net: sparx5: Add multicast support
Subsystems (3)
cover net: sparx5: Add multicast support
Add multicast support to Sparx5.
--
I apologies, I accidentally forgot to add the netdev mailing list the first time I sent this. Here it comes to netdev as well.
Casper Andersson (2):
net: sparx5: Add arbiter for managing PGID table
net: sparx5: Add mdb handlers
.../net/ethernet/microchip/sparx5/Makefile | 2 +- .../microchip/sparx5/sparx5_mactable.c | 33 ++++-- .../ethernet/microchip/sparx5/sparx5_main.c | 3 + .../ethernet/microchip/sparx5/sparx5_main.h | 23 ++++ .../ethernet/microchip/sparx5/sparx5_pgid.c | 60 ++++++++++ .../microchip/sparx5/sparx5_switchdev.c | 111 ++++++++++++++++++ 6 files changed, 221 insertions(+), 11 deletions(-) create mode 100644 drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c
Signature · 1 line
2.30.2
PATCH 1/2 net: sparx5: Add arbiter for managing PGID table
The PGID (Port Group ID) table holds port masks for different purposes. The first 72 are reserved for port destination masks, flood masks, and CPU forwarding. The rest are shared between multicast, link aggregation, and virtualization profiles. The GLAG area is reserved to not be used by anything else, since it is a subset of the MCAST area.
The arbiter keeps track of which entries are in
use. You can ask for a free ID or give back one
you are done using.
--- .../net/ethernet/microchip/sparx5/Makefile | 2 +- .../ethernet/microchip/sparx5/sparx5_main.c | 3 + .../ethernet/microchip/sparx5/sparx5_main.h | 21 +++++++ .../ethernet/microchip/sparx5/sparx5_pgid.c | 60 +++++++++++++++++++ 4 files changed, 85 insertions(+), 1 deletion(-) create mode 100644 drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c
drivers/net/ethernet/microchip/sparx5/Makefile +1 -1
@@ -8,4 +8,4 @@ obj-$(CONFIG_SPARX5_SWITCH) += sparx5-switch.o sparx5-switch-objs := sparx5_main.o sparx5_packet.o \ sparx5_netdev.o sparx5_phylink.o sparx5_port.o sparx5_mactable.o sparx5_vlan.o \ sparx5_switchdev.o sparx5_calendar.o sparx5_ethtool.o sparx5_fdma.o \ - sparx5_ptp.o + sparx5_ptp.o sparx5_pgid.o
drivers/net/ethernet/microchip/sparx5/sparx5_main.c +3 -0
@@ -626,6 +626,9 @@ static int sparx5_start(struct sparx5 *sparx5)/* Init MAC table, ageing */ sparx5_mact_init (sparx5); +/* Init PGID table arbitrator */ +sparx5_pgid_init (sparx5); +/* Setup VLANs */ sparx5_vlan_init (sparx5);
drivers/net/ethernet/microchip/sparx5/sparx5_main.h +21 -0
@@ -66,6 +66,12 @@ enum sparx5_vlan_port_type {#define PGID_BCAST (PGID_BASE + 6)#define PGID_CPU (PGID_BASE + 7) +#define PGID_TABLE_SIZE 3290 + +#define PGID_MCAST_START 65 +#define PGID_GLAG_START 833 +#define PGID_GLAG_END 1088 +#define IFH_LEN 9/* 36 bytes */ #define NULL_VID 0#define SPX5_MACT_PULL_DELAY (2 * HZ)
@@ -271,6 +277,8 @@ struct sparx5 {struct mutex ptp_lock;/* lock for ptp interface state */ u16 ptp_skbs;int ptp_irq; +/* PGID allocation map */ +u8 pgid_map[PGID_TABLE_SIZE ]; };/* sparx5_switchdev.c */
@@ -359,6 +367,19 @@ void sparx5_ptp_txtstamp_release(struct sparx5_port *port, struct sk_buff *skb);irqreturn_t sparx5_ptp_irq_handler (int irq,void *args); +/* sparx5_pgid.c */ +enum sparx5_pgid_type { +SPX5_PGID_FREE , +SPX5_PGID_RESERVED , +SPX5_PGID_MULTICAST , +SPX5_PGID_GLAG +}; + +void sparx5_pgid_init (struct sparx5 *spx5); +int sparx5_pgid_alloc_glag (struct sparx5 *spx5,u16 *idx); +int sparx5_pgid_alloc_mcast (struct sparx5 *spx5,u16 *idx); +int sparx5_pgid_free (struct sparx5 *spx5,u16 idx); +/* Clock period in picoseconds */ static inline u32 sparx5_clk_period (enum sparx5_core_clockfreq cclock) {
drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c new file +60 -0
60 lines not on this page · raw message
Signature · 1 line
2.30.2
Steen Hegelund
maintainerRe: [PATCH net-next 1/2] net: sparx5: Add arbiter for managing PGID table
Hi Casper,
Quoted 55 lines · Casper Andersson
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
The PGID (Port Group ID) table holds port masks for different purposes. The first 72 are reserved for port destination masks, flood masks, and CPU forwarding. The rest are shared between multicast, link aggregation, and virtualization profiles. The GLAG area is reserved to not be used by anything else, since it is a subset of the MCAST area.
