From 692b0399a22530b2de8490bea75a7d20d59391d0 Mon Sep 17 00:00:00 2001 From: Hamdan Igbaria Date: Mon, 24 Feb 2020 14:41:29 +0200 Subject: net/mlx5: DR, Fix postsend actions write length Fix the send info write length to be (actions x action) size in bytes. Fixes: 297cccebdc5a ("net/mlx5: DR, Expose an internal API to issue RDMA operations") Signed-off-by: Hamdan Igbaria Reviewed-by: Alex Vesker Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/steering/dr_action.c | 1 - drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c | 3 ++- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_action.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_action.c index 6dec2a550a10..2d93228ff633 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_action.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_action.c @@ -933,7 +933,6 @@ static int dr_actions_l2_rewrite(struct mlx5dr_domain *dmn, action->rewrite.data = (void *)ops; action->rewrite.num_of_actions = i; - action->rewrite.chunk->byte_size = i * sizeof(*ops); ret = mlx5dr_send_postsend_action(dmn, action); if (ret) { diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c index c7f10d4f8f8d..095ec7b1399d 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c @@ -558,7 +558,8 @@ int mlx5dr_send_postsend_action(struct mlx5dr_domain *dmn, int ret; send_info.write.addr = (uintptr_t)action->rewrite.data; - send_info.write.length = action->rewrite.chunk->byte_size; + send_info.write.length = action->rewrite.num_of_actions * + DR_MODIFY_ACTION_SIZE; send_info.write.lkey = 0; send_info.remote_addr = action->rewrite.chunk->mr_addr; send_info.rkey = action->rewrite.chunk->rkey; -- cgit v1.2.3 From 56917766def72f5afdf4235adb91b6897ff26d9d Mon Sep 17 00:00:00 2001 From: Tariq Toukan Date: Thu, 20 Feb 2020 13:40:24 +0200 Subject: net/mlx5e: kTLS, Fix TCP seq off-by-1 issue in TX resync flow We have an off-by-1 issue in the TCP seq comparison. The last sequence number that belongs to the TCP packet's payload is not "start_seq + len", but one byte before it. Fix it so the 'ends_before' is evaluated properly. This fixes a bug that results in error completions in the kTLS HW offload flows. Fixes: ffbd9ca94e2e ("net/mlx5e: kTLS, Fix corner-case checks in TX resync flow") Signed-off-by: Tariq Toukan Reviewed-by: Boris Pismenny Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c index f260dd96873b..52a56622034a 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls_tx.c @@ -218,7 +218,7 @@ tx_sync_info_get(struct mlx5e_ktls_offload_context_tx *priv_tx, * this packet was already acknowledged and its record info * was released. */ - ends_before = before(tcp_seq + datalen, tls_record_start_seq(record)); + ends_before = before(tcp_seq + datalen - 1, tls_record_start_seq(record)); if (unlikely(tls_record_is_start_marker(record))) { ret = ends_before ? MLX5E_KTLS_SYNC_SKIP_NO_DATA : MLX5E_KTLS_SYNC_FAIL; -- cgit v1.2.3 From f28ca65efa87b3fb8da3d69ca7cb1ebc0448de66 Mon Sep 17 00:00:00 2001 From: Tariq Toukan Date: Mon, 24 Feb 2020 13:56:53 +0200 Subject: net/mlx5e: kTLS, Fix wrong value in record tracker enum Fix to match the HW spec: TRACKING state is 1, SEARCHING is 2. No real issue for now, as these values are not currently