The main change is to move secondary TLB invalidation mmu notifier
callbacks into the architecture specific TLB flushing functions. This
makes secondary TLB invalidation mostly match CPU invalidation while
still allowing efficient range based invalidations based on the
existing TLB batching code.
Changes for v4:
- Fixed a NULL pointer dereference when registering the first
notifier with mmu_interval_notifier_insert() instead of
mmu_notifier_register() - Thanks Dan and Chaitanya for the bug
reports.
- Collected Acked/Reviewed tags.
- Don't call the notifier from radix__local_flush_tlb_page() on
PowerPC.
Changes for v3:
- On x86 call the invalidation when adding pending TLB invalidates
rather than when flushing the batch. This is because the mm is
required. It also matches what happens on ARM64. Fixes a bug
reported by SeongJae Park (thanks!)
Changes for v2:
- Rebased on linux-next commit 906fa30154ef ("mm/rmap: correct stale
comment of rmap_walk_anon and rmap_walk_file") to fix a minor
integration conflict with "arm64: support batched/deferred tlb
shootdown during page reclamation/migration". This series will need
to be applied after the conflicting patch.
- Reordered the function rename until the end of the series as many
places that were getting renamed ended up being removed anyway.
- Fixed a couple of build issues which broke bisection.
- Added a minor patch to fix up a stale/incorrect comment.
==========
Background
==========
The arm64 architecture specifies TLB permission bits may be cached and
therefore the TLB must be invalidated during permission upgrades. For
the CPU this currently occurs in the architecture specific
ptep_set_access_flags() routine.
Secondary TLBs such as implemented by the SMMU IOMMU match the CPU
architecture specification and may also cache permission bits and
require the same TLB invalidations. This may be achieved in one of two
ways.
Some SMMU implementations implement broadcast TLB maintenance
(BTM). This snoops CPU TLB invalidates and will invalidate any
secondary TLB at the same time as the CPU. However implementations are
not required to implement BTM.
Implementations without BTM rely on mmu notifier callbacks to send
explicit TLB invalidation commands to invalidate SMMU TLB. Therefore
either generic kernel code or architecture specific code needs to call
the mmu notifier on permission upgrade.
Currently that doesn't happen so devices will fault indefinitely when
writing to a PTE that was previously read-only as nothing invalidates
the SMMU TLB.
========
Solution
========
To fix this the series first renames the .invalidate_range() callback
to .arch_invalidate_secondary_tlbs() as suggested by Jason and Sean to
make it clear this callback is only used for secondary TLBs. That was
made possible thanks to Sean's series [1] to remove KVM's incorrect
usage.
Based on feedback from Jason [2] the proposed solution to the bug is
to move the calls to mmu_notifier_arch_invalidate_secondary_tlbs()
closer to the architecture specific TLB invalidation code. This
ensures the secondary TLB won't miss invalidations, including the
existing invalidation in the ARM64 code to deal with permission
upgrade.
Currently only ARM64, PowerPC and x86 have IOMMU with secondary TLBs
requiring SW invalidation so the notifier is only called for those
architectures. It is also not called for invalidation of kernel
mappings as no secondary IOMMU implementations can access those and
hence it is not required.
[1] - https://lore.kernel.org/all/[email protected]/
[2] - https://lore.kernel.org/linux-mm/[email protected]/
Alistair Popple (5):
arm64/smmu: Use TLBI ASID when invalidating entire range
mmu_notifiers: Fixup comment in mmu_interval_read_begin()
mmu_notifiers: Call invalidate_range() when invalidating TLBs
mmu_notifiers: Don't invalidate secondary TLBs as part of mmu_notifier_invalidate_range_end()
mmu_notifiers: Rename invalidate_range notifier
arch/arm64/include/asm/tlbflush.h | 5 +-
arch/powerpc/include/asm/book3s/64/tlbflush.h | 1 +-
arch/powerpc/mm/book3s64/radix_hugetlbpage.c | 1 +-
arch/powerpc/mm/book3s64/radix_tlb.c | 4 +-
arch/x86/include/asm/tlbflush.h | 2 +-
arch/x86/mm/tlb.c | 2 +-
drivers/iommu/amd/iommu_v2.c | 10 +-
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c | 29 +++--
drivers/iommu/intel/svm.c | 8 +-
drivers/misc/ocxl/link.c | 8 +-
include/asm-generic/tlb.h | 1 +-
include/linux/mmu_notifier.h | 104 ++++-------------
kernel/events/uprobes.c | 2 +-
mm/huge_memory.c | 29 +----
mm/hugetlb.c | 8 +-
mm/memory.c | 8 +-
mm/migrate_device.c | 9 +-
mm/mmu_notifier.c | 50 +++-----
mm/rmap.c | 40 +-------
19 files changed, 110 insertions(+), 211 deletions(-)
base-commit: 906fa30154ef42f93d28d7322860e76c6ae392ac
--
git-series 0.9.1
Secondary TLBs are now invalidated from the architecture specific TLB
invalidation functions. Therefore there is no need to explicitly
notify or invalidate as part of the range end functions. This means we
can remove mmu_notifier_invalidate_range_end_only() and some of the
ptep_*_notify() functions.
