed, Aug 2, 2023 at 4:28?PM Raghavendra Rao Ananta <[email protected]>…
Sure, I'll change it to
kvm_arch_flush_vm_tlbs()in v8.While working on the renaming, I realized that since this function is called from kvm_main.c's
kvm_flush_remote_tlbs(). Do we want to rename this and the other kvm_flush_*() functions that the series introduces to match their kvm_arch_flush_*() counterparts?Hmm, if we're going to rename one arch hook, then yes, I think it makes sense to rename all the common APIs and arch hooks to match.
However, x86 is rife with the "
remote_tlbs" nomenclature, and renaming the common APIs will just push the inconsistencies into x86. While I 100% agree that the current naming is flawed, I am not willing to end up with x86 being partially converted.
Quoted 5 lines
I think I'm ok renaming all of x86's many hooks? But I'd definitely want input from more x86 folks, and the size and scope of this series would explode. Unless Marc objects and/or has a better idea, the least awful option is probably to ignore the poor "
remote_tlbs" naming and tackle it in a separate series.
Sure, I think it's better to do it in a separate series as well. I'm happy to carry out the task after this one gets merged. But, let's wait for Marc and others' opinion on the matter.
Thank you.
Raghavendra
Quoted 9 lines
Sorry for not noticiing this earlier, I didn't realize just how much x86 uses
remote_tlbs.Quoted 2 lines
(spiraling more into this, we also have the '
remote_tlb_flush_requests' and 'remote_tlb_flush' stats)Regardless of what we decide for the APIs, definitely leave the stats alone. The names are ABI. We could preserve the names and changes the struct fields, but that would be a net negative IMO.
Marc Zyngier
maintainerRe: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 34 lines · Raghavendra Rao Ananta
Quoted 24 lines · On Tue, Aug 8, 2023 at 8:07?AM Sean Christopherson <[email protected]> wrote:
Quoted 9 lines · Raghavendra Rao Ananta
Quoted 3 lines · On Wed, Aug 2, 2023 at 4:28?PM Raghavendra Rao Ananta <[email protected]>…
Sure, I'll change it to
kvm_arch_flush_vm_tlbs()in v8.While working on the renaming, I realized that since this function is called from kvm_main.c's
kvm_flush_remote_tlbs(). Do we want to rename this and the other kvm_flush_*() functions that the series introduces to match their kvm_arch_flush_*() counterparts?Hmm, if we're going to rename one arch hook, then yes, I think it makes sense to
rename all the common APIs and arch hooks to match.However, x86 is rife with the "
remote_tlbs" nomenclature, and renaming the common
APIs will just push the inconsistencies into x86. While I 100% agree that the
current naming is flawed, I am not willing to end up with x86 being partially
converted.Quoted 8 lines
I think I'm ok renaming all of x86's many hooks? But I'd definitely want input
from more x86 folks, and the size and scope of this series would explode. Unless
Marc objects and/or has a better idea, the least awful option is probably to ignore
the poor "remote_tlbs" naming and tackle it in a separate series.
Quoted 3 lines
Sure, I think it's better to do it in a separate series as well. I'm happy to carry out the task after this one gets merged. But, let's wait for Marc and others' opinion on the matter.
Yeah, let's punt that to a separate series. I'm more interested in getting this code merged than in the inevitable bike-shedding that will result from such a proposal.
Raghavendra, any chance you could respin the series this week? I'd really like it to spend some quality time in -next...
Thanks,
M.
Signature · 1 line
Jazz is not dead. It just smells funny...
Re: [PATCH v7 02/12] KVM: arm64: Use kvm_arch_flush_remote_tlbs()
Quoted 40 lines · On Tue, Aug 8, 2023 at 9:43?AM Marc Zyngier <[email protected]> wrote:
Quoted 34 lines · Raghavendra Rao Ananta
Quoted 24 lines · On Tue, Aug 8, 2023 at 8:07?AM Sean Christopherson <[email protected]> wrote:
Quoted 9 lines · Raghavendra Rao Ananta
Quoted 3 lines · On Wed, Aug 2, 2023 at 4:28?PM Raghavendra Rao Ananta <[email protected]>…
Sure, I'll change it to
kvm_arch_flush_vm_tlbs()in v8.While working on the renaming, I realized that since this function is called from kvm_main.c's
kvm_flush_remote_tlbs(). Do we want to rename this and the other kvm_flush_*() functions that the series introduces to match their kvm_arch_flush_*() counterparts?Hmm, if we're going to rename one arch hook, then yes, I think it makes sense to
rename all the common APIs and arch hooks to match.However, x86 is rife with the "
remote_tlbs" nomenclature, and renaming the common
APIs will just push the inconsistencies into x86. While I 100% agree that the
current naming is flawed, I am not willing to end up with x86 being partially
converted.Quoted 8 lines
I think I'm ok renaming all of x86's many hooks? But I'd definitely want input
from more x86 folks, and the size and scope of this series would explode. Unless
Marc objects and/or has a better idea, the least awful option is probably to ignore
the poor "remote_tlbs" naming and tackle it in a separate series.Quoted 3 lines
Sure, I think it's better to do it in a separate series as well. I'm happy to carry out the task after this one gets merged. But, let's wait for Marc and others' opinion on the matter.
Quoted 7 lines
Yeah, let's punt that to a separate series. I'm more interested in getting this code merged than in the inevitable bike-shedding that will result from such a proposal.
Raghavendra, any chance you could respin the series this week? I'd really like it to spend some quality time in -next...
No problem. I'll send out v8 by today or tomorrow.
Thank you.
Raghavendra
Quoted 5 lines
Thanks,
M.--
Jazz is not dead. It just smells funny...
PATCH v7 3/12 KVM: Remove CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
In reply to Raghavendra Rao Anantakvm_arch_flush_remote_tlbs() or CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL are two mechanisms to solve the same problem, allowing architecture-specific code to provide a non-IPI implementation of remote TLB flushing.
Dropping CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL allows KVM to standardize all architectures on kvm_arch_flush_remote_tlbs() instead of maintaining two mechanisms.
--- virt/kvm/Kconfig | 3 --- virt/kvm/kvm_main.c | 2 -- 2 files changed, 5 deletions(-)
virt/kvm/Kconfig +0 -3
@@ -62,9 +62,6 @@ config HAVE_KVM_CPU_RELAX_INTERCEPT config KVM_VFIO bool -config HAVE_KVM_ARCH_TLB_FLUSH_ALL - bool - config HAVE_KVM_INVALID_WAKEUPS bool
virt/kvm/kvm_main.c +0 -2
@@ -345,7 +345,6 @@ bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req) }EXPORT_SYMBOL_GPL (kvm_make_all_cpus_request); -#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL void kvm_flush_remote_tlbs (struct kvm *kvm) { ++kvm->stat.generic.remote_tlb_flush_requests;
@@ -366,7 +365,6 @@ void kvm_flush_remote_tlbs(struct kvm *kvm) ++kvm->stat.generic.remote_tlb_flush; }EXPORT_SYMBOL_GPL (kvm_flush_remote_tlbs); -#endif static void kvm_flush_shadow_all (struct kvm *kvm) {
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Re: [PATCH v7 03/12] KVM: Remove CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL
Quoted 10 lines · Raghavendra Rao Ananta
kvm_arch_flush_remote_tlbs()orCONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLare two mechanisms to solve the same problem, allowing architecture-specific code to provide a non-IPI implementation of remote TLB flushing.Dropping
CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALLallows KVM to standardize all architectures onkvm_arch_flush_remote_tlbs()instead of maintaining two mechanisms.Signed-off-by Raghavendra Rao Ananta <[email protected]>
Quoted 39 lines
--- virt/kvm/Kconfig | 3 --- virt/kvm/kvm_main.c | 2 -- 2 files changed, 5 deletions(-)virt/kvm/Kconfig +0
-3@@ -62,9 +62,6 @@ config HAVE_KVM_CPU_RELAX_INTERCEPT config KVM_VFIO bool -config HAVE_KVM_ARCH_TLB_FLUSH_ALL - bool - config HAVE_KVM_INVALID_WAKEUPS boolvirt/kvm/kvm_main.c +0
-2@@ -345,7 +345,6 @@ bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req) }EXPORT_SYMBOL_GPL (kvm_make_all_cpus_request); -#ifndef CONFIG_HAVE_KVM_ARCH_TLB_FLUSH_ALL void kvm_flush_remote_tlbs (struct kvm *kvm) { ++kvm->stat.generic.remote_tlb_flush_requests;@@ -366,7 +365,6 @@ void kvm_flush_remote_tlbs(struct kvm *kvm) ++kvm->stat.generic.remote_tlb_flush; }EXPORT_SYMBOL_GPL (kvm_flush_remote_tlbs); -#endif static void kvm_flush_shadow_all (struct kvm *kvm) {
Signature · 1 line
Shaoqin
PATCH v7 4/12 KVM: Allow range-based TLB invalidation from common code
In reply to Raghavendra Rao AnantaFrom: David Matlack <[email protected]>
Make kvm_flush_remote_tlbs_range() visible in common code and create a default implementation that just invalidates the whole TLB.
This paves the way for several future features/cleanups:
- Introduction of range-based TLBI on ARM.
- Eliminating kvm_arch_flush_remote_tlbs_memslot()
- Moving the KVM/x86 TDP MMU to common code.
No functional change intended.
--- arch/x86/include/asm/kvm_host.h | 3 +++ arch/x86/kvm/mmu/mmu.c | 9 ++++----- arch/x86/kvm/mmu/mmu_internal.h | 3 --- include/linux/kvm_host.h | 9 +++++++++ virt/kvm/kvm_main.c | 13 +++++++++++++ 5 files changed, 29 insertions(+), 8 deletions(-)
arch/x86/include/asm/kvm_host.h +3 -0
@@ -1804,6 +1804,9 @@ static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm)return -ENOTSUPP ; } +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages); +#define kvm_arch_pmi_in_guest (vcpu) \ ((vcpu) && (vcpu)->arch.handling_intr_from_guest)
arch/x86/kvm/mmu/mmu.c +4 -5
@@ -278,16 +278,15 @@ static inline bool kvm_available_flush_remote_tlbs_range(void)return kvm_x86_ops.flush_remote_tlbs_range; } -void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages) +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages) {int ret = -EOPNOTSUPP ;if (kvm_x86_ops.flush_remote_tlbs_range) ret =static_call (kvm_x86_flush_remote_tlbs_range)(kvm, start_gfn, - nr_pages); -if (ret) -kvm_flush_remote_tlbs (kvm); + pages); + +return ret; }static gfn_t kvm_mmu_page_get_gfn (struct kvm_mmu_page *sp,int index);
arch/x86/kvm/mmu/mmu_internal.h +0 -3
@@ -170,9 +170,6 @@ bool kvm_mmu_slot_gfn_write_protect(struct kvm *kvm, struct kvm_memory_slot *slot, u64 gfn, int min_level); -void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages); -/* Flush the given page (huge or not) of guest memory. */ static inline void kvm_flush_remote_tlbs_gfn (struct kvm *kvm,gfn_t gfn,int level) {
include/linux/kvm_host.h +9 -0
@@ -1359,6 +1359,7 @@ int kvm_vcpu_yield_to(struct kvm_vcpu *target);void kvm_vcpu_on_spin (struct kvm_vcpu *vcpu,bool yield_to_kernel_mode);void kvm_flush_remote_tlbs (struct kvm *kvm); +void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages);#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE int kvm_mmu_topup_memory_cache (struct kvm_mmu_memory_cache *mc,int min);
@@ -1486,6 +1487,14 @@ static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm) }#endif +#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +static inline int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm, +gfn_t gfn,u64 pages) +{ +return -EOPNOTSUPP ; +} +#endif +#ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMAvoid kvm_arch_register_noncoherent_dma (struct kvm *kvm);void kvm_arch_unregister_noncoherent_dma (struct kvm *kvm);
virt/kvm/kvm_main.c +13 -0
@@ -366,6 +366,19 @@ void kvm_flush_remote_tlbs(struct kvm *kvm) }EXPORT_SYMBOL_GPL (kvm_flush_remote_tlbs); +void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages) +{ +if (!kvm_arch_flush_remote_tlbs_range (kvm, gfn, pages)) +return ; + +/* +* Fall back to a flushing entire TLBs if the architecture range-based +* TLB invalidation is unsupported or can't be performed for whatever +* reason. +*/ +kvm_flush_remote_tlbs (kvm); +} +static void kvm_flush_shadow_all (struct kvm *kvm) {kvm_arch_flush_shadow_all (kvm);
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Sean Christopherson
maintainerRe: [PATCH v7 04/12] KVM: Allow range-based TLB invalidation from common code
arch/x86/kvm/mmu/mmu.c static inline bool kvm_available_flush_remote_tlbs_range(void) · line 281
Quoted 11 lines · Raghavendra Rao Ananta
arch/x86/kvm/mmu/mmu.c +1
-2@@ -278,16 +278,15 @@ static inline bool kvm_available_flush_remote_tlbs_range(void)return kvm_x86_ops.flush_remote_tlbs_range; } -void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages) +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages)
Please keep "nr_pages", I have a very strong preference for that over just "pages" as the "nr_" makes it super obvious that it's a single number, as opposed to an array of pages or something.
