* [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios
@ 2023-03-30 13:40 Peng Zhang
2023-03-30 13:40 ` [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio Peng Zhang
` (5 more replies)
0 siblings, 6 replies; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
This patch series converts several userfaultfd functions to use folios.
Change log:
v3->v4:
- Rebase onto mm-unstable per Andrew Morton. Update commit description
because some function names are changed. (patch 1,4,6)
v2->v3:
- Split patch 2 into three patches, suggested by Mike Kravetz. (patch
2-4)
- Add a new patch to convert copy_user_huge_page to copy_user_folio,
suggested by Mike Kravetz. (patch 5)
- Fix two uninitialized bugs, thanks to Dan Carpenter. (patch 6)
- Do some indenting cleanups.
v1->v2:
Modified patch 2, suggested by Matthew Wilcox:
- Rename copy_large_folio_from_user() to copy_folio_from_user().
- Delete the inner_folio.
- kmap() and kmap_atomic() are converted to kmap_local_page(). Use
pagefault_disable() to ensure that a deadlock will not occur.
- flush_dcache_folio() is placed outside the loop.
ZhangPeng (6):
userfaultfd: convert mfill_atomic_pte_copy() to use a folio
userfaultfd: use kmap_local_page() in copy_huge_page_from_user()
userfaultfd: convert copy_huge_page_from_user() to
copy_folio_from_user()
userfaultfd: convert mfill_atomic_hugetlb() to use a folio
mm: convert copy_user_huge_page() to copy_user_folio()
userfaultfd: convert mfill_atomic() to use a folio
include/linux/hugetlb.h | 4 +--
include/linux/mm.h | 15 ++++----
include/linux/shmem_fs.h | 4 +--
mm/hugetlb.c | 41 +++++++++++-----------
mm/memory.c | 64 ++++++++++++++++------------------
mm/shmem.c | 16 ++++-----
mm/userfaultfd.c | 75 ++++++++++++++++++++--------------------
7 files changed, 106 insertions(+), 113 deletions(-)
--
2.25.1
^ permalink raw reply [flat|nested] 15+ messages in thread
* [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
2023-03-30 17:02 ` Sidhartha Kumar
2023-03-30 13:40 ` [PATCH v4 2/6] userfaultfd: use kmap_local_page() in copy_huge_page_from_user() Peng Zhang
` (4 subsequent siblings)
5 siblings, 1 reply; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
Call vma_alloc_folio() directly instead of alloc_page_vma(). Add an
assertion that this is a single-page folio and removes several calls to
compound_head().
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
mm/userfaultfd.c | 25 +++++++++++++------------
1 file changed, 13 insertions(+), 12 deletions(-)
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 7f1b5f8b712c..efa9e1d681ee 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -137,15 +137,15 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
{
void *page_kaddr;
int ret;
- struct page *page;
+ struct folio *folio;
if (!*pagep) {
ret = -ENOMEM;
- page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
- if (!page)
+ folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, dst_vma, dst_addr, false);
+ if (!folio)
goto out;
- page_kaddr = kmap_local_page(page);
+ page_kaddr = kmap_local_folio(folio, 0);
/*
* The read mmap_lock is held here. Despite the
* mmap_lock being read recursive a deadlock is still
@@ -171,36 +171,37 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
/* fallback to copy_from_user outside mmap_lock */
if (unlikely(ret)) {
ret = -ENOENT;
- *pagep = page;
+ *pagep = &folio->page;
/* don't free the page */
goto out;
}
- flush_dcache_page(page);
+ flush_dcache_folio(folio);
} else {
- page = *pagep;
+ folio = page_folio(*pagep);
+ VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
*pagep = NULL;
}
/*
- * The memory barrier inside __SetPageUptodate makes sure that
+ * The memory barrier inside __folio_mark_uptodate makes sure that
* preceding stores to the page contents become visible before
* the set_pte_at() write.
*/
- __SetPageUptodate(page);
+ __folio_mark_uptodate(folio);
ret = -ENOMEM;
- if (mem_cgroup_charge(page_folio(page), dst_vma->vm_mm, GFP_KERNEL))
+ if (mem_cgroup_charge(folio, dst_vma->vm_mm, GFP_KERNEL))
goto out_release;
ret = mfill_atomic_install_pte(dst_pmd, dst_vma, dst_addr,
- page, true, flags);
+ &folio->page, true, flags);
if (ret)
goto out_release;
out:
return ret;
out_release:
- put_page(page);
+ folio_put(folio);
goto out;
}
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH v4 2/6] userfaultfd: use kmap_local_page() in copy_huge_page_from_user()
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
2023-03-30 13:40 ` [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
2023-03-30 13:40 ` [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user() Peng Zhang
` (3 subsequent siblings)
5 siblings, 0 replies; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
kmap() and kmap_atomic() are being deprecated in favor of
kmap_local_page() which is appropriate for any thread local context.[1]
Let's replace the kmap() and kmap_atomic() with kmap_local_page() in
copy_huge_page_from_user(). When allow_pagefault is false, disable page
faults to prevent potential deadlock.[2]
[1] https://lore.kernel.org/all/20220813220034.806698-1-ira.weiny@intel.com/
[2] https://lkml.kernel.org/r/20221025220136.2366143-1-ira.weiny@intel.com
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
mm/memory.c | 14 ++++++--------
1 file changed, 6 insertions(+), 8 deletions(-)
diff --git a/mm/memory.c b/mm/memory.c
index f77fccb5310c..c47b8991410a 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -5866,16 +5866,14 @@ long copy_huge_page_from_user(struct page *dst_page,
for (i = 0; i < pages_per_huge_page; i++) {
subpage = nth_page(dst_page, i);
- if (allow_pagefault)
- page_kaddr = kmap(subpage);
- else
- page_kaddr = kmap_atomic(subpage);
+ page_kaddr = kmap_local_page(subpage);
+ if (!allow_pagefault)
+ pagefault_disable();
rc = copy_from_user(page_kaddr,
usr_src + i * PAGE_SIZE, PAGE_SIZE);
- if (allow_pagefault)
- kunmap(subpage);
- else
- kunmap_atomic(page_kaddr);
+ if (!allow_pagefault)
+ pagefault_enable();
+ kunmap_local(page_kaddr);
ret_val -= (PAGE_SIZE - rc);
if (rc)
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user()
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
2023-03-30 13:40 ` [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio Peng Zhang
2023-03-30 13:40 ` [PATCH v4 2/6] userfaultfd: use kmap_local_page() in copy_huge_page_from_user() Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
2023-03-30 17:18 ` Sidhartha Kumar
2023-03-30 13:40 ` [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio Peng Zhang
` (2 subsequent siblings)
5 siblings, 1 reply; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
Replace copy_huge_page_from_user() with copy_folio_from_user().
copy_folio_from_user() does the same as copy_huge_page_from_user(), but
takes in a folio instead of a page. Convert page_kaddr to kaddr in
copy_folio_from_user() to do indenting cleanup.