Signed-off-by: Alistair Popple <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]>
---
include/linux/mmu_notifier.h | 56 +------------------------------------
kernel/events/uprobes.c | 2 +-
mm/huge_memory.c | 25 ++---------------
mm/hugetlb.c | 1 +-
mm/memory.c | 8 +----
mm/migrate_device.c | 9 +-----
mm/mmu_notifier.c | 25 ++---------------
mm/rmap.c | 40 +--------------------------
8 files changed, 14 insertions(+), 152 deletions(-)
diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h
index 64a3e05..f2e9edc 100644
--- a/include/linux/mmu_notifier.h
+++ b/include/linux/mmu_notifier.h
@@ -395,8 +395,7 @@ extern int __mmu_notifier_test_young(struct mm_struct *mm,
extern void __mmu_notifier_change_pte(struct mm_struct *mm,
unsigned long address, pte_t pte);
extern int __mmu_notifier_invalidate_range_start(struct mmu_notifier_range *r);
-extern void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *r,
- bool only_end);
+extern void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *r);
extern void __mmu_notifier_invalidate_range(struct mm_struct *mm,
unsigned long start, unsigned long end);
extern bool
@@ -481,14 +480,7 @@ mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
might_sleep();
if (mm_has_notifiers(range->mm))
- __mmu_notifier_invalidate_range_end(range, false);
-}
-
-static inline void
-mmu_notifier_invalidate_range_only_end(struct mmu_notifier_range *range)
-{
- if (mm_has_notifiers(range->mm))
- __mmu_notifier_invalidate_range_end(range, true);
+ __mmu_notifier_invalidate_range_end(range);
}
static inline void mmu_notifier_invalidate_range(struct mm_struct *mm,
@@ -582,45 +574,6 @@ static inline void mmu_notifier_range_init_owner(
__young; \
})
-#define ptep_clear_flush_notify(__vma, __address, __ptep) \
-({ \
- unsigned long ___addr = __address & PAGE_MASK; \
- struct mm_struct *___mm = (__vma)->vm_mm; \
- pte_t ___pte; \
- \
- ___pte = ptep_clear_flush(__vma, __address, __ptep); \
- mmu_notifier_invalidate_range(___mm, ___addr, \
- ___addr + PAGE_SIZE); \
- \
- ___pte; \
-})
-
-#define pmdp_huge_clear_flush_notify(__vma, __haddr, __pmd) \
-({ \
- unsigned long ___haddr = __haddr & HPAGE_PMD_MASK; \
- struct mm_struct *___mm = (__vma)->vm_mm; \
- pmd_t ___pmd; \
- \
- ___pmd = pmdp_huge_clear_flush(__vma, __haddr, __pmd); \
- mmu_notifier_invalidate_range(___mm, ___haddr, \
- ___haddr + HPAGE_PMD_SIZE); \
- \
- ___pmd; \
-})
-
-#define pudp_huge_clear_flush_notify(__vma, __haddr, __pud) \
-({ \
- unsigned long ___haddr = __haddr & HPAGE_PUD_MASK; \
- struct mm_struct *___mm = (__vma)->vm_mm; \
- pud_t ___pud; \
- \
- ___pud = pudp_huge_clear_flush(__vma, __haddr, __pud); \
- mmu_notifier_invalidate_range(___mm, ___haddr, \
- ___haddr + HPAGE_PUD_SIZE); \
- \
- ___pud; \
-})
-
/*
* set_pte_at_notify() sets the pte _after_ running the notifier.
* This is safe to start by updating the secondary MMUs, because the primary MMU
@@ -711,11 +664,6 @@ void mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
{
}
-static inline void
-mmu_notifier_invalidate_range_only_end(struct mmu_notifier_range *range)
-{
-}
-
static inline void mmu_notifier_invalidate_range(struct mm_struct *mm,
unsigned long start, unsigned long end)
{
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index f0ac5b8..3048589 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -193,7 +193,7 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr,
}
flush_cache_page(vma, addr, pte_pfn(ptep_get(pvmw.pte)));
- ptep_clear_flush_notify(vma, addr, pvmw.pte);
+ ptep_clear_flush(vma, addr, pvmw.pte);
if (new_page)
set_pte_at_notify(mm, addr, pvmw.pte,
mk_pte(new_page, vma->vm_page_prot));
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 762be2f..3ece117 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2003,7 +2003,7 @@ static void __split_huge_pud_locked(struct vm_area_struct *vma, pud_t *pud,
count_vm_event(THP_SPLIT_PUD);
- pudp_huge_clear_flush_notify(vma, haddr, pud);
+ pudp_huge_clear_flush(vma, haddr, pud);
}
void __split_huge_pud(struct vm_area_struct *vma, pud_t *pud,
@@ -2023,11 +2023,7 @@ void __split_huge_pud(struct vm_area_struct *vma, pud_t *pud,
out:
spin_unlock(ptl);
- /*
- * No need to double call mmu_notifier->invalidate_range() callback as
- * the above pudp_huge_clear_flush_notify() did already call it.