The arbiter keeps track of which entries are in use. You can ask for a free ID or give back one you are done using.
Signed-off-by Casper Andersson <[email protected]>---
.../net/ethernet/microchip/sparx5/Makefile | 2 +-.../ethernet/microchip/sparx5/sparx5_main.c | 3 + .../ethernet/microchip/sparx5/sparx5_main.h | 21 +++++++ .../ethernet/microchip/sparx5/sparx5_pgid.c | 60 +++++++++++++++++++ 4 files changed, 85 insertions(+), 1 deletion(-)create mode 100644
drivers/net/ethernet/microchip/sparx5/sparx5_pgid.cdrivers/net/ethernet/microchip/sparx5/Makefile +1
-1@@ -8,4 +8,4 @@ obj-$(CONFIG_SPARX5_SWITCH) += sparx5-switch.o sparx5-switch-objs := sparx5_main.o sparx5_packet.o \ sparx5_netdev.o sparx5_phylink.o sparx5_port.o sparx5_mactable.o sparx5_vlan.o \ sparx5_switchdev.o sparx5_calendar.o sparx5_ethtool.o sparx5_fdma.o \ - sparx5_ptp.o + sparx5_ptp.o sparx5_pgid.odrivers/net/ethernet/microchip/sparx5/sparx5_main.c +3
-0@@ -626,6 +626,9 @@ static int sparx5_start(struct sparx5 *sparx5) /* Init MAC table, ageing */ sparx5_mact_init(sparx5); + /* Init PGID table arbitrator */ + sparx5_pgid_init(sparx5); + /* Setup VLANs */ sparx5_vlan_init(sparx5);drivers/net/ethernet/microchip/sparx5/sparx5_main.h +0
-0
drivers/net/ethernet/microchip/sparx5/sparx5_main.h enum sparx5_vlan_port_type { · line 71
Quoted 6 lines
drivers/net/ethernet/microchip/sparx5/sparx5_main.h +3
-0@@ -66,6 +66,12 @@ enum sparx5_vlan_port_type { #define PGID_BCAST (PGID_BASE + 6) #define PGID_CPU (PGID_BASE + 7) +#define PGID_TABLE_SIZE 3290 + +#define PGID_MCAST_START 65
This overlaps with PGID_UC_FLOOD above. You should drop this. Please see this description:
https://microchip-ung.github.io/sparx-5_reginfo/reginfo_sparx-5.html?select=ana_ac,pgid
drivers/net/ethernet/microchip/sparx5/sparx5_main.h enum sparx5_vlan_port_type { · line 73
Quoted 3 lines
drivers/net/ethernet/microchip/sparx5/sparx5_main.h +2
-0+#define PGID_GLAG_START 833 +#define PGID_GLAG_END 1088
You do not appear to put the GLAG feature into use so you should remove these for now.