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
include/linux/mm.h | 7 +++----
mm/hugetlb.c | 5 ++---
mm/memory.c | 26 ++++++++++++--------------
mm/userfaultfd.c | 6 ++----
4 files changed, 19 insertions(+), 25 deletions(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index e249208f8fbe..cf4d773ca506 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -3682,10 +3682,9 @@ extern void copy_user_huge_page(struct page *dst, struct page *src,
unsigned long addr_hint,
struct vm_area_struct *vma,
unsigned int pages_per_huge_page);
-extern long copy_huge_page_from_user(struct page *dst_page,
- const void __user *usr_src,
- unsigned int pages_per_huge_page,
- bool allow_pagefault);
+long copy_folio_from_user(struct folio *dst_folio,
+ const void __user *usr_src,
+ bool allow_pagefault);
/**
* vma_is_special_huge - Are transhuge page-table entries considered special?
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 7e4a80769c9e..aade1b513474 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6217,9 +6217,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
goto out;
}
- ret = copy_huge_page_from_user(&folio->page,
- (const void __user *) src_addr,
- pages_per_huge_page(h), false);
+ ret = copy_folio_from_user(folio, (const void __user *) src_addr,
+ false);
/* fallback to copy_from_user outside mmap_lock */
if (unlikely(ret)) {
diff --git a/mm/memory.c b/mm/memory.c
index c47b8991410a..9d59dad319b3 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -5854,35 +5854,33 @@ void copy_user_huge_page(struct page *dst, struct page *src,
process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
}
-long copy_huge_page_from_user(struct page *dst_page,
- const void __user *usr_src,
- unsigned int pages_per_huge_page,
- bool allow_pagefault)
+long copy_folio_from_user(struct folio *dst_folio,
+ const void __user *usr_src,
+ bool allow_pagefault)
{
- void *page_kaddr;
+ void *kaddr;
unsigned long i, rc = 0;
- unsigned long ret_val = pages_per_huge_page * PAGE_SIZE;
+ unsigned int nr_pages = folio_nr_pages(dst_folio);
+ unsigned long ret_val = nr_pages * PAGE_SIZE;
struct page *subpage;
- for (i = 0; i < pages_per_huge_page; i++) {
- subpage = nth_page(dst_page, i);
- page_kaddr = kmap_local_page(subpage);
+ for (i = 0; i < nr_pages; i++) {
+ subpage = folio_page(dst_folio, i);
+ kaddr = kmap_local_page(subpage);
if (!allow_pagefault)
pagefault_disable();
- rc = copy_from_user(page_kaddr,
- usr_src + i * PAGE_SIZE, PAGE_SIZE);
+ rc = copy_from_user(kaddr, usr_src + i * PAGE_SIZE, PAGE_SIZE);
if (!allow_pagefault)
pagefault_enable();
- kunmap_local(page_kaddr);
+ kunmap_local(kaddr);
ret_val -= (PAGE_SIZE - rc);
if (rc)
break;
- flush_dcache_page(subpage);
-
cond_resched();
}
+ flush_dcache_folio(dst_folio);
return ret_val;
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index efa9e1d681ee..b453a4d2a0d3 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -422,10 +422,8 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
mmap_read_unlock(dst_mm);
BUG_ON(!page);
- err = copy_huge_page_from_user(page,
- (const void __user *)src_addr,
- vma_hpagesize / PAGE_SIZE,
- true);
+ err = copy_folio_from_user(page_folio(page),
+ (const void __user *)src_addr, true);
if (unlikely(err)) {
err = -EFAULT;
goto out;
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
` (2 preceding siblings ...)
2023-03-30 13:40 ` [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user() Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
2023-03-30 17:29 ` Sidhartha Kumar
2023-03-30 13:40 ` [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio() Peng Zhang
2023-03-30 13:40 ` [PATCH v4 6/6] userfaultfd: convert mfill_atomic() to use a folio Peng Zhang
5 siblings, 1 reply; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
Convert hugetlb_mfill_atomic_pte() to take in a folio pointer instead of
a page pointer. Convert mfill_atomic_hugetlb() to use a folio.
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
include/linux/hugetlb.h | 4 ++--
mm/hugetlb.c | 26 +++++++++++++-------------
mm/userfaultfd.c | 16 ++++++++--------
3 files changed, 23 insertions(+), 23 deletions(-)
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 2a758bcd6719..28703fe22386 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -163,7 +163,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep);
+ struct folio **foliop);
#endif /* CONFIG_USERFAULTFD */
bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
struct vm_area_struct *vma,
@@ -397,7 +397,7 @@ static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep)
+ struct folio **foliop)
{
BUG();
return 0;
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index aade1b513474..c88f856ec2e2 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6178,7 +6178,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep)
+ struct folio **foliop)
{
struct mm_struct *dst_mm = dst_vma->vm_mm;
bool is_continue = uffd_flags_mode_is(flags, MFILL_ATOMIC_CONTINUE);
@@ -6201,8 +6201,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
if (IS_ERR(folio))
goto out;
folio_in_pagecache = true;
- } else if (!*pagep) {
- /* If a page already exists, then it's UFFDIO_COPY for
+ } else if (!*foliop) {
+ /* If a folio already exists, then it's UFFDIO_COPY for
* a non-missing case. Return -EEXIST.
*/
if (vm_shared &&
@@ -6237,33 +6237,33 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
ret = -ENOMEM;
goto out;
}
- *pagep = &folio->page;
- /* Set the outparam pagep and return to the caller to
+ *foliop = folio;
+ /* Set the outparam foliop and return to the caller to
* copy the contents outside the lock. Don't free the
- * page.
+ * folio.