- */
- mmu_notifier_invalidate_range_only_end(&range);
+ mmu_notifier_invalidate_range_end(&range);
}
#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
@@ -2094,7 +2090,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
count_vm_event(THP_SPLIT_PMD);
if (!vma_is_anonymous(vma)) {
- old_pmd = pmdp_huge_clear_flush_notify(vma, haddr, pmd);
+ old_pmd = pmdp_huge_clear_flush(vma, haddr, pmd);
/*
* We are going to unmap this huge page. So
* just go ahead and zap it
@@ -2304,20 +2300,7 @@ void __split_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
out:
spin_unlock(ptl);
- /*
- * No need to double call mmu_notifier->invalidate_range() callback.
- * They are 3 cases to consider inside __split_huge_pmd_locked():
- * 1) pmdp_huge_clear_flush_notify() call invalidate_range() obvious
- * 2) __split_huge_zero_page_pmd() read only zero page and any write
- * fault will trigger a flush_notify before pointing to a new page
- * (it is fine if the secondary mmu keeps pointing to the old zero
- * page in the meantime)
- * 3) Split a huge pmd into pte pointing to the same page. No need
- * to invalidate secondary tlb entry they are all still valid.
- * any further changes to individual pte will notify. So no need
- * to call mmu_notifier->invalidate_range()
- */
- mmu_notifier_invalidate_range_only_end(&range);
+ mmu_notifier_invalidate_range_end(&range);
}
void split_huge_pmd_address(struct vm_area_struct *vma, unsigned long address,
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index dc1ec19..9c6e431 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5715,7 +5715,6 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
/* Break COW or unshare */
huge_ptep_clear_flush(vma, haddr, ptep);
- mmu_notifier_invalidate_range(mm, range.start, range.end);
page_remove_rmap(&old_folio->page, vma, true);
hugepage_add_new_anon_rmap(new_folio, vma, haddr);
if (huge_pte_uffd_wp(pte))
diff --git a/mm/memory.c b/mm/memory.c
index ad79039..8dca544 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -3158,7 +3158,7 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
* that left a window where the new PTE could be loaded into
* some TLBs while the old PTE remains in others.
*/
- ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
+ ptep_clear_flush(vma, vmf->address, vmf->pte);
folio_add_new_anon_rmap(new_folio, vma, vmf->address);
folio_add_lru_vma(new_folio, vma);
/*
@@ -3204,11 +3204,7 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
pte_unmap_unlock(vmf->pte, vmf->ptl);
}
- /*
- * No need to double call mmu_notifier->invalidate_range() callback as
- * the above ptep_clear_flush_notify() did already call it.
- */
- mmu_notifier_invalidate_range_only_end(&range);
+ mmu_notifier_invalidate_range_end(&range);
if (new_folio)
folio_put(new_folio);
diff --git a/mm/migrate_device.c b/mm/migrate_device.c
index e29626e..6c556b5 100644
--- a/mm/migrate_device.c
+++ b/mm/migrate_device.c
@@ -658,7 +658,7 @@ static void migrate_vma_insert_page(struct migrate_vma *migrate,
if (flush) {
flush_cache_page(vma, addr, pte_pfn(orig_pte));
- ptep_clear_flush_notify(vma, addr, ptep);
+ ptep_clear_flush(vma, addr, ptep);
set_pte_at_notify(mm, addr, ptep, entry);
update_mmu_cache(vma, addr, ptep);
} else {
@@ -763,13 +763,8 @@ static void __migrate_device_pages(unsigned long *src_pfns,
src_pfns[i] &= ~MIGRATE_PFN_MIGRATE;
}
- /*
- * No need to double call mmu_notifier->invalidate_range() callback as
- * the above ptep_clear_flush_notify() inside migrate_vma_insert_page()
- * did already call it.
- */
if (notified)
- mmu_notifier_invalidate_range_only_end(&range);
+ mmu_notifier_invalidate_range_end(&range);
}
/**
diff --git a/mm/mmu_notifier.c b/mm/mmu_notifier.c
index b7ad155..453a156 100644
--- a/mm/mmu_notifier.c
+++ b/mm/mmu_notifier.c
@@ -551,7 +551,7 @@ int __mmu_notifier_invalidate_range_start(struct mmu_notifier_range *range)
static void
mn_hlist_invalidate_end(struct mmu_notifier_subscriptions *subscriptions,
- struct mmu_notifier_range *range, bool only_end)
+ struct mmu_notifier_range *range)
{
struct mmu_notifier *subscription;
int id;
@@ -559,24 +559,6 @@ mn_hlist_invalidate_end(struct mmu_notifier_subscriptions *subscriptions,
id = srcu_read_lock(&srcu);
hlist_for_each_entry_rcu(subscription, &subscriptions->list, hlist,
srcu_read_lock_held(&srcu)) {
- /*
- * Call invalidate_range here too to avoid the need for the
- * subsystem of having to register an invalidate_range_end
- * call-back when there is invalidate_range already. Usually a
- * subsystem registers either invalidate_range_start()/end() or
- * invalidate_range(), so this will be no additional overhead
- * (besides the pointer check).
