Reference counters are preferred to use refcount_t instead of
atomic_t.
This is because the implementation of refcount_t can prevent
overflows and detect possible use-after-free.
First convert the refcount field to refcount_t in mlx5/driver.h.
Then convert the uses to refcount_() APIs.
Chuhong Yuan (3):
mlx5: Use refcount_t for refcount
net/mlx5: Use refcount_() APIs
IB/mlx5: Use refcount_() APIs
drivers/infiniband/hw/mlx5/srq_cmd.c | 6 +++---
drivers/net/ethernet/mellanox/mlx5/core/qp.c | 6 +++---
include/linux/mlx5/driver.h | 2 +-
3 files changed, 7 insertions(+), 7 deletions(-)
--
2.20.1
On Fri, Aug 02, 2019 at 08:10:35PM +0800, Chuhong Yuan wrote:
> Reference counters are preferred to use refcount_t instead of
> atomic_t.
> This is because the implementation of refcount_t can prevent
> overflows and detect possible use-after-free.
>
> First convert the refcount field to refcount_t in mlx5/driver.h.
> Then convert the uses to refcount_() APIs.
You can't do it, because you need to ensure that driver compiles and
works between patches. By converting driver.h alone to refcount_t, you
simply broke mlx5 driver.
NAK, to be clear.
And please don't sent series of patches as standalone patches.
Thanks,
On Sun, Aug 4, 2019 at 8:48 PM Leon Romanovsky <[email protected]> wrote:
>
> On Fri, Aug 02, 2019 at 08:10:35PM +0800, Chuhong Yuan wrote:
> > Reference counters are preferred to use refcount_t instead of
> > atomic_t.
> > This is because the implementation of refcount_t can prevent
> > overflows and detect possible use-after-free.
> >
> > First convert the refcount field to refcount_t in mlx5/driver.h.
> > Then convert the uses to refcount_() APIs.
>
> You can't do it, because you need to ensure that driver compiles and
> works between patches. By converting driver.h alone to refcount_t, you
> simply broke mlx5 driver.
>
It is my fault... I am not clear how to send patches which cross
several subsystems, so I sent them in series.
Maybe I should merge these patches together?
> NAK, to be clear.
>
> And please don't sent series of patches as standalone patches.
>
Due to the reason mentioned above, I sent them seperately.
> Thanks,
On Sun, Aug 04, 2019 at 10:58:19PM +0800, Chuhong Yuan wrote:
> On Sun, Aug 4, 2019 at 8:48 PM Leon Romanovsky <[email protected]> wrote:
> >
> > On Fri, Aug 02, 2019 at 08:10:35PM +0800, Chuhong Yuan wrote:
> > > Reference counters are preferred to use refcount_t instead of
> > > atomic_t.
> > > This is because the implementation of refcount_t can prevent
> > > overflows and detect possible use-after-free.
> > >
> > > First convert the refcount field to refcount_t in mlx5/driver.h.
> > > Then convert the uses to refcount_() APIs.
> >
> > You can't do it, because you need to ensure that driver compiles and
> > works between patches. By converting driver.h alone to refcount_t, you
> > simply broke mlx5 driver.
> >
>
> It is my fault... I am not clear how to send patches which cross
> several subsystems, so I sent them in series.
> Maybe I should merge these patches together?
In case of conversion patches, yes, you need to perform such change
in one shot.
>
>
> > NAK, to be clear.
> >
> > And please don't sent series of patches as standalone patches.
> >
>
> Due to the reason mentioned above, I sent them seperately.
Create patch, run ./scripts/get_maintainer.pl on it and send according
to generated output. You are not doing kernel core changes and there is
no need to worry about cross subsystem complexity as long as you will
put relevant maintainers in TO: field.
Thanks
>
> > Thanks,