*/
goto out;
}
} else {
if (vm_shared &&
hugetlbfs_pagecache_present(h, dst_vma, dst_addr)) {
- put_page(*pagep);
+ folio_put(*foliop);
ret = -EEXIST;
- *pagep = NULL;
+ *foliop = NULL;
goto out;
}
folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0);
if (IS_ERR(folio)) {
- put_page(*pagep);
+ folio_put(*foliop);
ret = -ENOMEM;
- *pagep = NULL;
+ *foliop = NULL;
goto out;
}
- copy_user_huge_page(&folio->page, *pagep, dst_addr, dst_vma,
+ copy_user_huge_page(&folio->page, &(*foliop)->page, dst_addr, dst_vma,
pages_per_huge_page(h));
- put_page(*pagep);
- *pagep = NULL;
+ folio_put(*foliop);
+ *foliop = NULL;
}
/*
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index b453a4d2a0d3..9e95af5ce2dd 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -322,7 +322,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
pte_t *dst_pte;
unsigned long src_addr, dst_addr;
long copied;
- struct page *page;
+ struct folio *folio;
unsigned long vma_hpagesize;
pgoff_t idx;
u32 hash;
@@ -342,7 +342,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
src_addr = src_start;
dst_addr = dst_start;
copied = 0;
- page = NULL;
+ folio = NULL;
vma_hpagesize = vma_kernel_pagesize(dst_vma);
/*
@@ -411,7 +411,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
}
err = hugetlb_mfill_atomic_pte(dst_pte, dst_vma, dst_addr,
- src_addr, flags, &page);
+ src_addr, flags, &folio);
hugetlb_vma_unlock_read(dst_vma);
mutex_unlock(&hugetlb_fault_mutex_table[hash]);
@@ -420,9 +420,9 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
if (unlikely(err == -ENOENT)) {
mmap_read_unlock(dst_mm);
- BUG_ON(!page);
+ BUG_ON(!folio);
- err = copy_folio_from_user(page_folio(page),
+ err = copy_folio_from_user(folio,
(const void __user *)src_addr, true);
if (unlikely(err)) {
err = -EFAULT;
@@ -433,7 +433,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
dst_vma = NULL;
goto retry;
} else
- BUG_ON(page);
+ BUG_ON(folio);
if (!err) {
dst_addr += vma_hpagesize;
@@ -450,8 +450,8 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
out_unlock:
mmap_read_unlock(dst_mm);
out:
- if (page)
- put_page(page);
+ if (folio)
+ folio_put(folio);
BUG_ON(copied < 0);
BUG_ON(err > 0);
BUG_ON(!copied && !err);
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio()
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
` (3 preceding siblings ...)
2023-03-30 13:40 ` [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
2023-03-30 18:15 ` Matthew Wilcox
2023-03-30 13:40 ` [PATCH v4 6/6] userfaultfd: convert mfill_atomic() to use a folio Peng Zhang
5 siblings, 1 reply; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
Replace copy_user_huge_page() with copy_user_folio(). copy_user_folio()
does the same as copy_user_huge_page(), but takes in folios instead of
pages. Convert copy_user_gigantic_page() to take in folios.
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
include/linux/mm.h | 8 ++++----
mm/hugetlb.c | 12 ++++++------
mm/memory.c | 28 ++++++++++++++--------------
3 files changed, 24 insertions(+), 24 deletions(-)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index cf4d773ca506..7b8fb88fb14d 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -3678,10 +3678,10 @@ extern const struct attribute_group memory_failure_attr_group;
extern void clear_huge_page(struct page *page,
unsigned long addr_hint,
unsigned int pages_per_huge_page);
-extern void copy_user_huge_page(struct page *dst, struct page *src,
- unsigned long addr_hint,
- struct vm_area_struct *vma,
- unsigned int pages_per_huge_page);
+void copy_user_folio(struct folio *dst, struct folio *src,
+ unsigned long addr_hint,
+ struct vm_area_struct *vma,
+ unsigned int pages_per_huge_page);
long copy_folio_from_user(struct folio *dst_folio,
const void __user *usr_src,
bool allow_pagefault);
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index c88f856ec2e2..fc3017a6a818 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5097,8 +5097,8 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
ret = PTR_ERR(new_folio);
break;
}
- copy_user_huge_page(&new_folio->page, ptepage, addr, dst_vma,
- npages);
+ copy_user_folio(new_folio, page_folio(ptepage), addr, dst_vma,
+ npages);
put_page(ptepage);
/* Install the new hugetlb folio if src pte stable */
@@ -5616,8 +5616,8 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
goto out_release_all;
}
- copy_user_huge_page(&new_folio->page, old_page, address, vma,
- pages_per_huge_page(h));
+ copy_user_folio(new_folio, page_folio(old_page), address, vma,
+ pages_per_huge_page(h));
__folio_mark_uptodate(new_folio);
mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, mm, haddr,
@@ -6260,8 +6260,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
*foliop = NULL;
goto out;
}
- copy_user_huge_page(&folio->page, &(*foliop)->page, dst_addr, dst_vma,
- pages_per_huge_page(h));
+ copy_user_folio(folio, *foliop, dst_addr, dst_vma,
+ pages_per_huge_page(h));
folio_put(*foliop);
*foliop = NULL;
}
diff --git a/mm/memory.c b/mm/memory.c
index 9d59dad319b3..8bddbb927806 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -5801,21 +5801,21 @@ void clear_huge_page(struct page *page,
process_huge_page(addr_hint, pages_per_huge_page, clear_subpage, page);
}
-static void copy_user_gigantic_page(struct page *dst, struct page *src,
- unsigned long addr,
- struct vm_area_struct *vma,
- unsigned int pages_per_huge_page)
+static void copy_user_gigantic_page(struct folio *dst, struct folio *src,
+ unsigned long addr,
+ struct vm_area_struct *vma,
+ unsigned int pages_per_huge_page)
{
int i;
- struct page *dst_base = dst;
- struct page *src_base = src;
+ struct page *dst_page;
+ struct page *src_page;
for (i = 0; i < pages_per_huge_page; i++) {
- dst = nth_page(dst_base, i);
- src = nth_page(src_base, i);
+ dst_page = folio_page(dst, i);
+ src_page = folio_page(src, i);
cond_resched();
- copy_user_highpage(dst, src, addr + i*PAGE_SIZE, vma);
+ copy_user_highpage(dst_page, src_page, addr + i*PAGE_SIZE, vma);
}
}
@@ -5833,15 +5833,15 @@ static void copy_subpage(unsigned long addr, int idx, void *arg)
addr, copy_arg->vma);
}
-void copy_user_huge_page(struct page *dst, struct page *src,
- unsigned long addr_hint, struct vm_area_struct *vma,
- unsigned int pages_per_huge_page)
+void copy_user_folio(struct folio *dst, struct folio *src,
+ unsigned long addr_hint, struct vm_area_struct *vma,
+ unsigned int pages_per_huge_page)
{
unsigned long addr = addr_hint &
~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
struct copy_subpage_arg arg = {
- .dst = dst,
- .src = src,
+ .dst = &dst->page,
+ .src = &src->page,
.vma = vma,
};
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH v4 6/6] userfaultfd: convert mfill_atomic() to use a folio
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
` (4 preceding siblings ...)
