Add Multicast Filtering support for VLAN interface
MD Danish Anwar
reviewerv3 cover Add Multicast Filtering support for VLAN interface
This series adds Multicast filtering support for VLAN interfaces in dual EMAC and HSR offload mode for ICSSG driver.
Patch 1/3 - Adds support for VLAN in dual EMAC mode Patch 2/3 - Adds MC filtering support for VLAN in dual EMAC mode Patch 3/3 - Adds MC filtering support for VLAN in HSR mode
Changes from v2 to v3:
*) Rebased on latest net-next and re-spun the series as net-next is now open. *) No functional change.
Changes from v1 to v2:
*) Changed netdev_err to netdev_dbg in emac_ndo_vlan_rx_del_vid() in patch 1/3 as suggested by Michal Swiatkowski <[email protected]> *) Dropped patch [1] from previous version as the patch created issue [2]. Will send out a separate patch to set HSR=m in arch/arm64/configs/defconfig. Once the defconfig patch gets merged, I will add depends on HSR in Kconfig for TI_ICSSG_PRUETH as suggested by Larysa Zaremba <[email protected]>
[1] https://lore.kernel.org/all/[email protected]/
[2] https://lore.kernel.org/all/[email protected]/#t
v1 https://lore.kernel.org/all/[email protected]/
v2 https://lore.kernel.org/all/[email protected]/
MD Danish Anwar (3):
net: ti: icssg-prueth: Add VLAN support in EMAC mode
net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC
mode
net: ti: icssg-prueth: Add Support for Multicast filtering with VLAN
in HSR mode
drivers/net/ethernet/ti/icssg/icssg_prueth.c | 175 ++++++++++++++----- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 8 + include/linux/if_hsr.h | 18 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- net/hsr/hsr_device.c | 13 ++ net/hsr/hsr_main.h | 9 - 7 files changed, 174 insertions(+), 59 deletions(-)
base-commit: 94c16fd4df9089931f674fb9aaec41ea20b0fd7a
Signature · 1 line
2.34.1
MD Danish Anwar
reviewerPATCH v3 1/3 net: ti: icssg-prueth: Add VLAN support in EMAC mode
Add support for vlan filtering in dual EMAC mode.
--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 29 +++++++++----------- 1 file changed, 13 insertions(+), 16 deletions(-)
drivers/net/ethernet/ti/icssg/icssg_prueth.c +13 -16
@@ -822,19 +822,18 @@ static int emac_ndo_vlan_rx_add_vid(struct net_device *ndev, {struct prueth_emac *emac =netdev_priv (ndev);struct prueth *prueth = emac->prueth; +int port_mask =BIT (emac->port_id);int untag_mask =0 ; -int port_mask; -if (prueth->is_hsr_offload_mode) { - port_mask =BIT (PRUETH_PORT_HOST ) |BIT (emac->port_id); - untag_mask =0 ; +if (prueth->is_hsr_offload_mode) + port_mask |=BIT (PRUETH_PORT_HOST ); -netdev_dbg (emac->ndev,"VID add vid:%u port_mask:%X untag_mask %X\n" , - vid, port_mask, untag_mask); +netdev_dbg (emac->ndev,"VID add vid:%u port_mask:%X untag_mask %X\n" , + vid, port_mask, untag_mask); + +icssg_vtbl_modify (emac, vid, port_mask, untag_mask, true); +icssg_set_pvid (emac->prueth, vid, emac->port_id); -icssg_vtbl_modify (emac, vid, port_mask, untag_mask, true); -icssg_set_pvid (emac->prueth, vid, emac->port_id); - }return 0 ; }
@@ -843,18 +842,16 @@ static int emac_ndo_vlan_rx_del_vid(struct net_device *ndev, {struct prueth_emac *emac =netdev_priv (ndev);struct prueth *prueth = emac->prueth; +int port_mask =BIT (emac->port_id);int untag_mask =0 ; -int port_mask; -if (prueth->is_hsr_offload_mode) { +if (prueth->is_hsr_offload_mode) port_mask =BIT (PRUETH_PORT_HOST ); - untag_mask =0 ; -netdev_dbg (emac->ndev,"VID del vid:%u port_mask:%X untag_mask %X\n" , - vid, port_mask, untag_mask); +netdev_dbg (emac->ndev,"VID del vid:%u port_mask:%X untag_mask %X\n" , + vid, port_mask, untag_mask); +icssg_vtbl_modify (emac, vid, port_mask, untag_mask, false); -icssg_vtbl_modify (emac, vid, port_mask, untag_mask, false); - }return 0 ; }
Signature · 1 line
2.34.1
Re: [PATCH net-next v3 1/3] net: ti: icssg-prueth: Add VLAN support in EMAC mode
Quoted 1 line · MD Danish Anwar
Add support for vlan filtering in dual EMAC mode.
vlan/VLAN
Quoted 2 lines
Signed-off-by MD Danish Anwar <[email protected]>
MD Danish Anwar
reviewerPATCH v3 2/3 net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
In reply to MD Danish AnwarAdd multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses vlan_for_each() API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using __hw_addr_sync_multiple() API.
The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports __hw_addr_sync_multiple() in order to use it from the ICSSG driver.