Quoted 60 lines
drivers/net/ethernet/microchip/sparx5/sparx5_main.h +1
-0+ #define IFH_LEN 9 /* 36 bytes */ #define NULL_VID 0 #define SPX5_MACT_PULL_DELAY (2 * HZ)@@ -271,6 +277,8 @@ struct sparx5 {struct mutex
ptp_lock; /* lock for ptp interface state */ u16ptp_skbs;
intptp_irq;
+ /* PGID allocation map */
+ u8 pgid_map[PGID_TABLE_SIZE];
};/*
sparx5_switchdev.c*/drivers/net/ethernet/microchip/sparx5/sparx5_main.h +13
-0@@ -359,6 +367,19 @@ void sparx5_ptp_txtstamp_release(struct sparx5_port *port, struct sk_buff *skb); irqreturn_t sparx5_ptp_irq_handler(int irq, void *args); +/* sparx5_pgid.c */ +enum sparx5_pgid_type { + SPX5_PGID_FREE, + SPX5_PGID_RESERVED, + SPX5_PGID_MULTICAST, + SPX5_PGID_GLAG +}; + +void sparx5_pgid_init (struct sparx5 *spx5); +int sparx5_pgid_alloc_glag (struct sparx5 *spx5,u16 *idx); +int sparx5_pgid_alloc_mcast (struct sparx5 *spx5,u16 *idx); +int sparx5_pgid_free (struct sparx5 *spx5,u16 idx); + /* Clock period in picoseconds */ static inline u32 sparx5_clk_period(enum sparx5_core_clockfreq cclock) {drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c new file +20
-0@@ -0,0 +1,60 @@ +// SPDX-License-Identifier: GPL-2.0+ +#include "sparx5_main.h" + +void sparx5_pgid_init (struct sparx5 *spx5) +{ + int i; + + for (i = 0; i < PGID_TABLE_SIZE; i++) + spx5->pgid_map[i] = SPX5_PGID_FREE; + + /* Reserved for unicast, flood control, broadcast, and CPU. + * These cannot be freed. + */ + for (i = 0; i <= PGID_CPU; i++) + spx5->pgid_map[i] = SPX5_PGID_RESERVED; +} + +int sparx5_pgid_alloc_glag (struct sparx5 *spx5,u16 *idx) +{ + int i;
drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c line 30
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c new file +10
-0+ + for (i = PGID_GLAG_START; i <= PGID_GLAG_END; i++) + if (spx5->pgid_map[i] == SPX5_PGID_FREE) { + spx5->pgid_map[i] = SPX5_PGID_GLAG; + *idx = i; + return 0; + } + + return -EBUSY; +}
You do not appear to put the GLAG feature into use so you should remove this function for now.
Quoted 33 lines
drivers/net/ethernet/microchip/sparx5/sparx5_pgid.c new file +30
-1+ +int sparx5_pgid_alloc_mcast (struct sparx5 *spx5,u16 *idx) +{ + int i; + + for (i = PGID_MCAST_START; i < PGID_TABLE_SIZE; i++) { + if (i == PGID_GLAG_START) + i = PGID_GLAG_END + 1; + + if (spx5->pgid_map[i] == SPX5_PGID_FREE) { + spx5->pgid_map[i] = SPX5_PGID_MULTICAST; + *idx = i; + return 0; + } + } + + return -EBUSY; +} + +int sparx5_pgid_free (struct sparx5 *spx5,u16 idx) +{ + if (idx <= PGID_CPU || idx >= PGID_TABLE_SIZE) + return -EINVAL; + + if (spx5->pgid_map[idx] == SPX5_PGID_FREE) + return -EINVAL; + + spx5->pgid_map[idx] = SPX5_PGID_FREE; + return 0; +} --2.30.2
PATCH 2/2 net: sparx5: Add mdb handlers
In reply to Casper AnderssonAdds mdb handlers. Uses the PGID arbiter to
find a free entry in the PGID table for the
multicast group port mask.
