From: Baolin Wang <baolin.wang@linux.alibaba.com>
To: akpm@linux-foundation.org, songmuchun@bytedance.com,
mike.kravetz@oracle.com
Cc: baolin.wang@linux.alibaba.com, linux-mm@kvack.org,
linux-kernel@vger.kernel.org
Subject: [PATCH v2 1/5] mm/hugetlb: fix races when looking up a CONT-PTE size hugetlb page
Date: Tue, 23 Aug 2022 15:50:01 +0800 [thread overview]
Message-ID: <0e5d92da043d147a867f634b17acbcc97a7f0e64.1661240170.git.baolin.wang@linux.alibaba.com> (raw)
In-Reply-To: <cover.1661240170.git.baolin.wang@linux.alibaba.com>
In-Reply-To: <cover.1661240170.git.baolin.wang@linux.alibaba.com>
On some architectures (like ARM64), it can support CONT-PTE/PMD size
hugetlb, which means it can support not only PMD/PUD size hugetlb
(2M and 1G), but also CONT-PTE/PMD size(64K and 32M) if a 4K page size
specified.
So when looking up a CONT-PTE size hugetlb page by follow_page(), it
will use pte_offset_map_lock() to get the pte entry lock for the CONT-PTE
size hugetlb in follow_page_pte(). However this pte entry lock is incorrect
for the CONT-PTE size hugetlb, since we should use huge_pte_lock() to
get the correct lock, which is mm->page_table_lock.
That means the pte entry of the CONT-PTE size hugetlb under current
pte lock is unstable in follow_page_pte(), we can continue to migrate
or poison the pte entry of the CONT-PTE size hugetlb, which can cause
some potential race issues, and following pte_xxx() validation is also
unstable in follow_page_pte(), even though they are under the 'pte lock'.
Moreover we should use huge_ptep_get() to get the pte entry value of
the CONT-PTE size hugetlb, which already takes into account the subpages'
dirty or young bits in case we missed the dirty or young state of the
CONT-PTE size hugetlb.
To fix above issues, introducing a new helper follow_huge_pte() to look
up a CONT-PTE size hugetlb page, which uses huge_pte_lock() to get the
correct pte entry lock to make the pte entry stable, as well as
supporting non-present pte handling.
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
---
include/linux/hugetlb.h | 8 ++++++++
mm/gup.c | 11 ++++++++++
mm/hugetlb.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 72 insertions(+)
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 3ec981a..d491138 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -207,6 +207,8 @@ struct page *follow_huge_addr(struct mm_struct *mm, unsigned long address,
struct page *follow_huge_pd(struct vm_area_struct *vma,
unsigned long address, hugepd_t hpd,
int flags, int pdshift);
+struct page *follow_huge_pte(struct vm_area_struct *vma, unsigned long address,
+ pmd_t *pmd, int flags);
struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
pmd_t *pmd, int flags);
struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
@@ -312,6 +314,12 @@ static inline struct page *follow_huge_pd(struct vm_area_struct *vma,
return NULL;
}
+static inline struct page *follow_huge_pte(struct vm_area_struct *vma,
+ unsigned long address, pmd_t *pmd, int flags)
+{
+ return NULL;
+}
+
static inline struct page *follow_huge_pmd(struct mm_struct *mm,
unsigned long address, pmd_t *pmd, int flags)
{
diff --git a/mm/gup.c b/mm/gup.c
index 3b656b7..87a94f5 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -534,6 +534,17 @@ static struct page *follow_page_pte(struct vm_area_struct *vma,
if (unlikely(pmd_bad(*pmd)))
return no_page_table(vma, flags);
+ /*
+ * Considering PTE level hugetlb, like continuous-PTE hugetlb on
+ * ARM64 architecture.
+ */
+ if (is_vm_hugetlb_page(vma)) {
+ page = follow_huge_pte(vma, address, pmd, flags);
+ if (page)
+ return page;
+ return no_page_table(vma, flags);
+ }
+
ptep = pte_offset_map_lock(mm, pmd, address, &ptl);
pte = *ptep;
if (!pte_present(pte)) {
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 6c00ba1..cf742d1 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6981,6 +6981,59 @@ struct page * __weak
return NULL;
}
+/* Support looking up a CONT-PTE size hugetlb page. */
+struct page * __weak
+follow_huge_pte(struct vm_area_struct *vma, unsigned long address,
+ pmd_t *pmd, int flags)
+{
+ struct mm_struct *mm = vma->vm_mm;
+ struct hstate *hstate = hstate_vma(vma);
+ unsigned long size = huge_page_size(hstate);
+ struct page *page = NULL;
+ spinlock_t *ptl;
+ pte_t *ptep, pte;
+
+ /*
+ * FOLL_PIN is not supported for follow_page(). Ordinary GUP goes via
+ * follow_hugetlb_page().
