2019-09-05 02:24:53

by zhong jiang

Subject: [PATCH v2] mm: Unsigned 'nr_pages' always larger than zero

With the help of unsigned_lesser_than_zero.cocci. Unsigned 'nr_pages'
compare with zero. And __gup_longterm_locked pass an long local variant
'rc' to check_and_migrate_cma_pages. Hence it is nicer to change the
parameter to long to fix the issue.

Fixes: 932f4a630a69 ("mm/gup: replace get_user_pages_longterm() with FOLL_LONGTERM")
Signed-off-by: zhong jiang <[email protected]>
---
mm/gup.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/mm/gup.c b/mm/gup.c
index 23a9f9c..ee0b71f 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1433,13 +1433,13 @@ static struct page *new_non_cma_page(struct page *page, unsigned long private)
static long check_and_migrate_cma_pages(struct task_struct *tsk,
struct mm_struct *mm,
unsigned long start,
- unsigned long nr_pages,
+ long nr_pages,
struct page **pages,
struct vm_area_struct **vmas,
unsigned int gup_flags)
{
- unsigned long i;
- unsigned long step;
+ long i;
+ long step;
bool drain_allow = true;
bool migrate_allow = true;
LIST_HEAD(cma_page_list);
@@ -1520,7 +1520,7 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
static long check_and_migrate_cma_pages(struct task_struct *tsk,
struct mm_struct *mm,
unsigned long start,
- unsigned long nr_pages,
+ long nr_pages,
struct page **pages,
struct vm_area_struct **vmas,
unsigned int gup_flags)
--
1.7.12.4

2019-09-05 03:46:38

by Matthew Wilcox

Subject: Re: [PATCH v2] mm: Unsigned 'nr_pages' always larger than zero

On Thu, Sep 05, 2019 at 10:17:51AM +0800, zhong jiang wrote:
> With the help of unsigned_lesser_than_zero.cocci. Unsigned 'nr_pages'
> compare with zero. And __gup_longterm_locked pass an long local variant
> 'rc' to check_and_migrate_cma_pages. Hence it is nicer to change the
> parameter to long to fix the issue.

I think this patch is right, but I have concerns about this cocci grep.

The code says:

if ((nr_pages > 0) && migrate_allow) {

There's nothing wrong with this (... other than the fact that nr_pages might
happen to be a negative errno). nr_pages might be 0, and this would be
exactly the right test for that situation. I suppose some might argue
that this should be != 0 instead of > 0, but it depends on the situation
which one would read better.

So please don't blindly make these changes; you're right this time.

2019-09-05 07:05:50

by zhong jiang

Subject: Re: [PATCH v2] mm: Unsigned 'nr_pages' always larger than zero

On 2019/9/5 11:12, Matthew Wilcox wrote:
> On Thu, Sep 05, 2019 at 10:17:51AM +0800, zhong jiang wrote:
>> With the help of unsigned_lesser_than_zero.cocci. Unsigned 'nr_pages'
>> compare with zero. And __gup_longterm_locked pass an long local variant
>> 'rc' to check_and_migrate_cma_pages. Hence it is nicer to change the
>> parameter to long to fix the issue.
> I think this patch is right, but I have concerns about this cocci grep.
>
> The code says:
>
> if ((nr_pages > 0) && migrate_allow) {
>
> There's nothing wrong with this (... other than the fact that nr_pages might
> happen to be a negative errno). nr_pages might be 0, and this would be
> exactly the right test for that situation. I suppose some might argue
> that this should be != 0 instead of > 0, but it depends on the situation
> which one would read better.
>
> So please don't blindly make these changes; you're right this time.
Thanks for your affirmation. but Andrew come up with anther fix, using an local long variant
to store the nr_pages. which one do you prefer ?

Thanks,
zhong jiang