And it doesn't truly matter, but IMO the gfn_t type is more appropriate since the gfn and the number of pages need to have the same type.
Re: [PATCH v7 04/12] KVM: Allow range-based TLB invalidation from common code
Quoted 10 lines · On Mon, Jul 31, 2023 at 2:55?PM Sean Christopherson <[email protected]> wrote:
Quoted 8 lines · Raghavendra Rao Ananta
arch/x86/kvm/mmu/mmu.c +0
-0@@ -278,16 +278,15 @@ static inline bool kvm_available_flush_remote_tlbs_range(void)return kvm_x86_ops.flush_remote_tlbs_range; }
Quoted 8 lines
arch/x86/kvm/mmu/mmu.c static inline bool kvm_available_flush_remote_tlbs_range(void) · line 281
Quoted 3 lines
arch/x86/kvm/mmu/mmu.c +1
-2-void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages) +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages)Please keep "
nr_pages", I have a very strong preference for that over just "pages" as the "nr_" makes it super obvious that it's a single number, as opposed to an array of pages or something.
Sure, I'll revert back to 'nr_pages'.
- Raghavendra
Quoted 2 lines
And it doesn't truly matter, but IMO the
gfn_ttype is more appropriate since the gfn and the number of pages need to have the same type.
Re: [PATCH v7 04/12] KVM: Allow range-based TLB invalidation from common code
In reply to Raghavendra Rao AnantaQuoted 12 lines · On Sat, Jul 22, 2023 at 7:52?AM Raghavendra Rao Ananta <[email protected]>…
From: David Matlack <[email protected]>
Make
kvm_flush_remote_tlbs_range()visible in common code and create a default implementation that just invalidates the whole TLB.This paves the way for several future features/cleanups:
- Introduction of range-based TLBI on ARM.
- Eliminatingkvm_arch_flush_remote_tlbs_memslot()
- Moving the KVM/x86 TDP MMU to common code.
Quoted 6 lines
No functional change intended.
Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>
For KVM RISC-V:
Regards,
Anup
Quoted 115 lines
--- arch/x86/include/asm/kvm_host.h | 3 +++ arch/x86/kvm/mmu/mmu.c | 9 ++++----- arch/x86/kvm/mmu/mmu_internal.h | 3 --- include/linux/kvm_host.h | 9 +++++++++ virt/kvm/kvm_main.c | 13 +++++++++++++ 5 files changed, 29 insertions(+), 8 deletions(-)arch/x86/include/asm/kvm_host.h +3
-0@@ -1804,6 +1804,9 @@ static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm)return -ENOTSUPP ; } +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages); +#define kvm_arch_pmi_in_guest (vcpu) \ ((vcpu) && (vcpu)->arch.handling_intr_from_guest)arch/x86/kvm/mmu/mmu.c +4
-5@@ -278,16 +278,15 @@ static inline bool kvm_available_flush_remote_tlbs_range(void)return kvm_x86_ops.flush_remote_tlbs_range; } -void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages) +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages) {int ret = -EOPNOTSUPP ;if (kvm_x86_ops.flush_remote_tlbs_range) ret =static_call (kvm_x86_flush_remote_tlbs_range)(kvm, start_gfn, - nr_pages); -if (ret) -kvm_flush_remote_tlbs (kvm); + pages); + +return ret; }static gfn_t kvm_mmu_page_get_gfn (struct kvm_mmu_page *sp,int index);arch/x86/kvm/mmu/mmu_internal.h +0
-3@@ -170,9 +170,6 @@ bool kvm_mmu_slot_gfn_write_protect(struct kvm *kvm, struct kvm_memory_slot *slot, u64 gfn, int min_level); -void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn, -gfn_t nr_pages); -/* Flush the given page (huge or not) of guest memory. */ static inline void kvm_flush_remote_tlbs_gfn (struct kvm *kvm,gfn_t gfn,int level) {include/linux/kvm_host.h +9
-0@@ -1359,6 +1359,7 @@ int kvm_vcpu_yield_to(struct kvm_vcpu *target);void kvm_vcpu_on_spin (struct kvm_vcpu *vcpu,bool yield_to_kernel_mode);void kvm_flush_remote_tlbs (struct kvm *kvm); +void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages);#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE int kvm_mmu_topup_memory_cache (struct kvm_mmu_memory_cache *mc,int min);@@ -1486,6 +1487,14 @@ static inline int kvm_arch_flush_remote_tlbs(struct kvm *kvm) }#endif +#ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +static inline int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm, +gfn_t gfn,u64 pages) +{ +return -EOPNOTSUPP ; +} +#endif +#ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMAvoid kvm_arch_register_noncoherent_dma (struct kvm *kvm);void kvm_arch_unregister_noncoherent_dma (struct kvm *kvm);virt/kvm/kvm_main.c +13
-1@@ -366,6 +366,19 @@ void kvm_flush_remote_tlbs(struct kvm *kvm) }EXPORT_SYMBOL_GPL (kvm_flush_remote_tlbs); +void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages) +{ +if (!kvm_arch_flush_remote_tlbs_range (kvm, gfn, pages)) +return ; + +/* +* Fall back to a flushing entire TLBs if the architecture range-based +* TLB invalidation is unsupported or can't be performed for whatever +* reason. +*/ +kvm_flush_remote_tlbs (kvm); +} +static void kvm_flush_shadow_all (struct kvm *kvm) {kvm_arch_flush_shadow_all (kvm); --2.41.0.487.g6d72f3e995-goog
PATCH v7 5/12 KVM: Move kvm_arch_flush_remote_tlbs_memslot() to common code
In reply to Raghavendra Rao AnantaFrom: David Matlack <[email protected]>
Move kvm_arch_flush_remote_tlbs_memslot() to common code and drop "arch_" from the name. kvm_arch_flush_remote_tlbs_memslot() is just a range-based TLB invalidation where the range is defined by the memslot. Now that kvm_flush_remote_tlbs_range() can be called from common code we can just use that and drop a bunch of duplicate code from the arch directories.
Note this adds a lockdep assertion for slots_lock being held when calling kvm_flush_remote_tlbs_memslot(), which was previously only asserted on x86. MIPS has calls to kvm_flush_remote_tlbs_memslot(), but they all hold the slots_lock, so the lockdep assertion continues to hold true.
Also drop the CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT ifdef gating kvm_flush_remote_tlbs_memslot(), since it is no longer necessary.
--- arch/arm64/kvm/arm.c | 6 ------ arch/mips/kvm/mips.c | 10 ++-------- arch/riscv/kvm/mmu.c | 6 ------ arch/x86/kvm/mmu/mmu.c | 16 +--------------- arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 7 +++---- virt/kvm/kvm_main.c | 18 ++++++++++++++++-- 7 files changed, 23 insertions(+), 42 deletions(-)
arch/arm64/kvm/arm.c +0 -6
@@ -1525,12 +1525,6 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -static int kvm_vm_ioctl_set_device_addr (struct kvm *kvm,struct kvm_arm_device_addr *dev_addr) {
arch/mips/kvm/mips.c +2 -8
@@ -199,7 +199,7 @@ void kvm_arch_flush_shadow_memslot(struct kvm *kvm,/* Flush slot from GPA */ kvm_mips_flush_gpa_pt (kvm, slot->base_gfn, slot->base_gfn + slot->npages -1 ); -kvm_arch_flush_remote_tlbs_memslot (kvm, slot); +kvm_flush_remote_tlbs_memslot (kvm, slot);spin_unlock (&kvm->mmu_lock); }
@@ -235,7 +235,7 @@ void kvm_arch_commit_memory_region(struct kvm *kvm, needs_flush =kvm_mips_mkclean_gpa_pt (kvm, new->base_gfn, new->base_gfn + new->npages -1 );if (needs_flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, new); +kvm_flush_remote_tlbs_memslot (kvm, new);spin_unlock (&kvm->mmu_lock); } }
@@ -987,12 +987,6 @@ int kvm_arch_flush_remote_tlbs(struct kvm *kvm)return 1 ; } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -int kvm_arch_vm_ioctl (struct file *filp,unsigned int ioctl,unsigned long arg) {int r;
arch/riscv/kvm/mmu.c +0 -6
@@ -406,12 +406,6 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) { } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -void kvm_arch_free_memslot (struct kvm *kvm,struct kvm_memory_slot *free) { }
arch/x86/kvm/mmu/mmu.c +1 -15
@@ -6669,7 +6669,7 @@ static void kvm_rmap_zap_collapsible_sptes(struct kvm *kvm,*/ if (walk_slot_rmaps (kvm, slot, kvm_mmu_zap_collapsible_spte,PG_LEVEL_4K ,KVM_MAX_HUGEPAGE_LEVEL -1 , true)) -kvm_arch_flush_remote_tlbs_memslot (kvm, slot); +kvm_flush_remote_tlbs_memslot (kvm, slot); } void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,
@@ -6688,20 +6688,6 @@ void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm, } } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -/* -* All current use cases for flushing the TLBs for a specific memslot -* related to dirty logging, and many do the TLB flush out of mmu_lock. -* The interaction between the various operations on memslot must be -* serialized by slots_locks to ensure the TLB flush from one operation -* is observed by any other operation on the same memslot. -*/ -lockdep_assert_held (&kvm->slots_lock); -kvm_flush_remote_tlbs_range (kvm, memslot->base_gfn, memslot->npages); -} -void kvm_mmu_slot_leaf_clear_dirty (struct kvm *kvm,const struct kvm_memory_slot *memslot) {
arch/x86/kvm/x86.c +1 -1
@@ -12751,7 +12751,7 @@ static void kvm_mmu_slot_apply_flags(struct kvm *kvm,* See is_writable_pte() for more details (the case involving * access-tracked SPTEs is particularly relevant). */ -kvm_arch_flush_remote_tlbs_memslot (kvm, new); +kvm_flush_remote_tlbs_memslot (kvm, new); } }
include/linux/kvm_host.h +3 -4
@@ -1360,6 +1360,8 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool yield_to_kernel_mode);void kvm_flush_remote_tlbs (struct kvm *kvm);void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages); +void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot);#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE int kvm_mmu_topup_memory_cache (struct kvm_mmu_memory_cache *mc,int min);
@@ -1388,10 +1390,7 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm, unsigned long mask);void kvm_arch_sync_dirty_log (struct kvm *kvm,struct kvm_memory_slot *memslot); -#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot); -#else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */ +#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT int kvm_vm_ioctl_get_dirty_log (struct kvm *kvm,struct kvm_dirty_log *log);int kvm_get_dirty_log (struct kvm *kvm,struct kvm_dirty_log *log,int *is_dirty,struct kvm_memory_slot **memslot);
virt/kvm/kvm_main.c +16 -2
@@ -379,6 +379,20 @@ void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages)kvm_flush_remote_tlbs (kvm); } +void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot) +{ +/* +* All current use cases for flushing the TLBs for a specific memslot +* related to dirty logging, and many do the TLB flush out of mmu_lock. +* The interaction between the various operations on memslot must be +* serialized by slots_locks to ensure the TLB flush from one operation +* is observed by any other operation on the same memslot. +*/ +lockdep_assert_held (&kvm->slots_lock); +kvm_flush_remote_tlbs_range (kvm, memslot->base_gfn, memslot->npages); +} +static void kvm_flush_shadow_all (struct kvm *kvm) {kvm_arch_flush_shadow_all (kvm);
@@ -2191,7 +2205,7 @@ static int kvm_get_dirty_log_protect(struct kvm *kvm, struct kvm_dirty_log *log) }if (flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, memslot); +kvm_flush_remote_tlbs_memslot (kvm, memslot);if (copy_to_user (log->dirty_bitmap, dirty_bitmap_buffer, n))return -EFAULT ;
@@ -2308,7 +2322,7 @@ static int kvm_clear_dirty_log_protect(struct kvm *kvm,KVM_MMU_UNLOCK (kvm);if (flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, memslot); +kvm_flush_remote_tlbs_memslot (kvm, memslot);return 0 ; }
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Marc Zyngier
maintainerRe: [PATCH v7 05/12] KVM: Move kvm_arch_flush_remote_tlbs_memslot() to common code
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Move
kvm_arch_flush_remote_tlbs_memslot()to common code and drop "arch_" from the name.kvm_arch_flush_remote_tlbs_memslot()is just a range-based TLB invalidation where the range is defined by the memslot. Now thatkvm_flush_remote_tlbs_range()can be called from common code we can just use that and drop a bunch of duplicate code from the arch directories.Note this adds a lockdep assertion for
slots_lockbeing held when callingkvm_flush_remote_tlbs_memslot(), which was previously only asserted on x86. MIPS has calls tokvm_flush_remote_tlbs_memslot(), but they all hold theslots_lock, so the lockdep assertion continues to hold true.Also drop the
CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECTifdef gatingkvm_flush_remote_tlbs_memslot(), since it is no longer necessary.Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/kvm/arm.c | 6 ------ arch/mips/kvm/mips.c | 10 ++-------- arch/riscv/kvm/mmu.c | 6 ------ arch/x86/kvm/mmu/mmu.c | 16 +--------------- arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 7 +++---- virt/kvm/kvm_main.c | 18 ++++++++++++++++-- 7 files changed, 23 insertions(+), 42 deletions(-)
[...]