--- .../microchip/sparx5/sparx5_mactable.c | 33 ++++-- .../ethernet/microchip/sparx5/sparx5_main.h | 2 + .../microchip/sparx5/sparx5_switchdev.c | 111 ++++++++++++++++++ 3 files changed, 136 insertions(+), 10 deletions(-)
drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c +23 -10
@@ -186,11 +186,11 @@ bool sparx5_mact_getnext(struct sparx5 *sparx5,return ret ==0 ; } -static int sparx5_mact_lookup (struct sparx5 *sparx5, -const unsigned char mac[ETH_ALEN ], -u16 vid) +bool sparx5_mact_find (struct sparx5 *sparx5, +const unsigned char mac[ETH_ALEN ],u16 vid,u32 *pcfg2) {int ret; +u32 cfg2;mutex_lock (&sparx5->lock);
@@ -202,16 +202,29 @@ static int sparx5_mact_lookup(struct sparx5 *sparx5, sparx5, LRN_COMMON_ACCESS_CTRL); ret =sparx5_mact_wait_for_completion (sparx5); -if (ret) - goto out; - - ret =LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_VLD_GET - (spx5_rd (sparx5,LRN_MAC_ACCESS_CFG_2 )); +if (ret ==0 ) { + cfg2 =spx5_rd (sparx5,LRN_MAC_ACCESS_CFG_2 ); +if (LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_VLD_GET (cfg2)) + *pcfg2 = cfg2; +else + ret = -ENOENT ; + } -out:mutex_unlock (&sparx5->lock); -return ret; +return ret ==0 ; +} + +static int sparx5_mact_lookup (struct sparx5 *sparx5, +const unsigned char mac[ETH_ALEN ], +u16 vid) +{ +u32 pcfg2; + +if (sparx5_mact_find (sparx5, mac, vid, &pcfg2)) +return 1 ; + +return 0 ; } int sparx5_mact_forget(struct sparx5 *sparx5,
drivers/net/ethernet/microchip/sparx5/sparx5_main.h +2 -0
@@ -310,6 +310,8 @@ int sparx5_mact_learn(struct sparx5 *sparx5, int port, const unsigned char mac[ETH_ALEN], u16 vid);bool sparx5_mact_getnext (struct sparx5 *sparx5,unsigned char mac[ETH_ALEN ],u16 *vid,u32 *pcfg2); +bool sparx5_mact_find (struct sparx5 *sparx5, +const unsigned char mac[ETH_ALEN ],u16 vid,u32 *pcfg2);int sparx5_mact_forget (struct sparx5 *sparx5,const unsigned char mac[ETH_ALEN ],u16 vid); int sparx5_add_mact_entry(struct sparx5 *sparx5,
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +111 -0
129 lines not on this page · raw message
Signature · 1 line
2.30.2
Steen Hegelund
maintainerRe: [PATCH net-next 2/2] net: sparx5: Add mdb handlers
Quoted 58 lines · Casper Andersson
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Adds mdb handlers. Uses the PGID arbiter to
find a free entry in the PGID table for the
multicast group port mask.Signed-off-by Casper Andersson <[email protected]>---
.../microchip/sparx5/sparx5_mactable.c | 33 ++++-- .../ethernet/microchip/sparx5/sparx5_main.h | 2 + .../microchip/sparx5/sparx5_switchdev.c | 111 ++++++++++++++++++3 files changed, 136 insertions(+), 10 deletions(-)drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c +14
-10@@ -186,11 +186,11 @@ bool sparx5_mact_getnext(struct sparx5 *sparx5, return ret == 0; } -static int sparx5_mact_lookup (struct sparx5 *sparx5, - const unsigned char mac[ETH_ALEN], - u16 vid) +bool sparx5_mact_find (struct sparx5 *sparx5, + const unsigned char mac[ETH_ALEN], u16 vid, u32 *pcfg2) { int ret; + u32 cfg2; mutex_lock(&sparx5->lock);@@ -202,16 +202,29 @@ static int sparx5_mact_lookup(struct sparx5 *sparx5, sparx5, LRN_COMMON_ACCESS_CTRL); ret = sparx5_mact_wait_for_completion(sparx5); - if (ret) - goto out; - - ret = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_VLD_GET - (spx5_rd(sparx5, LRN_MAC_ACCESS_CFG_2)); + if (ret == 0) { + cfg2 = spx5_rd(sparx5, LRN_MAC_ACCESS_CFG_2); + if (LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_VLD_GET(cfg2)) + *pcfg2 = cfg2; + else + ret = -ENOENT; + } -out: mutex_unlock(&sparx5->lock); - return ret; + return ret == 0; +} + +static int sparx5_mact_lookup(struct sparx5 *sparx5,
drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c static int sparx5_mact_lookup(struct sparx5 *sparx5, · line 228
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c +9
-0+ const unsigned char mac[ETH_ALEN], + u16 vid) +{ + u32 pcfg2; + + if (sparx5_mact_find(sparx5, mac, vid, &pcfg2)) + return 1; + + return 0; }
I suggest to drop this and only use your new function (or at least let it return a real error code like -ENOENT in case of an error).