- *
- * We skip call to invalidate_range() if we know it is safe ie
- * call site use mmu_notifier_invalidate_range_only_end() which
- * is safe to do when we know that a call to invalidate_range()
- * already happen under page table lock.
- */
- if (!only_end && subscription->ops->invalidate_range)
- subscription->ops->invalidate_range(subscription,
- range->mm,
- range->start,
- range->end);
if (subscription->ops->invalidate_range_end) {
if (!mmu_notifier_range_blockable(range))
non_block_start();
@@ -589,8 +571,7 @@ mn_hlist_invalidate_end(struct mmu_notifier_subscriptions *subscriptions,
srcu_read_unlock(&srcu, id);
}
-void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range,
- bool only_end)
+void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
{
struct mmu_notifier_subscriptions *subscriptions =
range->mm->notifier_subscriptions;
@@ -600,7 +581,7 @@ void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range,
mn_itree_inv_end(subscriptions);
if (!hlist_empty(&subscriptions->list))
- mn_hlist_invalidate_end(subscriptions, range, only_end);
+ mn_hlist_invalidate_end(subscriptions, range);
lock_map_release(&__mmu_notifier_invalidate_range_start_map);
}
diff --git a/mm/rmap.c b/mm/rmap.c
index 1355bf6..51ec8aa 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -985,13 +985,6 @@ static int page_vma_mkclean_one(struct page_vma_mapped_walk *pvmw)
#endif
}
- /*
- * No need to call mmu_notifier_invalidate_range() as we are
- * downgrading page table protection not changing it to point
- * to a new page.
- *
- * See Documentation/mm/mmu_notifier.rst
- */
if (ret)
cleaned++;
}
@@ -1549,8 +1542,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
hugetlb_vma_unlock_write(vma);
flush_tlb_range(vma,
range.start, range.end);
- mmu_notifier_invalidate_range(mm,
- range.start, range.end);
/*
* The ref count of the PMD page was
* dropped which is part of the way map
@@ -1623,9 +1614,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
* copied pages.
*/
dec_mm_counter(mm, mm_counter(&folio->page));
- /* We have to invalidate as we cleared the pte */
- mmu_notifier_invalidate_range(mm, address,
- address + PAGE_SIZE);
} else if (folio_test_anon(folio)) {
swp_entry_t entry = { .val = page_private(subpage) };
pte_t swp_pte;
@@ -1637,9 +1625,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
folio_test_swapcache(folio))) {
WARN_ON_ONCE(1);
ret = false;
- /* We have to invalidate as we cleared the pte */
- mmu_notifier_invalidate_range(mm, address,
- address + PAGE_SIZE);
page_vma_mapped_walk_done(&pvmw);
break;
}
@@ -1670,9 +1655,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
*/
if (ref_count == 1 + map_count &&
!folio_test_dirty(folio)) {
- /* Invalidate as we cleared the pte */
- mmu_notifier_invalidate_range(mm,
- address, address + PAGE_SIZE);
dec_mm_counter(mm, MM_ANONPAGES);
goto discard;
}
@@ -1727,9 +1709,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
if (pte_uffd_wp(pteval))
swp_pte = pte_swp_mkuffd_wp(swp_pte);
set_pte_at(mm, address, pvmw.pte, swp_pte);
- /* Invalidate as we cleared the pte */
- mmu_notifier_invalidate_range(mm, address,
- address + PAGE_SIZE);
} else {
/*
* This is a locked file-backed folio,
@@ -1745,13 +1724,6 @@ static bool try_to_unmap_one(struct folio *folio, struct vm_area_struct *vma,
dec_mm_counter(mm, mm_counter_file(&folio->page));
}
discard:
- /*
- * No need to call mmu_notifier_invalidate_range() it has be
- * done above for all cases requiring it to happen under page
- * table lock before mmu_notifier_invalidate_range_end()
- *
- * See Documentation/mm/mmu_notifier.rst
- */
page_remove_rmap(subpage, vma, folio_test_hugetlb(folio));
if (vma->vm_flags & VM_LOCKED)
mlock_drain_local();
@@ -1930,8 +1902,6 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
hugetlb_vma_unlock_write(vma);
flush_tlb_range(vma,
range.start, range.end);
- mmu_notifier_invalidate_range(mm,
- range.start, range.end);
/*
* The ref count of the PMD page was
@@ -2036,9 +2006,6 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
* copied pages.