Quoted 5 lines
virt/kvm/kvm_main.c +0
-0
virt/kvm/kvm_main.c void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages) · line 387
Quoted 9 lines
virt/kvm/kvm_main.c +6
-0@@ -379,6 +379,20 @@ void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages)kvm_flush_remote_tlbs (kvm); } +void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot) +{ +/* +* All current use cases for flushing the TLBs for a specific memslot +* related to dirty logging, and many do the TLB flush out of mmu_lock.
I appreciate this is a copy paste of an existing comment, but I can't parse it. My command of the English language is notoriously approximate, but it feels that something is missing in the first sentence, such as a verb.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 05/12] KVM: Move kvm_arch_flush_remote_tlbs_memslot() to common code
Quoted 47 lines · On Thu, Jul 27, 2023 at 3:53?AM Marc Zyngier <[email protected]> wrote:
Quoted 33 lines · Raghavendra Rao Ananta
From: David Matlack <[email protected]>
Move
kvm_arch_flush_remote_tlbs_memslot()to common code and drop "arch_" from the name.kvm_arch_flush_remote_tlbs_memslot()is just a range-based TLB invalidation where the range is defined by the memslot. Now thatkvm_flush_remote_tlbs_range()can be called from common code we can just use that and drop a bunch of duplicate code from the arch directories.Note this adds a lockdep assertion for
slots_lockbeing held when callingkvm_flush_remote_tlbs_memslot(), which was previously only asserted on x86. MIPS has calls tokvm_flush_remote_tlbs_memslot(), but they all hold theslots_lock, so the lockdep assertion continues to hold true.Also drop the
CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECTifdef gatingkvm_flush_remote_tlbs_memslot(), since it is no longer necessary.Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/kvm/arm.c | 6 ------ arch/mips/kvm/mips.c | 10 ++-------- arch/riscv/kvm/mmu.c | 6 ------ arch/x86/kvm/mmu/mmu.c | 16 +--------------- arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 7 +++---- virt/kvm/kvm_main.c | 18 ++++++++++++++++-- 7 files changed, 23 insertions(+), 42 deletions(-)[...]
Quoted 8 lines
virt/kvm/kvm_main.c +0
-0@@ -379,6 +379,20 @@ void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages)kvm_flush_remote_tlbs (kvm); }
Quoted 12 lines
virt/kvm/kvm_main.c void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages) · line 387
Quoted 6 lines
virt/kvm/kvm_main.c +6
-0+void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot) +{ +/* +* All current use cases for flushing the TLBs for a specific memslot +* related to dirty logging, and many do the TLB flush out of mmu_lock. I appreciate this is a copy paste of an existing comment, but I can't parse it. My command of the English language is notoriously approximate, but it feels that something is missing in the first sentence, such as a verb.
No, you are right. The sentence is broken, probably a missing "are" at the end of the first line. I'll fix it.
- Raghavendra
Quoted 4 lines
M.--
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 05/12] KVM: Move kvm_arch_flush_remote_tlbs_memslot() to common code
In reply to Raghavendra Rao AnantaQuoted 16 lines · On Sat, Jul 22, 2023 at 7:53?AM Raghavendra Rao Ananta <[email protected]>…
From: David Matlack <[email protected]>
Move
kvm_arch_flush_remote_tlbs_memslot()to common code and drop "arch_" from the name.kvm_arch_flush_remote_tlbs_memslot()is just a range-based TLB invalidation where the range is defined by the memslot. Now thatkvm_flush_remote_tlbs_range()can be called from common code we can just use that and drop a bunch of duplicate code from the arch directories.Note this adds a lockdep assertion for
slots_lockbeing held when callingkvm_flush_remote_tlbs_memslot(), which was previously only asserted on x86. MIPS has calls tokvm_flush_remote_tlbs_memslot(), but they all hold theslots_lock, so the lockdep assertion continues to hold true.
Quoted 7 lines
Also drop the
CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECTifdef gatingkvm_flush_remote_tlbs_memslot(), since it is no longer necessary.Signed-off-by David Matlack <[email protected]>Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>
For KVM RISC-V:
Thanks,
Anup
Quoted 197 lines
--- arch/arm64/kvm/arm.c | 6 ------ arch/mips/kvm/mips.c | 10 ++-------- arch/riscv/kvm/mmu.c | 6 ------ arch/x86/kvm/mmu/mmu.c | 16 +--------------- arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 7 +++---- virt/kvm/kvm_main.c | 18 ++++++++++++++++-- 7 files changed, 23 insertions(+), 42 deletions(-)arch/arm64/kvm/arm.c +0
-6@@ -1525,12 +1525,6 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -static int kvm_vm_ioctl_set_device_addr (struct kvm *kvm,struct kvm_arm_device_addr *dev_addr) {arch/mips/kvm/mips.c +2
-8@@ -199,7 +199,7 @@ void kvm_arch_flush_shadow_memslot(struct kvm *kvm,/* Flush slot from GPA */ kvm_mips_flush_gpa_pt (kvm, slot->base_gfn, slot->base_gfn + slot->npages -1 ); -kvm_arch_flush_remote_tlbs_memslot (kvm, slot); +kvm_flush_remote_tlbs_memslot (kvm, slot);spin_unlock (&kvm->mmu_lock); }@@ -235,7 +235,7 @@ void kvm_arch_commit_memory_region(struct kvm *kvm, needs_flush =kvm_mips_mkclean_gpa_pt (kvm, new->base_gfn, new->base_gfn + new->npages -1 );if (needs_flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, new); +kvm_flush_remote_tlbs_memslot (kvm, new);spin_unlock (&kvm->mmu_lock); } }@@ -987,12 +987,6 @@ int kvm_arch_flush_remote_tlbs(struct kvm *kvm)return 1 ; } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -int kvm_arch_vm_ioctl (struct file *filp,unsigned int ioctl,unsigned long arg) {int r;arch/riscv/kvm/mmu.c +0
-6@@ -406,12 +406,6 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) { } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -kvm_flush_remote_tlbs (kvm); -} -void kvm_arch_free_memslot (struct kvm *kvm,struct kvm_memory_slot *free) { }arch/x86/kvm/mmu/mmu.c +1
-15@@ -6669,7 +6669,7 @@ static void kvm_rmap_zap_collapsible_sptes(struct kvm *kvm,*/ if (walk_slot_rmaps (kvm, slot, kvm_mmu_zap_collapsible_spte,PG_LEVEL_4K ,KVM_MAX_HUGEPAGE_LEVEL -1 , true)) -kvm_arch_flush_remote_tlbs_memslot (kvm, slot); +kvm_flush_remote_tlbs_memslot (kvm, slot); } void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm,@@ -6688,20 +6688,6 @@ void kvm_mmu_zap_collapsible_sptes(struct kvm *kvm, } } -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot) -{ -/* -* All current use cases for flushing the TLBs for a specific memslot -* related to dirty logging, and many do the TLB flush out of mmu_lock. -* The interaction between the various operations on memslot must be -* serialized by slots_locks to ensure the TLB flush from one operation -* is observed by any other operation on the same memslot. -*/ -lockdep_assert_held (&kvm->slots_lock); -kvm_flush_remote_tlbs_range (kvm, memslot->base_gfn, memslot->npages); -} -void kvm_mmu_slot_leaf_clear_dirty (struct kvm *kvm,const struct kvm_memory_slot *memslot) {arch/x86/kvm/x86.c +1
-1@@ -12751,7 +12751,7 @@ static void kvm_mmu_slot_apply_flags(struct kvm *kvm,* See is_writable_pte() for more details (the case involving * access-tracked SPTEs is particularly relevant). */ -kvm_arch_flush_remote_tlbs_memslot (kvm, new); +kvm_flush_remote_tlbs_memslot (kvm, new); } }include/linux/kvm_host.h +3
-4@@ -1360,6 +1360,8 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool yield_to_kernel_mode);void kvm_flush_remote_tlbs (struct kvm *kvm);void kvm_flush_remote_tlbs_range (struct kvm *kvm,gfn_t gfn,u64 pages); +void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot);#ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE int kvm_mmu_topup_memory_cache (struct kvm_mmu_memory_cache *mc,int min);@@ -1388,10 +1390,7 @@ void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm, unsigned long mask);void kvm_arch_sync_dirty_log (struct kvm *kvm,struct kvm_memory_slot *memslot); -#ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT -void kvm_arch_flush_remote_tlbs_memslot (struct kvm *kvm, -const struct kvm_memory_slot *memslot); -#else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */ +#ifndef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT int kvm_vm_ioctl_get_dirty_log (struct kvm *kvm,struct kvm_dirty_log *log);int kvm_get_dirty_log (struct kvm *kvm,struct kvm_dirty_log *log,int *is_dirty,struct kvm_memory_slot **memslot);virt/kvm/kvm_main.c +16
-3@@ -379,6 +379,20 @@ void kvm_flush_remote_tlbs_range(struct kvm *kvm, gfn_t gfn, u64 pages)kvm_flush_remote_tlbs (kvm); } +void kvm_flush_remote_tlbs_memslot (struct kvm *kvm, +const struct kvm_memory_slot *memslot) +{ +/* +* All current use cases for flushing the TLBs for a specific memslot +* related to dirty logging, and many do the TLB flush out of mmu_lock. +* The interaction between the various operations on memslot must be +* serialized by slots_locks to ensure the TLB flush from one operation +* is observed by any other operation on the same memslot. +*/ +lockdep_assert_held (&kvm->slots_lock); +kvm_flush_remote_tlbs_range (kvm, memslot->base_gfn, memslot->npages); +} +static void kvm_flush_shadow_all (struct kvm *kvm) {kvm_arch_flush_shadow_all (kvm);@@ -2191,7 +2205,7 @@ static int kvm_get_dirty_log_protect(struct kvm *kvm, struct kvm_dirty_log *log) }if (flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, memslot); +kvm_flush_remote_tlbs_memslot (kvm, memslot);if (copy_to_user (log->dirty_bitmap, dirty_bitmap_buffer, n))return -EFAULT ;@@ -2308,7 +2322,7 @@ static int kvm_clear_dirty_log_protect(struct kvm *kvm,KVM_MMU_UNLOCK (kvm);if (flush) -kvm_arch_flush_remote_tlbs_memslot (kvm, memslot); +kvm_flush_remote_tlbs_memslot (kvm, memslot);return 0 ; } --2.41.0.487.g6d72f3e995-goog
PATCH v7 6/12 arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
In reply to Raghavendra Rao AnantaCurrently, the core TLB flush functionality of __flush_tlb_range() hardcodes vae1is (and variants) for the flush operation. In the upcoming patches, the KVM code reuses this core algorithm with ipas2e1is for range based TLB invalidations based on the IPA. Hence, extract the core flush functionality of __flush_tlb_range() into its own macro that accepts an 'op' argument to pass any TLBI operation, such that other callers (KVM) can benefit.
No functional changes intended.
