From ebbd17ce297a3f367ca20058272063eaeeced63a Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:02 +0200 Subject: mlxsw: spectrum_router: Add extack argument to mlxsw_sp_lb_rif_init() The extack will be handy in later patches. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/e87ba300121010d580b80a281877573a7b1377ca.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index 43e8f19c7a0a..0b1c17819388 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -10561,7 +10561,8 @@ static void __mlxsw_sp_router_fini(struct mlxsw_sp *mlxsw_sp) mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(rgcr), rgcr_pl); } -static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp) +static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, + struct netlink_ext_ack *extack) { u16 lb_rif_index; int err; @@ -10674,7 +10675,7 @@ int mlxsw_sp_router_init(struct mlxsw_sp *mlxsw_sp, if (err) goto err_vrs_init; - err = mlxsw_sp_lb_rif_init(mlxsw_sp); + err = mlxsw_sp_lb_rif_init(mlxsw_sp, extack); if (err) goto err_lb_rif_init; -- cgit v1.2.3 From f3c85eed1ac364ae2cb2729959c6150813cc9c20 Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:03 +0200 Subject: mlxsw: spectrum_router: Use mlxsw_sp_ul_rif_get() to get main VRF LB RIF The current function, mlxsw_sp_router_ul_rif_get(), is a wrapper around the function mentioned in the subject. As such it forms an external interface of the router code. In future patches we will want to maintain connection between RIFs and the CRIFs (introduced in the next patch) that back them. That will not hold for the VRF-based loopback netdevices, so the whole CRIF business can be kept hidden from the rest of mlxsw. But for the main VRF loopback RIF we do want to keep the RIF-CRIF connection, because that RIF is used for blackhole next hops, and the next hop code can be kept simpler for assuming rif->crif is valid. Hence, instead, call mlxsw_sp_ul_rif_get() to create the main VRF loopback RIF. This being an internal function will take the CRIF argument anyway. Furthermore, the function does not lock, which is not necessary at this point in code yet. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/7a39a011a02a84164cd7f5da7985ec5b2ae01ba5.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index 0b1c17819388..15ce0d557f39 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -10564,19 +10564,20 @@ static void __mlxsw_sp_router_fini(struct mlxsw_sp *mlxsw_sp) static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, struct netlink_ext_ack *extack) { - u16 lb_rif_index; + struct mlxsw_sp_rif *lb_rif; int err; /* Create a generic loopback RIF associated with the main table * (default VRF). Any table can be used, but the main table exists * anyway, so we do not waste resources. */ - err = mlxsw_sp_router_ul_rif_get(mlxsw_sp, RT_TABLE_MAIN, - &lb_rif_index); - if (err) + lb_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, RT_TABLE_MAIN, extack); + if (IS_ERR(lb_rif)) { + err = PTR_ERR(lb_rif); return err; + } - mlxsw_sp->router->lb_rif_index = lb_rif_index; + mlxsw_sp->router->lb_rif_index = lb_rif->rif_index; return 0; } -- cgit v1.2.3 From 4796c287b70a0f60fbc6a5df2ab33d92e8971732 Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:04 +0200 Subject: mlxsw: spectrum_router: Maintain a hash table of CRIFs CRIFs are objects that mlxsw maintains for netdevices that may not have an associated RIF (i.e. they may not have been instantiated in the ASIC), but if indeed they do not, it is quite possible they will in the future. These netdevices are candidate RIFs, hence CRIFs. Netdevices for which CRIFs are created include e.g. bridges, LAGs, or front panel ports. The idea is that next hops would be kept at CRIFs, not RIFs, and thus it would be easier to offload and unoffload the entities that have been added before the RIF was created. In this patch, add the code for low-level CRIF maintenance: create and destroy, and keep in a table keyed by the netdevice pointer for easy recall. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/186d44e399c475159da20689f2c540719f2d1ed0.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- .../net/ethernet/mellanox/mlxsw/spectrum_router.c | 175 +++++++++++++++++++++ .../net/ethernet/mellanox/mlxsw/spectrum_router.h | 1 + 2 files changed, 176 insertions(+) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index 15ce0d557f39..d251a926d140 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -51,6 +51,21 @@ struct mlxsw_sp_vr; struct mlxsw_sp_lpm_tree; struct mlxsw_sp_rif_ops; +struct mlxsw_sp_crif_key { + struct net_device *dev; +}; + +struct mlxsw_sp_crif { + struct mlxsw_sp_crif_key key; + struct rhash_head ht_node; +}; + +static const struct rhashtable_params mlxsw_sp_crif_ht_params = { + .key_offset = offsetof(struct mlxsw_sp_crif, key), + .key_len = sizeof_field(struct mlxsw_sp_crif, key), + .head_offset = offsetof(struct