used. Fixes: d2ead1f360e8 ("net/mlx5e: Add kTLS TX HW offload support") Signed-off-by: Tariq Toukan Reviewed-by: Boris Pismenny Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls.h b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls.h index a3efa29a4629..63116be6b1d6 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ktls.h @@ -38,8 +38,8 @@ enum { enum { MLX5E_TLS_PROGRESS_PARAMS_RECORD_TRACKER_STATE_START = 0, - MLX5E_TLS_PROGRESS_PARAMS_RECORD_TRACKER_STATE_SEARCHING = 1, - MLX5E_TLS_PROGRESS_PARAMS_RECORD_TRACKER_STATE_TRACKING = 2, + MLX5E_TLS_PROGRESS_PARAMS_RECORD_TRACKER_STATE_TRACKING = 1, + MLX5E_TLS_PROGRESS_PARAMS_RECORD_TRACKER_STATE_SEARCHING = 2, }; struct mlx5e_ktls_offload_context_tx { -- cgit v1.2.3 From 404402abd5f90aa90a134eb9604b1750c1941529 Mon Sep 17 00:00:00 2001 From: Sebastian Hense Date: Thu, 20 Feb 2020 08:11:36 +0100 Subject: net/mlx5e: Fix endianness handling in pedit mask The mask value is provided as 64 bit and has to be casted in either 32 or 16 bit. On big endian systems the wrong half was casted which resulted in an all zero mask. Fixes: 2b64beba0251 ("net/mlx5e: Support header re-write of partial fields in TC pedit offload") Signed-off-by: Sebastian Hense Reviewed-by: Roi Dayan Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c index 74091f72c9a8..ec5fc52bf572 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c @@ -2476,10 +2476,11 @@ static int offload_pedit_fields(struct pedit_headers_action *hdrs, continue; if (f->field_bsize == 32) { - mask_be32 = *(__be32 *)&mask; + mask_be32 = (__be32)mask; mask = (__force unsigned long)cpu_to_le32(be32_to_cpu(mask_be32)); } else if (f->field_bsize == 16) { - mask_be16 = *(__be16 *)&mask; + mask_be32 = (__be32)mask; + mask_be16 = *(__be16 *)&mask_be32; mask = (__force unsigned long)cpu_to_le16(be16_to_cpu(mask_be16)); } -- cgit v1.2.3 From 0b136454741b2f6cb18d55e355d396db9248b2ab Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Wed, 19 Feb 2020 09:03:28 +0200 Subject: net/mlx5: Clear LAG notifier pointer after unregister After returning from unregister_netdevice_notifier_dev_net(), set the notifier_call field to NULL so successive call to mlx5_lag_add() will function as expected. Fixes: 7907f23adc18 ("net/mlx5: Implement RoCE LAG feature") Signed-off-by: Eli Cohen Reviewed-by: Vlad Buslov Reviewed-by: Raed Salem Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/lag.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag.c b/drivers/net/ethernet/mellanox/mlx5/core/lag.c index 8e19f6ab8393..93052b07c76c 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/lag.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/lag.c @@ -615,8 +615,10 @@ void mlx5_lag_remove(struct mlx5_core_dev *dev) break; if (i == MLX5_MAX_PORTS) { - if (ldev->nb.notifier_call) + if (ldev->nb.notifier_call) { unregister_netdevice_notifier_net(&init_net, &ldev->nb); + ldev->nb.notifier_call = NULL; + } mlx5_lag_mp_cleanup(ldev); cancel_delayed_work_sync(&ldev->bond_work); mlx5_lag_dev_free(ldev); -- cgit v1.2.3 From cb851c01b51bb610a9093d01624565ce1d4e38fa Mon Sep 17 00:00:00 2001 From: Ido Schimmel Date: Sun, 15 Mar 2020 10:07:35 +0200 Subject: mlxsw: reg: Increase register field length to 31 bits The