| Message ID | 20210111174316.3515736-1-olteanv@gmail.com |
|---|---|
| Headers | show |
| Series | Configuring congestion watermarks on ocelot switch using devlink-sb | expand |
On Mon, 11 Jan 2021 19:43:06 +0200 Vladimir Oltean wrote: > In some applications, it is important to create resource reservations in > the Ethernet switches, to prevent background traffic, or deliberate > attacks, from inducing denial of service into the high-priority traffic. > > These patches give the user some knobs to turn. The ocelot switches > support per-port and per-port-tc reservations, on ingress and on egress. > The resources that are monitored are packet buffers (in cells of 60 > bytes each) and frame references. > > The frames that exceed the reservations can optionally consume from > sharing watermarks which are not per-port but global across the switch. > There are 10 sharing watermarks, 8 of them are per traffic class and 2 > are per drop priority. > > I am configuring the hardware using the best of my knowledge, and mostly > through trial and error. Same goes for devlink-sb integration. Feedback > is welcome. This no longer applies.
On Mon, 11 Jan 2021 19:43:14 +0200 Vladimir Oltean wrote: > +struct ocelot_devlink_private { > + struct ocelot *ocelot; > +}; I don't think you ever explained to me why you don't put struct ocelot in the priv. - ocelot = devm_kzalloc(&pdev->dev, sizeof(*ocelot), GFP_KERNEL); - if (!ocelot) + devlink = devlink_alloc(&ocelot_devlink_ops, sizeof(*ocelot)); + if (!devlink) return -ENOMEM; + ocelot = devlink_priv(ocelot->devlink);
On Wed, Jan 13, 2021 at 07:25:52PM -0800, Jakub Kicinski wrote: > On Mon, 11 Jan 2021 19:43:06 +0200 Vladimir Oltean wrote: > > In some applications, it is important to create resource reservations in > > the Ethernet switches, to prevent background traffic, or deliberate > > attacks, from inducing denial of service into the high-priority traffic. > > > > These patches give the user some knobs to turn. The ocelot switches > > support per-port and per-port-tc reservations, on ingress and on egress. > > The resources that are monitored are packet buffers (in cells of 60 > > bytes each) and frame references. > > > > The frames that exceed the reservations can optionally consume from > > sharing watermarks which are not per-port but global across the switch. > > There are 10 sharing watermarks, 8 of them are per traffic class and 2 > > are per drop priority. > > > > I am configuring the hardware using the best of my knowledge, and mostly > > through trial and error. Same goes for devlink-sb integration. Feedback > > is welcome. > > This no longer applies. I was not expecting you to apply it, giving the feedback.
On Wed, Jan 13, 2021 at 07:30:33PM -0800, Jakub Kicinski wrote: > On Mon, 11 Jan 2021 19:43:14 +0200 Vladimir Oltean wrote: > > +struct ocelot_devlink_private { > > + struct ocelot *ocelot; > > +}; > > I don't think you ever explained to me why you don't put struct ocelot > in the priv. > > - ocelot = devm_kzalloc(&pdev->dev, sizeof(*ocelot), GFP_KERNEL); > - if (!ocelot) > + devlink = devlink_alloc(&ocelot_devlink_ops, sizeof(*ocelot)); > + if (!devlink) > return -ENOMEM; > + ocelot = devlink_priv(ocelot->devlink); Because that's not going to be all? The error path handling and teardown all need to change, because I no longer use device-managed allocation, and I wanted to avoid that.
On Thu, 14 Jan 2021 12:27:43 +0200 Vladimir Oltean wrote: > On Wed, Jan 13, 2021 at 07:30:33PM -0800, Jakub Kicinski wrote: > > On Mon, 11 Jan 2021 19:43:14 +0200 Vladimir Oltean wrote: > > > +struct ocelot_devlink_private { > > > + struct ocelot *ocelot; > > > +}; > > > > I don't think you ever explained to me why you don't put struct ocelot > > in the priv. > > > > - ocelot = devm_kzalloc(&pdev->dev, sizeof(*ocelot), GFP_KERNEL); > > - if (!ocelot) > > + devlink = devlink_alloc(&ocelot_devlink_ops, sizeof(*ocelot)); > > + if (!devlink) > > return -ENOMEM; > > + ocelot = devlink_priv(ocelot->devlink); > > Because that's not going to be all? The error path handling and teardown > all need to change, because I no longer use device-managed allocation, > and I wanted to avoid that. Come on, is it really hard enough to warrant us exchanging multiple emails? Having driver structure in devlink priv is the standard way of handling this, there's value in uniformity.
