On Mon, Jan 08, 2018 at 11:06:28PM -0800, Saeed Mahameed wrote:
>
>
> On 01/08/2018 10:13 PM, Andy Gospodarek wrote:
> > From: Andy Gospodarek <go...@broadcom.com>
> >
> > Change all appropriate mlx5_am* and MLX5_AM* references to net_dim and
> > NET_DIM, respectively, in code that handles dynamic interrupt
> > moderation. Also change all references from 'am' to 'dim' when used as
> > local variables and add generic profile references.
> >
> > Signed-off-by: Andy Gospodarek <go...@broadcom.com>
> > Acked-by: Tal Gilboa <ta...@mellanox.com>
> > Acked-by: Saeed Mahameed <sae...@mellanox.com>
> > ---
> > drivers/net/ethernet/mellanox/mlx5/core/en.h | 9 +-
> > drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 14 +-
> > .../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 6 +-
> > drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 40 ++-
> > drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 8 +-
> > drivers/net/ethernet/mellanox/mlx5/core/net_dim.c | 286
> > ++++++++++-----------
> > drivers/net/ethernet/mellanox/mlx5/core/net_dim.h | 63 ++---
> > 7 files changed, 225 insertions(+), 201 deletions(-)
> >
>
> [...]
>
> > #define IS_SIGNIFICANT_DIFF(val, ref) \
> > (((100 * abs((val) - (ref))) / (ref)) > 10) /* more than 10% difference
> > */
> > -static int mlx5e_am_stats_compare(struct mlx5e_rx_am_stats *curr,
> > - struct mlx5e_rx_am_stats *prev)
> > +static int net_dim_stats_compare(struct net_dim_stats *curr,
> > + struct net_dim_stats *prev)
> > {
> > if (!prev->bpms)
> > - return curr->bpms ? MLX5E_AM_STATS_BETTER :
> > - MLX5E_AM_STATS_SAME;
> > + return curr->bpms ? NET_DIM_STATS_BETTER :
> > + NET_DIM_STATS_SAME;
> > if (IS_SIGNIFICANT_DIFF(curr->bpms, prev->bpms))
> > - return (curr->bpms > prev->bpms) ? MLX5E_AM_STATS_BETTER :
> > - MLX5E_AM_STATS_WORSE;
> > + return (curr->bpms > prev->bpms) ? NET_DIM_STATS_BETTER :
> > + NET_DIM_STATS_WORSE;
>
> Hey Andy,
>
> I am currently reviewing a patch internally that fixes a bug in this area,
> prev->ppms can be 0 and could cause IS_SIGNIFICANT_DIFF ouch !
> same goes for prev->eppm, for some reason we had a broken assumption that if
> ppms is 0 for some reason then the bpms is 0 and the above condition will
> cover us.
>
> Anyway the patch will go to net, which means when this series gets accepted
> then net-next will fail to merge with net and we need to manually push the
> fix to the new DIM library.
>
> But for now I don't think anything is required for this series other than
> bringing this division by 0 issue and the future merge conflict to your
> attention.
>
Thanks for bringing that to everyone's attention. I agree there is
probably not much that should be done at this point -- hopefully the
merge should go pretty smoothly, since net_dim.h is seen as a rename
from en_rx_am.c.
> > if (IS_SIGNIFICANT_DIFF(curr->ppms, prev->ppms))
> > - return (curr->ppms > prev->ppms) ? MLX5E_AM_STATS_BETTER :
> > - MLX5E_AM_STATS_WORSE;
> > + return (curr->ppms > prev->ppms) ? NET_DIM_STATS_BETTER :
> > + NET_DIM_STATS_WORSE;
> > if (IS_SIGNIFICANT_DIFF(curr->epms, prev->epms))
> > - return (curr->epms < prev->epms) ? MLX5E_AM_STATS_BETTER :
> > - MLX5E_AM_STATS_WORSE;
> > + return (curr->epms < prev->epms) ? NET_DIM_STATS_BETTER :
> > + NET_DIM_STATS_WORSE;
> > - return MLX5E_AM_STATS_SAME;
> > + return NET_DIM_STATS_SAME;
> > }
