[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANn89i+Vs63kwJZXXHTvhnNgDLPsPmXzJ99pSD5GimXd5Qt0EA@mail.gmail.com>
Date: Fri, 17 Apr 2020 05:09:23 -0700
From: Eric Dumazet <edumazet@...gle.com>
To: Saeed Mahameed <saeedm@...lanox.com>
Cc: "davem@...emloft.net" <davem@...emloft.net>,
"willemb@...gle.com" <willemb@...gle.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"eric.dumazet@...il.com" <eric.dumazet@...il.com>,
Tariq Toukan <tariqt@...lanox.com>
Subject: Re: [PATCH net-next] net/mlx4_en: avoid indirect call in TX completion
On Thu, Apr 16, 2020 at 8:55 PM Saeed Mahameed <saeedm@...lanox.com> wrote:
>
> On Wed, 2020-04-15 at 09:46 -0700, Eric Dumazet wrote:
> > Commit 9ecc2d86171a ("net/mlx4_en: add xdp forwarding and data write
> > support")
> > brought another indirect call in fast path.
> >
> > Use INDIRECT_CALL_2() helper to avoid the cost of the indirect call
> > when/if CONFIG_RETPOLINE=y
> >
> > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > Cc: Tariq Toukan <tariqt@...lanox.com>
> > Cc: Willem de Bruijn <willemb@...gle.com>
> > ---
>
> Hi Eric, I believe net-next is still closed.
>
> But FWIW,
>
> Reviewed-by: Saeed Mahameed <saeedm@...lanox.com>
>
Well, this can be pushed to net then, since this is a trivial patch
that helps performance.
With this COVID-19 thing, we need more capacity from the serving fleet
(Youtube and all..)
distributed all over the world and using a high number of CX-3 NIC.
Thanks.
>
> > drivers/net/ethernet/mellanox/mlx4/en_tx.c | 14 +++++++++++++-
> > 1 file changed, 13 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/mellanox/mlx4/en_tx.c
> > b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
> > index
> > 4d5ca302c067126b8627cb4809485b45c10e2460..a30edb436f4af11526e04c09623
> > 840288ebe4a29 100644
> > --- a/drivers/net/ethernet/mellanox/mlx4/en_tx.c
> > +++ b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
> > @@ -43,6 +43,7 @@
> > #include <linux/ip.h>
> > #include <linux/ipv6.h>
> > #include <linux/moduleparam.h>
> > +#include <linux/indirect_call_wrapper.h>
> >
> > #include "mlx4_en.h"
> >
> > @@ -261,6 +262,10 @@ static void mlx4_en_stamp_wqe(struct
> > mlx4_en_priv *priv,
> > }
> > }
> >
> > +INDIRECT_CALLABLE_DECLARE(u32 mlx4_en_free_tx_desc(struct
> > mlx4_en_priv *priv,
> > + struct
> > mlx4_en_tx_ring *ring,
> > + int index, u64
> > timestamp,
> > + int napi_mode));
> >
> > u32 mlx4_en_free_tx_desc(struct mlx4_en_priv *priv,
> > struct mlx4_en_tx_ring *ring,
> > @@ -329,6 +334,11 @@ u32 mlx4_en_free_tx_desc(struct mlx4_en_priv
> > *priv,
> > return tx_info->nr_txbb;
> > }
> >
> > +INDIRECT_CALLABLE_DECLARE(u32 mlx4_en_recycle_tx_desc(struct
> > mlx4_en_priv *priv,
> > + struct
> > mlx4_en_tx_ring *ring,
> > + int index, u64
> > timestamp,
> > + int napi_mode));
> > +
> > u32 mlx4_en_recycle_tx_desc(struct mlx4_en_priv *priv,
> > struct mlx4_en_tx_ring *ring,
> > int index, u64 timestamp,
> > @@ -449,7 +459,9 @@ bool mlx4_en_process_tx_cq(struct net_device
> > *dev,
> > timestamp = mlx4_en_get_cqe_ts(cqe);
> >
> > /* free next descriptor */
> > - last_nr_txbb = ring->free_tx_desc(
> > + last_nr_txbb = INDIRECT_CALL_2(ring-
> > >free_tx_desc,
> > + mlx4_en_free_tx_
> > desc,
> > + mlx4_en_recycle_
> > tx_desc,
> > priv, ring, ring_index,
> > timestamp, napi_budget);
> >
Powered by blists - more mailing lists