Quoted 40 lines
drivers/net/ethernet/microchip/sparx5/sparx5_mactable.c +0
-0int sparx5_mact_forget(struct sparx5 *sparx5,drivers/net/ethernet/microchip/sparx5/sparx5_main.h +2
-0@@ -310,6 +310,8 @@ int sparx5_mact_learn(struct sparx5 *sparx5, int port, const unsigned char mac[ETH_ALEN], u16 vid); bool sparx5_mact_getnext(struct sparx5 *sparx5, unsigned char mac[ETH_ALEN], u16 *vid, u32 *pcfg2); +bool sparx5_mact_find (struct sparx5 *sparx5, + const unsigned char mac[ETH_ALEN], u16 vid, u32 *pcfg2); int sparx5_mact_forget(struct sparx5 *sparx5, const unsigned char mac[ETH_ALEN], u16 vid); int sparx5_add_mact_entry(struct sparx5 *sparx5,drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +15
-0@@ -366,6 +366,109 @@ static int sparx5_handle_port_vlan_add(struct net_device *dev, v->flags & BRIDGE_VLAN_INFO_UNTAGGED); } +static int sparx5_handle_port_mdb_add (struct net_device *dev, + struct notifier_block *nb, + const struct switchdev_obj_port_mdb *v) +{ + struct sparx5_port *port = netdev_priv(dev); + struct sparx5 *spx5 = port->sparx5; + u16 pgid_idx, vid; + u32 mact_entry; + int res, err; + + /* When VLAN unaware the vlan value is not parsed and we receive vid 0. + * Fall back to bridge vid 1. + */ + if (!br_vlan_enabled(spx5->hw_bridge_dev)) + vid = 1;
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c static int sparx5_handle_port_vlan_add(struct net_device *dev, · line 393
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +10
-0+ else + vid = v->vid; + + res = sparx5_mact_find(spx5, v->addr, vid, &mact_entry); + + if (res) { + pgid_idx = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_ADDR_GET(mact_entry); + + /* MC_IDX has an offset of 65 in the PGID table. */ + pgid_idx += PGID_MCAST_START;
This will overlap some of the first ports with the flood masks according to:
https://microchip-ung.github.io/sparx-5_reginfo/reginfo_sparx-5.html?select=ana_ac,pgid
You should use the custom area (PGID_BASE + 8 and onwards) for this new feature.
Quoted 58 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +58
-0+ sparx5_pgid_update_mask(port, pgid_idx, true); + } else { + err = sparx5_pgid_alloc_mcast(spx5, &pgid_idx); + if (err) { + netdev_warn(dev, "multicast pgid table full\n"); + return err; + } + sparx5_pgid_update_mask(port, pgid_idx, true); + err = sparx5_mact_learn(spx5, pgid_idx, v->addr, vid); + if (err) { + netdev_warn(dev, "could not learn mac address %pM\n", v->addr); + sparx5_pgid_update_mask(port, pgid_idx, false); + return err; + } + } + + return 0; +} + +static int sparx5_mdb_del_entry (struct net_device *dev, + struct sparx5 *spx5, + const unsigned char mac[ETH_ALEN], + const u16 vid, + u16 pgid_idx) +{ + int err; + + err = sparx5_mact_forget(spx5, mac, vid); + if (err) { + netdev_warn(dev, "could not forget mac address %pM", mac); + return err; + } + err = sparx5_pgid_free(spx5, pgid_idx); + if (err) { + netdev_err(dev, "attempted to free already freed pgid\n"); + return err; + } + return 0; +} + +static int sparx5_handle_port_mdb_del (struct net_device *dev, + struct notifier_block *nb, + const struct switchdev_obj_port_mdb *v) +{ + struct sparx5_port *port = netdev_priv(dev); + struct sparx5 *spx5 = port->sparx5; + u16 pgid_idx, vid; + u32 mact_entry, res, pgid_entry[3]; + int err; + + if (!br_vlan_enabled(spx5->hw_bridge_dev)) + vid = 1; + else + vid = v->vid; + + res = sparx5_mact_find(spx5, v->addr, vid, &mact_entry); + + if (res) {
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c static int sparx5_handle_port_vlan_add(struct net_device *dev, · line 461
