[PATCH v2] mm/mremap: Don't account pages in vma_to_resize()

From: Dmitry Safonov
Date: Wed Jul 21 2021 - 09:13:30 EST


All this vm_unacct_memory(charged) dance seems to complicate the life
without a good reason. Furthermore, it seems not always done right on
error-pathes in mremap_to().
And worse than that: this `charged' difference is sometimes
double-accounted for growing MREMAP_DONTUNMAP mremap()s in move_vma():
: if (security_vm_enough_memory_mm(mm, new_len >> PAGE_SHIFT))

Let's not do this.
Account memory in mremap() fast-path for growing VMAs or in move_vma()
for actually moving things. The same simpler way as it's done by
vm_stat_account(), but with a difference to call
security_vm_enough_memory_mm() before copying/adjusting VMA.

Originally noticed by Chen Wandun:
https://lkml.kernel.org/r/20210717101942.120607-1-chenwandun@xxxxxxxxxx

Cc: Alexander Viro <viro@xxxxxxxxxxxxxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxx>
Cc: Brian Geffon <bgeffon@xxxxxxxxxx>
Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
Cc: Chen Wandun <chenwandun@xxxxxxxxxx>
Cc: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
Cc: Dan Williams <dan.j.williams@xxxxxxxxx>
Cc: Dave Jiang <dave.jiang@xxxxxxxxx>
Cc: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Jason Gunthorpe <jgg@xxxxxxxx>
Cc: John Hubbard <jhubbard@xxxxxxxxxx>
Cc: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
Cc: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
Cc: Minchan Kim <minchan@xxxxxxxxxx>
Cc: Ralph Campbell <rcampbell@xxxxxxxxxx>
Cc: Russell King <linux@xxxxxxxxxxxxxxx>
Cc: Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Vishal Verma <vishal.l.verma@xxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Cc: Wei Yongjun <weiyongjun1@xxxxxxxxxx>
Cc: Will Deacon <will@xxxxxxxxxx>
Fixes: e346b3813067 ("mm/mremap: add MREMAP_DONTUNMAP to mremap()")
Signed-off-by: Dmitry Safonov <dima@xxxxxxxxxx>
---
mm/mremap.c | 50 ++++++++++++++++++++++----------------------------
1 file changed, 22 insertions(+), 28 deletions(-)

diff --git a/mm/mremap.c b/mm/mremap.c
index 5989d3990020..b90c8e732e29 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -565,6 +565,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
bool *locked, unsigned long flags,
struct vm_userfaultfd_ctx *uf, struct list_head *uf_unmap)
{
+ long to_account = new_len - old_len;
struct mm_struct *mm = vma->vm_mm;
struct vm_area_struct *new_vma;
unsigned long vm_flags = vma->vm_flags;
@@ -583,6 +584,9 @@ static unsigned long move_vma(struct vm_area_struct *vma,
if (mm->map_count >= sysctl_max_map_count - 3)
return -ENOMEM;

+ if (unlikely(flags & MREMAP_DONTUNMAP))
+ to_account = new_len;
+
if (vma->vm_ops && vma->vm_ops->may_split) {
if (vma->vm_start != old_addr)
err = vma->vm_ops->may_split(vma, old_addr);
@@ -604,8 +608,8 @@ static unsigned long move_vma(struct vm_area_struct *vma,
if (err)
return err;

- if (unlikely(flags & MREMAP_DONTUNMAP && vm_flags & VM_ACCOUNT)) {
- if (security_vm_enough_memory_mm(mm, new_len >> PAGE_SHIFT))
+ if (vm_flags & VM_ACCOUNT) {
+ if (security_vm_enough_memory_mm(mm, to_account >> PAGE_SHIFT))
return -ENOMEM;
}

@@ -613,8 +617,8 @@ static unsigned long move_vma(struct vm_area_struct *vma,
new_vma = copy_vma(&vma, new_addr, new_len, new_pgoff,
&need_rmap_locks);
if (!new_vma) {
- if (unlikely(flags & MREMAP_DONTUNMAP && vm_flags & VM_ACCOUNT))
- vm_unacct_memory(new_len >> PAGE_SHIFT);
+ if (vm_flags & VM_ACCOUNT)
+ vm_unacct_memory(to_account >> PAGE_SHIFT);
return -ENOMEM;
}