+ */
+ if (WARN_ON_ONCE(flags & FOLL_PIN))
+ return NULL;
+
+ ptep = huge_pte_offset(mm, address, size);
+ if (!ptep)
+ return NULL;
+
+retry:
+ ptl = huge_pte_lock(hstate, mm, ptep);
+ pte = huge_ptep_get(ptep);
+ if (pte_present(pte)) {
+ page = pte_page(pte);
+ if (WARN_ON_ONCE(!try_grab_page(page, flags))) {
+ page = NULL;
+ goto out;
+ }
+ } else {
+ if (!(flags & FOLL_MIGRATION)) {
+ page = NULL;
+ goto out;
+ }
+
+ if (is_hugetlb_entry_migration(pte)) {
+ spin_unlock(ptl);
+ __migration_entry_wait_huge(ptep, ptl);
+ goto retry;
+ }
+ /*
+ * hwpoisoned entry is treated as no_page_table in
+ * follow_page_mask().
+ */
+ }
+out:
+ spin_unlock(ptl);
+ return page;
+}
+
struct page * __weak
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
pmd_t *pmd, int flags)
--
1.8.3.1
next prev parent reply other threads:[~2022-08-23 7:50 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-23 7:50 [PATCH v2 0/5] Fix some issues when looking up hugetlb page Baolin Wang
2022-08-23 7:50 ` Baolin Wang [this message]
2022-08-23 8:29 ` [PATCH v2 1/5] mm/hugetlb: fix races when looking up a CONT-PTE size " David Hildenbrand
2022-08-23 10:02 ` Baolin Wang
2022-08-23 10:23 ` David Hildenbrand
2022-08-23 23:55 ` Mike Kravetz
2022-08-24 2:06 ` Baolin Wang
2022-08-24 7:31 ` David Hildenbrand
2022-08-24 9:41 ` Baolin Wang
2022-08-24 11:55 ` David Hildenbrand
2022-08-24 14:30 ` Baolin Wang
2022-08-24 14:33 ` David Hildenbrand
2022-08-24 15:06 ` Baolin Wang
2022-08-24 15:13 ` David Hildenbrand
2022-08-24 15:23 ` Baolin Wang
2022-08-24 23:34 ` Mike Kravetz
2022-08-25 1:43 ` Baolin Wang
2022-08-25 7:10 ` David Hildenbrand
2022-08-25 7:58 ` Baolin Wang
2022-08-25 18:30 ` Mike Kravetz
2022-08-25 7:25 ` David Hildenbrand
2022-08-25 10:54 ` Baolin Wang
2022-08-25 21:13 ` Mike Kravetz
2022-08-26 22:40 ` Mike Kravetz
2022-08-27 13:59 ` Aneesh Kumar K.V
2022-08-29 18:30 ` Mike Kravetz
2022-08-23 7:50 ` [PATCH v2 2/5] mm/hugetlb: use PTE page lock to protect CONT-PTE entries Baolin Wang
2022-08-23 7:50 ` [PATCH v2 3/5] mm/hugetlb: fix races when looking up a CONT-PMD size hugetlb page Baolin Wang
2022-08-23 7:50 ` [PATCH v2 4/5] mm/hugetlb: use PMD page lock to protect CONT-PTE entries Baolin Wang
2022-08-23 8:14 ` David Hildenbrand
2022-08-23 10:12 ` Baolin Wang
2022-08-23 7:50 ` [PATCH v2 5/5] mm/hugetlb: add FOLL_MIGRATION validation before waiting for a migration entry Baolin Wang
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0e5d92da043d147a867f634b17acbcc97a7f0e64.1661240170.git.baolin.wang@linux.alibaba.com \
--to=baolin.wang@linux.alibaba.com \
--cc=akpm@linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mike.kravetz@oracle.com \
--cc=songmuchun@bytedance.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).