--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)
drivers/net/ethernet/ti/icssg/icssg_prueth.c +53 -14
@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;
@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args) +{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev); + +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast); + +return 0 ; +} +/** * emac_ndo_open - EMAC device open * @ndev: network adapter device
@@ -772,12 +805,17 @@ static void emac_ndo_set_rx_mode_work(struct work_struct *work)return ; } -if (emac->prueth->is_hsr_offload_mode) +if (emac->prueth->is_hsr_offload_mode) {__dev_mc_sync (ndev, icssg_prueth_hsr_add_mcast, icssg_prueth_hsr_del_mcast); -else + }else {__dev_mc_sync (ndev, icssg_prueth_add_mcast, icssg_prueth_del_mcast); +if (rtnl_trylock ()) { +vlan_for_each (ndev, icssg_update_vlan_mcast, emac); +rtnl_unlock (); + } + } }/**
@@ -828,6 +866,7 @@ static int emac_ndo_vlan_rx_add_vid(struct net_device *ndev,if (prueth->is_hsr_offload_mode) port_mask |=BIT (PRUETH_PORT_HOST ); +__hw_addr_init (&emac->vlan_mcast_list[vid]);netdev_dbg (emac->ndev,"VID add vid:%u port_mask:%X untag_mask %X\n" , vid, port_mask, untag_mask);
drivers/net/ethernet/ti/icssg/icssg_prueth.h +6 -0
@@ -83,6 +83,10 @@#define ICSS_CMD_ADD_FILTER 0x7#define ICSS_CMD_ADD_MAC 0x8 +/* VLAN Filtering Related MACROs */ +#define PRUETH_DFLT_VLAN_MAC 0 +#define MAX_VLAN_ID 256 +/* In switch mode there are 3 real ports i.e. 3 mac addrs. * however Linux sees only the host side port. The other 2 ports * are the switch ports.
@@ -201,6 +205,8 @@ struct prueth_emac {/* RX IRQ Coalescing Related */ struct hrtimer rx_hrtimer;unsigned long rx_pace_timeout_ns; + +struct netdev_hw_addr_list vlan_mcast_list[MAX_VLAN_ID ]; };/**
include/linux/netdevice.h +3 -0
@@ -4689,6 +4689,9 @@ int devm_register_netdev(struct device *dev, struct net_device *ndev);/* General hardware address lists handling functions */ int __hw_addr_sync (struct netdev_hw_addr_list *to_list,struct netdev_hw_addr_list *from_list,int addr_len); +int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, +struct netdev_hw_addr_list *from_list, +int addr_len);void __hw_addr_unsync (struct netdev_hw_addr_list *to_list,struct netdev_hw_addr_list *from_list,int addr_len); int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
net/core/dev_addr_lists.c +4 -3
@@ -242,9 +242,9 @@ static void __hw_addr_unsync_one(struct netdev_hw_addr_list *to_list,__hw_addr_del_entry (from_list, ha, false, false); } -static int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, -struct netdev_hw_addr_list *from_list, -int addr_len) +int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, +struct netdev_hw_addr_list *from_list, +int addr_len) {int err =0 ;struct netdev_hw_addr *ha, *tmp;
@@ -260,6 +260,7 @@ static int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, }return err; } +EXPORT_SYMBOL (__hw_addr_sync_multiple);/* This function only works where there is a strict 1-1 relationship * between source and destination of they synch. If you ever need to
Signature · 1 line
2.34.1
Paolo Abeni
maintainerRe: [PATCH net-next v3 2/3] net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
Quoted 90 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses
vlan_for_each()API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports
__hw_addr_sync_multiple()in order to use it from the ICSSG driver.Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)drivers/net/ethernet/ti/icssg/icssg_prueth.c +28
-12@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args)
drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 559
Quoted 10 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +10
-0+{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev);
At this point, isn't emac->vlan_mcast_list[vid] == vdev->mc?
drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 562
Quoted 3 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +3
-0+ +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast);
If so, can this function be reduced to just:
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);
?
Quoted 18 lines
net/core/dev_addr_lists.c +3
-3@@ -242,9 +242,9 @@ static void __hw_addr_unsync_one(struct netdev_hw_addr_list *to_list,__hw_addr_del_entry (from_list, ha, false, false); } -static int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, -struct netdev_hw_addr_list *from_list, -int addr_len) +int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, +struct netdev_hw_addr_list *from_list, +int addr_len) {int err =0 ;struct netdev_hw_addr *ha, *tmp;
net/core/dev_addr_lists.c static int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, · line 263
Quoted 4 lines
net/core/dev_addr_lists.c +1
-0@@ -260,6 +260,7 @@ static int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, }return err; } +EXPORT_SYMBOL (__hw_addr_sync_multiple);
I'm asking because this additional export looks suspect. How other drivers cope with similar situation?
Thanks!
Paolo
MD Danish Anwar
reviewerRe: [PATCH net-next v3 2/3] net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
Hi Paolo,
Quoted 110 lines · Paolo Abeni
Quoted 90 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses
vlan_for_each()API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports
__hw_addr_sync_multiple()in order to use it from the ICSSG driver.Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)drivers/net/ethernet/ti/icssg/icssg_prueth.c +28
-12@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args)drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 559
Quoted 10 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +10
-0+{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev);At this point, isn't emac->vlan_mcast_list[vid] ==
vdev->mc?drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 562
Quoted 3 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +3
-0+ +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast);If so, can this function be reduced to just:
Quoted 4 lines
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);?
I don't know but for some reason __dev_mc_sync() doesn't work here. My initial approach was to use __dev_mc_sync(vdev, sync, unsync) however it didn't work.
When I use __dev_mc_sync() and print the vlan_id in function icssg_prueth_add_mcast(). It always prints vlan_id as 0 implying __dev_mc_sync from here never gets called. Whereas when using __hw_addr_sync_dev() I see the appropriate vlan_id in icssg_prueth_add_mcast()
Anyways, Even if I use __dev_mc_sync(), we will still need the export. I am exporting __hw_addr_sync_multiple() not __hw_addr_sync_dev(). The API being used by me __hw_addr_sync_dev() is already exported.