*/
dec_mm_counter(mm, mm_counter(&folio->page));
- /* We have to invalidate as we cleared the pte */
- mmu_notifier_invalidate_range(mm, address,
- address + PAGE_SIZE);
} else {
swp_entry_t entry;
pte_t swp_pte;
@@ -2102,13 +2069,6 @@ static bool try_to_migrate_one(struct folio *folio, struct vm_area_struct *vma,
*/
}
- /*
- * No need to call mmu_notifier_invalidate_range() it has be
- * done above for all cases requiring it to happen under page
- * table lock before mmu_notifier_invalidate_range_end()
- *
- * See Documentation/mm/mmu_notifier.rst
- */
page_remove_rmap(subpage, vma, folio_test_hugetlb(folio));
if (vma->vm_flags & VM_LOCKED)
mlock_drain_local();
--
git-series 0.9.1
The ARM SMMU has a specific command for invalidating the TLB for an
entire ASID. Currently this is used for the IO_PGTABLE API but not for
ATS when called from the MMU notifier.
The current implementation of notifiers does not attempt to invalidate
such a large address range, instead walking each VMA and invalidating
each range individually during mmap removal. However in future SMMU
TLB invalidations are going to be sent as part of the normal
flush_tlb_*() kernel calls. To better deal with that add handling to
use TLBI ASID when invalidating the entire address space.
Signed-off-by: Alistair Popple <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]>
---
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c | 16 +++++++++++++---
1 file changed, 13 insertions(+), 3 deletions(-)
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
index a5a63b1..2a19784 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
@@ -200,10 +200,20 @@ static void arm_smmu_mm_invalidate_range(struct mmu_notifier *mn,
* range. So do a simple translation here by calculating size correctly.
*/
size = end - start;
+ if (size == ULONG_MAX)
+ size = 0;
+
+ if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_BTM)) {
+ if (!size)
+ arm_smmu_tlb_inv_asid(smmu_domain->smmu,
+ smmu_mn->cd->asid);
+ else
+ arm_smmu_tlb_inv_range_asid(start, size,
+ smmu_mn->cd->asid,
+ PAGE_SIZE, false,
+ smmu_domain);
+ }
- if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_BTM))
- arm_smmu_tlb_inv_range_asid(start, size, smmu_mn->cd->asid,
- PAGE_SIZE, false, smmu_domain);
arm_smmu_atc_inv_domain(smmu_domain, mm->pasid, start, size);
}
--
git-series 0.9.1
There are two main use cases for mmu notifiers. One is by KVM which
uses mmu_notifier_invalidate_range_start()/end() to manage a software
TLB.
The other is to manage hardware TLBs which need to use the
invalidate_range() callback because HW can establish new TLB entries
at any time. Hence using start/end() can lead to memory corruption as
these callbacks happen too soon/late during page unmap.
mmu notifier users should therefore either use the start()/end()
callbacks or the invalidate_range() callbacks. To make this usage
clearer rename the invalidate_range() callback to
arch_invalidate_secondary_tlbs() and update documention.
Signed-off-by: Alistair Popple <[email protected]>
Suggested-by: Jason Gunthorpe <[email protected]>
Acked-by: Catalin Marinas <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]>
---
arch/arm64/include/asm/tlbflush.h | 6 +-
arch/powerpc/mm/book3s64/radix_hugetlbpage.c | 2 +-
arch/powerpc/mm/book3s64/radix_tlb.c | 8 +--
arch/x86/include/asm/tlbflush.h | 2 +-
arch/x86/mm/tlb.c | 2 +-
drivers/iommu/amd/iommu_v2.c | 10 ++--
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c | 13 ++---
drivers/iommu/intel/svm.c | 8 +--
drivers/misc/ocxl/link.c | 8 +--
include/linux/mmu_notifier.h | 48 +++++++++---------
mm/huge_memory.c | 4 +-
mm/hugetlb.c | 7 +--
mm/mmu_notifier.c | 21 ++++++--
13 files changed, 76 insertions(+), 63 deletions(-)
diff --git a/arch/arm64/include/asm/tlbflush.h b/arch/arm64/include/asm/tlbflush.h
index a99349d..84a05a0 100644
--- a/arch/arm64/include/asm/tlbflush.h
+++ b/arch/arm64/include/asm/tlbflush.h
@@ -253,7 +253,7 @@ static inline void flush_tlb_mm(struct mm_struct *mm)
__tlbi(aside1is, asid);
__tlbi_user(aside1is, asid);
dsb(ish);
- mmu_notifier_invalidate_range(mm, 0, -1UL);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, 0, -1UL);
}
static inline void __flush_tlb_page_nosync(struct mm_struct *mm,
@@ -265,7 +265,7 @@ static inline void __flush_tlb_page_nosync(struct mm_struct *mm,
addr = __TLBI_VADDR(uaddr, ASID(mm));
__tlbi(vale1is, addr);
__tlbi_user(vale1is, addr);
- mmu_notifier_invalidate_range(mm, uaddr & PAGE_MASK,
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, uaddr & PAGE_MASK,