2023-03-30 13:40 ` [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio() Peng Zhang
@ 2023-03-30 13:40 ` Peng Zhang
5 siblings, 0 replies; 15+ messages in thread
From: Peng Zhang @ 2023-03-30 13:40 UTC (permalink / raw)
To: linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, sidhartha.kumar, wangkefeng.wang,
sunnanyong, ZhangPeng
From: ZhangPeng <zhangpeng362@huawei.com>
Convert mfill_atomic_pte_copy(), shmem_mfill_atomic_pte() and
mfill_atomic_pte() to take in a folio pointer.
Convert mfill_atomic() to use a folio.
Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
---
include/linux/shmem_fs.h | 4 ++--
mm/shmem.c | 16 ++++++++--------
mm/userfaultfd.c | 34 +++++++++++++++++-----------------
3 files changed, 27 insertions(+), 27 deletions(-)
diff --git a/include/linux/shmem_fs.h b/include/linux/shmem_fs.h
index 3bb8d21edbb3..9e151ba45068 100644
--- a/include/linux/shmem_fs.h
+++ b/include/linux/shmem_fs.h
@@ -158,10 +158,10 @@ extern int shmem_mfill_atomic_pte(pmd_t *dst_pmd,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep);
+ struct folio **foliop);
#else /* !CONFIG_SHMEM */
#define shmem_mfill_atomic_pte(dst_pmd, dst_vma, dst_addr, \
- src_addr, flags, pagep) ({ BUG(); 0; })
+ src_addr, flags, foliop) ({ BUG(); 0; })
#endif /* CONFIG_SHMEM */
#endif /* CONFIG_USERFAULTFD */
diff --git a/mm/shmem.c b/mm/shmem.c
index 6c08f5a75d3a..9218c955f482 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2548,7 +2548,7 @@ int shmem_mfill_atomic_pte(pmd_t *dst_pmd,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep)
+ struct folio **foliop)
{
struct inode *inode = file_inode(dst_vma->vm_file);
struct shmem_inode_info *info = SHMEM_I(inode);
@@ -2566,14 +2566,14 @@ int shmem_mfill_atomic_pte(pmd_t *dst_pmd,
* and now we find ourselves with -ENOMEM. Release the page, to
* avoid a BUG_ON in our caller.
*/
- if (unlikely(*pagep)) {
- put_page(*pagep);
- *pagep = NULL;
+ if (unlikely(*foliop)) {
+ folio_put(*foliop);
+ *foliop = NULL;
}
return -ENOMEM;
}
- if (!*pagep) {
+ if (!*foliop) {
ret = -ENOMEM;
folio = shmem_alloc_folio(gfp, info, pgoff);
if (!folio)
@@ -2605,7 +2605,7 @@ int shmem_mfill_atomic_pte(pmd_t *dst_pmd,
/* fallback to copy_from_user outside mmap_lock */
if (unlikely(ret)) {
- *pagep = &folio->page;
+ *foliop = folio;
ret = -ENOENT;
/* don't free the page */
goto out_unacct_blocks;
@@ -2616,9 +2616,9 @@ int shmem_mfill_atomic_pte(pmd_t *dst_pmd,
clear_user_highpage(&folio->page, dst_addr);
}
} else {
- folio = page_folio(*pagep);
+ folio = *foliop;
VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
- *pagep = NULL;
+ *foliop = NULL;
}
VM_BUG_ON(folio_test_locked(folio));
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 9e95af5ce2dd..74b122887901 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -133,13 +133,13 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep)
+ struct folio **foliop)
{
void *page_kaddr;
int ret;
struct folio *folio;
- if (!*pagep) {
+ if (!*foliop) {
ret = -ENOMEM;
folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, dst_vma, dst_addr, false);
if (!folio)
@@ -171,16 +171,16 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
/* fallback to copy_from_user outside mmap_lock */
if (unlikely(ret)) {
ret = -ENOENT;
- *pagep = &folio->page;
+ *foliop = folio;
/* don't free the page */
goto out;
}
flush_dcache_folio(folio);
} else {
- folio = page_folio(*pagep);
+ folio = *foliop;
VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
- *pagep = NULL;
+ *foliop = NULL;
}
/*
@@ -471,7 +471,7 @@ static __always_inline ssize_t mfill_atomic_pte(pmd_t *dst_pmd,
unsigned long dst_addr,
unsigned long src_addr,
uffd_flags_t flags,
- struct page **pagep)
+ struct folio **foliop)
{
ssize_t err;
@@ -494,14 +494,14 @@ static __always_inline ssize_t mfill_atomic_pte(pmd_t *dst_pmd,
if (uffd_flags_mode_is(flags, MFILL_ATOMIC_COPY))
err = mfill_atomic_pte_copy(dst_pmd, dst_vma,
dst_addr, src_addr,
- flags, pagep);
+ flags, foliop);
else
err = mfill_atomic_pte_zeropage(dst_pmd,
dst_vma, dst_addr);
} else {
err = shmem_mfill_atomic_pte(dst_pmd, dst_vma,
dst_addr, src_addr,
- flags, pagep);
+ flags, foliop);
}
return err;
@@ -519,7 +519,7 @@ static __always_inline ssize_t mfill_atomic(struct mm_struct *dst_mm,
pmd_t *dst_pmd;
unsigned long src_addr, dst_addr;
long copied;
- struct page *page;
+ struct folio *folio;
/*
* Sanitize the command parameters:
@@ -534,7 +534,7 @@ static __always_inline ssize_t mfill_atomic(struct mm_struct *dst_mm,
src_addr = src_start;
dst_addr = dst_start;
copied = 0;
- page = NULL;
+ folio = NULL;
retry:
mmap_read_lock(dst_mm);
@@ -630,16 +630,16 @@ static __always_inline ssize_t mfill_atomic(struct mm_struct *dst_mm,
BUG_ON(pmd_trans_huge(*dst_pmd));
err = mfill_atomic_pte(dst_pmd, dst_vma, dst_addr,
- src_addr, flags, &page);
+ src_addr, flags, &folio);
cond_resched();
if (unlikely(err == -ENOENT)) {
void *page_kaddr;
mmap_read_unlock(dst_mm);
- BUG_ON(!page);
+ BUG_ON(!folio);
- page_kaddr = kmap_local_page(page);
+ page_kaddr = kmap_local_folio(folio, 0);
err = copy_from_user(page_kaddr,
(const void __user *) src_addr,
PAGE_SIZE);