mlxsw_sp_crif, ht_node), +}; + struct mlxsw_sp_rif { struct list_head nexthop_list; struct list_head neigh_list; @@ -1060,6 +1075,56 @@ u32 mlxsw_sp_ipip_dev_ul_tb_id(const struct net_device *ol_dev) return tb_id; } +static void +mlxsw_sp_crif_init(struct mlxsw_sp_crif *crif, struct net_device *dev) +{ + crif->key.dev = dev; +} + +static struct mlxsw_sp_crif * +mlxsw_sp_crif_alloc(struct net_device *dev) +{ + struct mlxsw_sp_crif *crif; + + crif = kzalloc(sizeof(*crif), GFP_KERNEL); + if (!crif) + return NULL; + + mlxsw_sp_crif_init(crif, dev); + return crif; +} + +static void mlxsw_sp_crif_free(struct mlxsw_sp_crif *crif) +{ + kfree(crif); +} + +static int mlxsw_sp_crif_insert(struct mlxsw_sp_router *router, + struct mlxsw_sp_crif *crif) +{ + return rhashtable_insert_fast(&router->crif_ht, &crif->ht_node, + mlxsw_sp_crif_ht_params); +} + +static void mlxsw_sp_crif_remove(struct mlxsw_sp_router *router, + struct mlxsw_sp_crif *crif) +{ + rhashtable_remove_fast(&router->crif_ht, &crif->ht_node, + mlxsw_sp_crif_ht_params); +} + +static struct mlxsw_sp_crif * +mlxsw_sp_crif_lookup(struct mlxsw_sp_router *router, + const struct net_device *dev) +{ + struct mlxsw_sp_crif_key key = { + .dev = (struct net_device *)dev, + }; + + return rhashtable_lookup_fast(&router->crif_ht, &key, + mlxsw_sp_crif_ht_params); +} + static struct mlxsw_sp_rif * mlxsw_sp_rif_create(struct mlxsw_sp *mlxsw_sp, const struct mlxsw_sp_rif_params *params, @@ -9148,6 +9213,95 @@ static int mlxsw_sp_router_port_pre_changeaddr_event(struct mlxsw_sp_rif *rif, return -ENOBUFS; } +static bool mlxsw_sp_router_netdevice_interesting(struct mlxsw_sp *mlxsw_sp, + struct net_device *dev) +{ + struct vlan_dev_priv *vlan; + + if (netif_is_lag_master(dev) || + netif_is_bridge_master(dev) || + mlxsw_sp_port_dev_check(dev) || + mlxsw_sp_netdev_is_ipip_ol(mlxsw_sp, dev) || + netif_is_l3_master(dev)) + return true; + + if (!is_vlan_dev(dev)) + return false; + + vlan = vlan_dev_priv(dev); + return netif_is_lag_master(vlan->real_dev) || + netif_is_bridge_master(vlan->real_dev) || + mlxsw_sp_port_dev_check(vlan->real_dev); +} + +static struct mlxsw_sp_crif * +mlxsw_sp_crif_register(struct mlxsw_sp_router *router, struct net_device *dev) +{ + struct mlxsw_sp_crif *crif; + int err; + + if (WARN_ON(mlxsw_sp_crif_lookup(router, dev))) + return NULL; + + crif = mlxsw_sp_crif_alloc(dev); + if (!crif) + return ERR_PTR(-ENOMEM); + + err = mlxsw_sp_crif_insert(router, crif); + if (err) + goto err_netdev_insert; + + return crif; + +err_netdev_insert: + mlxsw_sp_crif_free(crif); + return ERR_PTR(err); +} + +static void mlxsw_sp_crif_unregister(struct mlxsw_sp_router *router, + struct mlxsw_sp_crif *crif) +{ + mlxsw_sp_crif_remove(router, crif); + mlxsw_sp_crif_free(crif); +} + +static int mlxsw_sp_netdevice_register(struct mlxsw_sp_router *router, + struct net_device *dev) +{ + struct mlxsw_sp_crif *crif; + + if (!mlxsw_sp_router_netdevice_interesting(router->mlxsw_sp, dev)) + return 0; + + crif = mlxsw_sp_crif_register(router, dev); + return PTR_ERR_OR_ZERO(crif); +} + +static void mlxsw_sp_netdevice_unregister(struct mlxsw_sp_router *router, + struct net_device *dev) +{ + struct mlxsw_sp_crif *crif; + + if (!mlxsw_sp_router_netdevice_interesting(router->mlxsw_sp, dev)) + return; + + /* netdev_run_todo(), by way of netdev_wait_allrefs_any(), rebroadcasts + * the NETDEV_UNREGISTER message, so we can get here twice. If that's + * what happened, the netdevice state is NETREG_UNREGISTERED. In that + * case, we expect to have collected the CRIF already, and warn if it + * still exists. Otherwise we expect the CRIF to exist. + */ + crif = mlxsw_sp_crif_lookup(router, dev); + if (dev->reg_state == NETREG_UNREGISTERED) { + if (!WARN_ON(crif)) + return; + } + if (WARN_ON(!crif)) + return; + + mlxsw_sp_crif_unregister(router, crif); +} + static bool mlxsw_sp_is_offload_xstats_event(unsigned long event) { switch (event) { @@ -9367,6 +9521,15 @@ static int mlxsw_sp_router_netdevice_event(struct notifier_block *nb, mutex_lock(&mlxsw_sp->router->lock); + if (event == NETDEV_REGISTER) { + err = mlxsw_sp_netdevice_register(router, dev); + if (err) + /* No need to roll this back, UNREGISTER will collect it + * anyhow. + */ + goto out; + } + if (mlxsw_sp_is_offload_xstats_event(event)) err = mlxsw_sp_netdevice_offload_xstats_cmd(mlxsw_sp, dev, event, ptr); @@ -9381,6 +9544,10 @@ static int mlxsw_sp_router_netdevice_event(struct notifier_block *nb, else if (mlxsw_sp_is_vrf_event(event, ptr)) err = mlxsw_sp_netdevice_vrf_event(dev, event, ptr); + if (event == NETDEV_UNREGISTER) + mlxsw_sp_netdevice_unregister(router, dev); + +out: mutex_unlock(&mlxsw_sp->router->lock); return notifier_from_errno(err); @@ -10649,6 +10816,11 @@ int mlxsw_sp_router_init(struct mlxsw_sp *mlxsw_sp, if (err) goto err_ipips_init; + err = rhashtable_init(&mlxsw_sp->router->crif_ht, + &mlxsw_sp_crif_ht_params); + if (err) + goto err_crif_ht_init; + err = mlxsw_sp_rifs_init(mlxsw_sp); if (err) goto err_rifs_init; @@ -10780,6 +10952,8 @@ err_nexthop_group_ht_init: err_nexthop_ht_init: mlxsw_sp_rifs_fini(mlxsw_sp); err_rifs_init: + rhashtable_destroy(&mlxsw_sp->router->crif_ht); +err_crif_ht_init: mlxsw_sp_ipips_fini(mlxsw_sp); err_ipips_init: __mlxsw_sp_router_fini(mlxsw_sp); @@ -10815,6 +10989,7 @@ void mlxsw_sp_router_fini(struct mlxsw_sp *mlxsw_sp) rhashtable_destroy(&router->nexthop_group_ht); rhashtable_destroy(&router->nexthop_ht); mlxsw_sp_rifs_fini(mlxsw_sp); + rhashtable_destroy(&mlxsw_sp->router->crif_ht); mlxsw_sp_ipips_fini(mlxsw_sp); __mlxsw_sp_router_fini(mlxsw_sp); cancel_delayed_work_sync(&router->nh_grp_activity_dw); diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h index 5a0babc614b4..b223e80303f5 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h @@ -20,6 +20,7 @@ struct mlxsw_sp_router_nve_decap { struct mlxsw_sp_router { struct mlxsw_sp *mlxsw_sp; + struct rhashtable crif_ht; struct gen_pool *rifs_table; struct mlxsw_sp_rif **rifs; struct idr rif_mac_profiles_idr; -- cgit v1.2.3 From 78126cfd5dc97da7baf66415374452e3f8805faf Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:05 +0200 Subject: mlxsw: spectrum_router: Maintain CRIF for fallback loopback RIF CRIFs are generally not maintained for loopback RIFs. However, the RIF for the default VRF is used for offloading of blackhole nexthops. Nexthops expect to have a valid CRIF. Therefore in this patch, add code to maintain CRIF for the loopback RIF as well. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/7f2b2fcc98770167ed1254a904c3f7f585ba43f0.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 12 +++++++++++- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h | 1 + 2 files changed, 12 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index d251a926d140..c4d538e0169e 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -10731,9 +10731,14 @@ static void __mlxsw_sp_router_fini(struct mlxsw_sp *mlxsw_sp) static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, struct netlink_ext_ack *extack) { + struct mlxsw_sp_router *router = mlxsw_sp->router; struct mlxsw_sp_rif *lb_rif; int err; + router->lb_crif = mlxsw_sp_crif_alloc(NULL); + if (IS_ERR(router->lb_crif)) + return PTR_ERR(router->lb_crif); + /* Create a generic loopback RIF associated with the main table * (default VRF). Any table can be used, but the main table exists * anyway, so we do not waste resources. @@ -10741,17 +10746,22 @@ static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, lb_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, RT_TABLE_MAIN, extack); if (IS_ERR(lb_rif)) { err = PTR_ERR(lb_rif); - return err; + goto err_ul_rif_get; } mlxsw_sp->router->lb_rif_index = lb_rif->rif_index; return 0; + +err_ul_rif_get: + mlxsw_sp_crif_free(router->lb_crif); + return err; } static void mlxsw_sp_lb_rif_fini(struct mlxsw_sp *mlxsw_sp) { mlxsw_sp_router_ul_rif_put(mlxsw_sp, mlxsw_sp->router->lb_rif_index); + mlxsw_sp_crif_free(mlxsw_sp->router->lb_crif); } static int mlxsw_sp1_router_init(struct mlxsw_sp *mlxsw_sp) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h index b223e80303f5..0909cf229c86 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h @@ -61,6 +61,7 @@ struct mlxsw_sp_router { struct mutex lock; /* Protects shared router resources */ struct mlxsw_sp_fib_entry_op_ctx *ll_op_ctx; u16 lb_rif_index; + struct mlxsw_sp_crif *lb_crif; const struct mlxsw_sp_adj_grp_size_range *adj_grp_size_ranges; size_t adj_grp_size_ranges_count; struct delayed_work nh_grp_activity_dw; -- cgit v1.2.3 From aa21242b07a8cde689bb6aedcbc224eda9646d9f Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:06 +0200 Subject: mlxsw: spectrum_router: Link CRIFs to RIFs When a RIF is about to be created, the registration of the netdevice that it should be associated with must have been seen in the past, and a CRIF created. Therefore make this a hard requirement by looking up the CRIF during RIF creation, and complaining loudly when there isn't one. This then allows to keep a link between a RIF and its corresponding CRIF (and back, as the relationship is one-to-at-most-one), which do. The CRIF will later be useful as the objects tracked there will be offloaded lazily as a result of RIF creation. CRIFs are created when an "interesting" netdevice is registered, and destroyed after such device is unregistered. CRIFs are supposed to already exist when a RIF creation request arises, and exist at least as long as that RIF exists. This makes for a simple invariant: it is always safe to dereference CRIF pointer from "its" RIF. To guarantee this, CRIFs cannot be removed immediately when the UNREGISTER event is delivered. The reason is that if a RIF's netdevices has an IPv6 address, removal of this address is notified in an atomic block. To remove the RIF, the IPv6 removal handler schedules a work item. It must be safe for this work item to access the associated CRIF as well. Thus when a netdevice that backs the CRIF is removed, if it still has a RIF, do not actually free the CRIF, only toggle its can_destroy flag, which this patch adds. Later on, mlxsw_sp_rif_destroy() collects the CRIF. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/68c8e33afa6b8c03c431b435e1685ffdff752e63.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- .../net/ethernet/mellanox/mlxsw/spectrum_router.c | 59 +++++++++++++++++----- 1 file changed, 47 insertions(+), 12 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index c4d538e0169e..daa59fc59d3b 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -58,6 +58,8 @@ struct mlxsw_sp_crif_key { struct mlxsw_sp_crif { struct mlxsw_sp_crif_key key; struct rhash_head ht_node; + bool can_destroy; + struct mlxsw_sp_rif *rif; }; static const struct rhashtable_params mlxsw_sp_crif_ht_params = { @@ -67,9 +69,9 @@ static const struct rhashtable_params mlxsw_sp_crif_ht_params = { }; struct mlxsw_sp_rif { + struct mlxsw_sp_crif *crif; /* NULL for underlay RIF */ struct list_head nexthop_list; struct list_head neigh_list; - struct net_device *dev; /* NULL for underlay RIF */ struct mlxsw_sp_fid *fid; unsigned char addr[ETH_ALEN]; int mtu; @@ -88,7 +90,9 @@ struct mlxsw_sp_rif { static struct net_device *mlxsw_sp_rif_dev(const struct mlxsw_sp_rif *rif) { - return rif->dev; + if (!rif->crif) + return NULL; + return rif->crif->key.dev; } struct mlxsw_sp_rif_params { @@ -1096,6 +1100,9 @@ mlxsw_sp_crif_alloc(struct net_device *dev) static void mlxsw_sp_crif_free(struct mlxsw_sp_crif *crif) { + if (WARN_ON(crif->rif)) + return; + kfree(crif); } @@ -7970,8 +7977,9 @@ static void mlxsw_sp_rif_index_free(struct mlxsw_sp *mlxsw_sp, u16 rif_index, static struct mlxsw_sp_rif *mlxsw_sp_rif_alloc(size_t rif_size, u16 rif_index, u16 vr_id, - struct net_device *l3_dev) + struct mlxsw_sp_crif *crif) { + struct net_device *l3_dev = crif ? crif->key.dev : NULL; struct mlxsw_sp_rif *rif; rif = kzalloc(rif_size, GFP_KERNEL); @@ -7983,10 +7991,13 @@ static struct mlxsw_sp_rif *mlxsw_sp_rif_alloc(size_t rif_size, u16 rif_index, if (l3_dev) { ether_addr_copy(rif->addr, l3_dev->dev_addr); rif->mtu = l3_dev->mtu; - rif->dev = l3_dev; } rif->vr_id = vr_id; rif->rif_index = rif_index; + if (crif) { + rif->crif = crif; + crif->rif = rif; + } return rif; } @@ -7995,6 +8006,9 @@ static void mlxsw_sp_rif_free(struct mlxsw_sp_rif *rif) { WARN_ON(!list_empty(&rif->neigh_list)); WARN_ON(!list_empty(&rif->nexthop_list)); + + if (rif->crif) + rif->crif->rif = NULL; kfree(rif); } @@ -8228,6 +8242,7 @@ mlxsw_sp_rif_create(struct mlxsw_sp *mlxsw_sp, const struct mlxsw_sp_rif_ops *ops; struct mlxsw_sp_fid *fid = NULL; enum mlxsw_sp_rif_type type; + struct mlxsw_sp_crif *crif; struct mlxsw_sp_rif *rif; struct mlxsw_sp_vr *vr; u16 rif_index; @@ -8247,7 +8262,13 @@ mlxsw_sp_rif_create(struct mlxsw_sp *mlxsw_sp, goto err_rif_index_alloc; } - rif = mlxsw_sp_rif_alloc(ops->rif_size, rif_index, vr->id, params->dev); + crif = mlxsw_sp_crif_lookup(mlxsw_sp->router, params->dev); + if (WARN_ON(!crif)) { + err = -ENOENT; + goto err_crif_lookup; + } + + rif = mlxsw_sp_rif_alloc(ops->rif_size, rif_index, vr->id, crif); if (!rif) { err = -ENOMEM; goto err_rif_alloc; @@ -8306,6 +8327,7 @@ err_fid_get: dev_put(params->dev); mlxsw_sp_rif_free(rif); err_rif_alloc: +err_crif_lookup: mlxsw_sp_rif_index_free(mlxsw_sp, rif_index, rif_entries); err_rif_index_alloc: vr->rif_count--; @@ -8318,6 +8340,7 @@ static void mlxsw_sp_rif_destroy(struct mlxsw_sp_rif *rif) struct net_device *dev = mlxsw_sp_rif_dev(rif); const struct mlxsw_sp_rif_ops *ops = rif->ops; struct mlxsw_sp *mlxsw_sp = rif->mlxsw_sp; + struct mlxsw_sp_crif *crif = rif->crif; struct mlxsw_sp_fid *fid = rif->fid; u8 rif_entries = rif->rif_entries; u16 rif_index = rif->rif_index; @@ -8348,6 +8371,9 @@ static void mlxsw_sp_rif_destroy(struct mlxsw_sp_rif *rif) mlxsw_sp_rif_index_free(mlxsw_sp, rif_index, rif_entries); vr->rif_count--; mlxsw_sp_vr_put(mlxsw_sp, vr); + + if (crif->can_destroy) + mlxsw_sp_crif_free(crif); } void mlxsw_sp_rif_destroy_by_dev(struct mlxsw_sp *mlxsw_sp, @@ -9262,7 +9288,10 @@ static void mlxsw_sp_crif_unregister(struct mlxsw_sp_router *router, struct mlxsw_sp_crif *crif) { mlxsw_sp_crif_remove(router, crif); - mlxsw_sp_crif_free(crif); + if (crif->rif) + crif->can_destroy = true; + else + mlxsw_sp_crif_free(crif); } static int mlxsw_sp_netdevice_register(struct mlxsw_sp_router *router, @@ -10068,6 +10097,7 @@ mlxsw_sp_rif_ipip_lb_ul_rif_op(struct mlxsw_sp_rif *ul_rif, bool enable) static struct mlxsw_sp_rif * mlxsw_sp_ul_rif_create(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_vr *vr, + struct mlxsw_sp_crif *ul_crif, struct netlink_ext_ack *extack) { struct mlxsw_sp_rif *ul_rif; @@ -10081,7 +10111,8 @@ mlxsw_sp_ul_rif_create(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_vr *vr, return ERR_PTR(err); } - ul_rif = mlxsw_sp_rif_alloc(sizeof(*ul_rif), rif_index, vr->id, NULL); + ul_rif = mlxsw_sp_rif_alloc(sizeof(*ul_rif), rif_index, vr->id, + ul_crif); if (!ul_rif) { err = -ENOMEM; goto err_rif_alloc; @@ -10120,6 +10151,7 @@ static void mlxsw_sp_ul_rif_destroy(struct mlxsw_sp_rif *ul_rif) static struct mlxsw_sp_rif * mlxsw_sp_ul_rif_get(struct mlxsw_sp *mlxsw_sp, u32 tb_id, + struct mlxsw_sp_crif *ul_crif, struct netlink_ext_ack *extack) { struct mlxsw_sp_vr *vr; @@ -10132,7 +10164,7 @@ mlxsw_sp_ul_rif_get(struct mlxsw_sp *mlxsw_sp, u32 tb_id, if (refcount_inc_not_zero(&vr->ul_rif_refcnt)) return vr->ul_rif; - vr->ul_rif = mlxsw_sp_ul_rif_create(mlxsw_sp, vr, extack); + vr->ul_rif = mlxsw_sp_ul_rif_create(mlxsw_sp, vr, ul_crif, extack); if (IS_ERR(vr->ul_rif)) { err = PTR_ERR(vr->ul_rif); goto err_ul_rif_create; @@ -10170,7 +10202,7 @@ int mlxsw_sp_router_ul_rif_get(struct mlxsw_sp *mlxsw_sp, u32 ul_tb_id, int err = 0; mutex_lock(&mlxsw_sp->router->lock); - ul_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, ul_tb_id, NULL); + ul_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, ul_tb_id, NULL, NULL); if (IS_ERR(ul_rif)) { err = PTR_ERR(ul_rif); goto out; @@ -10206,7 +10238,7 @@ mlxsw_sp2_rif_ipip_lb_configure(struct mlxsw_sp_rif *rif, struct mlxsw_sp_rif *ul_rif; int err; - ul_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, ul_tb_id, extack); + ul_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, ul_tb_id, NULL, extack); if (IS_ERR(ul_rif)) return PTR_ERR(ul_rif); @@ -10741,9 +10773,12 @@ static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, /* Create a generic loopback RIF associated with the main table * (default VRF). Any table can be used, but the main table exists - * anyway, so we do not waste resources. + * anyway, so we do not waste resources. Loopback RIFs are usually + * created with a NULL CRIF, but this RIF is used as a fallback RIF + * for blackhole nexthops, and nexthops expect to have a valid CRIF. */ - lb_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, RT_TABLE_MAIN, extack); + lb_rif = mlxsw_sp_ul_rif_get(mlxsw_sp, RT_TABLE_MAIN, router->lb_crif, + extack); if (IS_ERR(lb_rif)) { err = PTR_ERR(lb_rif); goto err_ul_rif_get; -- cgit v1.2.3 From bdc0b78e79a641fbbb928a9e5da56dbdd42ff674 Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:07 +0200 Subject: mlxsw: spectrum_router: Use router.lb_crif instead of .lb_rif_index A previous patch added a pointer to loopback CRIF to the router data structure. That makes the loopback RIF index redundant, as everything necessary can be derived from the CRIF. Drop the field and adjust the code accordingly. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/8637bf959bc5b6c9d5184b9bd8a0cd53c5132835.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 12 ++++-------- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h | 1 - 2 files changed, 4 insertions(+), 9 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index daa59fc59d3b..acd6f1b5eef9 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -3563,7 +3563,7 @@ static int __mlxsw_sp_nexthop_eth_update(struct mlxsw_sp *mlxsw_sp, u16 rif_index; rif_index = nh->rif ? nh->rif->rif_index : - mlxsw_sp->router->lb_rif_index; + mlxsw_sp->router->lb_crif->rif->rif_index; op = force ? MLXSW_REG_RATR_OP_WRITE_WRITE_ENTRY : MLXSW_REG_RATR_OP_WRITE_WRITE_ENTRY_ON_ACTIVITY; mlxsw_reg_ratr_pack(ratr_pl, op, true, MLXSW_REG_RATR_TYPE_ETHERNET, @@ -4530,7 +4530,7 @@ static int mlxsw_sp_adj_trap_entry_init(struct mlxsw_sp *mlxsw_sp) mlxsw_reg_ratr_pack(ratr_pl, MLXSW_REG_RATR_OP_WRITE_WRITE_ENTRY, true, MLXSW_REG_RATR_TYPE_ETHERNET, mlxsw_sp->router->adj_trap_index, - mlxsw_sp->router->lb_rif_index); + mlxsw_sp->router->lb_crif->rif->rif_index); mlxsw_reg_ratr_trap_action_set(ratr_pl, trap_action); mlxsw_reg_ratr_trap_id_set(ratr_pl, MLXSW_TRAP_ID_RTR_EGRESS0); err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ratr), ratr_pl); @@ -4846,15 +4846,13 @@ static bool mlxsw_sp_nexthop_obj_is_gateway(struct mlxsw_sp *mlxsw_sp, static void mlxsw_sp_nexthop_obj_blackhole_init(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop *nh) { - u16 lb_rif_index = mlxsw_sp->router->lb_rif_index; - nh->action = MLXSW_SP_NEXTHOP_ACTION_DISCARD; nh->should_offload = 1; /* While nexthops that discard packets do not forward packets * via an egress RIF, they still need to be programmed using a * valid RIF, so use the loopback RIF created during init. */ - nh->rif = mlxsw_sp->router->rifs[lb_rif_index]; + nh->rif = mlxsw_sp->router->lb_crif->rif; } static void mlxsw_sp_nexthop_obj_blackhole_fini(struct mlxsw_sp *mlxsw_sp, @@ -10784,8 +10782,6 @@ static int mlxsw_sp_lb_rif_init(struct mlxsw_sp *mlxsw_sp, goto err_ul_rif_get; } - mlxsw_sp->router->lb_rif_index = lb_rif->rif_index; - return 0; err_ul_rif_get: @@ -10795,7 +10791,7 @@ err_ul_rif_get: static void mlxsw_sp_lb_rif_fini(struct mlxsw_sp *mlxsw_sp) { - mlxsw_sp_router_ul_rif_put(mlxsw_sp, mlxsw_sp->router->lb_rif_index); + mlxsw_sp_ul_rif_put(mlxsw_sp->router->lb_crif->rif); mlxsw_sp_crif_free(mlxsw_sp->router->lb_crif); } diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h index 0909cf229c86..9a2669a08480 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.h @@ -60,7 +60,6 @@ struct mlxsw_sp_router { struct mlxsw_sp_router_nve_decap nve_decap_config; struct mutex lock; /* Protects shared router resources */ struct mlxsw_sp_fib_entry_op_ctx *ll_op_ctx; - u16 lb_rif_index; struct mlxsw_sp_crif *lb_crif; const struct mlxsw_sp_adj_grp_size_range *adj_grp_size_ranges; size_t adj_grp_size_ranges_count; -- cgit v1.2.3 From a285d664236eb0655eea0ceb97095ad4b07fcbae Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:08 +0200 Subject: mlxsw: spectrum_router: Split nexthop finalization to two stages Nexthop finalization consists of two steps: the part where the offload is removed, because the backing RIF is now gone; and the part where the association to the RIF is severed. Extract from mlxsw_sp_nexthop_type_fini() a helper that covers the unoffloading part, mlxsw_sp_nexthop_type_rif_gone(), so that it can later be called independently. Note that this swaps around the ordering of mlxsw_sp_nexthop_ipip_fini() vs. mlxsw_sp_nexthop_rif_fini(). The current ordering is more of a historical happenstance than a conscious decision. The two cleanups do not depend on each other, and this change should have no observable effects. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/7134559534c5f5c4807c3a1569fae56f8887e763.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index acd6f1b5eef9..6c9244c35192 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -4369,21 +4369,26 @@ err_neigh_init: return err; } -static void mlxsw_sp_nexthop_type_fini(struct mlxsw_sp *mlxsw_sp, - struct mlxsw_sp_nexthop *nh) +static void mlxsw_sp_nexthop_type_rif_gone(struct mlxsw_sp *mlxsw_sp, + struct mlxsw_sp_nexthop *nh) { switch (nh->type) { case MLXSW_SP_NEXTHOP_TYPE_ETH: mlxsw_sp_nexthop_neigh_fini(mlxsw_sp, nh); - mlxsw_sp_nexthop_rif_fini(nh); break; case MLXSW_SP_NEXTHOP_TYPE_IPIP: - mlxsw_sp_nexthop_rif_fini(nh); mlxsw_sp_nexthop_ipip_fini(mlxsw_sp, nh); break; } } +static void mlxsw_sp_nexthop_type_fini(struct mlxsw_sp *mlxsw_sp, + struct mlxsw_sp_nexthop *nh) +{ + mlxsw_sp_nexthop_type_rif_gone(mlxsw_sp, nh); + mlxsw_sp_nexthop_rif_fini(nh); +} + static int mlxsw_sp_nexthop4_init(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop_group *nh_grp, struct mlxsw_sp_nexthop *nh, -- cgit v1.2.3 From 9464a3d68ea99ccac7e3518e1dfd366ac80bbc90 Mon Sep 17 00:00:00 2001 From: Petr Machata Date: Thu, 22 Jun 2023 15:33:09 +0200 Subject: mlxsw: spectrum_router: Track next hops at CRIFs Move the list of next hops from struct mlxsw_sp_rif to mlxsw_sp_crif. The reason is that eventually, next hops for mlxsw uppers should be offloaded and unoffloaded on demand as a netdevice becomes an upper, or stops being one. Currently, next hops are tracked at RIFs, but RIFs do not exist when a netdevice is not an mlxsw uppers. CRIFs are kept track of throughout the netdevice lifetime. Correspondingly, track at each next hop not its RIF, but its CRIF (from which a RIF can always be deduced). Note that now that next hops are tracked at a CRIF, it is not necessary to move each over to a new RIF when it is necessary to edit a RIF. Therefore drop mlxsw_sp_nexthop_rif_migrate() and have mlxsw_sp_rif_migrate_destroy() call mlxsw_sp_nexthop_rif_update() directly. Signed-off-by: Petr Machata Reviewed-by: Danielle Ratson Link: https://lore.kernel.org/r/e7c1c0a7dd13883b0f09aeda12c4fcf4d63a70e3.1687438411.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski --- .../net/ethernet/mellanox/mlxsw/spectrum_router.c | 127 ++++++++++++--------- 1 file changed, 75 insertions(+), 52 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c index 6c9244c35192..445ba7fe3c40 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c @@ -59,6 +59,7 @@ struct mlxsw_sp_crif { struct mlxsw_sp_crif_key key; struct rhash_head ht_node; bool can_destroy; + struct list_head nexthop_list; struct mlxsw_sp_rif *rif; }; @@ -70,7 +71,6 @@ static const struct rhashtable_params mlxsw_sp_crif_ht_params = { struct mlxsw_sp_rif { struct mlxsw_sp_crif *crif; /* NULL for underlay RIF */ - struct list_head nexthop_list; struct list_head neigh_list; struct mlxsw_sp_fid *fid; unsigned char addr[ETH_ALEN]; @@ -1083,6 +1083,7 @@ static void mlxsw_sp_crif_init(struct mlxsw_sp_crif *crif, struct net_device *dev) { crif->key.dev = dev; + INIT_LIST_HEAD(&crif->nexthop_list); } static struct mlxsw_sp_crif * @@ -1103,6 +1104,7 @@ static void mlxsw_sp_crif_free(struct