(uaddr & PAGE_MASK) + PAGE_SIZE);
}
@@ -400,7 +400,7 @@ static inline void __flush_tlb_range(struct vm_area_struct *vma,
scale++;
}
dsb(ish);
- mmu_notifier_invalidate_range(vma->vm_mm, start, end);
+ mmu_notifier_arch_invalidate_secondary_tlbs(vma->vm_mm, start, end);
}
static inline void flush_tlb_range(struct vm_area_struct *vma,
diff --git a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
index f3fb49f..17075c7 100644
--- a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
+++ b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
@@ -39,7 +39,7 @@ void radix__flush_hugetlb_tlb_range(struct vm_area_struct *vma, unsigned long st
radix__flush_tlb_pwc_range_psize(vma->vm_mm, start, end, psize);
else
radix__flush_tlb_range_psize(vma->vm_mm, start, end, psize);
- mmu_notifier_invalidate_range(vma->vm_mm, start, end);
+ mmu_notifier_arch_invalidate_secondary_tlbs(vma->vm_mm, start, end);
}
void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
diff --git a/arch/powerpc/mm/book3s64/radix_tlb.c b/arch/powerpc/mm/book3s64/radix_tlb.c
index 4d44902..06e647e 100644
--- a/arch/powerpc/mm/book3s64/radix_tlb.c
+++ b/arch/powerpc/mm/book3s64/radix_tlb.c
@@ -987,7 +987,7 @@ void radix__flush_tlb_mm(struct mm_struct *mm)
}
}
preempt_enable();
- mmu_notifier_invalidate_range(mm, 0, -1UL);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, 0, -1UL);
}
EXPORT_SYMBOL(radix__flush_tlb_mm);
@@ -1021,7 +1021,7 @@ static void __flush_all_mm(struct mm_struct *mm, bool fullmm)
_tlbiel_pid_multicast(mm, pid, RIC_FLUSH_ALL);
}
preempt_enable();
- mmu_notifier_invalidate_range(mm, 0, -1UL);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, 0, -1UL);
}
void radix__flush_all_mm(struct mm_struct *mm)
@@ -1230,7 +1230,7 @@ static inline void __radix__flush_tlb_range(struct mm_struct *mm,
}
out:
preempt_enable();
- mmu_notifier_invalidate_range(mm, start, end);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, start, end);
}
void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
@@ -1395,7 +1395,7 @@ static void __radix__flush_tlb_range_psize(struct mm_struct *mm,
}
out:
preempt_enable();
- mmu_notifier_invalidate_range(mm, start, end);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, start, end);
}
void radix__flush_tlb_range_psize(struct mm_struct *mm, unsigned long start,
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 0a54323..6ab42ca 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -283,7 +283,7 @@ static inline void arch_tlbbatch_add_pending(struct arch_tlbflush_unmap_batch *b
{
inc_mm_tlb_gen(mm);
cpumask_or(&batch->cpumask, &batch->cpumask, mm_cpumask(mm));
- mmu_notifier_invalidate_range(mm, 0, -1UL);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, 0, -1UL);
}
static inline void arch_flush_tlb_batched_pending(struct mm_struct *mm)
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index 93b2f81..2d25391 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -1037,7 +1037,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
put_flush_tlb_info();
put_cpu();
- mmu_notifier_invalidate_range(mm, start, end);
+ mmu_notifier_arch_invalidate_secondary_tlbs(mm, start, end);
}
diff --git a/drivers/iommu/amd/iommu_v2.c b/drivers/iommu/amd/iommu_v2.c
index 261352a..2596466 100644
--- a/drivers/iommu/amd/iommu_v2.c
+++ b/drivers/iommu/amd/iommu_v2.c
@@ -355,9 +355,9 @@ static struct pasid_state *mn_to_state(struct mmu_notifier *mn)
return container_of(mn, struct pasid_state, mn);
}
-static void mn_invalidate_range(struct mmu_notifier *mn,
- struct mm_struct *mm,
- unsigned long start, unsigned long end)
+static void mn_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
+ struct mm_struct *mm,
+ unsigned long start, unsigned long end)
{
struct pasid_state *pasid_state;
struct device_state *dev_state;
@@ -391,8 +391,8 @@ static void mn_release(struct mmu_notifier *mn, struct mm_struct *mm)
}
static const struct mmu_notifier_ops iommu_mn = {
- .release = mn_release,
- .invalidate_range = mn_invalidate_range,
+ .release = mn_release,
+ .arch_invalidate_secondary_tlbs = mn_arch_invalidate_secondary_tlbs,
};
static void set_pri_tag_status(struct pasid_state *pasid_state,
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
index 2a19784..dbc812a 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
@@ -186,9 +186,10 @@ static void arm_smmu_free_shared_cd(struct arm_smmu_ctx_desc *cd)
}
}
-static void arm_smmu_mm_invalidate_range(struct mmu_notifier *mn,
- struct mm_struct *mm,
- unsigned long start, unsigned long end)
