| Message ID | 20220108181633.420433-1-dario.binacchi@amarulasolutions.com (mailing list archive) |
|---|---|
| State | RFC |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [RFC] can: flexcan: add ethtool support to get rx/tx ring parameters | expand |
| Context | Check | Description |
|---|---|---|
| netdev/tree_selection | success | Series ignored based on subject |
On 08.01.2022 19:16:33, Dario Binacchi wrote: > Adds ethtool support to get the number of message buffers configured for > reception/transmission, which may also depends on runtime configurations > such as the 'rx-rtr' flag state. > > Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com> > > --- > > drivers/net/can/flexcan/flexcan-ethtool.c | 17 +++++++++++++++++ > 1 file changed, 17 insertions(+) > > diff --git a/drivers/net/can/flexcan/flexcan-ethtool.c b/drivers/net/can/flexcan/flexcan-ethtool.c > index 5bb45653e1ac..d119bca584f6 100644 > --- a/drivers/net/can/flexcan/flexcan-ethtool.c > +++ b/drivers/net/can/flexcan/flexcan-ethtool.c > @@ -80,7 +80,24 @@ static int flexcan_set_priv_flags(struct net_device *ndev, u32 priv_flags) > return 0; > } > > +static void flexcan_get_ringparam(struct net_device *ndev, > + struct ethtool_ringparam *ring) This doesn't compile on net-next/master, as the prototype of the get_ringparam callback changed, fixed this while applying. > +{ > + struct flexcan_priv *priv = netdev_priv(ndev); > + > + ring->rx_max_pending = priv->mb_count; > + ring->tx_max_pending = priv->mb_count; > + > + if (priv->devtype_data.quirks & FLEXCAN_QUIRK_USE_RX_MAILBOX) > + ring->rx_pending = __sw_hweight64(priv->rx_mask); I've replaced the hamming weight calculation by the simpler: | ring->rx_pending = priv->offload.mb_last - | priv->offload.mb_first + 1; > + else > + ring->rx_pending = 6; > + > + ring->tx_pending = __sw_hweight64(priv->tx_mask); ...and here I added a hardcoded "1", as the driver currently only support on TX buffer. > +} > + > static const struct ethtool_ops flexcan_ethtool_ops = { > + .get_ringparam = flexcan_get_ringparam, > .get_sset_count = flexcan_get_sset_count, > .get_strings = flexcan_get_strings, > .get_priv_flags = flexcan_get_priv_flags, BTW: If you're looking for more TX performance, this can be done by using more than one TX buffer. It's also possible to configure the number of RX and TX buffers via ethtool during runtime. I'm currently preparing a patch set for the mcp251xfd to implement this. regards, Marc
On Sat, Jan 8, 2022 at 9:16 PM Marc Kleine-Budde <mkl@pengutronix.de> wrote: > > On 08.01.2022 19:16:33, Dario Binacchi wrote: > > Adds ethtool support to get the number of message buffers configured for > > reception/transmission, which may also depends on runtime configurations > > such as the 'rx-rtr' flag state. > > > > Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com> > > > > --- > > > > drivers/net/can/flexcan/flexcan-ethtool.c | 17 +++++++++++++++++ > > 1 file changed, 17 insertions(+) > > > > diff --git a/drivers/net/can/flexcan/flexcan-ethtool.c b/drivers/net/can/flexcan/flexcan-ethtool.c > > index 5bb45653e1ac..d119bca584f6 100644 > > --- a/drivers/net/can/flexcan/flexcan-ethtool.c > > +++ b/drivers/net/can/flexcan/flexcan-ethtool.c > > @@ -80,7 +80,24 @@ static int flexcan_set_priv_flags(struct net_device *ndev, u32 priv_flags) > > return 0; > > } > > > > +static void flexcan_get_ringparam(struct net_device *ndev, > > + struct ethtool_ringparam *ring) > > This doesn't compile on net-next/master, as the prototype of the > get_ringparam callback changed, fixed this while applying. > > > +{ > > + struct flexcan_priv *priv = netdev_priv(ndev); > > + > > + ring->rx_max_pending = priv->mb_count; > > + ring->tx_max_pending = priv->mb_count; > > + > > + if (priv->devtype_data.quirks & FLEXCAN_QUIRK_USE_RX_MAILBOX) > > + ring->rx_pending = __sw_hweight64(priv->rx_mask); > > I've replaced the hamming weight calculation by the simpler: > > | ring->rx_pending = priv->offload.mb_last - > | priv->offload.mb_first + 1; > > > + else > > + ring->rx_pending = 6; > > + > > + ring->tx_pending = __sw_hweight64(priv->tx_mask); > > ...and here I added a hardcoded "1", as the driver currently only > support on TX buffer. > > > +} > > + > > static const struct ethtool_ops flexcan_ethtool_ops = { > > + .get_ringparam = flexcan_get_ringparam, > > .get_sset_count = flexcan_get_sset_count, > > .get_strings = flexcan_get_strings, > > .get_priv_flags = flexcan_get_priv_flags, > > BTW: If you're looking for more TX performance, this can be done by > using more than one TX buffer. I didn't expect only one message buffer to be used for transmission thanks and regards, Dario > It's also possible to configure the > number of RX and TX buffers via ethtool during runtime. I'm currently > preparing a patch set for the mcp251xfd to implement this. > > regards, > Marc > > -- > Pengutronix e.K. | Marc Kleine-Budde | > Embedded Linux | https://www.pengutronix.de | > Vertretung West/Dortmund | Phone: +49-231-2826-924 | > Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
On 09.01.2022 12:22:54, Dario Binacchi wrote: > > > static const struct ethtool_ops flexcan_ethtool_ops = { > > > + .get_ringparam = flexcan_get_ringparam, > > > .get_sset_count = flexcan_get_sset_count, > > > .get_strings = flexcan_get_strings, > > > .get_priv_flags = flexcan_get_priv_flags, > > > > BTW: If you're looking for more TX performance, this can be done by > > using more than one TX buffer. > > I didn't expect only one message buffer to be used for transmission It was easier to implement, but now we've sorted it out how to implement multiple TX buffers race free and lock-less. Have a look at the mcp251xfd driver. regards, Marc
diff --git a/drivers/net/can/flexcan/flexcan-ethtool.c b/drivers/net/can/flexcan/flexcan-ethtool.c index 5bb45653e1ac..d119bca584f6 100644 --- a/drivers/net/can/flexcan/flexcan-ethtool.c +++ b/drivers/net/can/flexcan/flexcan-ethtool.c @@ -80,7 +80,24 @@ static int flexcan_set_priv_flags(struct net_device *ndev, u32 priv_flags) return 0; } +static void flexcan_get_ringparam(struct net_device *ndev, + struct ethtool_ringparam *ring) +{ + struct flexcan_priv *priv = netdev_priv(ndev); + + ring->rx_max_pending = priv->mb_count; + ring->tx_max_pending = priv->mb_count; + + if (priv->devtype_data.quirks & FLEXCAN_QUIRK_USE_RX_MAILBOX) + ring->rx_pending = __sw_hweight64(priv->rx_mask); + else + ring->rx_pending = 6; + + ring->tx_pending = __sw_hweight64(priv->tx_mask); +} + static const struct ethtool_ops flexcan_ethtool_ops = { + .get_ringparam = flexcan_get_ringparam, .get_sset_count = flexcan_get_sset_count, .get_strings = flexcan_get_strings, .get_priv_flags = flexcan_get_priv_flags,
Adds ethtool support to get the number of message buffers configured for reception/transmission, which may also depends on runtime configurations such as the 'rx-rtr' flag state. Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com> --- drivers/net/can/flexcan/flexcan-ethtool.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)