| Message ID | 6aabddaf4cae5ae2205c3a7df9b9e15dbd61b641.1652180088.git.baolin.wang@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Implement arm64 specific huge_ptep_get() | expand |
On Tue, May 10, 2022 at 07:12:52PM +0800, Baolin Wang wrote: > The original huge_ptep_get() on ARM64 is just a wrapper of ptep_get(), > which will not take into account any contig-PTEs dirty and access bits. > Meanwhile we will implement a new ARM64-specific huge_ptep_get() > interface in following patch, which will take into account any contig-PTEs > dirty and access bits and only be allowed to pass the head pte of > a contig-PTE/PMD size page. IIUC, the huge_ptep_get() you have implemented in patch 2 could handle non-head pte. It'll return the original pte without potential AD bit. I admit it is more efficeent to use ptep_get() directly, but the judgement here should be updated. With this update. Reviewed-by: Muchun Song <songmuchun@bytedance.com> Thanks. > Thus change to use ptep_get() to get the pte value of a huge page as > a preparation. > > Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com> > --- > arch/arm64/mm/hugetlbpage.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c > index ca8e65c..be5e2f3 100644 > --- a/arch/arm64/mm/hugetlbpage.c > +++ b/arch/arm64/mm/hugetlbpage.c > @@ -172,7 +172,7 @@ static pte_t get_clear_flush(struct mm_struct *mm, > unsigned long pgsize, > unsigned long ncontig) > { > - pte_t orig_pte = huge_ptep_get(ptep); > + pte_t orig_pte = ptep_get(ptep); > bool valid = pte_valid(orig_pte); > unsigned long i, saddr = addr; > > @@ -385,7 +385,7 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, > { > int ncontig; > size_t pgsize; > - pte_t orig_pte = huge_ptep_get(ptep); > + pte_t orig_pte = ptep_get(ptep); > > if (!pte_cont(orig_pte)) > return ptep_get_and_clear(mm, addr, ptep); > @@ -408,11 +408,11 @@ static int __cont_access_flags_changed(pte_t *ptep, pte_t pte, int ncontig) > { > int i; > > - if (pte_write(pte) != pte_write(huge_ptep_get(ptep))) > + if (pte_write(pte) != pte_write(ptep_get(ptep))) > return 1; > > for (i = 0; i < ncontig; i++) { > - pte_t orig_pte = huge_ptep_get(ptep + i); > + pte_t orig_pte = ptep_get(ptep + i); > > if (pte_dirty(pte) != pte_dirty(orig_pte)) > return 1; > -- > 1.8.3.1 > >
On 5/10/2022 11:55 PM, Muchun Song wrote: > On Tue, May 10, 2022 at 07:12:52PM +0800, Baolin Wang wrote: >> The original huge_ptep_get() on ARM64 is just a wrapper of ptep_get(), >> which will not take into account any contig-PTEs dirty and access bits. >> Meanwhile we will implement a new ARM64-specific huge_ptep_get() >> interface in following patch, which will take into account any contig-PTEs >> dirty and access bits and only be allowed to pass the head pte of >> a contig-PTE/PMD size page. > > IIUC, the huge_ptep_get() you have implemented in patch 2 could > handle non-head pte. It'll return the original pte without potential > AD bit. I admit it is more efficeent to use ptep_get() directly, > but the judgement here should be updated. Ah, right. I missed the 'ncontig' will be 0 if a non-head pte passed. Will update the commit message in next version. Thanks for reviewing. > > With this update. > > Reviewed-by: Muchun Song <songmuchun@bytedance.com>
diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c index ca8e65c..be5e2f3 100644 --- a/arch/arm64/mm/hugetlbpage.c +++ b/arch/arm64/mm/hugetlbpage.c @@ -172,7 +172,7 @@ static pte_t get_clear_flush(struct mm_struct *mm, unsigned long pgsize, unsigned long ncontig) { - pte_t orig_pte = huge_ptep_get(ptep); + pte_t orig_pte = ptep_get(ptep); bool valid = pte_valid(orig_pte); unsigned long i, saddr = addr; @@ -385,7 +385,7 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, { int ncontig; size_t pgsize; - pte_t orig_pte = huge_ptep_get(ptep); + pte_t orig_pte = ptep_get(ptep); if (!pte_cont(orig_pte)) return ptep_get_and_clear(mm, addr, ptep); @@ -408,11 +408,11 @@ static int __cont_access_flags_changed(pte_t *ptep, pte_t pte, int ncontig) { int i; - if (pte_write(pte) != pte_write(huge_ptep_get(ptep))) + if (pte_write(pte) != pte_write(ptep_get(ptep))) return 1; for (i = 0; i < ncontig; i++) { - pte_t orig_pte = huge_ptep_get(ptep + i); + pte_t orig_pte = ptep_get(ptep + i); if (pte_dirty(pte) != pte_dirty(orig_pte)) return 1;
The original huge_ptep_get() on ARM64 is just a wrapper of ptep_get(), which will not take into account any contig-PTEs dirty and access bits. Meanwhile we will implement a new ARM64-specific huge_ptep_get() interface in following patch, which will take into account any contig-PTEs dirty and access bits and only be allowed to pass the head pte of a contig-PTE/PMD size page. Thus change to use ptep_get() to get the pte value of a huge page as a preparation. Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com> --- arch/arm64/mm/hugetlbpage.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)