+static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
+ struct mm_struct *mm,
+ unsigned long start,
+ unsigned long end)
{
struct arm_smmu_mmu_notifier *smmu_mn = mn_to_smmu(mn);
struct arm_smmu_domain *smmu_domain = smmu_mn->domain;
@@ -247,9 +248,9 @@ static void arm_smmu_mmu_notifier_free(struct mmu_notifier *mn)
}
static const struct mmu_notifier_ops arm_smmu_mmu_notifier_ops = {
- .invalidate_range = arm_smmu_mm_invalidate_range,
- .release = arm_smmu_mm_release,
- .free_notifier = arm_smmu_mmu_notifier_free,
+ .arch_invalidate_secondary_tlbs = arm_smmu_mm_arch_invalidate_secondary_tlbs,
+ .release = arm_smmu_mm_release,
+ .free_notifier = arm_smmu_mmu_notifier_free,
};
/* Allocate or get existing MMU notifier for this {domain, mm} pair */
diff --git a/drivers/iommu/intel/svm.c b/drivers/iommu/intel/svm.c
index e95b339..8f6d680 100644
--- a/drivers/iommu/intel/svm.c
+++ b/drivers/iommu/intel/svm.c
@@ -219,9 +219,9 @@ static void intel_flush_svm_range(struct intel_svm *svm, unsigned long address,
}
/* Pages have been freed at this point */
-static void intel_invalidate_range(struct mmu_notifier *mn,
- struct mm_struct *mm,
- unsigned long start, unsigned long end)
+static void intel_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
+ struct mm_struct *mm,
+ unsigned long start, unsigned long end)
{
struct intel_svm *svm = container_of(mn, struct intel_svm, notifier);
@@ -256,7 +256,7 @@ static void intel_mm_release(struct mmu_notifier *mn, struct mm_struct *mm)
static const struct mmu_notifier_ops intel_mmuops = {
.release = intel_mm_release,
- .invalidate_range = intel_invalidate_range,
+ .arch_invalidate_secondary_tlbs = intel_arch_invalidate_secondary_tlbs,
};
static DEFINE_MUTEX(pasid_mutex);
diff --git a/drivers/misc/ocxl/link.c b/drivers/misc/ocxl/link.c
index 4cf4c55..c06c699 100644
--- a/drivers/misc/ocxl/link.c
+++ b/drivers/misc/ocxl/link.c
@@ -491,9 +491,9 @@ void ocxl_link_release(struct pci_dev *dev, void *link_handle)
}
EXPORT_SYMBOL_GPL(ocxl_link_release);
-static void invalidate_range(struct mmu_notifier *mn,
- struct mm_struct *mm,
- unsigned long start, unsigned long end)
+static void arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
+ struct mm_struct *mm,
+ unsigned long start, unsigned long end)
{
struct pe_data *pe_data = container_of(mn, struct pe_data, mmu_notifier);
struct ocxl_link *link = pe_data->link;
@@ -509,7 +509,7 @@ static void invalidate_range(struct mmu_notifier *mn,
}
static const struct mmu_notifier_ops ocxl_mmu_notifier_ops = {
- .invalidate_range = invalidate_range,
+ .arch_invalidate_secondary_tlbs = arch_invalidate_secondary_tlbs,
};
static u64 calculate_cfg_state(bool kernel)
diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h
index f2e9edc..6e3c857 100644
--- a/include/linux/mmu_notifier.h
+++ b/include/linux/mmu_notifier.h
@@ -187,27 +187,27 @@ struct mmu_notifier_ops {
const struct mmu_notifier_range *range);
/*
- * invalidate_range() is either called between
- * invalidate_range_start() and invalidate_range_end() when the
- * VM has to free pages that where unmapped, but before the
- * pages are actually freed, or outside of _start()/_end() when
- * a (remote) TLB is necessary.
+ * arch_invalidate_secondary_tlbs() is used to manage a non-CPU TLB
+ * which shares page-tables with the CPU. The
+ * invalidate_range_start()/end() callbacks should not be implemented as
+ * invalidate_secondary_tlbs() already catches the points in time when
+ * an external TLB needs to be flushed.
*
- * If invalidate_range() is used to manage a non-CPU TLB with
- * shared page-tables, it not necessary to implement the
- * invalidate_range_start()/end() notifiers, as
- * invalidate_range() already catches the points in time when an
- * external TLB range needs to be flushed. For more in depth
- * discussion on this see Documentation/mm/mmu_notifier.rst
+ * This requires arch_invalidate_secondary_tlbs() to be called while
+ * holding the ptl spin-lock and therefore this callback is not allowed
+ * to sleep.
*
- * Note that this function might be called with just a sub-range
- * of what was passed to invalidate_range_start()/end(), if
- * called between those functions.
+ * This is called by architecture code whenever invalidating a TLB
+ * entry. It is assumed that any secondary TLB has the same rules for
+ * when invalidations are required. If this is not the case architecture
+ * code will need to call this explicitly when required for secondary
+ * TLB invalidation.