Quoted 17 lines
Quoted 17 lines
net/core/dev_addr_lists.c +3
-3@@ -242,9 +242,9 @@ static void __hw_addr_unsync_one(struct netdev_hw_addr_list *to_list,__hw_addr_del_entry (from_list, ha, false, false); } -static int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, -struct netdev_hw_addr_list *from_list, -int addr_len) +int __hw_addr_sync_multiple (struct netdev_hw_addr_list *to_list, +struct netdev_hw_addr_list *from_list, +int addr_len) {int err =0 ;struct netdev_hw_addr *ha, *tmp;
Quoted 9 lines
net/core/dev_addr_lists.c static int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, · line 263
Quoted 5 lines
net/core/dev_addr_lists.c +1
-0@@ -260,6 +260,7 @@ static int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, }return err; } +EXPORT_SYMBOL (__hw_addr_sync_multiple);I'm asking because this additional export looks suspect. How other drivers cope with similar situation?
To avoid exporting I will need to use dev_mc_sync_multiple() which was in fact suggested by Michal Swiatkowski in [1]. However that won't work in this case as explained by me in the reply to Michal [2]
[1]
https://lore.kernel.org/all/[email protected]/#:~:text=Only%20question%2C%20why%20dev_mc_sync_multiple%20can%27t%20be%20used%20here%3F
[2]
https://lore.kernel.org/all/[email protected]/#:~:text=nice%20to%20have.%0A%3E-,I,-don%27t%20think%20that%27s
Quoted 4 lines
Thanks!
Paolo
Signature · 2 lines
Thanks and Regards, Danish
Paolo Abeni
maintainerRe: [PATCH net-next v3 2/3] net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
Hi,
Quoted 120 lines · MD Danish Anwar
Quoted 110 lines · Paolo Abeni
Quoted 90 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses
vlan_for_each()API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports
__hw_addr_sync_multiple()in order to use it from the ICSSG driver.Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)drivers/net/ethernet/ti/icssg/icssg_prueth.c +28
-12@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args)drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 559
Quoted 10 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +10
-0+{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev);At this point, isn't emac->vlan_mcast_list[vid] ==
vdev->mc?drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 562
Quoted 3 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +3
-0+ +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast);If so, can this function be reduced to just:
Quoted 4 lines
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);?
I don't know but for some reason
__dev_mc_sync()doesn't work here. My initial approach was to use__dev_mc_sync(vdev, sync, unsync)however it didn't work.
Quoted 5 lines
When I use
__dev_mc_sync()and print thevlan_idin functionicssg_prueth_add_mcast(). It always printsvlan_idas 0 implying__dev_mc_syncfrom here never gets called. Whereas when using__hw_addr_sync_dev()I see the appropriatevlan_idinicssg_prueth_add_mcast()
It look like the above needs more investigation, right? is vlan_mcast_list[vid] different from vdev->mc? why? At very least you need to provide a clear explaination of the above.
Quoted 3 lines
Anyways, Even if I use
__dev_mc_sync(), we will still need the export. I am exporting__hw_addr_sync_multiple()not__hw_addr_sync_dev(). The API being used by me__hw_addr_sync_dev()is already exported.
I fear there is a misunderstanding. I'm suggesting dropping __hw_addr_sync_multiple() usage entirely. If that is not possible, a clear and complete explaination of the reason/root cause must be provided.
Thanks,
Paolo
MD Danish Anwar
reviewerRe: [PATCH net-next v3 2/3] net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
Hi Paolo,
Quoted 129 lines · Paolo Abeni
Hi,
Quoted 120 lines · MD Danish Anwar
Quoted 110 lines · Paolo Abeni
Quoted 90 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses
vlan_for_each()API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports
__hw_addr_sync_multiple()in order to use it from the ICSSG driver.Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)drivers/net/ethernet/ti/icssg/icssg_prueth.c +28
-12@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args)drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 559
Quoted 10 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +10
-0+{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev);At this point, isn't emac->vlan_mcast_list[vid] ==
vdev->mc?drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 562
Quoted 3 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +3
-0+ +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast);If so, can this function be reduced to just:
Quoted 4 lines
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);?
I don't know but for some reason
__dev_mc_sync()doesn't work here. My initial approach was to use__dev_mc_sync(vdev, sync, unsync)however it didn't work.Quoted 5 lines
When I use
__dev_mc_sync()and print thevlan_idin functionicssg_prueth_add_mcast(). It always printsvlan_idas 0 implying__dev_mc_syncfrom here never gets called. Whereas when using__hw_addr_sync_dev()I see the appropriatevlan_idinicssg_prueth_add_mcast()
Quoted 4 lines
It look like the above needs more investigation, right? is vlan_mcast_list[vid] different from
vdev->mc? why? At very least you need to provide a clear explaination of the above.
I did further debug on this. At this point vlan_mcast_list[vid] is actually same as vdev->mc in terms of the multicast addresses.
However the sync_cnt and refcount of the addresses in both the lists are not same. Due to which vdev->mc doesn't work here. I will explain it.
__dev_mc_sync(vdev, sync, unsync) will call __hw_addr_sync_dev(&dev->mc, dev, sync, unsync)
Now in __hw_addr_sync_dev() sync is only called when ha->sync_cnt == 0 for the given mac address. If ha->sync_cnt is non zero, sync will not get called.
