lkml.org 
[lkml]   [2013]   [Sep]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH 1/2] hugetlbfs: support split page table lock
Date
Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> writes:

> Currently all of page table handling by hugetlbfs code are done under
> mm->page_table_lock. So when a process have many threads and they heavily
> access to the memory, lock contention happens and impacts the performance.
>
> This patch makes hugepage support split page table lock so that we use
> page->ptl of the leaf node of page table tree which is pte for normal pages
> but can be pmd and/or pud for hugepages of some architectures.
>
> ChangeLog v3:
> - disable split ptl for ppc with USE_SPLIT_PTLOCKS_HUGETLB.
> - remove replacement in some architecture dependent code. This is justified
> because an allocation of pgd/pud/pmd/pte entry can race with other
> allocation, not with read/write access, so we can use different locks.
> http://thread.gmane.org/gmane.linux.kernel.mm/106292/focus=106458
>
> ChangeLog v2:
> - add split ptl on other archs missed in v1
> - drop changes on arch/{powerpc,tile}/mm/hugetlbpage.c
>
> Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> ---
> include/linux/hugetlb.h | 20 +++++++++++
> include/linux/mm_types.h | 2 ++
> mm/hugetlb.c | 92 +++++++++++++++++++++++++++++-------------------
> mm/mempolicy.c | 5 +--
> mm/migrate.c | 4 +--
> mm/rmap.c | 2 +-
> 6 files changed, 84 insertions(+), 41 deletions(-)
>
> diff --git v3.11-rc3.orig/include/linux/hugetlb.h v3.11-rc3/include/linux/hugetlb.h
> index 0393270..5cb8a4e 100644
> --- v3.11-rc3.orig/include/linux/hugetlb.h
> +++ v3.11-rc3/include/linux/hugetlb.h
> @@ -80,6 +80,24 @@ extern const unsigned long hugetlb_zero, hugetlb_infinity;
> extern int sysctl_hugetlb_shm_group;
> extern struct list_head huge_boot_pages;
>
> +#if USE_SPLIT_PTLOCKS_HUGETLB
> +#define huge_pte_lockptr(mm, ptep) ({__pte_lockptr(virt_to_page(ptep)); })
> +#else /* !USE_SPLIT_PTLOCKS_HUGETLB */
> +#define huge_pte_lockptr(mm, ptep) ({&(mm)->page_table_lock; })
> +#endif /* USE_SPLIT_PTLOCKS_HUGETLB */
> +
> +#define huge_pte_offset_lock(mm, address, ptlp) \
> +({ \
> + pte_t *__pte = huge_pte_offset(mm, address); \
> + spinlock_t *__ptl = NULL; \
> + if (__pte) { \
> + __ptl = huge_pte_lockptr(mm, __pte); \
> + *(ptlp) = __ptl; \
> + spin_lock(__ptl); \
> + } \
> + __pte; \
> +})


why not a static inline function ?


> +
> /* arch callbacks */
>
> pte_t *huge_pte_alloc(struct mm_struct *mm,> @@ -164,6 +182,8 @@ static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
> BUG();
> }
>
> +#define huge_pte_lockptr(mm, ptep) 0
> +
> #endif /* !CONFIG_HUGETLB_PAGE */
>
> #define HUGETLB_ANON_FILE "anon_hugepage"
> diff --git v3.11-rc3.orig/include/linux/mm_types.h v3.11-rc3/include/linux/mm_types.h
> index fb425aa..cfb8c6f 100644
> --- v3.11-rc3.orig/include/linux/mm_types.h
> +++ v3.11-rc3/include/linux/mm_types.h
> @@ -24,6 +24,8 @@
> struct address_space;
>
> #define USE_SPLIT_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
> +#define USE_SPLIT_PTLOCKS_HUGETLB \
> + (USE_SPLIT_PTLOCKS && !defined(CONFIG_PPC))
>

Is that a common pattern ? Don't we generally use
HAVE_ARCH_SPLIT_PTLOCKS in arch config file ? Also are we sure this is
only an issue with PPC ?



-aneesh



\
 
 \ /
  Last update: 2013-09-08 19:41    [W:0.162 / U:0.208 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site