@@ -648,10 +648,10 @@ static __always_inline ssize_t mfill_atomic(struct mm_struct *dst_mm,
err = -EFAULT;
goto out;
}
- flush_dcache_page(page);
+ flush_dcache_folio(folio);
goto retry;
} else
- BUG_ON(page);
+ BUG_ON(folio);
if (!err) {
dst_addr += PAGE_SIZE;
@@ -668,8 +668,8 @@ static __always_inline ssize_t mfill_atomic(struct mm_struct *dst_mm,
out_unlock:
mmap_read_unlock(dst_mm);
out:
- if (page)
- put_page(page);
+ if (folio)
+ folio_put(folio);
BUG_ON(copied < 0);
BUG_ON(err > 0);
BUG_ON(!copied && !err);
--
2.25.1
^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio
2023-03-30 13:40 ` [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio Peng Zhang
@ 2023-03-30 17:02 ` Sidhartha Kumar
2023-03-30 18:02 ` Matthew Wilcox
0 siblings, 1 reply; 15+ messages in thread
From: Sidhartha Kumar @ 2023-03-30 17:02 UTC (permalink / raw)
To: Peng Zhang, linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, wangkefeng.wang, sunnanyong
On 3/30/23 6:40 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
>
> Call vma_alloc_folio() directly instead of alloc_page_vma(). Add an
> assertion that this is a single-page folio and removes several calls to
> compound_head().
>
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
> mm/userfaultfd.c | 25 +++++++++++++------------
> 1 file changed, 13 insertions(+), 12 deletions(-)
>
> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> index 7f1b5f8b712c..efa9e1d681ee 100644
> --- a/mm/userfaultfd.c
> +++ b/mm/userfaultfd.c
> @@ -137,15 +137,15 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
> {
> void *page_kaddr;
> int ret;
> - struct page *page;
> + struct folio *folio;
>
> if (!*pagep) {
> ret = -ENOMEM;
> - page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
> - if (!page)
> + folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, dst_vma, dst_addr, false);
> + if (!folio)
> goto out;
>
> - page_kaddr = kmap_local_page(page);
> + page_kaddr = kmap_local_folio(folio, 0);
Should this variable name be kept as page_kaddr or should it be changed
to something like folio_kaddr? kmap_local_folio() returns
page_address(), so maybe page_kaddr is better.
> /*
> * The read mmap_lock is held here. Despite the
> * mmap_lock being read recursive a deadlock is still
> @@ -171,36 +171,37 @@ static int mfill_atomic_pte_copy(pmd_t *dst_pmd,
> /* fallback to copy_from_user outside mmap_lock */
> if (unlikely(ret)) {
> ret = -ENOENT;
> - *pagep = page;
> + *pagep = &folio->page;
> /* don't free the page */
> goto out;
> }
>
> - flush_dcache_page(page);
> + flush_dcache_folio(folio);
> } else {
> - page = *pagep;
> + folio = page_folio(*pagep);
> + VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
> *pagep = NULL;
> }
>
> /*
> - * The memory barrier inside __SetPageUptodate makes sure that
> + * The memory barrier inside __folio_mark_uptodate makes sure that
> * preceding stores to the page contents become visible before
> * the set_pte_at() write.
> */
> - __SetPageUptodate(page);
> + __folio_mark_uptodate(folio);
>
> ret = -ENOMEM;
> - if (mem_cgroup_charge(page_folio(page), dst_vma->vm_mm, GFP_KERNEL))
> + if (mem_cgroup_charge(folio, dst_vma->vm_mm, GFP_KERNEL))
> goto out_release;
>
> ret = mfill_atomic_install_pte(dst_pmd, dst_vma, dst_addr,
> - page, true, flags);
> + &folio->page, true, flags);
> if (ret)
> goto out_release;
> out:
> return ret;
> out_release:
> - put_page(page);
> + folio_put(folio);
> goto out;
> }
Thanks,
Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>
>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user()
2023-03-30 13:40 ` [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user() Peng Zhang
@ 2023-03-30 17:18 ` Sidhartha Kumar
0 siblings, 0 replies; 15+ messages in thread
From: Sidhartha Kumar @ 2023-03-30 17:18 UTC (permalink / raw)
To: Peng Zhang, linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, wangkefeng.wang, sunnanyong
On 3/30/23 6:40 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
>
> Replace copy_huge_page_from_user() with copy_folio_from_user().
> copy_folio_from_user() does the same as copy_huge_page_from_user(), but
> takes in a folio instead of a page. Convert page_kaddr to kaddr in
> copy_folio_from_user() to do indenting cleanup.
>
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
> include/linux/mm.h | 7 +++----
> mm/hugetlb.c | 5 ++---
> mm/memory.c | 26 ++++++++++++--------------
> mm/userfaultfd.c | 6 ++----
> 4 files changed, 19 insertions(+), 25 deletions(-)
>
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index e249208f8fbe..cf4d773ca506 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -3682,10 +3682,9 @@ extern void copy_user_huge_page(struct page *dst, struct page *src,
> unsigned long addr_hint,
> struct vm_area_struct *vma,
> unsigned int pages_per_huge_page);
> -extern long copy_huge_page_from_user(struct page *dst_page,
> - const void __user *usr_src,
> - unsigned int pages_per_huge_page,
> - bool allow_pagefault);
> +long copy_folio_from_user(struct folio *dst_folio,
> + const void __user *usr_src,
> + bool allow_pagefault);
>
> /**
> * vma_is_special_huge - Are transhuge page-table entries considered special?