When we add a multicast address to the vlan interface using below command,
ip maddr add <mac_addr> dev eth1.6
ndo_set_rx_mode() gets called for the vlan interface i.e. vlan_dev_set_rx_mode() [net/8021q/vlan_dev.c] which syncs mc list from the vdev to the real_dev of vdev by calling dev_mc_sync(real_dev, vdev)
Now dev_mc_sync() sync addresses from vdev to real_dev using __hw_addr_sync().
__hw_addr_sync() calls __hw_addr_sync_one() which after successfully syncing an address from vdev to real_dev increment the ha->sync_cnt. As a result at this point the ha->sync_cnt == 1 for the above address passed by command. After this vlan_dev_set_rx_mode() calls the set_rx_mode() of the real_dev.
Now when icssg_update_vlan_mcast() calls __dev_mc_sync(vdev, sync, unsync), it calls _hw_addr_sync_dev(&dev->mc, dev, sync, unsync) and checks the ha->sync_cnt for the given address, since sync_cnt is already 1, it doesn't consider it as an newly added address and sync / unsync callbacks are not called. [1]
list_for_each_entry_safe(ha, tmp, &list->list, list) {
if (ha->sync_cnt)
continue;Since for addresses in vlan interfaces, sync_cnt will always be set as vlan_dev_set_rx_mode() will sync the mc list of vlan interface with the real dev. This will mean that the driver implemented sync / unsync APIs will only be called for the real_dev and the real_dev won't have any information about the vlan_id which I need in my sync / unsync APIs to populate the same in the hardware maintained FDB table.
To overcome this, I am maintaining a local copy of vdev->mc named emac->vlan_mcast_list[vid]. I will sync address from vdev->mc to this. and then call __hw_addr_sync_dev() on this list as the sync_cnt for address in this list will still be zero. Which will then trigger my sync() callback on the vdev which will help me obtain the vlan_id in the sync callback. I can now populate the same in the hardware maintained FDB table.
I hope this explains why I am using
__hw_addr_sync_dev(&emac->vlan_mcast_list[vid], vdev,
icssg_prueth_add_mcast, icssg_prueth_del_mcast)
instead of
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);
Quoted 8 lines
Quoted 3 lines
Anyways, Even if I use
__dev_mc_sync(), we will still need the export. I am exporting__hw_addr_sync_multiple()not__hw_addr_sync_dev(). The API being used by me__hw_addr_sync_dev()is already exported.I fear there is a misunderstanding. I'm suggesting dropping
__hw_addr_sync_multiple()usage entirely. If that is not possible, a clear and complete explaination of the reason/root cause must be provided.
As explained above, I need to be able to sync addresses from vdev->mc to my local list. Now this can be done using two APIs.
1. __hw_addr_sync() - This is already exported and is actually the first choice. However this will fail syncing address from vdev->mc to my local copy.
__hw_addr_sync() only syncs the address if ha->sync_cnt == 0. If the sync_cnt is non zero, this will skip the address. As explained above, for mc addresses of vlan interfaces, sync_cnt will always be set as vlan_dev_set_rx_mode() will sync the mc list of vlan interface with the real dev. As a result the addresses will be skipped and __hw_addr_sync() will not serve the purpose here.
2. __hw_addr_sync_multiple() - This actually works perfectly fine here as it sync address even if sync_cnt is non zero. As a result I am using this function in my implementation.
Since this is not public, I have to export it so that the driver can call this.
So I am afraid dropping __hw_addr_sync_multiple() is not possible here. I hope the explanation above makes sense to you.
Please let me know if this is OK with you. If you have some other way through which I can make this work please let me know.
Quoted 4 lines
Thanks,
Paolo
Signature · 2 lines
Thanks and Regards, Danish
Paolo Abeni
maintainerRe: [PATCH net-next v3 2/3] net: ti: icssg-prueth: Add Multicast Filtering support for VLAN in MAC mode
Quoted 224 lines · MD Danish Anwar
Quoted 127 lines · Paolo Abeni
Quoted 120 lines · MD Danish Anwar
Quoted 110 lines · Paolo Abeni
Quoted 90 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in dual EMAC mode for ICSSG driver.
The driver uses
vlan_for_each()API to get the list of available vlans. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks and based on whether the ndev is vlan or not, driver passes appropriate vid to FDB helper functions.
This commit also exports
__hw_addr_sync_multiple()in order to use it from the ICSSG driver.Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 67 ++++++++++++++++---- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 6 ++ include/linux/netdevice.h | 3 + net/core/dev_addr_lists.c | 7 +- 4 files changed, 66 insertions(+), 17 deletions(-)drivers/net/ethernet/ti/icssg/icssg_prueth.c +28
-12@@ -472,30 +472,44 @@ const struct icss_iep_clockops prueth_iep_clockops = {static int icssg_prueth_add_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac; +int port_mask; +u8 vlan_id; - port_mask |=icssg_fdb_lookup (emac, addr,0 ); -icssg_fdb_add_del (emac, addr,0 , port_mask, true); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, true); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); + + port_mask =BIT (emac->port_id) |icssg_fdb_lookup (emac, addr, vlan_id); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, true); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, true);return 0 ; }static int icssg_prueth_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -int port_mask =BIT (emac->port_id); +struct net_device *real_dev; +struct prueth_emac *emac;int other_port_mask; +int port_mask; +u8 vlan_id; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_MAC ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + emac =netdev_priv (real_dev); - other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr,0 ); + port_mask =BIT (emac->port_id); + other_port_mask = port_mask ^icssg_fdb_lookup (emac, addr, vlan_id); -icssg_fdb_add_del (emac, addr,0 , port_mask, false); -icssg_vtbl_modify (emac,0 , port_mask, port_mask, false); +icssg_fdb_add_del (emac, addr, vlan_id, port_mask, false); +icssg_vtbl_modify (emac, vlan_id, port_mask, port_mask, false);if (other_port_mask) { -icssg_fdb_add_del (emac, addr,0 , other_port_mask, true); -icssg_vtbl_modify (emac,0 , other_port_mask, other_port_mask, true); +icssg_fdb_add_del (emac, addr, vlan_id, other_port_mask, true); +icssg_vtbl_modify (emac, vlan_id, other_port_mask, + other_port_mask, true); }return 0 ;@@ -531,6 +545,25 @@ static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } +static int icssg_update_vlan_mcast (struct net_device *vdev,int vid, +void *args)drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 559
Quoted 10 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +10
-0+{ +struct prueth_emac *emac = args; + +if (!vdev || !vid) +return 0 ; + +netif_addr_lock_bh (vdev); +__hw_addr_sync_multiple (&emac->vlan_mcast_list[vid], &vdev->mc, + vdev->addr_len); +netif_addr_unlock_bh (vdev);At this point, isn't emac->vlan_mcast_list[vid] ==
vdev->mc?drivers/net/ethernet/ti/icssg/icssg_prueth.c static int icssg_prueth_hsr_del_mcast(struct net_device *ndev, const u8 *addr) · line 562
Quoted 3 lines
drivers/net/ethernet/ti/icssg/icssg_prueth.c +3
-0+ +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, icssg_prueth_del_mcast);If so, can this function be reduced to just:
Quoted 4 lines
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);?