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +10
-0+ pgid_idx = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_ADDR_GET(mact_entry); + + /* MC_IDX has an offset of 65 in the PGID table. */ + pgid_idx += PGID_MCAST_START; + sparx5_pgid_update_mask(port, pgid_idx, false); + + pgid_entry[0] = spx5_rd(spx5, ANA_AC_PGID_CFG(pgid_idx)); + pgid_entry[1] = spx5_rd(spx5, ANA_AC_PGID_CFG1(pgid_idx)); + pgid_entry[2] = spx5_rd(spx5, ANA_AC_PGID_CFG2(pgid_idx)); + if (pgid_entry[0] == 0 && pgid_entry[1] == 0 && pgid_entry[2] == 0) {
Looks like you could use a function that gets the pgid port mask (the inverse of the sparx5_pgid_update_mask() function)
Quoted 32 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +14
-0+ /* No ports are in MC group. Remove entry */ + err = sparx5_mdb_del_entry(dev, spx5, v->addr, vid, pgid_idx); + if (err) + return err; + } + } + + return 0; +} + static int sparx5_handle_port_obj_add(struct net_device *dev, struct notifier_block *nb, struct switchdev_notifier_port_obj_info *info)@@ -378,6 +481,10 @@ static int sparx5_handle_port_obj_add(struct net_device *dev, err = sparx5_handle_port_vlan_add(dev, nb, SWITCHDEV_OBJ_PORT_VLAN(obj)); break; + case SWITCHDEV_OBJ_ID_PORT_MDB: + err = sparx5_handle_port_mdb_add(dev, nb, + SWITCHDEV_OBJ_PORT_MDB(obj)); + break;default:
err =-EOPNOTSUPP;
break;drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +4
-0@@ -426,6 +533,10 @@ static int sparx5_handle_port_obj_del(struct net_device *dev, err = sparx5_handle_port_vlan_del(dev, nb, SWITCHDEV_OBJ_PORT_VLAN(obj)->vid); break; + case SWITCHDEV_OBJ_ID_PORT_MDB: + err = sparx5_handle_port_mdb_del(dev, nb, + SWITCHDEV_OBJ_PORT_MDB(obj)); + break;
Quoted 6 lines
default:
err =-EOPNOTSUPP;
break;
--
2.30.2
Best Regards
Steen
Re: [PATCH net-next 2/2] net: sparx5: Add mdb handlers
So this was already merged, but I have some comments on the feedback for the follow up patch.
Quoted 28 lines
Quoted 15 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +15
-0+static int sparx5_handle_port_mdb_add (struct net_device *dev, + struct notifier_block *nb, + const struct switchdev_obj_port_mdb *v) +{ + struct sparx5_port *port = netdev_priv(dev); + struct sparx5 *spx5 = port->sparx5; + u16 pgid_idx, vid; + u32 mact_entry; + int res, err; + + /* When VLAN unaware the vlan value is not parsed and we receive vid 0. + * Fall back to bridge vid 1. + */ + if (!br_vlan_enabled(spx5->hw_bridge_dev)) + vid = 1;drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c static int sparx5_handle_port_vlan_add(struct net_device *dev, · line 393
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +10
-0+ else + vid = v->vid; + + res = sparx5_mact_find(spx5, v->addr, vid, &mact_entry); + + if (res) { + pgid_idx = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_ADDR_GET(mact_entry); + + /* MC_IDX has an offset of 65 in the PGID table. */ + pgid_idx += PGID_MCAST_START;This will overlap some of the first ports with the flood masks according to:
Quoted 3 lines
https://microchip-ung.github.io/sparx-5_reginfo/reginfo_sparx-5.html?select=ana_ac,pgid
You should use the custom area (
PGID_BASE+ 8 and onwards) for this new feature.
I'm aware of the overlap, hence why the PGID table has those fields marked as reserved. But your datasheet says that the multicast index has an offset of 65 (ie. MC_IDX = 0 is at PGID = 65). This is already taken into account in the mact_learn function. I could set the allocation to start at PGID_BASE + 8, but the offset still needs to be 65, right?
