| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <20200706202615.32111-3-mike.kravetz@oracle.com>
Date: Mon, 6 Jul 2020 13:26:14 -0700
From: Mike Kravetz <mike.kravetz@...cle.com>
To: linux-mm@...ck.org, linux-kernel@...r.kernel.org
Cc: Michal Hocko <mhocko@...nel.org>, Hugh Dickins <hughd@...gle.com>,
Naoya Horiguchi <n-horiguchi@...jp.nec.com>,
"Aneesh Kumar K . V" <aneesh.kumar@...ux.vnet.ibm.com>,
Andrea Arcangeli <aarcange@...hat.com>,
"Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
Davidlohr Bueso <dave@...olabs.net>,
Prakash Sangappa <prakash.sangappa@...cle.com>,
Andrew Morton <akpm@...ux-foundation.org>,
Linus Torvalds <torvalds@...ux-foundation.org>,
Mike Kravetz <mike.kravetz@...cle.com>,
kernel test robot <rong.a.chen@...el.com>
Subject: [RFC PATCH 2/3] hugetlbfs: Only take i_mmap_rwsem when sharing is possible
Commit c0d0381ade79 added code to take i_mmap_rwsem in read mode during
fault processing. However, this was observed to increase fault processing
time by aprox 33%. Technically, i_mmap_rwsem only needs to be held when
pmd sharing is possible. pmd sharing depends on mapping flags, alignment
and size. The routine vma_shareable() already checks these conditions.
Therefore, use vma_shareable to determine if sharing is possible and if
taking i_mmap_rwsem is necessary. This is done during fault processing
and vma copying. Code in memory-failure, page migration and userfaultfd
continue to always take i_mmap_rwsem. These are not as performance
sensitive.
Reported-by: kernel test robot <rong.a.chen@...el.com>
Signed-off-by: Mike Kravetz <mike.kravetz@...cle.com>
---
mm/hugetlb.c | 96 ++++++++++++++++++++++++++++++++++------------------
1 file changed, 63 insertions(+), 33 deletions(-)
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 5349beda3658..6e9085464e78 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3656,6 +3656,21 @@ static int hugetlb_acct_memory(struct hstate *h, long delta)
return ret;
}
+#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
+{
+ unsigned long base = addr & PUD_MASK;
+ unsigned long end = base + PUD_SIZE;
+
+ /*
+ * check on proper vm_flags and page table alignment
+ */
+ if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
+ return true;
+ return false;
+}
+#endif
+
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
struct resv_map *resv = vma_resv_map(vma);
@@ -3807,6 +3822,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
unsigned long sz = huge_page_size(h);
struct address_space *mapping = vma->vm_file->f_mapping;
struct mmu_notifier_range range;
+ bool i_mmap_rwsem_held = false;
int ret = 0;
cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
@@ -3816,14 +3832,6 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
vma->vm_start,
vma->vm_end);
mmu_notifier_invalidate_range_start(&range);
- } else {
- /*
- * For shared mappings i_mmap_rwsem must be held to call
- * huge_pte_alloc, otherwise the returned ptep could go
- * away if part of a shared pmd and another thread calls
- * huge_pmd_unshare.
- */
- i_mmap_lock_read(mapping);
}
for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
@@ -3831,6 +3839,28 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
src_pte = huge_pte_offset(src, addr, sz);
if (!src_pte)
continue;
+
+ /*
+ * For shared mappings(non-cow) i_mmap_rwsem must be held to
+ * call huge_pte_alloc, otherwise the returned ptep could go
+ * away if part of a shared pmd and another thread calls
+ * huge_pmd_unshare. This is only necessary if the specific
+ * pmd can be shared. Acquire/drop semaphore as necessary.
