huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
hugetlb address.
Normally, it's always safe to walk the pgtable as long as we're with the
mmap lock held for either read or write, because that guarantees the
pgtable pages will always be valid during the process.
But it's not true for hugetlbfs: hugetlbfs has the pmd sharing feature, it
means that even with mmap lock held, the PUD pgtable page can still go away
from under us if pmd unsharing is possible during the walk.
It's not always the case, e.g.:
(1) If the mapping is private we're not prone to pmd sharing or
unsharing, so it's okay.
(2) If we're with the hugetlb vma lock held for either read/write, it's
okay too because pmd unshare cannot happen at all.
Document all these explicitly for huge_pte_offset(), because it's really
not that obvious. This also tells all the callers on what it needs to
guarantee huge_pte_offset() thread-safety.
Signed-off-by: Peter Xu <[email protected]>
---
arch/arm64/mm/hugetlbpage.c | 32 ++++++++++++++++++++++++++++++++
1 file changed, 32 insertions(+)
diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
index 35e9a468d13e..0bf930c75d4b 100644
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -329,6 +329,38 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
return ptep;
}
+/*
+ * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
+ * Returns the pte_t* if found, or NULL if the address is not mapped.
+ *
+ * NOTE: since this function will walk all the pgtable pages (including not
+ * only high-level pgtable page, but also PUD that can be unshared
+ * concurrently for VM_SHARED), the caller of this function should be
+ * responsible of its thread safety. One can follow this rule:
+ *
+ * (1) For private mappings: pmd unsharing is not possible, so it'll
+ * always be safe if we're with the mmap sem for either read or write.
+ * This is normally always the case, so IOW we don't need to do
+ * anything special.
+ *
+ * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged
+ * pgtable page can go away from under us! It can be done by a pmd
+ * unshare with a follow up munmap() on the other process), then we
+ * need either:
+ *
+ * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare
+ * won't happen upon the range (it also makes sure the pte_t we
+ * read is the right and stable one), or,
+ *
+ * (2.2) RCU read lock, to make sure even pmd unsharing happened, the
+ * old shared PUD page won't get freed from under us, so even of
+ * the pteval can be obsolete, at least it's still always safe to
+ * access the pgtable page (e.g., de-referencing pte_t* would not
+ * cause use-after-free).
+ *
+ * PS: from the regard of (2.2), it's the same logic of fast-gup being safe
+ * for generic mm, as long as RCU is used to free any pgtable page.
+ */
pte_t *huge_pte_offset(struct mm_struct *mm,
unsigned long addr, unsigned long sz)
{
--
2.37.3
On Oct 30, 2022, at 2:29 PM, Peter Xu <[email protected]> wrote:
> huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
> hugetlb address.
>
> Normally, it's always safe to walk the pgtable as long as we're with the
> mmap lock held for either read or write, because that guarantees the
> pgtable pages will always be valid during the process.
>
> But it's not true for hugetlbfs: hugetlbfs has the pmd sharing feature, it
> means that even with mmap lock held, the PUD pgtable page can still go away
> from under us if pmd unsharing is possible during the walk.
>
> It's not always the case, e.g.:
>
> (1) If the mapping is private we're not prone to pmd sharing or
> unsharing, so it's okay.
>
> (2) If we're with the hugetlb vma lock held for either read/write, it's
> okay too because pmd unshare cannot happen at all.
>
> Document all these explicitly for huge_pte_offset(), because it's really
> not that obvious. This also tells all the callers on what it needs to
> guarantee huge_pte_offset() thread-safety.
>
> Signed-off-by: Peter Xu <[email protected]>
> ---
> arch/arm64/mm/hugetlbpage.c | 32 ++++++++++++++++++++++++++++++++
Please excuse my ignorant question - is there something specific for arm64
code here? Other archs seem to have similar code, no?
On Mon, Oct 31, 2022 at 10:46:52PM -0700, Nadav Amit wrote:
> On Oct 30, 2022, at 2:29 PM, Peter Xu <[email protected]> wrote:
>
> > huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
> > hugetlb address.
> >
> > Normally, it's always safe to walk the pgtable as long as we're with the
> > mmap lock held for either read or write, because that guarantees the
> > pgtable pages will always be valid during the process.
> >
> > But it's not true for hugetlbfs: hugetlbfs has the pmd sharing feature, it
> > means that even with mmap lock held, the PUD pgtable page can still go away
> > from under us if pmd unsharing is possible during the walk.