I don't know but for some reason
__dev_mc_sync()doesn't work here. My initial approach was to use__dev_mc_sync(vdev, sync, unsync)however it didn't work.Quoted 5 lines
When I use
__dev_mc_sync()and print thevlan_idin functionicssg_prueth_add_mcast(). It always printsvlan_idas 0 implying__dev_mc_syncfrom here never gets called. Whereas when using__hw_addr_sync_dev()I see the appropriatevlan_idinicssg_prueth_add_mcast()Quoted 4 lines
It look like the above needs more investigation, right? is vlan_mcast_list[vid] different from
vdev->mc? why? At very least you need to provide a clear explaination of the above.I did further debug on this. At this point vlan_mcast_list[vid] is actually same as
vdev->mcin terms of the multicast addresses.However the
sync_cntand refcount of the addresses in both the lists are not same. Due to whichvdev->mcdoesn't work here. I will explain it.
__dev_mc_sync(vdev, sync, unsync)will call__hw_addr_sync_dev(&dev->mc, dev, sync, unsync)Now in
__hw_addr_sync_dev()sync is only called whenha->sync_cnt== 0 for the given mac address. Ifha->sync_cntis non zero, sync will not get called.When we add a multicast address to the vlan interface using below command,
ip maddr add <
mac_addr> dev eth1.6
ndo_set_rx_mode()gets called for the vlan interface i.e.vlan_dev_set_rx_mode()[net/8021q/vlan_dev.c] which syncs mc list from the vdev to thereal_devof vdev by callingdev_mc_sync(real_dev, vdev)Now
dev_mc_sync()sync addresses from vdev toreal_devusing__hw_addr_sync().
__hw_addr_sync()calls__hw_addr_sync_one()which after successfully syncing an address from vdev toreal_devincrement theha->sync_cnt. As a result at this point theha->sync_cnt== 1 for the above address passed by command. After thisvlan_dev_set_rx_mode()calls theset_rx_mode()of thereal_dev.Now when
icssg_update_vlan_mcast()calls__dev_mc_sync(vdev, sync, unsync), it calls_hw_addr_sync_dev(&dev->mc, dev, sync, unsync)and checks theha->sync_cntfor the given address, sincesync_cntis already 1, it doesn't consider it as an newly added address and sync / unsync callbacks are not called. [1]list_for_each_entry_safe(ha, tmp, &list->list, list) { if (ha->sync_cnt) continue;Since for addresses in vlan interfaces,
sync_cntwill always be set asvlan_dev_set_rx_mode()will sync the mc list of vlan interface with the real dev. This will mean that the driver implemented sync / unsync APIs will only be called for thereal_devand thereal_devwon't have any information about thevlan_idwhich I need in my sync / unsync APIs to populate the same in the hardware maintained FDB table.To overcome this, I am maintaining a local copy of
vdev->mcnamed emac->vlan_mcast_list[vid]. I will sync address fromvdev->mcto this. and then call__hw_addr_sync_dev()on this list as thesync_cntfor address in this list will still be zero. Which will then trigger mysync()callback on the vdev which will help me obtain thevlan_idin the sync callback. I can now populate the same in the hardware maintained FDB table.I hope this explains why I am using
__hw_addr_sync_dev(&emac->vlan_mcast_list[vid], vdev,
icssg_prueth_add_mcast,icssg_prueth_del_mcast)instead of
__dev_mc_sync(vdev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);Quoted 8 lines
Quoted 3 lines
Anyways, Even if I use
__dev_mc_sync(), we will still need the export. I am exporting__hw_addr_sync_multiple()not__hw_addr_sync_dev(). The API being used by me__hw_addr_sync_dev()is already exported.I fear there is a misunderstanding. I'm suggesting dropping
__hw_addr_sync_multiple()usage entirely. If that is not possible, a clear and complete explaination of the reason/root cause must be provided.As explained above, I need to be able to sync addresses from
vdev->mcto my local list. Now this can be done using two APIs.1.
__hw_addr_sync()- This is already exported and is actually the first choice. However this will fail syncing address fromvdev->mcto my local copy.
