Re: [PATCH rdma-next 1/4] RDMA/bnxt_re: Simplify bnxt_re_init_depth() callers and implementation

[Leon Romanovsky]

On Mon, Mar 23, 2026 at 11:46:43AM +0530, Selvin Xavier wrote:
> On Wed, Mar 18, 2026 at 3:39 PM Leon Romanovsky <leon@xxxxxxxxxx> wrote:
> >
> > From: Leon Romanovsky <leonro@xxxxxxxxxx>
> >
> > All callers of bnxt_re_init_depth() compute the minimum between its return
> > value and another internal variable, often mixing variable types in the
> > process. Clean this up by making the logic simpler and more readable.
> >
> > Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxx>
> > ---
> >  drivers/infiniband/hw/bnxt_re/ib_verbs.c | 81 ++++++++++++++------------------
> >  drivers/infiniband/hw/bnxt_re/ib_verbs.h |  9 ++--
> >  2 files changed, 42 insertions(+), 48 deletions(-)

<...>

> > -static inline u32 bnxt_re_init_depth(u32 ent, struct bnxt_re_ucontext *uctx)
> > +static inline u32 bnxt_re_init_depth(u32 ent, u32 max,
> > +                                    struct bnxt_re_ucontext *uctx)
> >  {
> > -       return uctx ? (uctx->cmask & BNXT_RE_UCNTX_CAP_POW2_DISABLED) ?
> > -               ent : roundup_pow_of_two(ent) : ent;
> > +       if (uctx && !(uctx->cmask & BNXT_RE_UCNTX_CAP_POW2_DISABLED))
> > +               return min(roundup_pow_of_two(ent), max);
> Looks like the min setting is missing in the else case. shouldn't we add that?

I did not add it because all callers of bnxt_re_init_depth() already check  
that ent is less than max. However, roundup_pow_of_two() requires a  
min() guard because it can overflow the prior range check.

Thanks

> 
> > +
> > +       return ent;
> >  }
> >
> >  static inline bool bnxt_re_is_var_size_supported(struct bnxt_re_dev *rdev,
> >
> > --
> > 2.53.0
> >