--- arch/arm64/include/asm/tlbflush.h | 109 +++++++++++++++--------------- 1 file changed, 56 insertions(+), 53 deletions(-)
arch/arm64/include/asm/tlbflush.h +56 -53
@@ -278,14 +278,62 @@ static inline void flush_tlb_page(struct vm_area_struct *vma,*/ #define MAX_TLBI_OPS PTRS_PER_PTE +/* When the CPU does not support TLB range operations, flush the TLB +* entries one by one at the granularity of 'stride'. If the TLB +* range ops are supported, then: +* +* 1. If 'pages' is odd, flush the first page through non-range +* operations; +* +* 2. For remaining pages: the minimum range granularity is decided +* by 'scale', so multiple range TLBI operations may be required. +* Start from scale = 0, flush the corresponding number of pages +* ((num+1)*2^(5*scale+1) starting from 'addr'), then increase it +* until no pages left. +* +* Note that certain ranges can be represented by either num = 31 and +* scale or num = 0 and scale + 1. The loop below favours the latter +* since num is limited to 30 by the __TLBI_RANGE_NUM() macro. +*/ +#define __flush_tlb_range_op (op, start, pages, stride, \ + asid, tlb_level, tlbi_user) \ +do { \ + int num = 0; \ + int scale = 0; \ + unsigned long addr; \ + \ + while (pages > 0) { \ + if (!system_supports_tlb_range() || \ + pages % 2 == 1) { \ + addr = __TLBI_VADDR(start, asid); \ + __tlbi_level(op, addr, tlb_level); \ + if (tlbi_user) \ + __tlbi_user_level(op, addr, tlb_level); \ + start += stride; \ + pages -= stride >> PAGE_SHIFT; \ + continue; \ + } \ + \ + num = __TLBI_RANGE_NUM(pages, scale); \ + if (num >= 0) { \ + addr = __TLBI_VADDR_RANGE(start, asid, scale, \ + num, tlb_level); \ + __tlbi(r##op, addr); \ + if (tlbi_user) \ + __tlbi_user(r##op, addr); \ + start += __TLBI_RANGE_PAGES(num, scale) << PAGE_SHIFT; \ + pages -= __TLBI_RANGE_PAGES(num, scale); \ + } \ + scale++; \ + } \ +} while (0) +static inline void __flush_tlb_range (struct vm_area_struct *vma,unsigned long start,unsigned long end,unsigned long stride,bool last_level,int tlb_level) { -int num =0 ; -int scale =0 ; -unsigned long asid, addr, pages; +unsigned long asid, pages; start =round_down (start, stride); end =round_up (end, stride);
@@ -307,56 +355,11 @@ static inline void __flush_tlb_range(struct vm_area_struct *vma,dsb (ishst); asid =ASID (vma->vm_mm); -/* -* When the CPU does not support TLB range operations, flush the TLB -* entries one by one at the granularity of 'stride'. If the TLB -* range ops are supported, then: -* -* 1. If 'pages' is odd, flush the first page through non-range -* operations; -* -* 2. For remaining pages: the minimum range granularity is decided -* by 'scale', so multiple range TLBI operations may be required. -* Start from scale = 0, flush the corresponding number of pages -* ((num+1)*2^(5*scale+1) starting from 'addr'), then increase it -* until no pages left. -* -* Note that certain ranges can be represented by either num = 31 and -* scale or num = 0 and scale + 1. The loop below favours the latter -* since num is limited to 30 by the __TLBI_RANGE_NUM() macro. -*/ -while (pages >0 ) { -if (!system_supports_tlb_range () || - pages %2 ==1 ) { - addr =__TLBI_VADDR (start, asid); -if (last_level) { -__tlbi_level (vale1is, addr, tlb_level); -__tlbi_user_level (vale1is, addr, tlb_level); - }else { -__tlbi_level (vae1is, addr, tlb_level); -__tlbi_user_level (vae1is, addr, tlb_level); - } - start += stride; - pages -= stride >>PAGE_SHIFT ; -continue ; - } - - num =__TLBI_RANGE_NUM (pages, scale); -if (num >=0 ) { - addr =__TLBI_VADDR_RANGE (start, asid, scale, - num, tlb_level); -if (last_level) { -__tlbi (rvale1is, addr); -__tlbi_user (rvale1is, addr); - }else { -__tlbi (rvae1is, addr); -__tlbi_user (rvae1is, addr); - } - start +=__TLBI_RANGE_PAGES (num, scale) <<PAGE_SHIFT ; - pages -=__TLBI_RANGE_PAGES (num, scale); - } - scale++; - } +if (last_level) +__flush_tlb_range_op (vale1is, start, pages, stride, asid, tlb_level, true); +else +__flush_tlb_range_op (vae1is, start, pages, stride, asid, tlb_level, true); +dsb (ish); }
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Marc Zyngier
maintainerRe: [PATCH v7 06/12] arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
Quoted 24 lines · Raghavendra Rao Ananta
Currently, the core TLB flush functionality of
__flush_tlb_range()hardcodes vae1is (and variants) for the flush operation. In the upcoming patches, the KVM code reuses this core algorithm with ipas2e1is for range based TLB invalidations based on the IPA. Hence, extract the core flush functionality of__flush_tlb_range()into its own macro that accepts an 'op' argument to pass any TLBI operation, such that other callers (KVM) can benefit.No functional changes intended.
Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Catalin Marinas <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/tlbflush.h | 109 +++++++++++++++--------------- 1 file changed, 56 insertions(+), 53 deletions(-)arch/arm64/include/asm/tlbflush.h +0
-0
arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 283
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h +3
-0@@ -278,14 +278,62 @@ static inline void flush_tlb_page(struct vm_area_struct *vma,*/ #define MAX_TLBI_OPS PTRS_PER_PTE +/* When the CPU does not support TLB range operations, flush the TLB +* entries one by one at the granularity of 'stride'. If the TLB +* range ops are supported, then:
Comment format (the original was correct).
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h +6
-0+* +* 1. If 'pages' is odd, flush the first page through non-range +* operations; +* +* 2. For remaining pages: the minimum range granularity is decided +* by 'scale', so multiple range TLBI operations may be required.
arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 299
Quoted 10 lines
arch/arm64/include/asm/tlbflush.h +10
-0+* Start from scale = 0, flush the corresponding number of pages +* ((num+1)*2^(5*scale+1) starting from 'addr'), then increase it +* until no pages left. +* +* Note that certain ranges can be represented by either num = 31 and +* scale or num = 0 and scale + 1. The loop below favours the latter +* since num is limited to 30 by the __TLBI_RANGE_NUM() macro. +*/ +#define __flush_tlb_range_op (op, start, pages, stride, \ + asid, tlb_level, tlbi_user) \
If you make this a common macro, please document the parameters, and what the constraints are. For example, what does tlbi_user mean for an IPA invalidation?
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 06/12] arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
Quoted 30 lines · On Thu, Jul 27, 2023 at 3:58?AM Marc Zyngier <[email protected]> wrote:
Quoted 27 lines · Raghavendra Rao Ananta
Currently, the core TLB flush functionality of
__flush_tlb_range()hardcodes vae1is (and variants) for the flush operation. In the upcoming patches, the KVM code reuses this core algorithm with ipas2e1is for range based TLB invalidations based on the IPA. Hence, extract the core flush functionality of__flush_tlb_range()into its own macro that accepts an 'op' argument to pass any TLBI operation, such that other callers (KVM) can benefit.No functional changes intended.
Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Catalin Marinas <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/tlbflush.h | 109 +++++++++++++++--------------- 1 file changed, 56 insertions(+), 53 deletions(-)arch/arm64/include/asm/tlbflush.h +0
-0@@ -278,14 +278,62 @@ static inline void flush_tlb_page(struct vm_area_struct *vma,*/ #define MAX_TLBI_OPS PTRS_PER_PTE
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 283
Quoted 3 lines
arch/arm64/include/asm/tlbflush.h +3
-0+/* When the CPU does not support TLB range operations, flush the TLB +* entries one by one at the granularity of 'stride'. If the TLB +* range ops are supported, then: Comment format (the original was correct).
Isn't the format the same as original? Or are you referring to the fact that it needs to be placed inside the macro definition?
Quoted 17 lines
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h +6
-0+* +* 1. If 'pages' is odd, flush the first page through non-range +* operations; +* +* 2. For remaining pages: the minimum range granularity is decided +* by 'scale', so multiple range TLBI operations may be required. arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 299
Quoted 10 lines
arch/arm64/include/asm/tlbflush.h +10
-0+* Start from scale = 0, flush the corresponding number of pages +* ((num+1)*2^(5*scale+1) starting from 'addr'), then increase it +* until no pages left. +* +* Note that certain ranges can be represented by either num = 31 and +* scale or num = 0 and scale + 1. The loop below favours the latter +* since num is limited to 30 by the __TLBI_RANGE_NUM() macro. +*/ +#define __flush_tlb_range_op (op, start, pages, stride, \ + asid, tlb_level, tlbi_user) \
Quoted 4 lines
If you make this a common macro, please document the parameters, and what the constraints are. For example, what does
tlbi_usermean for an IPA invalidation?
Sure, I'll document the parameters. That'll be helpful.
- Raghavendra
Quoted 4 lines
M.--
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH v7 06/12] arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
Quoted 32 lines · Raghavendra Rao Ananta
Quoted 30 lines · On Thu, Jul 27, 2023 at 3:58?AM Marc Zyngier <[email protected]> wrote:
Quoted 27 lines · Raghavendra Rao Ananta
Currently, the core TLB flush functionality of
__flush_tlb_range()hardcodes vae1is (and variants) for the flush operation. In the upcoming patches, the KVM code reuses this core algorithm with ipas2e1is for range based TLB invalidations based on the IPA. Hence, extract the core flush functionality of__flush_tlb_range()into its own macro that accepts an 'op' argument to pass any TLBI operation, such that other callers (KVM) can benefit.No functional changes intended.
Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Catalin Marinas <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/tlbflush.h | 109 +++++++++++++++--------------- 1 file changed, 56 insertions(+), 53 deletions(-)arch/arm64/include/asm/tlbflush.h +0
-0@@ -278,14 +278,62 @@ static inline void flush_tlb_page(struct vm_area_struct *vma,*/ #define MAX_TLBI_OPS PTRS_PER_PTE
Quoted 8 lines
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 283
Quoted 3 lines
arch/arm64/include/asm/tlbflush.h +3
-0+/* When the CPU does not support TLB range operations, flush the TLB +* entries one by one at the granularity of 'stride'. If the TLB +* range ops are supported, then: Comment format (the original was correct).
Isn't the format the same as original? Or are you referring to the fact that it needs to be placed inside the macro definition?
No, I'm referring to the multiline comment that starts with:
/* When the CPU does not support TLB range operations...
instead of the required:
/* * When the CPU does not support TLB range operations
which was correct before the coment was moved.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 06/12] arm64: tlb: Refactor the core flush algorithm of __flush_tlb_range
Quoted 46 lines · On Wed, Aug 2, 2023 at 8:58?AM Marc Zyngier <[email protected]> wrote:
Quoted 32 lines · Raghavendra Rao Ananta
Quoted 30 lines · On Thu, Jul 27, 2023 at 3:58?AM Marc Zyngier <[email protected]> wrote:
Quoted 27 lines · Raghavendra Rao Ananta
Currently, the core TLB flush functionality of
__flush_tlb_range()hardcodes vae1is (and variants) for the flush operation. In the upcoming patches, the KVM code reuses this core algorithm with ipas2e1is for range based TLB invalidations based on the IPA. Hence, extract the core flush functionality of__flush_tlb_range()into its own macro that accepts an 'op' argument to pass any TLBI operation, such that other callers (KVM) can benefit.No functional changes intended.
Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Catalin Marinas <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/tlbflush.h | 109 +++++++++++++++--------------- 1 file changed, 56 insertions(+), 53 deletions(-)arch/arm64/include/asm/tlbflush.h +0
-0@@ -278,14 +278,62 @@ static inline void flush_tlb_page(struct vm_area_struct *vma,*/ #define MAX_TLBI_OPS PTRS_PER_PTEQuoted 8 lines
Quoted 6 lines
arch/arm64/include/asm/tlbflush.h static inline void flush_tlb_page(struct vm_area_struct *vma, · line 283
Quoted 3 lines
arch/arm64/include/asm/tlbflush.h +3
-0+/* When the CPU does not support TLB range operations, flush the TLB +* entries one by one at the granularity of 'stride'. If the TLB +* range ops are supported, then: Comment format (the original was correct).
Isn't the format the same as original? Or are you referring to the fact that it needs to be placed inside the macro definition?
No, I'm referring to the multiline comment that starts with:
Quoted 9 lines
/* When the CPU does not support TLB range operations...instead of the required:
/*
* When the CPU does not support TLB range operationswhich was correct before the coment was moved.
Oh I see! I'll fix it in v8.
Thanks,
Raghavendra
Quoted 6 lines
Thanks,
M.--
Without deviation from the norm, progress is not possible.
PATCH v7 7/12 KVM: arm64: Implement __kvm_tlb_flush_vmid_range()
In reply to Raghavendra Rao AnantaDefine __kvm_tlb_flush_vmid_range() (for VHE and nVHE) to flush a range of stage-2 page-tables using IPA in one go. If the system supports FEAT_TLBIRANGE, the following patches would conviniently replace global TLBI such as vmalls12e1is in the map, unmap, and dirty-logging paths with ripas2e1is instead.