> >
> > It's not always the case, e.g.:
> >
> > (1) If the mapping is private we're not prone to pmd sharing or
> > unsharing, so it's okay.
> >
> > (2) If we're with the hugetlb vma lock held for either read/write, it's
> > okay too because pmd unshare cannot happen at all.
> >
> > Document all these explicitly for huge_pte_offset(), because it's really
> > not that obvious. This also tells all the callers on what it needs to
> > guarantee huge_pte_offset() thread-safety.
> >
> > Signed-off-by: Peter Xu <[email protected]>
> > ---
> > arch/arm64/mm/hugetlbpage.c | 32 ++++++++++++++++++++++++++++++++
>
> Please excuse my ignorant question - is there something specific for arm64
> code here? Other archs seem to have similar code, no?
Oops, sorry, I meant to add this to the version in mm/hugetlb.c. Or maybe
to linux/hugetlb.h would make more sense.
I should probably also mention that for any arch that does not support pmd
sharing at all (afaik, any arch outside arm, x86 and riscv), the shared
mapping locking rule should be the same as private.
--
Peter Xu
On 10/30/22 17:29, Peter Xu wrote:
> huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
> hugetlb address.
>
> Normally, it's always safe to walk the pgtable as long as we're with the
> mmap lock held for either read or write, because that guarantees the
> pgtable pages will always be valid during the process.
>
> But it's not true for hugetlbfs: hugetlbfs has the pmd sharing feature, it
> means that even with mmap lock held, the PUD pgtable page can still go away
> from under us if pmd unsharing is possible during the walk.
>
> It's not always the case, e.g.:
>
> (1) If the mapping is private we're not prone to pmd sharing or
> unsharing, so it's okay.
>
> (2) If we're with the hugetlb vma lock held for either read/write, it's
> okay too because pmd unshare cannot happen at all.
>
> Document all these explicitly for huge_pte_offset(), because it's really
> not that obvious. This also tells all the callers on what it needs to
> guarantee huge_pte_offset() thread-safety.
>
> Signed-off-by: Peter Xu <[email protected]>
> ---
> arch/arm64/mm/hugetlbpage.c | 32 ++++++++++++++++++++++++++++++++
> 1 file changed, 32 insertions(+)
>
> diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
> index 35e9a468d13e..0bf930c75d4b 100644
> --- a/arch/arm64/mm/hugetlbpage.c
> +++ b/arch/arm64/mm/hugetlbpage.c
> @@ -329,6 +329,38 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
> return ptep;
> }
>
> +/*
> + * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
> + * Returns the pte_t* if found, or NULL if the address is not mapped.
> + *
> + * NOTE: since this function will walk all the pgtable pages (including not
> + * only high-level pgtable page, but also PUD that can be unshared
> + * concurrently for VM_SHARED), the caller of this function should be
> + * responsible of its thread safety. One can follow this rule:
> + *
> + * (1) For private mappings: pmd unsharing is not possible, so it'll
> + * always be safe if we're with the mmap sem for either read or write.
> + * This is normally always the case, so IOW we don't need to do
> + * anything special.
Not sure if it is worth calling out that we are safe if the process owning the
page table being walked is single threaded? Although, a pmd can be 'unshared'
due to an operation in another process, the primary is when the pmd is cleared
which only happens when the unshare is initiated by a thread of the process
owning the page tables being walked.
--
Mike Kravetz
> + *
> + * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged
> + * pgtable page can go away from under us! It can be done by a pmd
> + * unshare with a follow up munmap() on the other process), then we
> + * need either:
> + *
> + * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare
> + * won't happen upon the range (it also makes sure the pte_t we
> + * read is the right and stable one), or,
> + *
> + * (2.2) RCU read lock, to make sure even pmd unsharing happened, the
> + * old shared PUD page won't get freed from under us, so even of
> + * the pteval can be obsolete, at least it's still always safe to
> + * access the pgtable page (e.g., de-referencing pte_t* would not
> + * cause use-after-free).
> + *
> + * PS: from the regard of (2.2), it's the same logic of fast-gup being safe
> + * for generic mm, as long as RCU is used to free any pgtable page.