__hw_addr_sync()only syncs the address ifha->sync_cnt== 0. If thesync_cntis non zero, this will skip the address. As explained above, for mc addresses of vlan interfaces,sync_cntwill always be set asvlan_dev_set_rx_mode()will sync the mc list of vlan interface with the real dev. As a result the addresses will be skipped and__hw_addr_sync()will not serve the purpose here.2.
__hw_addr_sync_multiple()- This actually works perfectly fine here as it sync address even ifsync_cntis non zero. As a result I am using this function in my implementation.
Quoted 7 lines
Since this is not public, I have to export it so that the driver can call this.
So I am afraid dropping
__hw_addr_sync_multiple()is not possible here. I hope the explanation above makes sense to you.Please let me know if this is OK with you. If you have some other way> through which I can make this work please let me know.
I think it makes sense and I can't think of simpler ways on top of my head. Please repost capturing a synopsis of the above in the commit message.
Thanks,
Paolo
MD Danish Anwar
reviewerPATCH v3 3/3 net: ti: icssg-prueth: Add Support for Multicast filtering with VLAN in HSR mode
In reply to MD Danish AnwarAdd multicast filtering support for VLAN interfaces in HSR offload mode for ICSSG driver.
The driver calls vlan_for_each() API on the hsr device's ndev to get the list of available vlans for the hsr device. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using __hw_addr_sync_multiple() API.
The driver then calls the sync / unsync callbacks.
In the sync / unsync call back, driver checks if the vdev's real dev is hsr device or not. If the real dev is hsr device, driver gets the per port device using hsr_get_port_ndev() and then driver passes appropriate vid to FDB helper functions.
This commit makes below changes in the hsr files.
- Move enum hsr_port_type from net/hsr/hsr_main.h to include/linux/if_hsr.h so that the enum can be accessed by drivers using hsr.
- Create hsr_get_port_ndev() API that can be used to get the ndev pointer to the slave port from ndev pointer to the hsr net device.
- Export hsr_get_port_ndev() API so that the API can be accessed by drivers using hsr.
--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 83 +++++++++++++++----- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 2 + include/linux/if_hsr.h | 18 +++++ net/hsr/hsr_device.c | 13 +++ net/hsr/hsr_main.h | 9 --- 5 files changed, 97 insertions(+), 28 deletions(-)
drivers/net/ethernet/ti/icssg/icssg_prueth.c +64 -19
@@ -515,32 +515,66 @@ static int icssg_prueth_del_mcast(struct net_device *ndev, const u8 *addr)return 0 ; } -static int icssg_prueth_hsr_add_mcast (struct net_device *ndev,const u8 *addr) +static void icssg_prueth_hsr_fdb_add_del (struct prueth_emac *emac, +const u8 *addr,u8 vid,bool add) { -struct prueth_emac *emac =netdev_priv (ndev); -struct prueth *prueth = emac->prueth; - -icssg_fdb_add_del (emac, addr, prueth->default_vlan, +icssg_fdb_add_del (emac, addr, vid,ICSSG_FDB_ENTRY_P0_MEMBERSHIP |ICSSG_FDB_ENTRY_P1_MEMBERSHIP |ICSSG_FDB_ENTRY_P2_MEMBERSHIP | -ICSSG_FDB_ENTRY_BLOCK , true); +ICSSG_FDB_ENTRY_BLOCK , add); + +if (add) +icssg_vtbl_modify (emac, vid,BIT (emac->port_id), +BIT (emac->port_id), add); +} + +static int icssg_prueth_hsr_add_mcast (struct net_device *ndev,const u8 *addr) +{ +struct net_device *real_dev; +struct prueth_emac *emac; +u8 vlan_id, i; + + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_HSR ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + +if (is_hsr_master (real_dev)) { +for (i =HSR_PT_SLAVE_A ; i <HSR_PT_INTERLINK ; i++) { + emac =netdev_priv (hsr_get_port_ndev (real_dev, i)); +if (!emac) +return -EINVAL ; +icssg_prueth_hsr_fdb_add_del (emac, addr, vlan_id, + true); + } + }else { + emac =netdev_priv (real_dev); +icssg_prueth_hsr_fdb_add_del (emac, addr, vlan_id, true); + } -icssg_vtbl_modify (emac, emac->port_vlan,BIT (emac->port_id), -BIT (emac->port_id), true);return 0 ; }static int icssg_prueth_hsr_del_mcast (struct net_device *ndev,const u8 *addr) { -struct prueth_emac *emac =netdev_priv (ndev); -struct prueth *prueth = emac->prueth; +struct net_device *real_dev; +struct prueth_emac *emac; +u8 vlan_id, i; -icssg_fdb_add_del (emac, addr, prueth->default_vlan, -ICSSG_FDB_ENTRY_P0_MEMBERSHIP | -ICSSG_FDB_ENTRY_P1_MEMBERSHIP | -ICSSG_FDB_ENTRY_P2_MEMBERSHIP | -ICSSG_FDB_ENTRY_BLOCK , false); + vlan_id =is_vlan_dev (ndev) ?vlan_dev_vlan_id (ndev) :PRUETH_DFLT_VLAN_HSR ; + real_dev =is_vlan_dev (ndev) ?vlan_dev_real_dev (ndev) : ndev; + +if (is_hsr_master (real_dev)) { +for (i =HSR_PT_SLAVE_A ; i <HSR_PT_INTERLINK ; i++) { + emac =netdev_priv (hsr_get_port_ndev (real_dev, i)); +if (!emac) +return -EINVAL ; +icssg_prueth_hsr_fdb_add_del (emac, addr, vlan_id, + false); + } + }else { + emac =netdev_priv (real_dev); +icssg_prueth_hsr_fdb_add_del (emac, addr, vlan_id, false); + }return 0 ; }
