[PATCH v2] mm/mmap.c: use IS_ERR_VALUE to check return value of get_unmapped_area

[PATCH v2] mm/mmap.c: use IS_ERR_VALUE to check return value of get_unmapped_area

[Gaowei Pu]

get_unmapped_area already cover the offset_in_page() check and returned
with error ptr. So replace offset_in_page() with IS_ERR_VALUE().

Signed-off-by: Gaowei Pu <pugaowei@gmail.com>

V2: fix compile warnning.
---
 mm/mmap.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/mm/mmap.c b/mm/mmap.c
index 7e8c3e8ae75f..6aebfeb6a486 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1430,7 +1430,7 @@ unsigned long do_mmap(struct file *file, unsigned long addr,
 	 * that it represents a valid section of the address space.
 	 */
 	addr = get_unmapped_area(file, addr, len, pgoff, flags);
-	if (offset_in_page(addr))
+	if (IS_ERR_VALUE(addr))
 		return addr;
 
 	if (flags & MAP_FIXED_NOREPLACE) {
@@ -2990,15 +2990,16 @@ static int do_brk_flags(unsigned long addr, unsigned long len, unsigned long fla
 	struct rb_node **rb_link, *rb_parent;
 	pgoff_t pgoff = addr >> PAGE_SHIFT;
 	int error;
+	unsigned long mapped_addr;
 
 	/* Until we need other flags, refuse anything except VM_EXEC. */
 	if ((flags & (~VM_EXEC)) != 0)
 		return -EINVAL;
 	flags |= VM_DATA_DEFAULT_FLAGS | VM_ACCOUNT | mm->def_flags;
 
-	error = get_unmapped_area(NULL, addr, len, 0, MAP_FIXED);
-	if (offset_in_page(error))
-		return error;
+	mapped_addr = get_unmapped_area(NULL, addr, len, 0, MAP_FIXED);
+	if (IS_ERR_VALUE(mapped_addr))
+		return mapped_addr;
 
 	error = mlock_future_check(mm, mm->def_flags, len);
 	if (error)
-- 
2.23.0

Re: [PATCH v2] mm/mmap.c: use IS_ERR_VALUE to check return value of get_unmapped_area

[Michal Hocko]

On Sat 12-10-19 18:25:12, Gaowei Pu wrote:
> get_unmapped_area already cover the offset_in_page() check and returned
> with error ptr. So replace offset_in_page() with IS_ERR_VALUE().

I have to say that I found offset_in_page check quite unintuitive. It is
a relict from the past I suspect. A newer code seems to use
IS_ERR_VALUE. If we want to make the code consistent then I am for it.
Could you check other users as well? git grep suggests checking
kernel/events/uprobes.c, mm/mmap.c and mm/mremap.c.

I would simply use the following for the changelog
"
get_unmapped_area returns an address or -errno on failure. Historically
we have checked for the failure by offset_in_page() which is correct but
quite hard to read. Newer code started using IS_ERR_VALUE which is much
easier to read. Convert remaining users of offset_in_page as well.

This shouldn't introduce any functional changes.
"

> Signed-off-by: Gaowei Pu <pugaowei@gmail.com>
> 
> V2: fix compile warnning.
> ---
>  mm/mmap.c | 9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/mmap.c b/mm/mmap.c
> index 7e8c3e8ae75f..6aebfeb6a486 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -1430,7 +1430,7 @@ unsigned long do_mmap(struct file *file, unsigned long addr,
>  	 * that it represents a valid section of the address space.
>  	 */
>  	addr = get_unmapped_area(file, addr, len, pgoff, flags);
> -	if (offset_in_page(addr))
> +	if (IS_ERR_VALUE(addr))
>  		return addr;
>  
>  	if (flags & MAP_FIXED_NOREPLACE) {
> @@ -2990,15 +2990,16 @@ static int do_brk_flags(unsigned long addr, unsigned long len, unsigned long fla
>  	struct rb_node **rb_link, *rb_parent;
>  	pgoff_t pgoff = addr >> PAGE_SHIFT;
>  	int error;
> +	unsigned long mapped_addr;
>  
>  	/* Until we need other flags, refuse anything except VM_EXEC. */
>  	if ((flags & (~VM_EXEC)) != 0)
>  		return -EINVAL;
>  	flags |= VM_DATA_DEFAULT_FLAGS | VM_ACCOUNT | mm->def_flags;
>  
> -	error = get_unmapped_area(NULL, addr, len, 0, MAP_FIXED);
> -	if (offset_in_page(error))
> -		return error;
> +	mapped_addr = get_unmapped_area(NULL, addr, len, 0, MAP_FIXED);
> +	if (IS_ERR_VALUE(mapped_addr))
> +		return mapped_addr;
>  
>  	error = mlock_future_check(mm, mm->def_flags, len);
>  	if (error)
> -- 
> 2.23.0
> 

-- 
Michal Hocko
SUSE Labs

Re: [PATCH v2] mm/mmap.c: use IS_ERR_VALUE to check return value of get_unmapped_area

[gaowei Pu]

[-- Attachment #1: Type: text/plain, Size: 1467 bytes --]

On Tue, Oct 15, 2019 at 8:33 PM Michal Hocko <mhocko@suse.com> wrote:

> On Sat 12-10-19 18:25:12, Gaowei Pu wrote:
> > get_unmapped_area already cover the offset_in_page() check and returned
> > with error ptr. So replace offset_in_page() with IS_ERR_VALUE().
>
> I have to say that I found offset_in_page check quite unintuitive. It is
> a relict from the past I suspect. A newer code seems to use
> IS_ERR_VALUE. If we want to make the code consistent then I am for it.
> Could you check other users as well? git grep suggests checking
> kernel/events/uprobes.c, mm/mmap.c and mm/mremap.c.
>
I only noticed the offset_in_page check after returned from get_unmapped_area
is redundant.
So I just replace the places where get_unmapped_area returned. As you
mentioned, IS_ERR_VALUE maybe
the newer code to check the error ptr compared to offset_in_page which is
really unintuitive and therefore I will
replace them in mm/mmap.c and mm/mremap.c.

> I would simply use the following for the changelog
> "
> get_unmapped_area returns an address or -errno on failure. Historically
> we have checked for the failure by offset_in_page() which is correct but
> quite hard to read. Newer code started using IS_ERR_VALUE which is much
> easier to read. Convert remaining users of offset_in_page as well.
>
> This shouldn't introduce any functional changes.
> "
>
> OK, I will use your changelog and resend a patch. Thanks.

>
> > 2.23.0
> >
>
> --
> Michal Hocko
> SUSE Labs
>

[-- Attachment #2: Type: text/html, Size: 2419 bytes --]