@@ -708,8 +712,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
}

static struct vm_area_struct *vma_to_resize(unsigned long addr,
- unsigned long old_len, unsigned long new_len, unsigned long flags,
- unsigned long *p)
+ unsigned long old_len, unsigned long new_len, unsigned long flags)
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
@@ -768,13 +771,6 @@ static struct vm_area_struct *vma_to_resize(unsigned long addr,
(new_len - old_len) >> PAGE_SHIFT))
return ERR_PTR(-ENOMEM);

- if (vma->vm_flags & VM_ACCOUNT) {
- unsigned long charged = (new_len - old_len) >> PAGE_SHIFT;
- if (security_vm_enough_memory_mm(mm, charged))
- return ERR_PTR(-ENOMEM);
- *p = charged;
- }
-
return vma;
}

@@ -787,7 +783,6 @@ static unsigned long mremap_to(unsigned long addr, unsigned long old_len,
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
unsigned long ret = -EINVAL;
- unsigned long charged = 0;
unsigned long map_flags = 0;

if (offset_in_page(new_addr))
@@ -830,7 +825,7 @@ static unsigned long mremap_to(unsigned long addr, unsigned long old_len,
old_len = new_len;
}

- vma = vma_to_resize(addr, old_len, new_len, flags, &charged);
+ vma = vma_to_resize(addr, old_len, new_len, flags);
if (IS_ERR(vma)) {
ret = PTR_ERR(vma);
goto out;
@@ -853,7 +848,7 @@ static unsigned long mremap_to(unsigned long addr, unsigned long old_len,
((addr - vma->vm_start) >> PAGE_SHIFT),
map_flags);
if (IS_ERR_VALUE(ret))
- goto out1;
+ goto out;

/* We got a new mapping */
if (!(flags & MREMAP_FIXED))
@@ -862,12 +857,6 @@ static unsigned long mremap_to(unsigned long addr, unsigned long old_len,
ret = move_vma(vma, addr, old_len, new_len, new_addr, locked, flags, uf,
uf_unmap);

- if (!(offset_in_page(ret)))
- goto out;
-
-out1:
- vm_unacct_memory(charged);
-
out:
return ret;
}
@@ -899,7 +888,6 @@ SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
unsigned long ret = -EINVAL;
- unsigned long charged = 0;
bool locked = false;
bool downgraded = false;
struct vm_userfaultfd_ctx uf = NULL_VM_UFFD_CTX;
@@ -981,7 +969,7 @@ SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
/*
* Ok, we need to grow..
*/
- vma = vma_to_resize(addr, old_len, new_len, flags, &charged);
+ vma = vma_to_resize(addr, old_len, new_len, flags);
if (IS_ERR(vma)) {
ret = PTR_ERR(vma);
goto out;
@@ -992,10 +980,18 @@ SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
if (old_len == vma->vm_end - addr) {
/* can we just expand the current mapping? */
if (vma_expandable(vma, new_len - old_len)) {
- int pages = (new_len - old_len) >> PAGE_SHIFT;
+ long pages = (new_len - old_len) >> PAGE_SHIFT;
+
+ if (vma->vm_flags & VM_ACCOUNT) {
+ if (security_vm_enough_memory_mm(mm, pages)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ }

if (vma_adjust(vma, vma->vm_start, addr + new_len,
vma->vm_pgoff, NULL)) {
+ vm_unacct_memory(pages);
ret = -ENOMEM;
goto out;
}
@@ -1034,10 +1030,8 @@ SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
&locked, flags, &uf, &uf_unmap);
}
out:
- if (offset_in_page(ret)) {
- vm_unacct_memory(charged);
+ if (offset_in_page(ret))
locked = false;
- }
if (downgraded)
mmap_read_unlock(current->mm);
else
--
2.32.0