mlxsw_sp_crif *crif) if (WARN_ON(crif->rif)) return; + WARN_ON(!list_empty(&crif->nexthop_list)); kfree(crif); } @@ -1720,17 +1722,26 @@ static void mlxsw_sp_netdevice_ipip_ol_down_event(struct mlxsw_sp *mlxsw_sp, mlxsw_sp_ipip_entry_ol_down_event(mlxsw_sp, ipip_entry); } -static void mlxsw_sp_nexthop_rif_migrate(struct mlxsw_sp *mlxsw_sp, - struct mlxsw_sp_rif *old_rif, - struct mlxsw_sp_rif *new_rif); +static void mlxsw_sp_nexthop_rif_update(struct mlxsw_sp *mlxsw_sp, + struct mlxsw_sp_rif *rif); + static void mlxsw_sp_rif_migrate_destroy(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_rif *old_rif, struct mlxsw_sp_rif *new_rif, bool migrate_nhs) { + struct mlxsw_sp_crif *crif = old_rif->crif; + struct mlxsw_sp_crif mock_crif = {}; + if (migrate_nhs) - mlxsw_sp_nexthop_rif_migrate(mlxsw_sp, old_rif, new_rif); + mlxsw_sp_nexthop_rif_update(mlxsw_sp, new_rif); + /* Plant a mock CRIF so that destroying the old RIF doesn't unoffload + * our nexthops and IPIP tunnels, and doesn't sever the crif->rif link. + */ + mlxsw_sp_crif_init(&mock_crif, crif->key.dev); + old_rif->crif = &mock_crif; + mock_crif.rif = old_rif; mlxsw_sp_rif_destroy(old_rif); } @@ -1756,9 +1767,6 @@ mlxsw_sp_ipip_entry_ol_lb_update(struct mlxsw_sp *mlxsw_sp, return 0; } -static void mlxsw_sp_nexthop_rif_update(struct mlxsw_sp *mlxsw_sp, - struct mlxsw_sp_rif *rif); - /** * __mlxsw_sp_ipip_entry_update_tunnel - Update offload related to IPIP entry. * @mlxsw_sp: mlxsw_sp. @@ -2987,7 +2995,7 @@ struct mlxsw_sp_nexthop_key { struct mlxsw_sp_nexthop { struct list_head neigh_list_node; /* member of neigh entry list */ - struct list_head rif_list_node; + struct list_head crif_list_node; struct list_head router_list_node; struct mlxsw_sp_nexthop_group_info *nhgi; /* pointer back to the group * this nexthop belongs to @@ -3000,7 +3008,7 @@ struct mlxsw_sp_nexthop { int nh_weight; int norm_nh_weight; int num_adj_entries; - struct mlxsw_sp_rif *rif; + struct mlxsw_sp_crif *crif; u8 should_offload:1, /* set indicates this nexthop should be written * to the adjacency table. */ @@ -3023,9 +3031,9 @@ struct mlxsw_sp_nexthop { static struct net_device * mlxsw_sp_nexthop_dev(const struct mlxsw_sp_nexthop *nh) { - if (nh->rif) - return mlxsw_sp_rif_dev(nh->rif); - return NULL; + if (!nh->crif) + return NULL; + return nh->crif->key.dev; } enum mlxsw_sp_nexthop_group_type { @@ -3050,7 +3058,11 @@ struct mlxsw_sp_nexthop_group_info { static struct mlxsw_sp_rif * mlxsw_sp_nhgi_rif(const struct mlxsw_sp_nexthop_group_info *nhgi) { - return nhgi->nexthops[0].rif; + struct mlxsw_sp_crif *crif = nhgi->nexthops[0].crif; + + if (!crif) + return NULL; + return crif->rif; } struct mlxsw_sp_nexthop_group_vr_key { @@ -3174,7 +3186,9 @@ int mlxsw_sp_nexthop_indexes(struct mlxsw_sp_nexthop *nh, u32 *p_adj_index, struct mlxsw_sp_rif *mlxsw_sp_nexthop_rif(struct mlxsw_sp_nexthop *nh) { - return nh->rif; + if (WARN_ON(!nh->crif)) + return NULL; + return nh->crif->rif; } bool mlxsw_sp_nexthop_group_has_ipip(struct mlxsw_sp_nexthop *nh) @@ -3559,11 +3573,12 @@ static int __mlxsw_sp_nexthop_eth_update(struct mlxsw_sp *mlxsw_sp, bool force, char *ratr_pl) { struct mlxsw_sp_neigh_entry *neigh_entry = nh->neigh_entry; + struct mlxsw_sp_rif *rif = mlxsw_sp_nexthop_rif(nh); enum mlxsw_reg_ratr_op op; u16 rif_index; - rif_index = nh->rif ? nh->rif->rif_index : - mlxsw_sp->router->lb_crif->rif->rif_index; + rif_index = rif ? rif->rif_index : + mlxsw_sp->router->lb_crif->rif->rif_index; op = force ? MLXSW_REG_RATR_OP_WRITE_WRITE_ENTRY : MLXSW_REG_RATR_OP_WRITE_WRITE_ENTRY_ON_ACTIVITY; mlxsw_reg_ratr_pack(ratr_pl, op, true, MLXSW_REG_RATR_TYPE_ETHERNET, @@ -4181,23 +4196,23 @@ mlxsw_sp_nexthop_neigh_update(struct mlxsw_sp *mlxsw_sp, } } -static void mlxsw_sp_nexthop_rif_init(struct mlxsw_sp_nexthop *nh, - struct mlxsw_sp_rif *rif) +static void mlxsw_sp_nexthop_crif_init(struct mlxsw_sp_nexthop *nh, + struct mlxsw_sp_crif *crif) { - if (nh->rif) + if (nh->crif) return; - nh->rif = rif; - list_add(&nh->rif_list_node, &rif->nexthop_list); + nh->crif = crif; + list_add(&nh->crif_list_node, &crif->nexthop_list); } -static void mlxsw_sp_nexthop_rif_fini(struct mlxsw_sp_nexthop *nh) +static void mlxsw_sp_nexthop_crif_fini(struct mlxsw_sp_nexthop *nh) { - if (!nh->rif) + if (!nh->crif) return; - list_del(&nh->rif_list_node); - nh->rif = NULL; + list_del(&nh->crif_list_node); + nh->crif = NULL; } static int mlxsw_sp_nexthop_neigh_init(struct mlxsw_sp *mlxsw_sp, @@ -4209,6 +4224,9 @@ static int mlxsw_sp_nexthop_neigh_init(struct mlxsw_sp *mlxsw_sp, u8 nud_state, dead; int err; + if (WARN_ON(!nh->crif->rif)) + return 0; + if (!nh->nhgi->gateway || nh->neigh_entry) return 0; dev = mlxsw_sp_nexthop_dev(nh); @@ -4299,15 +4317,20 @@ static void mlxsw_sp_nexthop_ipip_init(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop *nh, struct mlxsw_sp_ipip_entry *ipip_entry) { + struct