--- arch/arm64/include/asm/kvm_asm.h | 3 +++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++++++++++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++++++++++++++++++++++++ arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++++++++++++++++++++++ 4 files changed, 71 insertions(+)
arch/arm64/include/asm/kvm_asm.h +3 -0
@@ -70,6 +70,7 @@ enum __kvm_host_smccc_func { __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa_nsh, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid, + __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_range, __KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context, __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntvoff, __KVM_HOST_SMCCC_FUNC___vgic_v3_read_vmcr,
@@ -229,6 +230,8 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid_ipa_nsh (struct kvm_s2_mmu *mmu,phys_addr_t ipa,int level); +extern void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages);extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff);
arch/arm64/kvm/hyp/nvhe/hyp-main.c +11 -0
@@ -135,6 +135,16 @@ static void handle___kvm_tlb_flush_vmid_ipa_nsh(struct kvm_cpu_context *host_ctx__kvm_tlb_flush_vmid_ipa_nsh (kern_hyp_va (mmu), ipa, level); } +static void +handle___kvm_tlb_flush_vmid_range (struct kvm_cpu_context *host_ctxt) +{ + DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1); +DECLARE_REG (phys_addr_t, start, host_ctxt,2 ); + DECLARE_REG(unsigned long, pages, host_ctxt, 3); + +__kvm_tlb_flush_vmid_range (kern_hyp_va (mmu), start, pages); +} +static void handle___kvm_tlb_flush_vmid (struct kvm_cpu_context *host_ctxt) { DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);
@@ -327,6 +337,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa),HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa_nsh),HANDLE_FUNC (__kvm_tlb_flush_vmid), +HANDLE_FUNC (__kvm_tlb_flush_vmid_range),HANDLE_FUNC (__kvm_flush_cpu_context),HANDLE_FUNC (__kvm_timer_set_cntvoff), HANDLE_FUNC(__vgic_v3_read_vmcr),
arch/arm64/kvm/hyp/nvhe/tlb.c +30 -0
@@ -182,6 +182,36 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,__tlb_switch_to_host (&cxt); } +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/* +* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +/* Switch to requested VMID */ +__tlb_switch_to_guest (mmu, &cxt, false); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false); + +dsb (ish); +__tlbi (vmalle1is); +dsb (ish); +isb (); + +/* See the comment in __kvm_tlb_flush_vmid_ipa() */ +if (icache_is_vpipt ()) +icache_inval_all_pou (); + +__tlb_switch_to_host (&cxt); +} +void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu) {struct tlb_inv_context cxt;
arch/arm64/kvm/hyp/vhe/tlb.c +27 -0
@@ -138,6 +138,33 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,dsb (nsh);__tlbi (vmalle1);dsb (nsh); + +__tlb_switch_to_host (&cxt); +} + +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/* +* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +dsb (ishst); + +/* Switch to requested VMID */ +__tlb_switch_to_guest (mmu, &cxt); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false); + +dsb (ish); +__tlbi (vmalle1is); +dsb (ish);isb ();__tlb_switch_to_host (&cxt);
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Re: [PATCH v7 07/12] KVM: arm64: Implement __kvm_tlb_flush_vmid_range()
Quoted 9 lines · Raghavendra Rao Ananta
Define
__kvm_tlb_flush_vmid_range()(for VHE and nVHE) to flush a range of stage-2 page-tables using IPA in one go. If the system supportsFEAT_TLBIRANGE, the following patches would conviniently replace global TLBI such as vmalls12e1is in the map, unmap, and dirty-logging paths with ripas2e1is instead.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>
Quoted 136 lines
--- arch/arm64/include/asm/kvm_asm.h | 3 +++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++++++++++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++++++++++++++++++++++++ arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++++++++++++++++++++++ 4 files changed, 71 insertions(+)arch/arm64/include/asm/kvm_asm.h +3
-0@@ -70,6 +70,7 @@ enum __kvm_host_smccc_func { __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa_nsh, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid, + __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_range, __KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context, __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntvoff, __KVM_HOST_SMCCC_FUNC___vgic_v3_read_vmcr,@@ -229,6 +230,8 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid_ipa_nsh (struct kvm_s2_mmu *mmu,phys_addr_t ipa,int level); +extern void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages);extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff);arch/arm64/kvm/hyp/nvhe/hyp-main.c +11
-0@@ -135,6 +135,16 @@ static void handle___kvm_tlb_flush_vmid_ipa_nsh(struct kvm_cpu_context *host_ctx__kvm_tlb_flush_vmid_ipa_nsh (kern_hyp_va (mmu), ipa, level); } +static void +handle___kvm_tlb_flush_vmid_range (struct kvm_cpu_context *host_ctxt) +{ + DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1); +DECLARE_REG (phys_addr_t, start, host_ctxt,2 ); + DECLARE_REG(unsigned long, pages, host_ctxt, 3); + +__kvm_tlb_flush_vmid_range (kern_hyp_va (mmu), start, pages); +} +static void handle___kvm_tlb_flush_vmid (struct kvm_cpu_context *host_ctxt) { DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);@@ -327,6 +337,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa),HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa_nsh),HANDLE_FUNC (__kvm_tlb_flush_vmid), +HANDLE_FUNC (__kvm_tlb_flush_vmid_range),HANDLE_FUNC (__kvm_flush_cpu_context),HANDLE_FUNC (__kvm_timer_set_cntvoff), HANDLE_FUNC(__vgic_v3_read_vmcr),arch/arm64/kvm/hyp/nvhe/tlb.c +30
-0@@ -182,6 +182,36 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,__tlb_switch_to_host (&cxt); } +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/* +* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +/* Switch to requested VMID */ +__tlb_switch_to_guest (mmu, &cxt, false); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false); + +dsb (ish); +__tlbi (vmalle1is); +dsb (ish); +isb (); + +/* See the comment in __kvm_tlb_flush_vmid_ipa() */ +if (icache_is_vpipt ()) +icache_inval_all_pou (); + +__tlb_switch_to_host (&cxt); +} +void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu) {struct tlb_inv_context cxt;arch/arm64/kvm/hyp/vhe/tlb.c +27
-0@@ -138,6 +138,33 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,dsb (nsh);__tlbi (vmalle1);dsb (nsh); + +__tlb_switch_to_host (&cxt); +} + +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/* +* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +dsb (ishst); + +/* Switch to requested VMID */ +__tlb_switch_to_guest (mmu, &cxt); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false); + +dsb (ish); +__tlbi (vmalle1is); +dsb (ish);isb ();__tlb_switch_to_host (&cxt);
Signature · 1 line
Shaoqin
Marc Zyngier
maintainerRe: [PATCH v7 07/12] KVM: arm64: Implement __kvm_tlb_flush_vmid_range()
In reply to Raghavendra Rao AnantaQuoted 82 lines · Raghavendra Rao Ananta
Define
__kvm_tlb_flush_vmid_range()(for VHE and nVHE) to flush a range of stage-2 page-tables using IPA in one go. If the system supportsFEAT_TLBIRANGE, the following patches would conviniently replace global TLBI such as vmalls12e1is in the map, unmap, and dirty-logging paths with ripas2e1is instead.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>--- arch/arm64/include/asm/kvm_asm.h | 3 +++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++++++++++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++++++++++++++++++++++++ arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++++++++++++++++++++++ 4 files changed, 71 insertions(+)arch/arm64/include/asm/kvm_asm.h +3
-0@@ -70,6 +70,7 @@ enum __kvm_host_smccc_func { __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa_nsh, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid, + __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_range, __KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context, __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntvoff, __KVM_HOST_SMCCC_FUNC___vgic_v3_read_vmcr,@@ -229,6 +230,8 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid_ipa_nsh (struct kvm_s2_mmu *mmu,phys_addr_t ipa,int level); +extern void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages);extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff);arch/arm64/kvm/hyp/nvhe/hyp-main.c +11
-0@@ -135,6 +135,16 @@ static void handle___kvm_tlb_flush_vmid_ipa_nsh(struct kvm_cpu_context *host_ctx__kvm_tlb_flush_vmid_ipa_nsh (kern_hyp_va (mmu), ipa, level); } +static void +handle___kvm_tlb_flush_vmid_range (struct kvm_cpu_context *host_ctxt) +{ + DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1); +DECLARE_REG (phys_addr_t, start, host_ctxt,2 ); + DECLARE_REG(unsigned long, pages, host_ctxt, 3); + +__kvm_tlb_flush_vmid_range (kern_hyp_va (mmu), start, pages); +} +static void handle___kvm_tlb_flush_vmid (struct kvm_cpu_context *host_ctxt) { DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);@@ -327,6 +337,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa),HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa_nsh),HANDLE_FUNC (__kvm_tlb_flush_vmid), +HANDLE_FUNC (__kvm_tlb_flush_vmid_range),HANDLE_FUNC (__kvm_flush_cpu_context),HANDLE_FUNC (__kvm_timer_set_cntvoff), HANDLE_FUNC(__vgic_v3_read_vmcr),arch/arm64/kvm/hyp/nvhe/tlb.c +7
-0@@ -182,6 +182,36 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,__tlb_switch_to_host (&cxt); } +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/*
arch/arm64/kvm/hyp/nvhe/tlb.c void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu, · line 201
Quoted 10 lines
arch/arm64/kvm/hyp/nvhe/tlb.c +10
-0+* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +/* Switch to requested VMID */ +__tlb_switch_to_guest (mmu, &cxt, false); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false);
I really think we need an abstraction here. All this ASID and user nonsense shouldn't appear here. Something such as __flush_s2_tlb_range_op(), which would pass the correct parameters that this code shouldn't have to worry about.
I'm also a bit concerned by the fact we completely lose the level here. This is a massive fast-path for the CPU, and we don't make use of it. It'd be worth thinking of how we can make use of it if at all possible...
arch/arm64/kvm/hyp/nvhe/tlb.c void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu, · line 212
Quoted 11 lines
arch/arm64/kvm/hyp/nvhe/tlb.c +11
-0+ +dsb (ish); +__tlbi (vmalle1is); +dsb (ish); +isb (); + +/* See the comment in __kvm_tlb_flush_vmid_ipa() */ +if (icache_is_vpipt ()) +icache_inval_all_pou (); + +__tlb_switch_to_host (&cxt);
Another thing is that it is high time that some of this call gets refactored. All these helpers are basically the same sequence, only differing by a couple of lines. Not something we need to do immediately, but eventually we'll have to bite the bullet.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 07/12] KVM: arm64: Implement __kvm_tlb_flush_vmid_range()
Quoted 92 lines · On Thu, Jul 27, 2023 at 5:40?AM Marc Zyngier <[email protected]> wrote:
Quoted 89 lines · Raghavendra Rao Ananta
Define
__kvm_tlb_flush_vmid_range()(for VHE and nVHE) to flush a range of stage-2 page-tables using IPA in one go. If the system supportsFEAT_TLBIRANGE, the following patches would conviniently replace global TLBI such as vmalls12e1is in the map, unmap, and dirty-logging paths with ripas2e1is instead.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>--- arch/arm64/include/asm/kvm_asm.h | 3 +++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++++++++++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++++++++++++++++++++++++ arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++++++++++++++++++++++ 4 files changed, 71 insertions(+)arch/arm64/include/asm/kvm_asm.h +3
-0@@ -70,6 +70,7 @@ enum __kvm_host_smccc_func { __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa_nsh, __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid, + __KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_range, __KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context, __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntvoff, __KVM_HOST_SMCCC_FUNC___vgic_v3_read_vmcr,@@ -229,6 +230,8 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid_ipa_nsh (struct kvm_s2_mmu *mmu,phys_addr_t ipa,int level); +extern void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages);extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff);arch/arm64/kvm/hyp/nvhe/hyp-main.c +11
-0@@ -135,6 +135,16 @@ static void handle___kvm_tlb_flush_vmid_ipa_nsh(struct kvm_cpu_context *host_ctx__kvm_tlb_flush_vmid_ipa_nsh (kern_hyp_va (mmu), ipa, level); } +static void +handle___kvm_tlb_flush_vmid_range (struct kvm_cpu_context *host_ctxt) +{ + DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1); +DECLARE_REG (phys_addr_t, start, host_ctxt,2 ); + DECLARE_REG(unsigned long, pages, host_ctxt, 3); + +__kvm_tlb_flush_vmid_range (kern_hyp_va (mmu), start, pages); +} +static void handle___kvm_tlb_flush_vmid (struct kvm_cpu_context *host_ctxt) { DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);@@ -327,6 +337,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa),HANDLE_FUNC (__kvm_tlb_flush_vmid_ipa_nsh),HANDLE_FUNC (__kvm_tlb_flush_vmid), +HANDLE_FUNC (__kvm_tlb_flush_vmid_range),HANDLE_FUNC (__kvm_flush_cpu_context),HANDLE_FUNC (__kvm_timer_set_cntvoff), HANDLE_FUNC(__vgic_v3_read_vmcr),arch/arm64/kvm/hyp/nvhe/tlb.c +14
-0@@ -182,6 +182,36 @@ void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,__tlb_switch_to_host (&cxt); } +void __kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t start,unsigned long pages) +{ +struct tlb_inv_context cxt; +unsigned long stride; + +/* +* Since the range of addresses may not be mapped at +* the same level, assume the worst case as PAGE_SIZE +*/ + stride =PAGE_SIZE ; + start =round_down (start, stride); + +/* Switch to requested VMID */
Quoted 9 lines
arch/arm64/kvm/hyp/nvhe/tlb.c void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu, · line 201
Quoted 3 lines
arch/arm64/kvm/hyp/nvhe/tlb.c +3
-0+__tlb_switch_to_guest (mmu, &cxt, false); + +__flush_tlb_range_op (ipas2e1is, start, pages, stride,0 ,0 , false);I really think we need an abstraction here. All this ASID and user nonsense shouldn't appear here. Something such as
__flush_s2_tlb_range_op(), which would pass the correct parameters that this code shouldn't have to worry about.
Yes, a simple wrapper would be nice. I'll implement this in v8.
Quoted 5 lines
I'm also a bit concerned by the fact we completely lose the level here. This is a massive fast-path for the CPU, and we don't make use of it. It'd be worth thinking of how we can make use of it if at all possible...
Initial implementation of the series included the 'level', but had some complexities [1], and so we had to get rid of it for things to at least be correct.
But, we can think about it and include the 'level' as needed.
- Raghavendra
[1]: https://lore.kernel.org/all/ZCTjirkCgBkT65eP at linux.dev/
Quoted 21 lines
arch/arm64/kvm/hyp/nvhe/tlb.c void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu, · line 212
Quoted 11 lines
arch/arm64/kvm/hyp/nvhe/tlb.c +11
-0+ +dsb (ish); +__tlbi (vmalle1is); +dsb (ish); +isb (); + +/* See the comment in __kvm_tlb_flush_vmid_ipa() */ +if (icache_is_vpipt ()) +icache_inval_all_pou (); + +__tlb_switch_to_host (&cxt);Another thing is that it is high time that some of this call gets refactored. All these helpers are basically the same sequence, only differing by a couple of lines. Not something we need to do immediately, but eventually we'll have to bite the bullet.
M.--
Without deviation from the norm, progress is not possible.
PATCH v7 8/12 KVM: arm64: Define kvm_tlb_flush_vmid_range()
In reply to Raghavendra Rao AnantaImplement the helper kvm_tlb_flush_vmid_range() that acts as a wrapper for range-based TLB invalidations. For the given VMID, use the range-based TLBI instructions to do the job or fallback to invalidating all the TLB entries.