cited commit set a value of 2^31-1 in order to "disable" the shaper on a given a port. However, the length of the maximum shaper rate field was not updated from 28 bits to 31 bits, which means ports are still limited to ~268Gbps despite supporting speeds of 400Gbps. Fix this by increasing the field's length. Fixes: 92afbfedb77d ("mlxsw: reg: Increase MLXSW_REG_QEEC_MAS_DIS") Signed-off-by: Ido Schimmel Reviewed-by: Jiri Pirko Reviewed-by: Petr Machata Signed-off-by: David S. Miller --- drivers/net/ethernet/mellanox/mlxsw/reg.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h index dd6685156396..e05d1d1be2fd 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/reg.h +++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h @@ -3572,7 +3572,7 @@ MLXSW_ITEM32(reg, qeec, mase, 0x10, 31, 1); * When in bytes mode, value is specified in units of 1000bps. * Access: RW */ -MLXSW_ITEM32(reg, qeec, max_shaper_rate, 0x10, 0, 28); +MLXSW_ITEM32(reg, qeec, max_shaper_rate, 0x10, 0, 31); /* reg_qeec_de * DWRR configuration enable. Enables configuration of the dwrr and -- cgit v1.2.3 From 4a348601eb9131893c22b6ed2d3b6ba2bafc2391 Mon Sep 17 00:00:00 2001 From: Takashi Iwai Date: Sun, 15 Mar 2020 10:34:59 +0100 Subject: net: mlx4: Use scnprintf() for avoiding potential buffer overflow Since snprintf() returns the would-be-output size instead of the actual output size, the succeeding calls may go beyond the given buffer limit. Fix it by replacing with scnprintf(). Cc: "David S . Miller" Cc: Tariq Toukan To: netdev@vger.kernel.org Signed-off-by: Takashi Iwai Signed-off-by: David S. Miller --- drivers/net/ethernet/mellanox/mlx4/mcg.c | 62 ++++++++++++++++---------------- 1 file changed, 31 insertions(+), 31 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx4/mcg.c b/drivers/net/ethernet/mellanox/mlx4/mcg.c index 9c481823b3e8..9486caecfbdc 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mcg.c +++ b/drivers/net/ethernet/mellanox/mlx4/mcg.c @@ -906,59 +906,59 @@ static void mlx4_err_rule(struct mlx4_dev *dev, char *str, int len = 0; mlx4_err(dev, "%s", str); - len += snprintf(buf + len, BUF_SIZE - len, - "port = %d prio = 0x%x qp = 0x%x ", - rule->port, rule->priority, rule->qpn); + len += scnprintf(buf + len, BUF_SIZE - len, + "port = %d prio = 0x%x qp = 0x%x ", + rule->port, rule->priority, rule->qpn); list_for_each_entry(cur, &rule->list, list) { switch (cur->id) { case MLX4_NET_TRANS_RULE_ID_ETH: - len += snprintf(buf + len, BUF_SIZE - len, - "dmac = %pM ", &cur->eth.dst_mac); + len += scnprintf(buf + len, BUF_SIZE - len, + "dmac = %pM ", &cur->eth.dst_mac); if (cur->eth.ether_type) - len += snprintf(buf + len, BUF_SIZE - len, - "ethertype = 0x%x ", - be16_to_cpu(cur->eth.ether_type)); + len += scnprintf(buf + len, BUF_SIZE - len, + "ethertype = 0x%x ", + be16_to_cpu(cur->eth.ether_type)); if (cur->eth.vlan_id) - len += snprintf(buf + len, BUF_SIZE - len, - "vlan-id = %d ", - be16_to_cpu(cur->eth.vlan_id)); + len += scnprintf(buf + len, BUF_SIZE - len, + "vlan-id = %d ", + be16_to_cpu(cur->eth.vlan_id)); break; case MLX4_NET_TRANS_RULE_ID_IPV4: if (cur->ipv4.src_ip) - len += snprintf(buf + len, BUF_SIZE - len, - "src-ip = %pI4 ", - &cur->ipv4.src_ip); + len += scnprintf(buf + len, BUF_SIZE - len, + "src-ip = %pI4 ", + &cur->ipv4.src_ip); if (cur->ipv4.dst_ip) - len += snprintf(buf + len, BUF_SIZE - len, - "dst-ip = %pI4 ", - &cur->ipv4.dst_ip); + len += scnprintf(buf + len, BUF_SIZE - len, + "dst-ip = %pI4 ", + &cur->ipv4.dst_ip); break; case MLX4_NET_TRANS_RULE_ID_TCP: case MLX4_NET_TRANS_RULE_ID_UDP: if (cur->tcp_udp.src_port) - len += snprintf(buf + len, BUF_SIZE - len, - "src-port = %d ", - be16_to_cpu(cur->tcp_udp.src_port)); + len += scnprintf(buf + len, BUF_SIZE - len, + "src-port = %d ", + be16_to_cpu(cur->tcp_udp.src_port)); if (cur->tcp_udp.dst_port) - len += snprintf(buf + len, BUF_SIZE - len, - "dst-port = %d ", - be16_to_cpu(cur->tcp_udp.dst_port)); + len += scnprintf(buf + len, BUF_SIZE - len, + "dst-port = %d ", + be16_to_cpu(cur->tcp_udp.dst_port)); break; case MLX4_NET_TRANS_RULE_ID_IB: - len += snprintf(buf + len, BUF_SIZE - len, - "dst-gid = %pI6\n", cur->ib.dst_gid); - len += snprintf(buf + len, BUF_SIZE - len, - "dst-gid-mask = %pI6\n", - cur->ib.dst_gid_msk); + len += scnprintf(buf + len, BUF_SIZE - len, + "dst-gid = %pI6\n", cur->ib.dst_gid); + len += scnprintf(buf + len, BUF_SIZE - len, + "dst-gid-mask = %pI6\n", + cur->ib.dst_gid_msk); break; case MLX4_NET_TRANS_RULE_ID_VXLAN: - len += snprintf(buf + len, BUF_SIZE - len, - "VNID = %d ", be32_to_cpu(cur->vxlan.vni)); + len += scnprintf(buf + len, BUF_SIZE - len, + "VNID = %d ", be32_to_cpu(cur->vxlan.vni)); break; case MLX4_NET_TRANS_RULE_ID_IPV6: break; @@ -967,7 +967,7 @@ static void mlx4_err_rule(struct mlx4_dev *dev, char *str, break; } } - len += snprintf(buf + len, BUF_SIZE - len, "\n"); + len += scnprintf(buf + len, BUF_SIZE - len, "\n"); mlx4_err(dev, "%s", buf); if (len >= BUF_SIZE) -- cgit v1.2.3 From 6002059d7882c3512e6ac52fa82424272ddfcd5c Mon Sep 17 00:00:00 2001 From: Ido Schimmel Date: Thu, 19 Mar 2020 13:25:39 +0200 Subject: mlxsw: pci: Only issue reset when system is ready During initialization the driver issues a software reset command and then waits for the system status to change back to "ready" state. However, before issuing the reset command the driver does not check that the system is actually in "ready" state. On Spectrum-{1,2} systems this was always the case as the hardware initialization time is very short. On Spectrum-3 systems this is no longer the case. This results in the software reset command timing-out and the driver failing to load: [ 6.347591] mlxsw_spectrum3 0000:06:00.0: Cmd exec timed-out (opcode=40(ACCESS_REG),opcode_mod=0,in_mod=0) [ 6.358382] mlxsw_spectrum3 0000:06:00.0: Reg cmd access failed (reg_id=9023(mrsr),type=write) [ 6.368028] mlxsw_spectrum3 0000:06:00.0: cannot register bus device [ 6.375274] mlxsw_spectrum3: probe of 0000:06:00.0 failed with error -110 Fix this by waiting for the system to become ready both before issuing the reset command and afterwards. In case of failure, print the last system status to aid in debugging. Fixes: da382875c616 ("mlxsw: spectrum: Extend to support Spectrum-3 ASIC") Signed-off-by: Ido Schimmel Reviewed-by: Jiri Pirko Signed-off-by: David S. Miller --- drivers/net/ethernet/mellanox/mlxsw/pci.c | 50 ++++++++++++++++++++++++------- 1 file changed, 39 insertions(+), 11 