@@ -558,8 +592,14 @@ static int icssg_update_vlan_mcast(struct net_device *vdev, int vid, vdev->addr_len);netif_addr_unlock_bh (vdev); -__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, - icssg_prueth_add_mcast, icssg_prueth_del_mcast); +if (emac->prueth->is_hsr_offload_mode) +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_hsr_add_mcast, + icssg_prueth_hsr_del_mcast); +else +__hw_addr_sync_dev (&emac->vlan_mcast_list[vid], vdev, + icssg_prueth_add_mcast, + icssg_prueth_del_mcast);return 0 ; }
@@ -808,6 +848,11 @@ static void emac_ndo_set_rx_mode_work(struct work_struct *work)if (emac->prueth->is_hsr_offload_mode) {__dev_mc_sync (ndev, icssg_prueth_hsr_add_mcast, icssg_prueth_hsr_del_mcast); +if (rtnl_trylock ()) { +vlan_for_each (emac->prueth->hsr_dev, + icssg_update_vlan_mcast, emac); +rtnl_unlock (); + } }else {__dev_mc_sync (ndev, icssg_prueth_add_mcast, icssg_prueth_del_mcast);
@@ -1194,7 +1239,7 @@ static int prueth_netdevice_port_link(struct net_device *ndev,if (prueth->br_members &BIT (PRUETH_PORT_MII0 ) && prueth->br_members &BIT (PRUETH_PORT_MII1 )) { prueth->is_switch_mode = true; - prueth->default_vlan =1 ; + prueth->default_vlan =PRUETH_DFLT_VLAN_SW ; emac->port_vlan = prueth->default_vlan;icssg_change_mode (prueth); }
@@ -1247,7 +1292,7 @@ static int prueth_hsr_port_link(struct net_device *ndev) NETIF_PRUETH_HSR_OFFLOAD_FEATURES))return -EOPNOTSUPP ; prueth->is_hsr_offload_mode = true; - prueth->default_vlan =1 ; + prueth->default_vlan =PRUETH_DFLT_VLAN_HSR ; emac0->port_vlan = prueth->default_vlan; emac1->port_vlan = prueth->default_vlan;icssg_change_mode (prueth);
drivers/net/ethernet/ti/icssg/icssg_prueth.h +2 -0
@@ -84,6 +84,8 @@#define ICSS_CMD_ADD_MAC 0x8/* VLAN Filtering Related MACROs */ +#define PRUETH_DFLT_VLAN_HSR 1 +#define PRUETH_DFLT_VLAN_SW 1#define PRUETH_DFLT_VLAN_MAC 0#define MAX_VLAN_ID 256
include/linux/if_hsr.h +18 -0
@@ -13,6 +13,15 @@ enum hsr_version { PRP_V1, }; +enum hsr_port_type { +HSR_PT_NONE =0 ,/* Must be 0, used by framereg */ +HSR_PT_SLAVE_A , +HSR_PT_SLAVE_B , +HSR_PT_INTERLINK , +HSR_PT_MASTER , +HSR_PT_PORTS ,/* This must be the last item in the enum */ +}; +/* HSR Tag. * As defined in IEC-62439-3:2010, the HSR tag is really { ethertype = 0x88FB, * path, LSDU_size, sequence Nr }. But we let eth_header() create { h_dest,
@@ -32,6 +41,8 @@ struct hsr_tag {#if IS_ENABLED (CONFIG_HSR )extern bool is_hsr_master (struct net_device *dev);extern int hsr_get_version (struct net_device *dev,enum hsr_version *ver); +struct net_device *hsr_get_port_ndev (struct net_device *ndev, +enum hsr_port_type pt);#else static inline bool is_hsr_master (struct net_device *dev) {
@@ -42,6 +53,13 @@ static inline int hsr_get_version(struct net_device *dev, {return -EINVAL ; } + +static inline struct net_device *hsr_get_port_ndev (struct net_device *ndev, +enum hsr_port_type pt) +{ +return ERR_PTR (-EINVAL ); +} +#endif /* CONFIG_HSR */ #endif /*_LINUX_IF_HSR_H_*/
net/hsr/hsr_device.c +13 -0
@@ -663,6 +663,19 @@ bool is_hsr_master(struct net_device *dev) }EXPORT_SYMBOL (is_hsr_master); +struct net_device *hsr_get_port_ndev (struct net_device *ndev, +enum hsr_port_type pt) +{ +struct hsr_priv *hsr =netdev_priv (ndev); +struct hsr_port *port; + +hsr_for_each_port (hsr, port) +if (port->type == pt) +return port->dev; +return NULL ; +} +EXPORT_SYMBOL (hsr_get_port_ndev); +/* Default multicast address for HSR Supervision frames */ static const unsigned char def_multicast_addr[ETH_ALEN ]__aligned(2) = {0x01 ,0x15 ,0x4e ,0x00 ,0x01 ,0x00
net/hsr/hsr_main.h +0 -9
@@ -121,15 +121,6 @@ struct hsrv1_ethhdr_sp {struct hsr_sup_tag hsr_sup; } __packed; -enum hsr_port_type { -HSR_PT_NONE =0 ,/* Must be 0, used by framereg */ -HSR_PT_SLAVE_A , -HSR_PT_SLAVE_B , -HSR_PT_INTERLINK , -HSR_PT_MASTER , -HSR_PT_PORTS ,/* This must be the last item in the enum */ -}; -/* PRP Redunancy Control Trailor (RCT). * As defined in IEC-62439-4:2012, the PRP RCT is really { sequence Nr, * Lan indentifier (LanId), LSDU_size and PRP_suffix = 0x88FB }.