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 7e4a80769c9e..aade1b513474 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -6217,9 +6217,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> goto out;
> }
>
> - ret = copy_huge_page_from_user(&folio->page,
> - (const void __user *) src_addr,
> - pages_per_huge_page(h), false);
> + ret = copy_folio_from_user(folio, (const void __user *) src_addr,
> + false);
>
> /* fallback to copy_from_user outside mmap_lock */
> if (unlikely(ret)) {
> diff --git a/mm/memory.c b/mm/memory.c
> index c47b8991410a..9d59dad319b3 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -5854,35 +5854,33 @@ void copy_user_huge_page(struct page *dst, struct page *src,
> process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
> }
>
> -long copy_huge_page_from_user(struct page *dst_page,
> - const void __user *usr_src,
> - unsigned int pages_per_huge_page,
> - bool allow_pagefault)
> +long copy_folio_from_user(struct folio *dst_folio,
> + const void __user *usr_src,
> + bool allow_pagefault)
> {
> - void *page_kaddr;
> + void *kaddr;
> unsigned long i, rc = 0;
> - unsigned long ret_val = pages_per_huge_page * PAGE_SIZE;
> + unsigned int nr_pages = folio_nr_pages(dst_folio);
> + unsigned long ret_val = nr_pages * PAGE_SIZE;
> struct page *subpage;
>
> - for (i = 0; i < pages_per_huge_page; i++) {
> - subpage = nth_page(dst_page, i);
> - page_kaddr = kmap_local_page(subpage);
> + for (i = 0; i < nr_pages; i++) {
> + subpage = folio_page(dst_folio, i);
> + kaddr = kmap_local_page(subpage);
> if (!allow_pagefault)
> pagefault_disable();
> - rc = copy_from_user(page_kaddr,
> - usr_src + i * PAGE_SIZE, PAGE_SIZE);
> + rc = copy_from_user(kaddr, usr_src + i * PAGE_SIZE, PAGE_SIZE);
> if (!allow_pagefault)
> pagefault_enable();
> - kunmap_local(page_kaddr);
> + kunmap_local(kaddr);
>
> ret_val -= (PAGE_SIZE - rc);
> if (rc)
> break;
>
> - flush_dcache_page(subpage);
> -
> cond_resched();
> }
> + flush_dcache_folio(dst_folio);
> return ret_val;
> }
> #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> index efa9e1d681ee..b453a4d2a0d3 100644
> --- a/mm/userfaultfd.c
> +++ b/mm/userfaultfd.c
> @@ -422,10 +422,8 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> mmap_read_unlock(dst_mm);
> BUG_ON(!page);
>
> - err = copy_huge_page_from_user(page,
> - (const void __user *)src_addr,
> - vma_hpagesize / PAGE_SIZE,
> - true);
> + err = copy_folio_from_user(page_folio(page),
> + (const void __user *)src_addr, true);
> if (unlikely(err)) {
> err = -EFAULT;
> goto out;
Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio
2023-03-30 13:40 ` [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio Peng Zhang
@ 2023-03-30 17:29 ` Sidhartha Kumar
0 siblings, 0 replies; 15+ messages in thread
From: Sidhartha Kumar @ 2023-03-30 17:29 UTC (permalink / raw)
To: Peng Zhang, linux-mm, linux-kernel, akpm, willy, mike.kravetz
Cc: vishal.moola, muchun.song, wangkefeng.wang, sunnanyong
On 3/30/23 6:40 AM, Peng Zhang wrote:
> From: ZhangPeng <zhangpeng362@huawei.com>
>
> Convert hugetlb_mfill_atomic_pte() to take in a folio pointer instead of
> a page pointer. Convert mfill_atomic_hugetlb() to use a folio.
>
> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
> ---
> include/linux/hugetlb.h | 4 ++--
> mm/hugetlb.c | 26 +++++++++++++-------------
> mm/userfaultfd.c | 16 ++++++++--------
> 3 files changed, 23 insertions(+), 23 deletions(-)
>
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index 2a758bcd6719..28703fe22386 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -163,7 +163,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> unsigned long dst_addr,
> unsigned long src_addr,
> uffd_flags_t flags,
> - struct page **pagep);
> + struct folio **foliop);
> #endif /* CONFIG_USERFAULTFD */
> bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
> struct vm_area_struct *vma,
> @@ -397,7 +397,7 @@ static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> unsigned long dst_addr,
> unsigned long src_addr,
> uffd_flags_t flags,
> - struct page **pagep)
> + struct folio **foliop)
> {
> BUG();
> return 0;
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index aade1b513474..c88f856ec2e2 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -6178,7 +6178,7 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> unsigned long dst_addr,
> unsigned long src_addr,
> uffd_flags_t flags,
> - struct page **pagep)
> + struct folio **foliop)
> {
> struct mm_struct *dst_mm = dst_vma->vm_mm;
> bool is_continue = uffd_flags_mode_is(flags, MFILL_ATOMIC_CONTINUE);
> @@ -6201,8 +6201,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> if (IS_ERR(folio))
> goto out;
> folio_in_pagecache = true;
> - } else if (!*pagep) {
> - /* If a page already exists, then it's UFFDIO_COPY for
> + } else if (!*foliop) {
> + /* If a folio already exists, then it's UFFDIO_COPY for
> * a non-missing case. Return -EEXIST.
> */
> if (vm_shared &&
> @@ -6237,33 +6237,33 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
> ret = -ENOMEM;
> goto out;
> }
> - *pagep = &folio->page;
> - /* Set the outparam pagep and return to the caller to
> + *foliop = folio;
> + /* Set the outparam foliop and return to the caller to
> * copy the contents outside the lock. Don't free the
> - * page.
> + * folio.