--- arch/arm64/include/asm/kvm_pgtable.h | 10 ++++++++++ arch/arm64/kvm/hyp/pgtable.c | 20 ++++++++++++++++++++ 2 files changed, 30 insertions(+)
arch/arm64/include/asm/kvm_pgtable.h +10 -0
@@ -754,4 +754,14 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);* kvm_pgtable_prot format. */ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot (kvm_pte_t pte); + +/** +* kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries +* +* @mmu: Stage-2 KVM MMU struct +* @addr: The base Intermediate physical address from which to invalidate +* @size: Size of the range from the base to invalidate +*/ +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size);#endif /* __ARM64_KVM_PGTABLE_H__ */
arch/arm64/kvm/hyp/pgtable.c +20 -0
@@ -670,6 +670,26 @@ static bool stage2_has_fwb(struct kvm_pgtable *pgt)return !(pgt->flags &KVM_PGTABLE_S2_NOFWB ); } +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size) +{ +unsigned long pages, inval_pages; + +if (!system_supports_tlb_range ()) { +kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +return ; + } + + pages = size >>PAGE_SHIFT ; +while (pages >0 ) { + inval_pages =min (pages,MAX_TLBI_RANGE_PAGES ); +kvm_call_hyp (__kvm_tlb_flush_vmid_range, mmu, addr, inval_pages); + + addr += inval_pages <<PAGE_SHIFT ; + pages -= inval_pages; + } +} +#define KVM_S2_MEMATTR (pgt, attr) PAGE_S2_MEMATTR(attr, stage2_has_fwb(pgt)) static int stage2_set_prot_attr(struct kvm_pgtable *pgt, enum kvm_pgtable_prot prot,
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Marc Zyngier
maintainerRe: [PATCH v7 08/12] KVM: arm64: Define kvm_tlb_flush_vmid_range()
Quoted 51 lines · Raghavendra Rao Ananta
Implement the helper
kvm_tlb_flush_vmid_range()that acts as a wrapper for range-based TLB invalidations. For the given VMID, use the range-based TLBI instructions to do the job or fallback to invalidating all the TLB entries.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/kvm_pgtable.h | 10 ++++++++++ arch/arm64/kvm/hyp/pgtable.c | 20 ++++++++++++++++++++ 2 files changed, 30 insertions(+)arch/arm64/include/asm/kvm_pgtable.h +10
-0@@ -754,4 +754,14 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);* kvm_pgtable_prot format. */ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot (kvm_pte_t pte); + +/** +* kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries +* +* @mmu: Stage-2 KVM MMU struct +* @addr: The base Intermediate physical address from which to invalidate +* @size: Size of the range from the base to invalidate +*/ +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size);#endif /* __ARM64_KVM_PGTABLE_H__ */ arch/arm64/kvm/hyp/pgtable.c +10
-0@@ -670,6 +670,26 @@ static bool stage2_has_fwb(struct kvm_pgtable *pgt)return !(pgt->flags &KVM_PGTABLE_S2_NOFWB ); } +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size) +{ +unsigned long pages, inval_pages; + +if (!system_supports_tlb_range ()) { +kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +return ; + } +
arch/arm64/kvm/hyp/pgtable.c static bool stage2_has_fwb(struct kvm_pgtable *pgt) · line 692
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +10
-0+ pages = size >>PAGE_SHIFT ; +while (pages >0 ) { + inval_pages =min (pages,MAX_TLBI_RANGE_PAGES ); +kvm_call_hyp (__kvm_tlb_flush_vmid_range, mmu, addr, inval_pages); + + addr += inval_pages <<PAGE_SHIFT ; + pages -= inval_pages; + } +} +
This really shouldn't live in pgtable.c. This code gets linked into the EL2 object. What do you think happens if, for some reason, this gets called *from EL2*?
Furthermore, this doesn't deal with page tables at all. Why isn't mmu.c a convenient place for it, as an integral part of kvm_arch_flush_remote_tlbs_range?
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH v7 08/12] KVM: arm64: Define kvm_tlb_flush_vmid_range()
Quoted 60 lines · Marc Zyngier
Quoted 57 lines · Raghavendra Rao Ananta
Implement the helper
kvm_tlb_flush_vmid_range()that acts as a wrapper for range-based TLB invalidations. For the given VMID, use the range-based TLBI instructions to do the job or fallback to invalidating all the TLB entries.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/kvm_pgtable.h | 10 ++++++++++ arch/arm64/kvm/hyp/pgtable.c | 20 ++++++++++++++++++++ 2 files changed, 30 insertions(+)arch/arm64/include/asm/kvm_pgtable.h +10
-0@@ -754,4 +754,14 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);* kvm_pgtable_prot format. */ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot (kvm_pte_t pte); + +/** +* kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries +* +* @mmu: Stage-2 KVM MMU struct +* @addr: The base Intermediate physical address from which to invalidate +* @size: Size of the range from the base to invalidate +*/ +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size);#endif /* __ARM64_KVM_PGTABLE_H__ */ arch/arm64/kvm/hyp/pgtable.c +16
-0@@ -670,6 +670,26 @@ static bool stage2_has_fwb(struct kvm_pgtable *pgt)return !(pgt->flags &KVM_PGTABLE_S2_NOFWB ); } +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size) +{ +unsigned long pages, inval_pages; + +if (!system_supports_tlb_range ()) { +kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +return ; + } + + pages = size >>PAGE_SHIFT ; +while (pages >0 ) { + inval_pages =min (pages,MAX_TLBI_RANGE_PAGES ); +kvm_call_hyp (__kvm_tlb_flush_vmid_range, mmu, addr, inval_pages); + + addr += inval_pages <<PAGE_SHIFT ;
Quoted 8 lines
arch/arm64/kvm/hyp/pgtable.c static bool stage2_has_fwb(struct kvm_pgtable *pgt) · line 692
Quoted 4 lines
arch/arm64/kvm/hyp/pgtable.c +4
-0+ pages -= inval_pages; + } +} +This really shouldn't live in
pgtable.c. This code gets linked into the EL2 object. What do you think happens if, for some reason, this gets called *from EL2*?
Ah, actually, nothing too bad would happen, as we convert the kvm_call_hyp() into a function call.
But still, we don't need two copies of this stuff, and it can live in mmu.c.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 08/12] KVM: arm64: Define kvm_tlb_flush_vmid_range()
Quoted 72 lines · On Thu, Jul 27, 2023 at 6:01?AM Marc Zyngier <[email protected]> wrote:
Quoted 60 lines · Marc Zyngier
Quoted 57 lines · Raghavendra Rao Ananta
Implement the helper
kvm_tlb_flush_vmid_range()that acts as a wrapper for range-based TLB invalidations. For the given VMID, use the range-based TLBI instructions to do the job or fallback to invalidating all the TLB entries.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/kvm_pgtable.h | 10 ++++++++++ arch/arm64/kvm/hyp/pgtable.c | 20 ++++++++++++++++++++ 2 files changed, 30 insertions(+)arch/arm64/include/asm/kvm_pgtable.h +10
-0@@ -754,4 +754,14 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);* kvm_pgtable_prot format. */ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot (kvm_pte_t pte); + +/** +* kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries +* +* @mmu: Stage-2 KVM MMU struct +* @addr: The base Intermediate physical address from which to invalidate +* @size: Size of the range from the base to invalidate +*/ +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size);#endif /* __ARM64_KVM_PGTABLE_H__ */ arch/arm64/kvm/hyp/pgtable.c +16
-0@@ -670,6 +670,26 @@ static bool stage2_has_fwb(struct kvm_pgtable *pgt)return !(pgt->flags &KVM_PGTABLE_S2_NOFWB ); } +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size) +{ +unsigned long pages, inval_pages; + +if (!system_supports_tlb_range ()) { +kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +return ; + } + + pages = size >>PAGE_SHIFT ; +while (pages >0 ) { + inval_pages =min (pages,MAX_TLBI_RANGE_PAGES ); +kvm_call_hyp (__kvm_tlb_flush_vmid_range, mmu, addr, inval_pages); + + addr += inval_pages <<PAGE_SHIFT ;Quoted 8 lines
arch/arm64/kvm/hyp/pgtable.c static bool stage2_has_fwb(struct kvm_pgtable *pgt) · line 692
Quoted 4 lines
arch/arm64/kvm/hyp/pgtable.c +4
-0+ pages -= inval_pages; + } +} +This really shouldn't live in
pgtable.c. This code gets linked into the EL2 object. What do you think happens if, for some reason, this gets called *from EL2*?
Quoted 6 lines
Ah, actually, nothing too bad would happen, as we convert the
kvm_call_hyp()into a function call.But still, we don't need two copies of this stuff, and it can live in
mmu.c.
But since we have a couple of references in pgtable.c to kvm_tlb_flush_vmid_range(), wouldn't that be an (linking) issue if we moved the definition to mmu.c?
ld: error: undefined symbol: __kvm_nvhe_kvm_tlb_flush_vmid_range
Quoted 2 lines
Quoted 2 lines
arch/arm64/kvm/hyp/pgtable.c static bool stage2_has_fwb(struct kvm_pgtable *pgt) · line 694
Quoted 2 lines
arch/arm64/kvm/hyp/pgtable.c +0
-0referenced by pgtable.c:1148 (./arch/arm64/kvm/hyp/nvhe/../pgtable.c:1148) arch/arm64/kvm/hyp/nvhe/kvm_nvhe.o:(__kvm_nvhe_kvm_pgtable_stage2_unmap) in archive vmlinux.a
...
Or is there some other way to make it work?
- Raghavendra
Quoted 4 lines
M.--
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH v7 08/12] KVM: arm64: Define kvm_tlb_flush_vmid_range()
Quoted 80 lines · Raghavendra Rao Ananta
Quoted 72 lines · On Thu, Jul 27, 2023 at 6:01?AM Marc Zyngier <[email protected]> wrote:
Quoted 60 lines · Marc Zyngier
Quoted 57 lines · Raghavendra Rao Ananta
Implement the helper
kvm_tlb_flush_vmid_range()that acts as a wrapper for range-based TLB invalidations. For the given VMID, use the range-based TLBI instructions to do the job or fallback to invalidating all the TLB entries.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/kvm_pgtable.h | 10 ++++++++++ arch/arm64/kvm/hyp/pgtable.c | 20 ++++++++++++++++++++ 2 files changed, 30 insertions(+)arch/arm64/include/asm/kvm_pgtable.h +10
-0@@ -754,4 +754,14 @@ enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);* kvm_pgtable_prot format. */ enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot (kvm_pte_t pte); + +/** +* kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries +* +* @mmu: Stage-2 KVM MMU struct +* @addr: The base Intermediate physical address from which to invalidate +* @size: Size of the range from the base to invalidate +*/ +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size);#endif /* __ARM64_KVM_PGTABLE_H__ */ arch/arm64/kvm/hyp/pgtable.c +16
-0@@ -670,6 +670,26 @@ static bool stage2_has_fwb(struct kvm_pgtable *pgt)return !(pgt->flags &KVM_PGTABLE_S2_NOFWB ); } +void kvm_tlb_flush_vmid_range (struct kvm_s2_mmu *mmu, +phys_addr_t addr,size_t size) +{ +unsigned long pages, inval_pages; + +if (!system_supports_tlb_range ()) { +kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +return ; + } + + pages = size >>PAGE_SHIFT ; +while (pages >0 ) { + inval_pages =min (pages,MAX_TLBI_RANGE_PAGES ); +kvm_call_hyp (__kvm_tlb_flush_vmid_range, mmu, addr, inval_pages); + + addr += inval_pages <<PAGE_SHIFT ;Quoted 8 lines
arch/arm64/kvm/hyp/pgtable.c static bool stage2_has_fwb(struct kvm_pgtable *pgt) · line 692
Quoted 4 lines
arch/arm64/kvm/hyp/pgtable.c +4
-0+ pages -= inval_pages; + } +} +This really shouldn't live in
pgtable.c. This code gets linked into the EL2 object. What do you think happens if, for some reason, this gets called *from EL2*?Quoted 6 lines
Ah, actually, nothing too bad would happen, as we convert the
kvm_call_hyp()into a function call.But still, we don't need two copies of this stuff, and it can live in
mmu.c.
Quoted 8 lines
But since we have a couple of references in
pgtable.ctokvm_tlb_flush_vmid_range(), wouldn't that be an (linking) issue if we moved the definition tommu.c?ld: error: undefined symbol:
__kvm_nvhe_kvm_tlb_flush_vmid_rangeQuoted 2 lines
Quoted 2 lines
Quoted 2 lines
arch/arm64/kvm/hyp/pgtable.c +0
-0referenced by pgtable.c:1148 (./arch/arm64/kvm/hyp/nvhe/../pgtable.c:1148) arch/arm64/kvm/hyp/nvhe/kvm_nvhe.o:(__kvm_nvhe_kvm_pgtable_stage2_unmap) in archive vmlinux.a...
Ah crap, I missed that. What a mess. Forget it then. It really is a shame that all the neat separation between mmu.c and pgtable.c that we were aiming for is ultimately lost.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v7 9/12 KVM: arm64: Implement kvm_arch_flush_remote_tlbs_range()
In reply to Raghavendra Rao AnantaImplement kvm_arch_flush_remote_tlbs_range() for arm64 to invalidate the given range in the TLB.
