On 28/07/2019 13:33, Anshuman Khandual wrote:
>
>
> On 07/22/2019 09:12 PM, Steven Price wrote:
>> pgd_entry() and pud_entry() were removed by commit 0b1fbfe50006c410
>> ("mm/pagewalk: remove pgd_entry() and pud_entry()") because there were
>> no users. We're about to add users so reintroduce them, along with
>> p4d_entry() as we now have 5 levels of tables.
>>
>> Note that commit a00cc7d9dd93d66a ("mm, x86: add support for
>> PUD-sized transparent hugepages") already re-added pud_entry() but with
>> different semantics to the other callbacks. Since there have never
>> been upstream users of this, revert the semantics back to match the
>> other callbacks. This means pud_entry() is called for all entries, not
>> just transparent huge pages.
>>
>> Signed-off-by: Steven Price <steven.price@xxxxxxx>
>> ---
>> include/linux/mm.h | 15 +++++++++------
>> mm/pagewalk.c | 27 ++++++++++++++++-----------
>> 2 files changed, 25 insertions(+), 17 deletions(-)
>>
>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>> index 0334ca97c584..b22799129128 100644
>> --- a/include/linux/mm.h
>> +++ b/include/linux/mm.h
>> @@ -1432,15 +1432,14 @@ void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
>>
>> /**
>> * mm_walk - callbacks for walk_page_range
>> - * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
>> - * this handler should only handle pud_trans_huge() puds.
>> - * the pmd_entry or pte_entry callbacks will be used for
>> - * regular PUDs.
>> - * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
>> + * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
>> + * @p4d_entry: if set, called for each non-empty P4D entry
>> + * @pud_entry: if set, called for each non-empty PUD entry
>> + * @pmd_entry: if set, called for each non-empty PMD entry
>> * this handler is required to be able to handle
>> * pmd_trans_huge() pmds. They may simply choose to
>> * split_huge_page() instead of handling it explicitly.
>> - * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
>> + * @pte_entry: if set, called for each non-empty PTE (lowest-level) entry
>> * @pte_hole: if set, called for each hole at all levels
>> * @hugetlb_entry: if set, called for each hugetlb entry
>> * @test_walk: caller specific callback function to determine whether
>> @@ -1455,6 +1454,10 @@ void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
>> * (see the comment on walk_page_range() for more details)
>> */
>> struct mm_walk {
>> + int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
>> + unsigned long next, struct mm_walk *walk);
>> + int (*p4d_entry)(p4d_t *p4d, unsigned long addr,
>> + unsigned long next, struct mm_walk *walk);
>> int (*pud_entry)(pud_t *pud, unsigned long addr,
>> unsigned long next, struct mm_walk *walk);
>> int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
>> diff --git a/mm/pagewalk.c b/mm/pagewalk.c
>> index c3084ff2569d..98373a9f88b8 100644
>> --- a/mm/pagewalk.c
>> +++ b/mm/pagewalk.c
>> @@ -90,15 +90,9 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
>> }
>>
>> if (walk->pud_entry) {
>> - spinlock_t *ptl = pud_trans_huge_lock(pud, walk->vma);
>> -
>> - if (ptl) {
>> - err = walk->pud_entry(pud, addr, next, walk);
>> - spin_unlock(ptl);
>> - if (err)
>> - break;
>> - continue;
>> - }
>> + err = walk->pud_entry(pud, addr, next, walk);
>> + if (err)
>> + break;
>
> But will not this still encounter possible THP entries when walking user
> page tables (valid walk->vma) in which case still needs to get a lock.
> OR will the callback take care of it ?
This is what I mean in the commit message by:
> Since there have never
> been upstream users of this, revert the semantics back to match the
> other callbacks. This means pud_entry() is called for all entries, not
> just transparent huge pages.
So the expectation is that the caller takes care of it.
However, having checked again, it appears that mm/hmm.c now does use
this callback (merged in v5.2-rc1).
Jérôme - are you happy with this change in semantics? It looks like
hmm_vma_walk_pud() should deal gracefully with both normal and large
pages - although I'm unsure whether you are relying on the lock from
pud_trans_huge_lock()?
Thanks,
Steve