*/
- void (*invalidate_range)(struct mmu_notifier *subscription,
- struct mm_struct *mm,
- unsigned long start,
- unsigned long end);
+ void (*arch_invalidate_secondary_tlbs)(
+ struct mmu_notifier *subscription,
+ struct mm_struct *mm,
+ unsigned long start,
+ unsigned long end);
/*
* These callbacks are used with the get/put interface to manage the
@@ -396,8 +396,8 @@ extern void __mmu_notifier_change_pte(struct mm_struct *mm,
unsigned long address, pte_t pte);
extern int __mmu_notifier_invalidate_range_start(struct mmu_notifier_range *r);
extern void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *r);
-extern void __mmu_notifier_invalidate_range(struct mm_struct *mm,
- unsigned long start, unsigned long end);
+extern void __mmu_notifier_arch_invalidate_secondary_tlbs(struct mm_struct *mm,
+ unsigned long start, unsigned long end);
extern bool
mmu_notifier_range_update_to_read_only(const struct mmu_notifier_range *range);
@@ -483,11 +483,11 @@ mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
__mmu_notifier_invalidate_range_end(range);
}
-static inline void mmu_notifier_invalidate_range(struct mm_struct *mm,
- unsigned long start, unsigned long end)
+static inline void mmu_notifier_arch_invalidate_secondary_tlbs(struct mm_struct *mm,
+ unsigned long start, unsigned long end)
{
if (mm_has_notifiers(mm))
- __mmu_notifier_invalidate_range(mm, start, end);
+ __mmu_notifier_arch_invalidate_secondary_tlbs(mm, start, end);
}
static inline void mmu_notifier_subscriptions_init(struct mm_struct *mm)
@@ -664,7 +664,7 @@ void mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
{
}
-static inline void mmu_notifier_invalidate_range(struct mm_struct *mm,
+static inline void mmu_notifier_arch_invalidate_secondary_tlbs(struct mm_struct *mm,
unsigned long start, unsigned long end)
{
}
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 3ece117..e0420de 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2120,8 +2120,8 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
if (is_huge_zero_pmd(*pmd)) {
/*
* FIXME: Do we want to invalidate secondary mmu by calling
- * mmu_notifier_invalidate_range() see comments below inside
- * __split_huge_pmd() ?
+ * mmu_notifier_arch_invalidate_secondary_tlbs() see comments below
+ * inside __split_huge_pmd() ?
*
* We are going from a zero huge page write protected to zero
* small page also write protected so it does not seems useful
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 9c6e431..e0028cb 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6676,8 +6676,9 @@ long hugetlb_change_protection(struct vm_area_struct *vma,
else
flush_hugetlb_tlb_range(vma, start, end);
/*
- * No need to call mmu_notifier_invalidate_range() we are downgrading
- * page table protection not changing it to point to a new page.
+ * No need to call mmu_notifier_arch_invalidate_secondary_tlbs() we are
+ * downgrading page table protection not changing it to point to a new
+ * page.
*
* See Documentation/mm/mmu_notifier.rst
*/
@@ -7321,7 +7322,7 @@ static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
i_mmap_unlock_write(vma->vm_file->f_mapping);
hugetlb_vma_unlock_write(vma);
/*
- * No need to call mmu_notifier_invalidate_range(), see
+ * No need to call mmu_notifier_arch_invalidate_secondary_tlbs(), see
* Documentation/mm/mmu_notifier.rst.
*/
mmu_notifier_invalidate_range_end(&range);
diff --git a/mm/mmu_notifier.c b/mm/mmu_notifier.c
index 453a156..ec3b068 100644
--- a/mm/mmu_notifier.c
+++ b/mm/mmu_notifier.c
@@ -585,8 +585,8 @@ void __mmu_notifier_invalidate_range_end(struct mmu_notifier_range *range)
lock_map_release(&__mmu_notifier_invalidate_range_start_map);
}
-void __mmu_notifier_invalidate_range(struct mm_struct *mm,
- unsigned long start, unsigned long end)
+void __mmu_notifier_arch_invalidate_secondary_tlbs(struct mm_struct *mm,
+ unsigned long start, unsigned long end)
{
struct mmu_notifier *subscription;
int id;
@@ -595,9 +595,10 @@ void __mmu_notifier_invalidate_range(struct mm_struct *mm,
hlist_for_each_entry_rcu(subscription,
&mm->notifier_subscriptions->list, hlist,
srcu_read_lock_held(&srcu)) {
- if (subscription->ops->invalidate_range)
- subscription->ops->invalidate_range(subscription, mm,
- start, end);
+ if (subscription->ops->arch_invalidate_secondary_tlbs)
+ subscription->ops->arch_invalidate_secondary_tlbs(
+ subscription, mm,
+ start, end);
}
srcu_read_unlock(&srcu, id);
}
@@ -616,6 +617,16 @@ int __mmu_notifier_register(struct mmu_notifier *subscription,
mmap_assert_write_locked(mm);
BUG_ON(atomic_read(&mm->mm_users) <= 0);
+ /*
+ * Subsystems should only register for invalidate_secondary_tlbs() or
+ * invalidate_range_start()/end() callbacks, not both.
+ */
+ if (WARN_ON_ONCE(subscription &&
+ (subscription->ops->arch_invalidate_secondary_tlbs &&
+ (subscription->ops->invalidate_range_start ||
+ subscription->ops->invalidate_range_end))))
+ return -EINVAL;
+
if (!mm->notifier_subscriptions) {
/*
* kmalloc cannot be called under mm_take_all_locks(), but we
--
git-series 0.9.1