--- arch/arm64/include/asm/kvm_host.h | 3 +++ arch/arm64/kvm/mmu.c | 7 +++++++ 2 files changed, 10 insertions(+)
arch/arm64/include/asm/kvm_host.h +3 -0
@@ -1114,6 +1114,9 @@ struct kvm *kvm_arch_alloc_vm(void);#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSint kvm_arch_flush_remote_tlbs (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages); +static inline bool kvm_vm_is_protected (struct kvm *kvm) {return false;
arch/arm64/kvm/mmu.c +7 -0
@@ -172,6 +172,13 @@ int kvm_arch_flush_remote_tlbs(struct kvm *kvm)return 0 ; } +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages) +{ +kvm_tlb_flush_vmid_range (&kvm->arch.mmu, + start_gfn <<PAGE_SHIFT , pages <<PAGE_SHIFT ); +return 0 ; +} +static bool kvm_is_device_pfn (unsigned long pfn) {return !pfn_is_map_memory (pfn);
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Marc Zyngier
maintainerRe: [PATCH v7 09/12] KVM: arm64: Implement kvm_arch_flush_remote_tlbs_range()
Quoted 17 lines · Raghavendra Rao Ananta
Implement
kvm_arch_flush_remote_tlbs_range()for arm64 to invalidate the given range in the TLB.Signed-off-by Raghavendra Rao Ananta <[email protected]>Reviewed-by Gavin Shan <[email protected]>Reviewed-by Shaoqin Huang <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 3 +++ arch/arm64/kvm/mmu.c | 7 +++++++ 2 files changed, 10 insertions(+)arch/arm64/include/asm/kvm_host.h +0
-0
arch/arm64/include/asm/kvm_host.h struct kvm *kvm_arch_alloc_vm(void); · line 1118
Quoted 5 lines
arch/arm64/include/asm/kvm_host.h +2
-0@@ -1114,6 +1114,9 @@ struct kvm *kvm_arch_alloc_vm(void);#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBSint kvm_arch_flush_remote_tlbs (struct kvm *kvm); +#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE +int kvm_arch_flush_remote_tlbs_range (struct kvm *kvm,gfn_t start_gfn,u64 pages);
As per the other patches, these prototypes should be global and not arch-specific.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v7 10/12 KVM: arm64: Flush only the memslot after write-protect
In reply to Raghavendra Rao AnantaAfter write-protecting the region, currently KVM invalidates the entire TLB entries using kvm_flush_remote_tlbs(). Instead, scope the invalidation only to the targeted memslot. If supported, the architecture would use the range-based TLBI instructions to flush the memslot or else fallback to flushing all of the TLBs.
--- arch/arm64/kvm/mmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kvm/mmu.c +1 -1
@@ -1082,7 +1082,7 @@ static void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot)write_lock (&kvm->mmu_lock);stage2_wp_range (&kvm->arch.mmu, start, end);write_unlock (&kvm->mmu_lock); -kvm_flush_remote_tlbs (kvm); +kvm_flush_remote_tlbs_memslot (kvm, memslot); }/**
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
PATCH v7 11/12 KVM: arm64: Invalidate the table entries upon a range
In reply to Raghavendra Rao AnantaCurrently, during the operations such as a hugepage collapse, KVM would flush the entire VM's context using 'vmalls12e1is' TLBI operation. Specifically, if the VM is faulting on many hugepages (say after dirty-logging), it creates a performance penalty for the guest whose pages have already been faulted earlier as they would have to refill their TLBs again.
Instead, leverage kvm_tlb_flush_vmid_range() for table entries. If the system supports it, only the required range will be flushed. Else, it'll fallback to the previous mechanism.
--- arch/arm64/kvm/hyp/pgtable.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm64/kvm/hyp/pgtable.c +2 -1
@@ -806,7 +806,8 @@ static bool stage2_try_break_pte(const struct kvm_pgtable_visit_ctx *ctx,* evicted pte value (if any). */ if (kvm_pte_table (ctx->old, ctx->level)) -kvm_call_hyp (__kvm_tlb_flush_vmid, mmu); +kvm_tlb_flush_vmid_range (mmu, ctx->addr, +kvm_granule_size (ctx->level));else if (kvm_pte_valid (ctx->old))kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level);
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
PATCH v7 12/12 KVM: arm64: Use TLBI range-based intructions for unmap
In reply to Raghavendra Rao AnantaThe current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.
Rename stage2_put_pte() to stage2_unmap_put_pte() as the function now serves the stage-2 unmap walker specifically, rather than acting generic.
--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)
arch/arm64/kvm/hyp/pgtable.c +58 -9
@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with -* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for +* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data)) +kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level); } mm_ops->put_page (ctx->ptep);
@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;
@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),
@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , }; -return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {
Signature · 1 line
2.41.0.487.g6d72f3e995-goog
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Hi Raghavendra,
Quoted 70 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +36
-3@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 879
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +5
-2-* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for +* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data))
Why not directly check (unmap_data->defer_tlb_flush_init) here?
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 881
Quoted 2 lines
arch/arm64/kvm/hyp/pgtable.c +2
-0+kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level);
Small indent hint. The ctx->addr can align with __kvm_tlb_flush_vmid_ipa.
Thanks,
Shaoqin
Quoted 48 lines
arch/arm64/kvm/hyp/pgtable.c +15
-4} mm_ops->put_page (ctx->ptep);@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , }; -return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {
Signature · 1 line
Shaoqin
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Quoted 77 lines · On Mon, Jul 24, 2023 at 2:35?AM Shaoqin Huang <[email protected]> wrote:
Hi Raghavendra,
Quoted 73 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +38
-4@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with -* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for
Quoted 9 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 879
Quoted 7 lines
arch/arm64/kvm/hyp/pgtable.c +3
-1+* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data))Why not directly check (
unmap_data->defer_tlb_flush_init) here?
(Re-sending the reply as the previous one was formatted as HTML and was blocked by many lists)
No particular reason per say, but I was just going with the logic of determining if we need to defer the flush and the WARN_ON() parts separate.
Any advantage if we directly check in stage2_unmap_put_pte() that I missed or is this purely for readability?
Quoted 4 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 881
Quoted 2 lines
arch/arm64/kvm/hyp/pgtable.c +2
-0+kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level);Small indent hint. The
ctx->addrcan align with__kvm_tlb_flush_vmid_ipa.
Ah, yes. I'll adjust this if I send out a v8.
Thank you.
Raghavendra
Quoted 54 lines
Thanks,
ShaoqinQuoted 38 lines
arch/arm64/kvm/hyp/pgtable.c +9
-3} mm_ops->put_page (ctx->ptep);@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , };arch/arm64/kvm/hyp/pgtable.c static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx, · line 1173
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +6
-1-return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {--
Shaoqin
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Quoted 87 lines · Raghavendra Rao Ananta
Quoted 77 lines · On Mon, Jul 24, 2023 at 2:35?AM Shaoqin Huang <[email protected]> wrote:
Hi Raghavendra,
Quoted 73 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +38
-4@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with -* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for Quoted 9 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 879
Quoted 7 lines
arch/arm64/kvm/hyp/pgtable.c +3
-1+* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data))Why not directly check (
unmap_data->defer_tlb_flush_init) here?
Quoted 8 lines
(Re-sending the reply as the previous one was formatted as HTML and was blocked by many lists)
No particular reason per say, but I was just going with the logic of determining if we need to defer the flush and the
WARN_ON()parts separate.
Any advantage if we directly check instage2_unmap_put_pte()that I missed or is this purely for readability?
Hi Raghavendra,
I just wondering if before the stage2 walk, we want to defer the tlb flush, but if when walk the stage2, the stage2_unmap_defer_tlb_flush() trigger the WARN_ON() and return don't defer the tlb flush, it will flush the ctx->addr's tlb. But before the WARN_ON() triggered, these tlb will not be flushed, since when walk stage2 done in the kvm_pgtable_stage2_unmap(), the stage2_unmap_defer_tlb_flush() still trigger the WARN_ON() and don't use the tlb range-based flush. Thus some of the tlb are not flushed.
If we directly check the (unmap_data->defer_tlb_flush_init), this isn't change during walking the stage2, so the WARN_ON() should only trigger in kvm_pgtable_stage2_unmap()->stage2_unmap_defer_tlb_flush().
I'm not sure if it's right since I just think once we set up use the TLBI range-based flush, the result of the __stage2_unmap_defer_tlb_flush() shouldn't change. Otherwise there must have some stale TLB entry.
Thanks,
Shaoqin
Quoted 64 lines
Quoted 5 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 881
Quoted 3 lines
arch/arm64/kvm/hyp/pgtable.c +2
-0+kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level);Small indent hint. The
ctx->addrcan align with__kvm_tlb_flush_vmid_ipa.Ah, yes. I'll adjust this if I send out a v8.
Thank you.
RaghavendraQuoted 54 lines
Thanks,
ShaoqinQuoted 38 lines
arch/arm64/kvm/hyp/pgtable.c +9
-3} mm_ops->put_page (ctx->ptep);@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , };arch/arm64/kvm/hyp/pgtable.c static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx, · line 1173
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +6
-1-return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {--
Shaoqin
Signature · 1 line
Shaoqin
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Hi Shaoqin,
Quoted 102 lines · On Mon, Jul 24, 2023 at 7:32?PM Shaoqin Huang <[email protected]> wrote:
Quoted 87 lines · Raghavendra Rao Ananta
Quoted 77 lines · On Mon, Jul 24, 2023 at 2:35?AM Shaoqin Huang <[email protected]> wrote:
Hi Raghavendra,
Quoted 73 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +38
-4@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with -* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for Quoted 9 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 879
Quoted 7 lines
arch/arm64/kvm/hyp/pgtable.c +3
-1+* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data))Why not directly check (
unmap_data->defer_tlb_flush_init) here?Quoted 8 lines
(Re-sending the reply as the previous one was formatted as HTML and was blocked by many lists)
No particular reason per say, but I was just going with the logic of determining if we need to defer the flush and the
WARN_ON()parts separate.
Any advantage if we directly check instage2_unmap_put_pte()that I missed or is this purely for readability?Hi Raghavendra,
Quoted 9 lines
I just wondering if before the stage2 walk, we want to defer the tlb flush, but if when walk the stage2, the
stage2_unmap_defer_tlb_flush()trigger theWARN_ON()and return don't defer the tlb flush, it will flush the ctx->addr's tlb. But before theWARN_ON()triggered, these tlb will not be flushed, since when walk stage2 done in thekvm_pgtable_stage2_unmap(), thestage2_unmap_defer_tlb_flush()still trigger theWARN_ON()and don't use the tlb range-based flush. Thus some of the tlb are not flushed.
Excellent point!
Quoted 9 lines
If we directly check the (
unmap_data->defer_tlb_flush_init), this isn't change during walking the stage2, so theWARN_ON()should only trigger in kvm_pgtable_stage2_unmap()->stage2_unmap_defer_tlb_flush().I'm not sure if it's right since I just think once we set up use the TLBI range-based flush, the result of the
__stage2_unmap_defer_tlb_flush()shouldn't change. Otherwise there must have some stale TLB entry.
One solution that I can think of is, if the code triggers the WARN_ON(), we flush the entire VM's TLB using kvm_call_hyp(__kvm_tlb_flush_vmid) after the entire walk is finished. In this special/rare situation, it'll be a little expensive, but we will at least be correct, leaving no stale TLBs behind. WDYT?
Thank you.
Raghavendra
Quoted 71 lines
Thanks,
ShaoqinQuoted 64 lines
Quoted 5 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 881
Quoted 3 lines
arch/arm64/kvm/hyp/pgtable.c +2
-0+kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level);Small indent hint. The
ctx->addrcan align with__kvm_tlb_flush_vmid_ipa.Ah, yes. I'll adjust this if I send out a v8.
Thank you.
RaghavendraQuoted 54 lines
Thanks,
ShaoqinQuoted 38 lines
arch/arm64/kvm/hyp/pgtable.c +9
-3} mm_ops->put_page (ctx->ptep);@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , };arch/arm64/kvm/hyp/pgtable.c static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx, · line 1173
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +6
-1-return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {--
Shaoqin--
Shaoqin
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Hi Raghavendra,
Quoted 124 lines · Raghavendra Rao Ananta
Hi Shaoqin,
Quoted 102 lines · On Mon, Jul 24, 2023 at 7:32?PM Shaoqin Huang <[email protected]> wrote:
Quoted 87 lines · Raghavendra Rao Ananta
Quoted 77 lines · On Mon, Jul 24, 2023 at 2:35?AM Shaoqin Huang <[email protected]> wrote:
Hi Raghavendra,
Quoted 73 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +38
-4@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data) +{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush; +} + +static void stage2_unmap_put_pte (const struct kvm_pgtable_visit_ctx *ctx, +struct kvm_s2_mmu *mmu, +struct kvm_pgtable_mm_ops *mm_ops) { +struct stage2_unmap_data *unmap_data = ctx->arg; +/* -* Clear the existing PTE, and perform break-before-make with -* TLB maintenance if it was valid. +* Clear the existing PTE, and perform break-before-make if it was +* valid. Depending on the system support, the TLB maintenance for Quoted 9 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 879
Quoted 7 lines
arch/arm64/kvm/hyp/pgtable.c +3
-1+* the same can be deferred until the entire unmap is completed. */ if (kvm_pte_valid (ctx->old)) {kvm_clear_pte (ctx->ptep); -kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level); + +if (!stage2_unmap_defer_tlb_flush (unmap_data))Why not directly check (
unmap_data->defer_tlb_flush_init) here?Quoted 8 lines
(Re-sending the reply as the previous one was formatted as HTML and was blocked by many lists)
No particular reason per say, but I was just going with the logic of determining if we need to defer the flush and the
WARN_ON()parts separate.
