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

[Alexander Duyck <alexander.duyck@gmail.com>]

On Fri, Oct 18, 2019 at 7:00 AM zhong jiang <zhongjiang@huawei.com> wrote:
>
> On 2019/10/18 4:42, John Hubbard wrote:
> > On 10/16/19 8:19 PM, zhong jiang wrote:
> >> With the help of unsigned_lesser_than_zero.cocci. Unsigned 'nr_pages"'
> >> compare with zero. And __get_user_pages_locked will return an long value.
> >>
> >> The patch use a new local variable to store the return value of
> >> __get_user_pages_locked(). Then use it to compare with zero.
> > Hi Zhong,
> >
> > The above are actually more like notes to yourself, and while those are
> > good to have, but it's not yet a perfect commit description: it talks
> > about how you got here, which is only part of the story. A higher level
> > summary is better, and let the code itself cover the details.
> >
> > Like this:
> >
> > First line (subject):
> >
> >    mm/gup: allow CMA migration to propagate errors back to caller
> >
> > Commit description:
> >
> > check_and_migrate_cma_pages() was recording the result of
> > __get_user_pages_locked() in an unsigned "nr_pages" variable. Because
> > __get_user_pages_locked() returns a signed value that can include
> > negative errno values, this had the effect of hiding errors.
> >
> > Change check_and_migrate_cma_pages() implementation so that it
> > uses a signed variable instead, and propagates the results back
> > to the caller just as other gup internal functions do.
> >
> > This was discovered with the help of unsigned_lesser_than_zero.cocci.
> >
> >> Suggested-by: Andrew Morton <akpm@linux-foundation.org>
> >> Signed-off-by: zhong jiang <zhongjiang@huawei.com>
> >> ---
> >>  mm/gup.c | 8 +++++---
> >>  1 file changed, 5 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/mm/gup.c b/mm/gup.c
> >> index 8f236a3..1fe0ceb 100644
> >> --- a/mm/gup.c
> >> +++ b/mm/gup.c
> >> @@ -1443,6 +1443,7 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >>      bool drain_allow = true;
> >>      bool migrate_allow = true;
> >>      LIST_HEAD(cma_page_list);
> >> +    long ret;
> > Ira pointed out that this needs initialization, see below.
> >
> >>
> >>  check_again:
> >>      for (i = 0; i < nr_pages;) {
> >> @@ -1504,17 +1505,18 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >>               * again migrating any new CMA pages which we failed to isolate
> >>               * earlier.
> >>               */
> >> -            nr_pages = __get_user_pages_locked(tsk, mm, start, nr_pages,
> >> +            ret = __get_user_pages_locked(tsk, mm, start, nr_pages,
> >>                                                 pages, vmas, NULL,
> >>                                                 gup_flags);
> >>
> >> -            if ((nr_pages > 0) && migrate_allow) {
> >> +            nr_pages = ret;
> > Although technically correct, it makes me feel odd to see the assignment
> > done from signed to unsigned, *before* checking for >0. And Ira is hinting
> > at the same thing, when he asks if we can return early here. See below...
> >
> >> +            if ((ret > 0) && migrate_allow) {
> >>                      drain_allow = true;
> >>                      goto check_again;
> >>              }
> >>      }
> >>
> >> -    return nr_pages;
> >> +    return ret;
> >>  }
> >>  #else
> >>  static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >>
> > So, overall, I'd recommend this, instead:
> >
> > diff --git a/mm/gup.c b/mm/gup.c
> > index 23a9f9c9d377..72bc027037fa 100644
> > --- a/mm/gup.c
> > +++ b/mm/gup.c
> > @@ -1443,6 +1443,7 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
> >         bool drain_allow = true;
> >         bool migrate_allow = true;
> >         LIST_HEAD(cma_page_list);
> > +       long ret = nr_pages;
> >
> I think the ret should be assigned as zero. we will return ret if cma_page_list is empty.
> I miss something?

That wasn't the behavior before. Before if the cma_page_list was empty
nr_pages would not be modified from what what passed. I believe that
is the intended behavior for this since the cma_page_list seems like a
workaround to migrate out any CMA pages if present in the longterm
mapping. If there are no CMA pages the return result should be
nr_pages.

Thanks.

- Alex