> */
> goto out;
> }
> } else {
> if (vm_shared &&
> hugetlbfs_pagecache_present(h, dst_vma, dst_addr)) {
> - put_page(*pagep);
> + folio_put(*foliop);
> ret = -EEXIST;
> - *pagep = NULL;
> + *foliop = NULL;
> goto out;
> }
>
> folio = alloc_hugetlb_folio(dst_vma, dst_addr, 0);
> if (IS_ERR(folio)) {
> - put_page(*pagep);
> + folio_put(*foliop);
> ret = -ENOMEM;
> - *pagep = NULL;
> + *foliop = NULL;
> goto out;
> }
> - copy_user_huge_page(&folio->page, *pagep, dst_addr, dst_vma,
> + copy_user_huge_page(&folio->page, &(*foliop)->page, dst_addr, dst_vma,
> pages_per_huge_page(h));
> - put_page(*pagep);
> - *pagep = NULL;
> + folio_put(*foliop);
> + *foliop = NULL;
> }
>
> /*
> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> index b453a4d2a0d3..9e95af5ce2dd 100644
> --- a/mm/userfaultfd.c
> +++ b/mm/userfaultfd.c
> @@ -322,7 +322,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> pte_t *dst_pte;
> unsigned long src_addr, dst_addr;
> long copied;
> - struct page *page;
> + struct folio *folio;
> unsigned long vma_hpagesize;
> pgoff_t idx;
> u32 hash;
> @@ -342,7 +342,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> src_addr = src_start;
> dst_addr = dst_start;
> copied = 0;
> - page = NULL;
> + folio = NULL;
> vma_hpagesize = vma_kernel_pagesize(dst_vma);
>
> /*
> @@ -411,7 +411,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> }
>
> err = hugetlb_mfill_atomic_pte(dst_pte, dst_vma, dst_addr,
> - src_addr, flags, &page);
> + src_addr, flags, &folio);
>
> hugetlb_vma_unlock_read(dst_vma);
> mutex_unlock(&hugetlb_fault_mutex_table[hash]);
> @@ -420,9 +420,9 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
>
> if (unlikely(err == -ENOENT)) {
> mmap_read_unlock(dst_mm);
> - BUG_ON(!page);
> + BUG_ON(!folio);
>
> - err = copy_folio_from_user(page_folio(page),
> + err = copy_folio_from_user(folio,
> (const void __user *)src_addr, true);
> if (unlikely(err)) {
> err = -EFAULT;
> @@ -433,7 +433,7 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> dst_vma = NULL;
> goto retry;
> } else
> - BUG_ON(page);
> + BUG_ON(folio);
>
> if (!err) {
> dst_addr += vma_hpagesize;
> @@ -450,8 +450,8 @@ static __always_inline ssize_t mfill_atomic_hugetlb(
> out_unlock:
> mmap_read_unlock(dst_mm);
> out:
> - if (page)
> - put_page(page);
> + if (folio)
> + folio_put(folio);
> BUG_ON(copied < 0);
> BUG_ON(err > 0);
> BUG_ON(!copied && !err);
Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio
2023-03-30 17:02 ` Sidhartha Kumar
@ 2023-03-30 18:02 ` Matthew Wilcox
2023-03-31 9:38 ` zhangpeng (AS)
2023-04-06 21:28 ` Mike Kravetz
0 siblings, 2 replies; 15+ messages in thread
From: Matthew Wilcox @ 2023-03-30 18:02 UTC (permalink / raw)
To: Sidhartha Kumar
Cc: Peng Zhang, linux-mm, linux-kernel, akpm, mike.kravetz,
vishal.moola, muchun.song, wangkefeng.wang, sunnanyong
On Thu, Mar 30, 2023 at 10:02:52AM -0700, Sidhartha Kumar wrote:
> On 3/30/23 6:40 AM, Peng Zhang wrote:
> > From: ZhangPeng <zhangpeng362@huawei.com>
> >
> > Call vma_alloc_folio() directly instead of alloc_page_vma(). Add an
> > assertion that this is a single-page folio and removes several calls to
> > compound_head().
There's no added assertion in this patch any more, so I'd drop that
part of the description.
> > if (!*pagep) {
> > ret = -ENOMEM;
> > - page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
> > - if (!page)
> > + folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, dst_vma, dst_addr, false);
> > + if (!folio)
> > goto out;
> > - page_kaddr = kmap_local_page(page);
> > + page_kaddr = kmap_local_folio(folio, 0);
>
> Should this variable name be kept as page_kaddr or should it be changed to
> something like folio_kaddr? kmap_local_folio() returns page_address(), so
> maybe page_kaddr is better.
I'd just call it 'kaddr'. Or 'addr'.
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio()
2023-03-30 13:40 ` [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio() Peng Zhang
@ 2023-03-30 18:15 ` Matthew Wilcox
2023-03-31 9:38 ` zhangpeng (AS)
0 siblings, 1 reply; 15+ messages in thread
From: Matthew Wilcox @ 2023-03-30 18:15 UTC (permalink / raw)
To: Peng Zhang
Cc: linux-mm, linux-kernel, akpm, mike.kravetz, vishal.moola,
muchun.song, sidhartha.kumar, wangkefeng.wang, sunnanyong
On Thu, Mar 30, 2023 at 09:40:44PM +0800, Peng Zhang wrote:
> +void copy_user_folio(struct folio *dst, struct folio *src,
> + unsigned long addr_hint,
> + struct vm_area_struct *vma,
> + unsigned int pages_per_huge_page);
Do we really want to pass in pages_per_huge_page here? We can get
that from folio_nr_pages(dst).
> +++ b/mm/hugetlb.c
> @@ -5097,8 +5097,8 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
> ret = PTR_ERR(new_folio);
> break;
> }
> - copy_user_huge_page(&new_folio->page, ptepage, addr, dst_vma,
> - npages);
> + copy_user_folio(new_folio, page_folio(ptepage), addr, dst_vma,
> + npages);
> put_page(ptepage);
This function should be converted to s/ptepage/pte_folio/, but that's
not for this patch series.
> - copy_user_huge_page(&new_folio->page, old_page, address, vma,
> - pages_per_huge_page(h));
> + copy_user_folio(new_folio, page_folio(old_page), address, vma,
> + pages_per_huge_page(h));
Likewise for s/old_page/old_folio/
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio
2023-03-30 18:02 ` Matthew Wilcox
@ 2023-03-31 9:38 ` zhangpeng (AS)
2023-04-06 21:28 ` Mike Kravetz
1 sibling, 0 replies; 15+ messages in thread
From: zhangpeng (AS) @ 2023-03-31 9:38 UTC (permalink / raw)
To: Matthew Wilcox, Sidhartha Kumar
Cc: linux-mm, linux-kernel, akpm, mike.kravetz, vishal.moola,
muchun.song, wangkefeng.wang, sunnanyong
On 2023/3/31 2:02, Matthew Wilcox wrote:
> On Thu, Mar 30, 2023 at 10:02:52AM -0700, Sidhartha Kumar wrote:
>> On 3/30/23 6:40 AM, Peng Zhang wrote:
>>> From: ZhangPeng <zhangpeng362@huawei.com>
>>>
>>> Call vma_alloc_folio() directly instead of alloc_page_vma(). Add an
>>> assertion that this is a single-page folio and removes several calls to
>>> compound_head().