Any advantage if we directly check instage2_unmap_put_pte()that I missed or is this purely for readability?Hi Raghavendra,
Quoted 9 lines
I just wondering if before the stage2 walk, we want to defer the tlb flush, but if when walk the stage2, the
stage2_unmap_defer_tlb_flush()trigger theWARN_ON()and return don't defer the tlb flush, it will flush the ctx->addr's tlb. But before theWARN_ON()triggered, these tlb will not be flushed, since when walk stage2 done in thekvm_pgtable_stage2_unmap(), thestage2_unmap_defer_tlb_flush()still trigger theWARN_ON()and don't use the tlb range-based flush. Thus some of the tlb are not flushed.Excellent point!
Quoted 9 lines
If we directly check the (
unmap_data->defer_tlb_flush_init), this isn't change during walking the stage2, so theWARN_ON()should only trigger in kvm_pgtable_stage2_unmap()->stage2_unmap_defer_tlb_flush().I'm not sure if it's right since I just think once we set up use the TLBI range-based flush, the result of the
__stage2_unmap_defer_tlb_flush()shouldn't change. Otherwise there must have some stale TLB entry.
Quoted 6 lines
One solution that I can think of is, if the code triggers the
WARN_ON(), we flush the entire VM's TLB usingkvm_call_hyp(__kvm_tlb_flush_vmid)after the entire walk is finished. In this special/rare situation, it'll be a little expensive, but we will at least be correct, leaving no stale TLBs behind. WDYT?
I think it will be good to have this handling.
Thanks,
Shaoqin
Quoted 74 lines
Thank you.
RaghavendraQuoted 71 lines
Thanks,
ShaoqinQuoted 64 lines
Quoted 5 lines
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 881
Quoted 3 lines
arch/arm64/kvm/hyp/pgtable.c +2
-0+kvm_call_hyp (__kvm_tlb_flush_vmid_ipa, mmu, + ctx->addr, ctx->level);Small indent hint. The
ctx->addrcan align with__kvm_tlb_flush_vmid_ipa.Ah, yes. I'll adjust this if I send out a v8.
Thank you.
RaghavendraQuoted 54 lines
Thanks,
ShaoqinQuoted 38 lines
arch/arm64/kvm/hyp/pgtable.c +9
-3} mm_ops->put_page (ctx->ptep);@@ -1070,7 +1108,8 @@ int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,static int stage2_unmap_walker (const struct kvm_pgtable_visit_ctx *ctx,enum kvm_pgtable_walk_flags visit) { -struct kvm_pgtable *pgt = ctx->arg; +struct stage2_unmap_data *unmap_data = ctx->arg; +struct kvm_pgtable *pgt = unmap_data->pgt;struct kvm_s2_mmu *mmu = pgt->mmu;struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;kvm_pte_t *childp =NULL ;@@ -1098,7 +1137,7 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,* block entry and rely on the remaining portions being faulted * back lazily. */ -stage2_put_pte (ctx, mmu, mm_ops); +stage2_unmap_put_pte (ctx, mmu, mm_ops);if (need_flush && mm_ops->dcache_clean_inval_poc) mm_ops->dcache_clean_inval_poc(kvm_pte_follow(ctx->old, mm_ops),@@ -1112,13 +1151,23 @@ static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx,int kvm_pgtable_stage2_unmap (struct kvm_pgtable *pgt,u64 addr,u64 size) { +int ret; +struct stage2_unmap_data unmap_data = { + .pgt = pgt, + .defer_tlb_flush_init =__stage2_unmap_defer_tlb_flush (pgt), + };struct kvm_pgtable_walker walker = { .cb = stage2_unmap_walker, - .arg = pgt, + .arg = &unmap_data, .flags =KVM_PGTABLE_WALK_LEAF |KVM_PGTABLE_WALK_TABLE_POST , };arch/arm64/kvm/hyp/pgtable.c static int stage2_unmap_walker(const struct kvm_pgtable_visit_ctx *ctx, · line 1173
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +6
-1-return kvm_pgtable_walk (pgt, addr, size, &walker); + ret =kvm_pgtable_walk (pgt, addr, size, &walker); +if (stage2_unmap_defer_tlb_flush (&unmap_data)) +/* Perform the deferred TLB invalidations */ +kvm_tlb_flush_vmid_range (pgt->mmu, addr, size); + +return ret; }struct stage2_attr_data {--
Shaoqin--
Shaoqin
Signature · 1 line
Shaoqin
Marc Zyngier
maintainerRe: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
In reply to Raghavendra Rao AnantaQuoted 51 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +19
-2@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data)
arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 862
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +10
-0+{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush;
I really don't understand what you're testing here. The ability to defer TLB invalidation is a function of the system capabilities (range+FWB) and a single flag that is only set on the host for pKVM.
How could that change in the middle of the life of the system? if further begs the question about the need for the unmap_data data structure.
It looks to me that we could simply pass the pgt pointer around and be done with it. Am I missing something obvious?
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Quoted 69 lines · On Thu, Jul 27, 2023 at 6:12?AM Marc Zyngier <[email protected]> wrote:
Quoted 51 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +19
-2@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data)arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 862
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +10
-0+{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush;I really don't understand what you're testing here. The ability to defer TLB invalidation is a function of the system capabilities (range+FWB) and a single flag that is only set on the host for pKVM.
Quoted 7 lines
How could that change in the middle of the life of the system? if further begs the question about the need for the
unmap_datadata structure.It looks to me that we could simply pass the pgt pointer around and be done with it. Am I missing something obvious?
From one of the previous comments [1] (used in a different context), I'm given to understand that since these feature checks are governed by alternative patching, they can potentially change (at runtime?). Is that not the case and I have misunderstood the idea in comment [1] entirely? Is it solely used for optimization purposes and set only once?
If that's the case, I can get rid of the WARN_ON() and unmap_data.
- Raghavendra
[1]: https://lore.kernel.org/all/ZGPPj1AXS0Uah2Ug at linux.dev/
Quoted 4 lines
M.--
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Quoted 78 lines · Raghavendra Rao Ananta
Quoted 69 lines · On Thu, Jul 27, 2023 at 6:12?AM Marc Zyngier <[email protected]> wrote:
Quoted 51 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +19
-2@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data)arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 862
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +10
-0+{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush;I really don't understand what you're testing here. The ability to defer TLB invalidation is a function of the system capabilities (range+FWB) and a single flag that is only set on the host for pKVM.
Quoted 7 lines
How could that change in the middle of the life of the system? if further begs the question about the need for the
unmap_datadata structure.It looks to me that we could simply pass the pgt pointer around and be done with it. Am I missing something obvious?
Quoted 6 lines
From one of the previous comments [1] (used in a different context), I'm given to understand that since these feature checks are governed by alternative patching, they can potentially change (at runtime?). Is that not the case and I have misunderstood the idea in comment [1] entirely? Is it solely used for optimization purposes and set only once?
Alternative patching, just like the static branches used to implement the capability stuff, is a one way street. At the point where KVM is initialised, these configurations are set in stone, and there is no going back.
Quoted 1 line
If that's the case, I can get rid of the
WARN_ON()andunmap_data.
yes, please.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v7 12/12] KVM: arm64: Use TLBI range-based intructions for unmap
Quoted 88 lines · On Wed, Aug 2, 2023 at 4:28?PM Marc Zyngier <[email protected]> wrote:
Quoted 78 lines · Raghavendra Rao Ananta
Quoted 69 lines · On Thu, Jul 27, 2023 at 6:12?AM Marc Zyngier <[email protected]> wrote:
Quoted 51 lines · Raghavendra Rao Ananta
The current implementation of the stage-2 unmap walker traverses the given range and, as a part of break-before-make, performs TLB invalidations with a DSB for every PTE. A multitude of this combination could cause a performance bottleneck on some systems.
Hence, if the system supports
FEAT_TLBIRANGE, defer the TLB invalidations until the entire walk is finished, and then use range-based instructions to invalidate the TLBs in one go. Condition deferred TLB invalidation on the system supporting FWB, as the optimization is entirely pointless when the unmap walker needs to perform CMOs.Rename
stage2_put_pte()tostage2_unmap_put_pte()as the function now serves the stage-2 unmap walker specifically, rather than acting generic.Signed-off-by Raghavendra Rao Ananta <[email protected]>--- arch/arm64/kvm/hyp/pgtable.c | 67 +++++++++++++++++++++++++++++++----- 1 file changed, 58 insertions(+), 9 deletions(-)arch/arm64/kvm/hyp/pgtable.c +19
-2@@ -831,16 +831,54 @@ static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t nsmp_store_release (ctx->ptep, new); } -static void stage2_put_pte (const struct kvm_pgtable_visit_ctx *ctx,struct kvm_s2_mmu *mmu, -struct kvm_pgtable_mm_ops *mm_ops) +struct stage2_unmap_data { +struct kvm_pgtable *pgt; +bool defer_tlb_flush_init; +}; + +static bool __stage2_unmap_defer_tlb_flush (struct kvm_pgtable *pgt) +{ +/* +* If FEAT_TLBIRANGE is implemented, defer the individual +* TLB invalidations until the entire walk is finished, and +* then use the range-based TLBI instructions to do the +* invalidations. Condition deferred TLB invalidation on the +* system supporting FWB, as the optimization is entirely +* pointless when the unmap walker needs to perform CMOs. +*/ +return system_supports_tlb_range () &&stage2_has_fwb (pgt); +} + +static bool stage2_unmap_defer_tlb_flush (struct stage2_unmap_data *unmap_data)arch/arm64/kvm/hyp/pgtable.c static void stage2_make_pte(const struct kvm_pgtable_visit_ctx *ctx, kvm_pte_t n · line 862
Quoted 10 lines
arch/arm64/kvm/hyp/pgtable.c +10
-0+{ +bool defer_tlb_flush =__stage2_unmap_defer_tlb_flush (unmap_data->pgt); + +/* +* Since __stage2_unmap_defer_tlb_flush() is based on alternative +* patching and the TLBIs' operations behavior depend on this, +* track if there's any change in the state during the unmap sequence. +*/ +WARN_ON (unmap_data->defer_tlb_flush_init != defer_tlb_flush); +return defer_tlb_flush;I really don't understand what you're testing here. The ability to defer TLB invalidation is a function of the system capabilities (range+FWB) and a single flag that is only set on the host for pKVM.
Quoted 7 lines
How could that change in the middle of the life of the system? if further begs the question about the need for the
unmap_datadata structure.It looks to me that we could simply pass the pgt pointer around and be done with it. Am I missing something obvious?
Quoted 6 lines
From one of the previous comments [1] (used in a different context), I'm given to understand that since these feature checks are governed by alternative patching, they can potentially change (at runtime?). Is that not the case and I have misunderstood the idea in comment [1] entirely? Is it solely used for optimization purposes and set only once?
Quoted 5 lines
Alternative patching, just like the static branches used to implement the capability stuff, is a one way street. At the point where KVM is initialised, these configurations are set in stone, and there is no going back.
Understood.
Quoted 4 lines
Quoted 1 line
If that's the case, I can get rid of the
WARN_ON()andunmap_data.yes, please.
Sure, I'll get rid of the WARN_ON and 'struct stage2_unmap_data' in v8.
Thanks,
Raghavendra
Quoted 6 lines
Thanks,
M.--
Without deviation from the norm, progress is not possible.
Sean Christopherson
maintainerQuoted 14 lines · Raghavendra Rao Ananta
arch/arm64/include/asm/kvm_asm.h | 3 + arch/arm64/include/asm/kvm_host.h | 6 ++ arch/arm64/include/asm/kvm_pgtable.h | 10 +++ arch/arm64/include/asm/tlbflush.h | 109 ++++++++++++++------------- arch/arm64/kvm/Kconfig | 1 - arch/arm64/kvm/arm.c | 6 -- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 11 +++ arch/arm64/kvm/hyp/nvhe/tlb.c | 30 ++++++++ arch/arm64/kvm/hyp/pgtable.c | 90 +++++++++++++++++++--- arch/arm64/kvm/hyp/vhe/tlb.c | 27 +++++++ arch/arm64/kvm/mmu.c | 15 +++- arch/mips/include/asm/kvm_host.h | 4 +- arch/mips/kvm/mips.c | 12 +-- arch/riscv/kvm/mmu.c | 6 --
Quoted 8 lines
arch/x86/include/asm/kvm_host.h | 7 +- arch/x86/kvm/mmu/mmu.c | 25 ++---- arch/x86/kvm/mmu/mmu_internal.h | 3 - arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 20 +++-- virt/kvm/Kconfig | 3 - virt/kvm/kvm_main.c | 35 +++++++-- 21 files changed, 294 insertions(+), 131 deletions(-)
Unless I've missed something, nothing in this series conflicts with anything that's on the horizon for x86, so feel free to take this through the ARM tree once we've emerged from behind the bikeshed :-)