mlxsw_sp_crif *crif; bool removing; if (!nh->nhgi->gateway || nh->ipip_entry) return; + crif = mlxsw_sp_crif_lookup(mlxsw_sp->router, ipip_entry->ol_dev); + if (WARN_ON(!crif)) + return; + nh->ipip_entry = ipip_entry; removing = !mlxsw_sp_ipip_netdev_ul_up(ipip_entry->ol_dev); __mlxsw_sp_nexthop_neigh_update(nh, removing); - mlxsw_sp_nexthop_rif_init(nh, &ipip_entry->ol_lb->common); + mlxsw_sp_nexthop_crif_init(nh, crif); } static void mlxsw_sp_nexthop_ipip_fini(struct mlxsw_sp *mlxsw_sp, @@ -4339,7 +4362,7 @@ static int mlxsw_sp_nexthop_type_init(struct mlxsw_sp *mlxsw_sp, { const struct mlxsw_sp_ipip_ops *ipip_ops; struct mlxsw_sp_ipip_entry *ipip_entry; - struct mlxsw_sp_rif *rif; + struct mlxsw_sp_crif *crif; int err; ipip_entry = mlxsw_sp_ipip_entry_find_by_ol_dev(mlxsw_sp, dev); @@ -4353,11 +4376,15 @@ static int mlxsw_sp_nexthop_type_init(struct mlxsw_sp *mlxsw_sp, } nh->type = MLXSW_SP_NEXTHOP_TYPE_ETH; - rif = mlxsw_sp_rif_find_by_dev(mlxsw_sp, dev); - if (!rif) + crif = mlxsw_sp_crif_lookup(mlxsw_sp->router, dev); + if (!crif) + return 0; + + mlxsw_sp_nexthop_crif_init(nh, crif); + + if (!crif->rif) return 0; - mlxsw_sp_nexthop_rif_init(nh, rif); err = mlxsw_sp_nexthop_neigh_init(mlxsw_sp, nh); if (err) goto err_neigh_init; @@ -4365,7 +4392,7 @@ static int mlxsw_sp_nexthop_type_init(struct mlxsw_sp *mlxsw_sp, return 0; err_neigh_init: - mlxsw_sp_nexthop_rif_fini(nh); + mlxsw_sp_nexthop_crif_fini(nh); return err; } @@ -4386,7 +4413,7 @@ static void mlxsw_sp_nexthop_type_fini(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop *nh) { mlxsw_sp_nexthop_type_rif_gone(mlxsw_sp, nh); - mlxsw_sp_nexthop_rif_fini(nh); + mlxsw_sp_nexthop_crif_fini(nh); } static int mlxsw_sp_nexthop4_init(struct mlxsw_sp *mlxsw_sp, @@ -4479,7 +4506,7 @@ static void mlxsw_sp_nexthop_rif_update(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop *nh; bool removing; - list_for_each_entry(nh, &rif->nexthop_list, rif_list_node) { + list_for_each_entry(nh, &rif->crif->nexthop_list, crif_list_node) { switch (nh->type) { case MLXSW_SP_NEXTHOP_TYPE_ETH: removing = false; @@ -4497,25 +4524,14 @@ static void mlxsw_sp_nexthop_rif_update(struct mlxsw_sp *mlxsw_sp, } } -static void mlxsw_sp_nexthop_rif_migrate(struct mlxsw_sp *mlxsw_sp, - struct mlxsw_sp_rif *old_rif, - struct mlxsw_sp_rif *new_rif) -{ - struct mlxsw_sp_nexthop *nh; - - list_splice_init(&old_rif->nexthop_list, &new_rif->nexthop_list); - list_for_each_entry(nh, &new_rif->nexthop_list, rif_list_node) - nh->rif = new_rif; - mlxsw_sp_nexthop_rif_update(mlxsw_sp, new_rif); -} - static void mlxsw_sp_nexthop_rif_gone_sync(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_rif *rif) { struct mlxsw_sp_nexthop *nh, *tmp; - list_for_each_entry_safe(nh, tmp, &rif->nexthop_list, rif_list_node) { - mlxsw_sp_nexthop_type_fini(mlxsw_sp, nh); + list_for_each_entry_safe(nh, tmp, &rif->crif->nexthop_list, + crif_list_node) { + mlxsw_sp_nexthop_type_rif_gone(mlxsw_sp, nh); mlxsw_sp_nexthop_group_refresh(mlxsw_sp, nh->nhgi->nh_grp); } } @@ -4857,13 +4873,13 @@ static void mlxsw_sp_nexthop_obj_blackhole_init(struct mlxsw_sp *mlxsw_sp, * via an egress RIF, they still need to be programmed using a * valid RIF, so use the loopback RIF created during init. */ - nh->rif = mlxsw_sp->router->lb_crif->rif; + nh->crif = mlxsw_sp->router->lb_crif; } static void mlxsw_sp_nexthop_obj_blackhole_fini(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_nexthop *nh) { - nh->rif = NULL; + nh->crif = NULL; nh->should_offload = 0; } @@ -7871,6 +7887,9 @@ static int mlxsw_sp_router_rif_disable(struct mlxsw_sp *mlxsw_sp, u16 rif) static void mlxsw_sp_router_rif_gone_sync(struct mlxsw_sp *mlxsw_sp, struct mlxsw_sp_rif *rif) { + /* Signal to nexthop cleanup that the RIF is going away. */ + rif->crif->rif = NULL; + mlxsw_sp_router_rif_disable(mlxsw_sp, rif->rif_index); mlxsw_sp_nexthop_rif_gone_sync(mlxsw_sp, rif); mlxsw_sp_neigh_rif_gone_sync(mlxsw_sp, rif); @@ -7989,7 +8008,6 @@ static struct mlxsw_sp_rif *mlxsw_sp_rif_alloc(size_t rif_size, u16 rif_index, if (!rif) return NULL; - INIT_LIST_HEAD(&rif->nexthop_list); INIT_LIST_HEAD(&rif->neigh_list); if (l3_dev) { ether_addr_copy(rif->addr, l3_dev->dev_addr); @@ -8008,7 +8026,6 @@ static struct mlxsw_sp_rif *mlxsw_sp_rif_alloc(size_t rif_size, u16 rif_index, static void mlxsw_sp_rif_free(struct mlxsw_sp_rif *rif) { WARN_ON(!list_empty(&rif->neigh_list)); - WARN_ON(!list_empty(&rif->nexthop_list)); if (rif->crif) rif->crif->rif = NULL; @@ -9290,7 +9307,13 @@ err_netdev_insert: static void mlxsw_sp_crif_unregister(struct mlxsw_sp_router *router, struct mlxsw_sp_crif *crif) { + struct mlxsw_sp_nexthop *nh, *tmp; + mlxsw_sp_crif_remove(router, crif); + + list_for_each_entry_safe(nh, tmp, &crif->nexthop_list, crif_list_node) + mlxsw_sp_nexthop_type_fini(router->mlxsw_sp, nh); + if (crif->rif) crif->can_destroy = true; else -- cgit v1.2.3