+ */
+ if (!cow) {
+ if (!i_mmap_rwsem_held) {
+ if (vma_shareable(vma, addr)) {
+ i_mmap_lock_read(mapping);
+ i_mmap_rwsem_held = true;
+ }
+ } else {
+ if (!vma_shareable(vma, addr)) {
+ i_mmap_unlock_read(mapping);
+ i_mmap_rwsem_held = false;
+ }
+ }
+ }
+
dst_pte = huge_pte_alloc(dst, addr, sz);
if (!dst_pte) {
ret = -ENOMEM;
@@ -3901,7 +3931,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
if (cow)
mmu_notifier_invalidate_range_end(&range);
- else
+ if (i_mmap_rwsem_held)
i_mmap_unlock_read(mapping);
return ret;
@@ -4357,9 +4387,11 @@ static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
*/
hash = hugetlb_fault_mutex_hash(mapping, idx);
mutex_unlock(&hugetlb_fault_mutex_table[hash]);
- i_mmap_unlock_read(mapping);
+ if (vma_shareable(vma, haddr))
+ i_mmap_unlock_read(mapping);
ret = handle_userfault(&vmf, VM_UFFD_MISSING);
- i_mmap_lock_read(mapping);
+ if (vma_shareable(vma, haddr))
+ i_mmap_lock_read(mapping);
mutex_lock(&hugetlb_fault_mutex_table[hash]);
goto out;
}
@@ -4543,19 +4575,22 @@ vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
}
/*
- * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
- * until finished with ptep. This prevents huge_pmd_unshare from
- * being called elsewhere and making the ptep no longer valid.
+ * If PMD sharing is possible, acquire i_mmap_rwsem before calling
+ * huge_pte_alloc and hold until finished with ptep. This prevents
+ * huge_pmd_unshare from being called elsewhere and making the ptep
+ * no longer valid.
*
* ptep could have already be assigned via huge_pte_offset. That
* is OK, as huge_pte_alloc will return the same value unless
* something has changed.
*/
mapping = vma->vm_file->f_mapping;
- i_mmap_lock_read(mapping);
+ if (vma_shareable(vma, haddr))
+ i_mmap_lock_read(mapping);
ptep = huge_pte_alloc(mm, haddr, huge_page_size(h));
if (!ptep) {
- i_mmap_unlock_read(mapping);
+ if (vma_shareable(vma, haddr))
+ i_mmap_unlock_read(mapping);
return VM_FAULT_OOM;
}
@@ -4652,7 +4687,8 @@ vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
}
out_mutex:
mutex_unlock(&hugetlb_fault_mutex_table[hash]);
- i_mmap_unlock_read(mapping);
+ if (vma_shareable(vma, haddr))
+ i_mmap_unlock_read(mapping);
/*
* Generally it's safe to hold refcount during waiting page lock. But
* here we just wait to defer the next page fault to avoid busy loop and
@@ -5287,19 +5323,6 @@ static unsigned long page_table_shareable(struct vm_area_struct *svma,
return saddr;
}
-static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
-{
- unsigned long base = addr & PUD_MASK;
- unsigned long end = base + PUD_SIZE;
-
- /*
- * check on proper vm_flags and page table alignment
- */
- if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
- return true;
- return false;
-}
-
/*
* Determine if start,end range within vma could be mapped by shared pmd.
* If yes, adjust start and end to cover range associated with possible
@@ -5335,9 +5358,12 @@ void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
* !shared pmd case because we can allocate the pmd later as well, it makes the
* code much cleaner.
*
- * This routine must be called with i_mmap_rwsem held in at least read mode.
- * For hugetlbfs, this prevents removal of any page table entries associated
- * with the address space. This is important as we are setting up sharing
+ * FIXME - If sharing is possible, this routine must be called with
+ * i_mmap_rwsem held in at least read mode. Leaving it up to the caller
+ * to determine if sharing is possible is asking for trouble. Right now, all
+ * calling code is correct. But, this needs to be cleaner. Holding
+ * i_mmap_rwsem prevents removal of any page table entries associated with
+ * the address space. This is important as we are setting up sharing
* based on existing page table entries (mappings).
*/
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
@@ -5355,6 +5381,10 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
if (!vma_shareable(vma, addr))
return (pte_t *)pmd_alloc(mm, pud, addr);
+ /*
+ * If we get here, caller should have acquired i_mmap_rwsem in
+ * at least read mode.
+ */
vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
if (svma == vma)
continue;
--
2.25.4
Powered by blists - more mailing lists