> There's no added assertion in this patch any more, so I'd drop that
> part of the description.
Thanks. I'll update this description.
>>> if (!*pagep) {
>>> ret = -ENOMEM;
>>> - page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
>>> - if (!page)
>>> + folio = vma_alloc_folio(GFP_HIGHUSER_MOVABLE, 0, dst_vma, dst_addr, false);
>>> + if (!folio)
>>> goto out;
>>> - page_kaddr = kmap_local_page(page);
>>> + page_kaddr = kmap_local_folio(folio, 0);
>> Should this variable name be kept as page_kaddr or should it be changed to
>> something like folio_kaddr? kmap_local_folio() returns page_address(), so
>> maybe page_kaddr is better.
> I'd just call it 'kaddr'. Or 'addr'.
Agreed. I'll change page_kaddr to kaddr.
Thanks for your review.
Best Regards,
Peng
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio()
2023-03-30 18:15 ` Matthew Wilcox
@ 2023-03-31 9:38 ` zhangpeng (AS)
0 siblings, 0 replies; 15+ messages in thread
From: zhangpeng (AS) @ 2023-03-31 9:38 UTC (permalink / raw)
To: Matthew Wilcox
Cc: linux-mm, linux-kernel, akpm, mike.kravetz, vishal.moola,
muchun.song, sidhartha.kumar, wangkefeng.wang, sunnanyong
On 2023/3/31 2:15, Matthew Wilcox wrote:
> On Thu, Mar 30, 2023 at 09:40:44PM +0800, Peng Zhang wrote:
>> +void copy_user_folio(struct folio *dst, struct folio *src,
>> + unsigned long addr_hint,
>> + struct vm_area_struct *vma,
>> + unsigned int pages_per_huge_page);
> Do we really want to pass in pages_per_huge_page here? We can get
> that from folio_nr_pages(dst).
Agreed. It's not necessary to pass in pages_per_huge_page. I'll remove
it in a v5.
>> +++ b/mm/hugetlb.c
>> @@ -5097,8 +5097,8 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
>> ret = PTR_ERR(new_folio);
>> break;
>> }
>> - copy_user_huge_page(&new_folio->page, ptepage, addr, dst_vma,
>> - npages);
>> + copy_user_folio(new_folio, page_folio(ptepage), addr, dst_vma,
>> + npages);
>> put_page(ptepage);
> This function should be converted to s/ptepage/pte_folio/, but that's
> not for this patch series.
Appreciate your suggestion. I'll convert this function in a future patch
set.
>> - copy_user_huge_page(&new_folio->page, old_page, address, vma,
>> - pages_per_huge_page(h));
>> + copy_user_folio(new_folio, page_folio(old_page), address, vma,
>> + pages_per_huge_page(h));
> Likewise for s/old_page/old_folio/
Will also convert this.
Best Regards,
Peng
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio
2023-03-30 18:02 ` Matthew Wilcox
2023-03-31 9:38 ` zhangpeng (AS)
@ 2023-04-06 21:28 ` Mike Kravetz
1 sibling, 0 replies; 15+ messages in thread
From: Mike Kravetz @ 2023-04-06 21:28 UTC (permalink / raw)
To: Matthew Wilcox
Cc: Sidhartha Kumar, Peng Zhang, linux-mm, linux-kernel, akpm,
vishal.moola, muchun.song, wangkefeng.wang, sunnanyong
On 03/30/23 19:02, Matthew Wilcox wrote:
> On Thu, Mar 30, 2023 at 10:02:52AM -0700, Sidhartha Kumar wrote:
> > On 3/30/23 6:40 AM, Peng Zhang wrote:
> > > From: ZhangPeng <zhangpeng362@huawei.com>
> > >
> > > Call vma_alloc_folio() directly instead of alloc_page_vma(). Add an
> > > assertion that this is a single-page folio and removes several calls to
> > > compound_head().
>
> There's no added assertion in this patch any more, so I'd drop that
> part of the description.
>
I thought it was this.
>
> - flush_dcache_page(page);
> + flush_dcache_folio(folio);
> } else {
> - page = *pagep;
> + folio = page_folio(*pagep);
> + VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
> *pagep = NULL;
> }
I was wondering what that VM_BUG_ON_FOLIO was added?
--
Mike Kravetz
^ permalink raw reply [flat|nested] 15+ messages in thread
end of thread, other threads:[~2023-04-06 21:29 UTC | newest]
Thread overview: 15+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-03-30 13:40 [PATCH v4 0/6] userfaultfd: convert userfaultfd functions to use folios Peng Zhang
2023-03-30 13:40 ` [PATCH v4 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio Peng Zhang
2023-03-30 17:02 ` Sidhartha Kumar
2023-03-30 18:02 ` Matthew Wilcox
2023-03-31 9:38 ` zhangpeng (AS)
2023-04-06 21:28 ` Mike Kravetz
2023-03-30 13:40 ` [PATCH v4 2/6] userfaultfd: use kmap_local_page() in copy_huge_page_from_user() Peng Zhang
2023-03-30 13:40 ` [PATCH v4 3/6] userfaultfd: convert copy_huge_page_from_user() to copy_folio_from_user() Peng Zhang
2023-03-30 17:18 ` Sidhartha Kumar
2023-03-30 13:40 ` [PATCH v4 4/6] userfaultfd: convert mfill_atomic_hugetlb() to use a folio Peng Zhang
2023-03-30 17:29 ` Sidhartha Kumar
2023-03-30 13:40 ` [PATCH v4 5/6] mm: convert copy_user_huge_page() to copy_user_folio() Peng Zhang
2023-03-30 18:15 ` Matthew Wilcox
2023-03-31 9:38 ` zhangpeng (AS)
2023-03-30 13:40 ` [PATCH v4 6/6] userfaultfd: convert mfill_atomic() to use a folio Peng Zhang
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.