BR
Casper
Steen Hegelund
maintainerRe: [PATCH net-next 2/2] net: sparx5: Add mdb handlers
Hi Casper
Quoted 35 lines · Casper Andersson
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
So this was already merged, but I have some comments on the feedback for the follow up patch.
Quoted 30 lines
Quoted 15 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +15
-0+static int sparx5_handle_port_mdb_add (struct net_device *dev, + struct notifier_block *nb, + const struct switchdev_obj_port_mdb *v) +{ + struct sparx5_port *port = netdev_priv(dev); + struct sparx5 *spx5 = port->sparx5; + u16 pgid_idx, vid; + u32 mact_entry; + int res, err; + + /* When VLAN unaware the vlan value is not parsed and we receive vid 0. + * Fall back to bridge vid 1. + */ + if (!br_vlan_enabled(spx5->hw_bridge_dev)) + vid = 1;drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c static int sparx5_handle_port_vlan_add(struct net_device *dev, · line 393
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +10
-0+ else + vid = v->vid; + + res = sparx5_mact_find(spx5, v->addr, vid, &mact_entry); + + if (res) { + pgid_idx = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_ADDR_GET(mact_entry); + + /* MC_IDX has an offset of 65 in the PGID table. */ + pgid_idx += PGID_MCAST_START;This will overlap some of the first ports with the flood masks according to:
https://microchip-ung.github.io/sparx-5_reginfo/reginfo_sparx-5.html?select=ana_ac,pgid
Quoted 8 lines
Quoted 1 line
You should use the custom area (
PGID_BASE+ 8 and onwards) for this new feature.I'm aware of the overlap, hence why the PGID table has those fields marked as reserved. But your datasheet says that the multicast index has an offset of 65 (ie.
MC_IDX= 0 is at PGID = 65). This is already taken into account in themact_learnfunction. I could set the allocation to start atPGID_BASE+ 8, but the offset still needs to be 65, right?
As I understand the PGID table functionality, you will need to start your custom table at PGID_BASE
+ 8 as the bitmasks at offset 65 to 70 are used as flood masks, so their purpose are already defined.
BR
Steen
Quoted 3 lines
BR
Casper
Re: [PATCH net-next 2/2] net: sparx5: Add mdb handlers
Quoted 47 lines · Steen Hegelund
Hi Casper
Quoted 35 lines · Casper Andersson
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
So this was already merged, but I have some comments on the feedback for the follow up patch.
Quoted 30 lines
Quoted 15 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +15
-0+static int sparx5_handle_port_mdb_add (struct net_device *dev, + struct notifier_block *nb, + const struct switchdev_obj_port_mdb *v) +{ + struct sparx5_port *port = netdev_priv(dev); + struct sparx5 *spx5 = port->sparx5; + u16 pgid_idx, vid; + u32 mact_entry; + int res, err; + + /* When VLAN unaware the vlan value is not parsed and we receive vid 0. + * Fall back to bridge vid 1. + */ + if (!br_vlan_enabled(spx5->hw_bridge_dev)) + vid = 1;drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c static int sparx5_handle_port_vlan_add(struct net_device *dev, · line 393
Quoted 10 lines
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +10
-0+ else + vid = v->vid; + + res = sparx5_mact_find(spx5, v->addr, vid, &mact_entry); + + if (res) { + pgid_idx = LRN_MAC_ACCESS_CFG_2_MAC_ENTRY_ADDR_GET(mact_entry); + + /* MC_IDX has an offset of 65 in the PGID table. */ + pgid_idx += PGID_MCAST_START;This will overlap some of the first ports with the flood masks according to:
https://microchip-ung.github.io/sparx-5_reginfo/reginfo_sparx-5.html?select=ana_ac,pgid
Quoted 8 lines
Quoted 1 line
You should use the custom area (
PGID_BASE+ 8 and onwards) for this new feature.I'm aware of the overlap, hence why the PGID table has those fields marked as reserved. But your datasheet says that the multicast index has an offset of 65 (ie.
MC_IDX= 0 is at PGID = 65). This is already taken into account in themact_learnfunction. I could set the allocation to start atPGID_BASE+ 8, but the offset still needs to be 65, right?
