2021-08-18 06:11:53

by Xiaoliang Yang

[permalink] [raw]
Subject: [RFC v2 net-next 3/8] net: mscc: ocelot: set vcap IS2 chain to goto PSFP chain

Some chips in the ocelot series such as VSC9959 support Per-Stream
Filtering and Policing(PSFP), which is processing after VCAP blocks.
We set this block on chain 30000 and set vcap IS2 chain to goto PSFP
chain if hardware support.

Signed-off-by: Xiaoliang Yang <[email protected]>
---
drivers/net/ethernet/mscc/ocelot_flower.c | 17 ++++++++++++++---
1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/mscc/ocelot_flower.c b/drivers/net/ethernet/mscc/ocelot_flower.c
index 8b843d3c9189..ce812194e44c 100644
--- a/drivers/net/ethernet/mscc/ocelot_flower.c
+++ b/drivers/net/ethernet/mscc/ocelot_flower.c
@@ -20,6 +20,9 @@
(1 * VCAP_BLOCK + (lookup) * VCAP_LOOKUP)
#define VCAP_IS2_CHAIN(lookup, pag) \
(2 * VCAP_BLOCK + (lookup) * VCAP_LOOKUP + (pag))
+/* PSFP chain and block ID */
+#define PSFP_BLOCK_ID OCELOT_NUM_VCAP_BLOCKS
+#define OCELOT_PSFP_CHAIN (3 * VCAP_BLOCK)

static int ocelot_chain_to_block(int chain, bool ingress)
{
@@ -46,6 +49,9 @@ static int ocelot_chain_to_block(int chain, bool ingress)
if (chain == VCAP_IS2_CHAIN(lookup, pag))
return VCAP_IS2;

+ if (chain == OCELOT_PSFP_CHAIN)
+ return PSFP_BLOCK_ID;
+
return -EOPNOTSUPP;
}

@@ -84,7 +90,8 @@ static bool ocelot_is_goto_target_valid(int goto_target, int chain,
goto_target == VCAP_IS1_CHAIN(1) ||
goto_target == VCAP_IS1_CHAIN(2) ||
goto_target == VCAP_IS2_CHAIN(0, 0) ||
- goto_target == VCAP_IS2_CHAIN(1, 0));
+ goto_target == VCAP_IS2_CHAIN(1, 0) ||
+ goto_target == OCELOT_PSFP_CHAIN);

if (chain == VCAP_IS1_CHAIN(0))
return (goto_target == VCAP_IS1_CHAIN(1));
@@ -111,7 +118,11 @@ static bool ocelot_is_goto_target_valid(int goto_target, int chain,
if (chain == VCAP_IS2_CHAIN(0, pag))
return (goto_target == VCAP_IS2_CHAIN(1, pag));

- /* VCAP IS2 lookup 1 cannot jump anywhere */
+ /* VCAP IS2 lookup 1 can goto to PSFP block if hardware support */
+ for (pag = 0; pag < VCAP_IS2_NUM_PAG; pag++)
+ if (chain == VCAP_IS2_CHAIN(1, pag))
+ return (goto_target == OCELOT_PSFP_CHAIN);
+
return false;
}

@@ -353,7 +364,7 @@ static int ocelot_flower_parse_action(struct ocelot *ocelot, int port,

if (filter->goto_target == -1) {
if ((filter->block_id == VCAP_IS2 && filter->lookup == 1) ||
- chain == 0) {
+ chain == 0 || filter->block_id == PSFP_BLOCK_ID) {
allow_missing_goto_target = true;
} else {
NL_SET_ERR_MSG_MOD(extack, "Missing GOTO action");
--
2.17.1


2021-08-18 14:55:46

by Vladimir Oltean

[permalink] [raw]
Subject: Re: [RFC v2 net-next 3/8] net: mscc: ocelot: set vcap IS2 chain to goto PSFP chain

On Wed, Aug 18, 2021 at 02:19:17PM +0800, Xiaoliang Yang wrote:
> @@ -353,7 +364,7 @@ static int ocelot_flower_parse_action(struct ocelot *ocelot, int port,
>
> if (filter->goto_target == -1) {
> if ((filter->block_id == VCAP_IS2 && filter->lookup == 1) ||
> - chain == 0) {
> + chain == 0 || filter->block_id == PSFP_BLOCK_ID) {
> allow_missing_goto_target = true;

I would like to not allow missing "goto" targets for new filter chains.

Due to legacy support we must keep support for VCAP IS2 on chain 0, but
ever since we added the ability to offload multiple chains corresponding
to multiple hardware blocks, we should really use that precise chain ID,
and chain 0 should just goto the first used chain in the pipeline.

Makes sense?

> } else {
> NL_SET_ERR_MSG_MOD(extack, "Missing GOTO action");
> --
> 2.17.1
>

2021-08-19 09:51:06

by Xiaoliang Yang

[permalink] [raw]
Subject: RE: [RFC v2 net-next 3/8] net: mscc: ocelot: set vcap IS2 chain to goto PSFP chain



On Wed, Aug 18, 2021 at 22:55:23PM +0800, Vladimir Oltean wrote:
> > @@ -353,7 +364,7 @@ static int ocelot_flower_parse_action(struct
> > ocelot *ocelot, int port,
> >
> > if (filter->goto_target == -1) {
> > if ((filter->block_id == VCAP_IS2 && filter->lookup == 1) ||
> > - chain == 0) {
> > + chain == 0 || filter->block_id == PSFP_BLOCK_ID) {
> > allow_missing_goto_target = true;
>
> I would like to not allow missing "goto" targets for new filter chains.
>
> Due to legacy support we must keep support for VCAP IS2 on chain 0, but ever
> since we added the ability to offload multiple chains corresponding to multiple
> hardware blocks, we should really use that precise chain ID, and chain 0 should
> just goto the first used chain in the pipeline.
>
> Makes sense?
>
Because the PSFP chain is the last chain now, I allow missing "goto" targets. Once we add a new chain like FRER, I will delete this allow missing "goto" for PSFP block chain.

Thanks,
Xiaoliang