Signature · 1 line
2.34.1
Roger Quadros
reviewerRe: [PATCH net-next v3 3/3] net: ti: icssg-prueth: Add Support for Multicast filtering with VLAN in HSR mode
Quoted 23 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in HSR offload mode for ICSSG driver.
The driver calls
vlan_for_each()API on the hsr device's ndev to get the list of available vlans for the hsr device. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks.
In the sync / unsync call back, driver checks if the vdev's real dev is hsr device or not. If the real dev is hsr device, driver gets the per port device using
hsr_get_port_ndev()and then driver passes appropriate vid to FDB helper functions.This commit makes below changes in the hsr files.
- Move enumhsr_port_typefromnet/hsr/hsr_main.htoinclude/linux/if_hsr.hso that the enum can be accessed by drivers using hsr.
- Createhsr_get_port_ndev()API that can be used to get the ndev pointer to the slave port from ndev pointer to the hsr net device.
- Exporthsr_get_port_ndev()API so that the API can be accessed by drivers using hsr.
Quoted 7 lines
Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 83 +++++++++++++++----- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 2 + include/linux/if_hsr.h | 18 +++++ net/hsr/hsr_device.c | 13 +++ net/hsr/hsr_main.h | 9 ---
Should we be splitting hsr core changes into separate patch first, then followed by a patch with icssg driver changes?
Quoted 1 line
5 files changed, 97 insertions(+), 28 deletions(-)
Signature · 2 lines
cheers, -roger
Re: [PATCH net-next v3 3/3] net: ti: icssg-prueth: Add Support for Multicast filtering with VLAN in HSR mode
Quoted 34 lines · Roger Quadros
Quoted 23 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in HSR offload mode for ICSSG driver.
The driver calls
vlan_for_each()API on the hsr device's ndev to get the list of available vlans for the hsr device. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks.
In the sync / unsync call back, driver checks if the vdev's real dev is hsr device or not. If the real dev is hsr device, driver gets the per port device using
hsr_get_port_ndev()and then driver passes appropriate vid to FDB helper functions.This commit makes below changes in the hsr files.
- Move enumhsr_port_typefromnet/hsr/hsr_main.htoinclude/linux/if_hsr.hso that the enum can be accessed by drivers using hsr.
- Createhsr_get_port_ndev()API that can be used to get the ndev pointer to the slave port from ndev pointer to the hsr net device.
- Exporthsr_get_port_ndev()API so that the API can be accessed by drivers using hsr.Quoted 7 lines
Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 83 +++++++++++++++----- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 2 + include/linux/if_hsr.h | 18 +++++ net/hsr/hsr_device.c | 13 +++ net/hsr/hsr_main.h | 9 ---
Quoted 3 lines
Should we be splitting hsr core changes into separate patch first, then followed by a patch with icssg driver changes?
I wanted to make sure that the changes to hsr core are done with the driver change so that any one looking at the commit can understand why these changes are done.
If we split the changes and someone looks at the commit modifying hsr core, they will not be able to understand why this is done. We will be creating and exporting API hsr_get_port_ndev() but there will be no caller for this.
Quoted 2 lines
Quoted 1 line
5 files changed, 97 insertions(+), 28 deletions(-)
Signature · 2 lines
Thanks and Regards, Md Danish Anwar
Roger Quadros
reviewerRe: [PATCH net-next v3 3/3] net: ti: icssg-prueth: Add Support for Multicast filtering with VLAN in HSR mode
Quoted 41 lines · Md Danish
Quoted 34 lines · Roger Quadros
Quoted 23 lines · MD Danish Anwar
Add multicast filtering support for VLAN interfaces in HSR offload mode for ICSSG driver.
The driver calls
vlan_for_each()API on the hsr device's ndev to get the list of available vlans for the hsr device. The driver then sync mc addr of vlan interface with a locally mainatined list emac->vlan_mcast_list[vid] using__hw_addr_sync_multiple()API.The driver then calls the sync / unsync callbacks.
In the sync / unsync call back, driver checks if the vdev's real dev is hsr device or not. If the real dev is hsr device, driver gets the per port device using
hsr_get_port_ndev()and then driver passes appropriate vid to FDB helper functions.This commit makes below changes in the hsr files.
- Move enumhsr_port_typefromnet/hsr/hsr_main.htoinclude/linux/if_hsr.hso that the enum can be accessed by drivers using hsr.
- Createhsr_get_port_ndev()API that can be used to get the ndev pointer to the slave port from ndev pointer to the hsr net device.
- Exporthsr_get_port_ndev()API so that the API can be accessed by drivers using hsr.Quoted 7 lines
Signed-off-by MD Danish Anwar <[email protected]>--- drivers/net/ethernet/ti/icssg/icssg_prueth.c | 83 +++++++++++++++----- drivers/net/ethernet/ti/icssg/icssg_prueth.h | 2 + include/linux/if_hsr.h | 18 +++++ net/hsr/hsr_device.c | 13 +++ net/hsr/hsr_main.h | 9 ---Quoted 3 lines
Should we be splitting hsr core changes into separate patch first, then followed by a patch with icssg driver changes?
Quoted 8 lines
I wanted to make sure that the changes to hsr core are done with the driver change so that any one looking at the commit can understand why these changes are done.
If we split the changes and someone looks at the commit modifying hsr core, they will not be able to understand why this is done. We will be creating and exporting API
hsr_get_port_ndev()but there will be no caller for this.
I think you can explain in the commit log why you need the API.
Quoted 4 lines
Quoted 3 lines
Quoted 2 lines
5 files changed, 97 insertions(+), 28 deletions(-)
Signature · 2 lines
cheers, -roger