Quoted 3 lines
As I understand the PGID table functionality, you will need to start your custom table at
PGID_BASE
+ 8 as the bitmasks at offset 65 to 70 are used as flood masks, so their purpose are already defined.
It is fine to start the allocation of multicast entries at PGID_BASE + 8, but the multicast index (MC_IDX) in the mactable (that points at the PGID table, to get the port mask) assumes that MC_IDX = 0 is at PGID = 65. Not sure if it's appropriate to reference the datasheet here on Netdev, but on figure 4-48 (PGID Layout) this is shown.
I tried setting the offset to base + 8 but it does not seem to find the right mask. Because if I say I put the mask on MC_IDX 0, and then place it on base + 8 (73), then it will not find that mask because the hardware will look for the mask at 65.
Even though the offset is 65, the PGID allocation will make sure that it never allocates anything below 73. Meaning that the lowest possible MC_IDX it can allocate will be 8.
I'm also a bit confused as to why the multicast offset is 65, but there are a lot of overlapping areas in the PGID table and, e.g., GLAG has another offset where its index 0 is at PGID = 833 (as can be seen in figure 4-48).
BR
Casper
Quoted 7 lines
BR
SteenQuoted 3 lines
BR
Casper
This series was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Quoted 8 lines · On Mon, 21 Mar 2022 11:14:44 +0100 you wrote:
Add multicast support to Sparx5.
--
I apologies, I accidentally forgot to add the netdev mailing list the first time I sent this. Here it comes to netdev as well.[...]
Here is the summary with links:
- [net-next,1/2] net: sparx5: Add arbiter for managing PGID table
https://git.kernel.org/netdev/net-next/c/af9b45d08eb4
- [net-next,2/2] net: sparx5: Add mdb handlers
https://git.kernel.org/netdev/net-next/c/3bacfccdcb2d
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Steen Hegelund
maintainerHi Dave,
That must be a mistake.
I have just added some comments to the series, and I have not had more than a few hours to look at it, so I do not think that you have given this enough time to mature.
BR
Steen
Quoted 29 lines · On Mon, 2022-03-21 at 13:30 +0000, [email protected] wrote:
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Hello:
This series was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:
Quoted 8 lines · On Mon, 21 Mar 2022 11:14:44 +0100 you wrote:
Add multicast support to Sparx5.
--
I apologies, I accidentally forgot to add the netdev mailing list the first time I sent this. Here it comes to netdev as well.[...]
Here is the summary with links:
- [net-next,1/2] net: sparx5: Add arbiter for managing PGID table
https://git.kernel.org/netdev/net-next/c/af9b45d08eb4
- [net-next,2/2] net: sparx5: Add mdb handlers
https://git.kernel.org/netdev/net-next/c/3bacfccdcb2dYou are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
Jakub Kicinski
maintainerQuoted 3 lines · On Mon, 21 Mar 2022 14:33:34 +0100 Steen Hegelund wrote:
I have just added some comments to the series, and I have not had more than a few hours to look at it, so I do not think that you have given this enough time to mature.
Sorry about that. Is it possible to fix the issues in a follow up or should we revert the patches?
Steen Hegelund
maintainerHi Jacub,
I guess that Casper can fix the issues in a follow up. What do you say Casper?
BR
Steen
Quoted 9 lines · Jakub Kicinski
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Quoted 3 lines · On Mon, 21 Mar 2022 14:33:34 +0100 Steen Hegelund wrote:
I have just added some comments to the series, and I have not had more than a few hours to look at it, so I do not think that you have given this enough time to mature.
Sorry about that. Is it possible to fix the issues in a follow up or should we revert the patches?
I will do a follow up. Thanks for the feedback, Steen.
BR
Casper
Quoted 19 lines · Steen Hegelund
Hi Jacub,
I guess that Casper can fix the issues in a follow up. What do you say Casper?
BR
SteenQuoted 9 lines · Jakub Kicinski
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Quoted 3 lines · On Mon, 21 Mar 2022 14:33:34 +0100 Steen Hegelund wrote:
I have just added some comments to the series, and I have not had more than a few hours to look at it, so I do not think that you have given this enough time to mature.
Sorry about that. Is it possible to fix the issues in a follow up or should we revert the patches?