> + */
> pte_t *huge_pte_offset(struct mm_struct *mm,
> unsigned long addr, unsigned long sz)
> {
> --
> 2.37.3
>
On 11/03/22 14:11, Peter Xu wrote:
> On Thu, Nov 03, 2022 at 08:42:01AM -0700, Mike Kravetz wrote:
> > On 10/30/22 17:29, Peter Xu wrote:
> > Not sure if it is worth calling out that we are safe if the process owning the
> > page table being walked is single threaded? Although, a pmd can be 'unshared'
> > due to an operation in another process, the primary is when the pmd is cleared
> > which only happens when the unshare is initiated by a thread of the process
> > owning the page tables being walked.
>
> Even if the process is single threaded, the pmd unshare can still trigger
> from other threads too, am I right?
>
> Looking at huge_pmd_unshare() callers, the major ones that doesn't need
> current mm context are:
>
> - __unmap_hugepage_range() (e.g. hole punch from other process on file?)
> - try_to_unmap_one()
> - try_to_migrate_one()
>
> So for example, even for a single thread process, if its pmd shared with
> another process, the other process can do (1) punch hole on pmd shared
> region, then (2) munmap() the pmd shared region, then it seems the single
> thread process can be still on risk of accessing freed pgtable.
Yes, you are correct. I was not thinking about an unmap initiated by another
process doing something like hole punch or truncation.
--
Mike Kravetz
On Thu, Nov 03, 2022 at 08:42:01AM -0700, Mike Kravetz wrote:
> On 10/30/22 17:29, Peter Xu wrote:
> > huge_pte_offset() is potentially a pgtable walker, looking up pte_t* for a
> > hugetlb address.
> >
> > Normally, it's always safe to walk the pgtable as long as we're with the
> > mmap lock held for either read or write, because that guarantees the
> > pgtable pages will always be valid during the process.
> >
> > But it's not true for hugetlbfs: hugetlbfs has the pmd sharing feature, it
> > means that even with mmap lock held, the PUD pgtable page can still go away
> > from under us if pmd unsharing is possible during the walk.
> >
> > It's not always the case, e.g.:
> >
> > (1) If the mapping is private we're not prone to pmd sharing or
> > unsharing, so it's okay.
> >
> > (2) If we're with the hugetlb vma lock held for either read/write, it's
> > okay too because pmd unshare cannot happen at all.
> >
> > Document all these explicitly for huge_pte_offset(), because it's really
> > not that obvious. This also tells all the callers on what it needs to
> > guarantee huge_pte_offset() thread-safety.
> >
> > Signed-off-by: Peter Xu <[email protected]>
> > ---
> > arch/arm64/mm/hugetlbpage.c | 32 ++++++++++++++++++++++++++++++++
> > 1 file changed, 32 insertions(+)
> >
> > diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
> > index 35e9a468d13e..0bf930c75d4b 100644
> > --- a/arch/arm64/mm/hugetlbpage.c
> > +++ b/arch/arm64/mm/hugetlbpage.c
> > @@ -329,6 +329,38 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
> > return ptep;
> > }
> >
> > +/*
> > + * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
> > + * Returns the pte_t* if found, or NULL if the address is not mapped.
> > + *
> > + * NOTE: since this function will walk all the pgtable pages (including not
> > + * only high-level pgtable page, but also PUD that can be unshared
> > + * concurrently for VM_SHARED), the caller of this function should be
> > + * responsible of its thread safety. One can follow this rule:
> > + *
> > + * (1) For private mappings: pmd unsharing is not possible, so it'll
> > + * always be safe if we're with the mmap sem for either read or write.
> > + * This is normally always the case, so IOW we don't need to do
> > + * anything special.
>
> Not sure if it is worth calling out that we are safe if the process owning the
> page table being walked is single threaded? Although, a pmd can be 'unshared'
> due to an operation in another process, the primary is when the pmd is cleared
> which only happens when the unshare is initiated by a thread of the process
> owning the page tables being walked.
Even if the process is single threaded, the pmd unshare can still trigger
from other threads too, am I right?
Looking at huge_pmd_unshare() callers, the major ones that doesn't need
current mm context are:
- __unmap_hugepage_range() (e.g. hole punch from other process on file?)
- try_to_unmap_one()
- try_to_migrate_one()
So for example, even for a single thread process, if its pmd shared with
another process, the other process can do (1) punch hole on pmd shared
region, then (2) munmap() the pmd shared region, then it seems the single
thread process can be still on risk of accessing freed pgtable.
Thanks,
--
Peter Xu