On Thu, Jan 14, 2021 at 09:19:43AM -0800, Jakub Kicinski wrote: > On Thu, 14 Jan 2021 12:27:43 +0200 Vladimir Oltean wrote: > > On Wed, Jan 13, 2021 at 07:30:33PM -0800, Jakub Kicinski wrote: > > > On Mon, 11 Jan 2021 19:43:14 +0200 Vladimir Oltean wrote: > > > > +struct ocelot_devlink_private { > > > > + struct ocelot *ocelot; > > > > +}; > > > > > > I don't think you ever explained to me why you don't put struct ocelot > > > in the priv. > > > > > > - ocelot = devm_kzalloc(&pdev->dev, sizeof(*ocelot), GFP_KERNEL); > > > - if (!ocelot) > > > + devlink = devlink_alloc(&ocelot_devlink_ops, sizeof(*ocelot)); > > > + if (!devlink) > > > return -ENOMEM; > > > + ocelot = devlink_priv(ocelot->devlink); > > > > Because that's not going to be all? The error path handling and teardown > > all need to change, because I no longer use device-managed allocation, > > and I wanted to avoid that. > > Come on, is it really hard enough to warrant us exchanging multiple > emails? Having driver structure in devlink priv is the standard way > of handling this, there's value in uniformity. I did as you requested in v5 anyway. It does not save me of having to keep a devlink pointer in struct ocelot though, due to the fact that the layout with struct devlink being a container of struct ocelot is not common between the DSA felix driver and the switchdev ocelot driver. So much for uniformity.
From: Vladimir Oltean <vladimir.oltean@nxp.com> In some applications, it is important to create resource reservations in the Ethernet switches, to prevent background traffic, or deliberate attacks, from inducing denial of service into the high-priority traffic. These patches give the user some knobs to turn. The ocelot switches support per-port and per-port-tc reservations, on ingress and on egress. The resources that are monitored are packet buffers (in cells of 60 bytes each) and frame references. The frames that exceed the reservations can optionally consume from sharing watermarks which are not per-port but global across the switch. There are 10 sharing watermarks, 8 of them are per traffic class and 2 are per drop priority. I am configuring the hardware using the best of my knowledge, and mostly through trial and error. Same goes for devlink-sb integration. Feedback is welcome. Vladimir Oltean (10): net: mscc: ocelot: auto-detect packet buffer size and number of frame references net: mscc: ocelot: add ops for decoding watermark threshold and occupancy net: dsa: add ops for devlink-sb net: dsa: felix: reindent struct dsa_switch_ops net: dsa: felix: perform teardown in reverse order of setup net: mscc: ocelot: export NUM_TC constant from felix to common switch lib net: mscc: ocelot: delete unused ocelot_set_cpu_port prototype net: mscc: ocelot: register devlink ports net: mscc: ocelot: initialize watermarks to sane defaults net: mscc: ocelot: configure watermarks using devlink-sb drivers/net/dsa/ocelot/felix.c | 209 +++-- drivers/net/dsa/ocelot/felix.h | 2 - drivers/net/dsa/ocelot/felix_vsc9959.c | 23 +- drivers/net/dsa/ocelot/seville_vsc9953.c | 20 +- drivers/net/ethernet/mscc/Makefile | 3 +- drivers/net/ethernet/mscc/ocelot.c | 18 +- drivers/net/ethernet/mscc/ocelot.h | 9 +- drivers/net/ethernet/mscc/ocelot_devlink.c | 885 +++++++++++++++++++++ drivers/net/ethernet/mscc/ocelot_net.c | 253 +++++- drivers/net/ethernet/mscc/ocelot_vsc7514.c | 108 ++- include/net/dsa.h | 34 + include/soc/mscc/ocelot.h | 55 +- include/soc/mscc/ocelot_qsys.h | 7 +- net/dsa/dsa2.c | 159 +++- 14 files changed, 1684 insertions(+), 101 deletions(-) create mode 100644 drivers/net/ethernet/mscc/ocelot_devlink.c