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c index 914c33e46fb4..e9ded1a6e131 100644 --- a/drivers/net/ethernet/mellanox/mlxsw/pci.c +++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c @@ -1322,36 +1322,64 @@ static void mlxsw_pci_mbox_free(struct mlxsw_pci *mlxsw_pci, mbox->mapaddr); } -static int mlxsw_pci_sw_reset(struct mlxsw_pci *mlxsw_pci, - const struct pci_device_id *id) +static int mlxsw_pci_sys_ready_wait(struct mlxsw_pci *mlxsw_pci, + const struct pci_device_id *id, + u32 *p_sys_status) { unsigned long end; - char mrsr_pl[MLXSW_REG_MRSR_LEN]; - int err; + u32 val; - mlxsw_reg_mrsr_pack(mrsr_pl); - err = mlxsw_reg_write(mlxsw_pci->core, MLXSW_REG(mrsr), mrsr_pl); - if (err) - return err; if (id->device == PCI_DEVICE_ID_MELLANOX_SWITCHX2) { msleep(MLXSW_PCI_SW_RESET_TIMEOUT_MSECS); return 0; } - /* We must wait for the HW to become responsive once again. */ + /* We must wait for the HW to become responsive. */ msleep(MLXSW_PCI_SW_RESET_WAIT_MSECS); end = jiffies + msecs_to_jiffies(MLXSW_PCI_SW_RESET_TIMEOUT_MSECS); do { - u32 val = mlxsw_pci_read32(mlxsw_pci, FW_READY); - + val = mlxsw_pci_read32(mlxsw_pci, FW_READY); if ((val & MLXSW_PCI_FW_READY_MASK) == MLXSW_PCI_FW_READY_MAGIC) return 0; cond_resched(); } while (time_before(jiffies, end)); + + *p_sys_status = val & MLXSW_PCI_FW_READY_MASK; + return -EBUSY; } +static int mlxsw_pci_sw_reset(struct mlxsw_pci *mlxsw_pci, + const struct pci_device_id *id) +{ + struct pci_dev *pdev = mlxsw_pci->pdev; + char mrsr_pl[MLXSW_REG_MRSR_LEN]; + u32 sys_status; + int err; + + err = mlxsw_pci_sys_ready_wait(mlxsw_pci, id, &sys_status); + if (err) { + dev_err(&pdev->dev, "Failed to reach system ready status before reset. Status is 0x%x\n", + sys_status); + return err; + } + + mlxsw_reg_mrsr_pack(mrsr_pl); + err = mlxsw_reg_write(mlxsw_pci->core, MLXSW_REG(mrsr), mrsr_pl); + if (err) + return err; + + err = mlxsw_pci_sys_ready_wait(mlxsw_pci, id, &sys_status); + if (err) { + dev_err(&pdev->dev, "Failed to reach system ready status after reset. Status is 0x%x\n", + sys_status); + return err; + } + + return 0; +} + static int mlxsw_pci_alloc_irq_vectors(struct mlxsw_pci *mlxsw_pci) { int err; -- cgit v1.2.3 From 306f354c67397b3138300cde875c5cab45b857f7 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Mon, 16 Mar 2020 09:31:03 +0200 Subject: net/mlx5_core: Set IB capability mask1 to fix ib_srpt connection failure The cap_mask1 isn't protected by field_select and not listed among RW fields, but it is required to be written to properly initialize ports in IB virtualization mode. Link: https://lore.kernel.org/linux-rdma/88bab94d2fd72f3145835b4518bc63dda587add6.camel@redhat.com Fixes: ab118da4c10a ("net/mlx5: Don't write read-only fields in MODIFY_HCA_VPORT_CONTEXT command") Signed-off-by: Leon Romanovsky Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/vport.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c index 1faac31f74d0..23f879da9104 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c @@ -1071,6 +1071,9 @@ int mlx5_core_modify_hca_vport_context(struct mlx5_core_dev *dev, MLX5_SET64(hca_vport_context, ctx, port_guid, req->port_guid); if (req->field_select & MLX5_HCA_VPORT_SEL_NODE_GUID) MLX5_SET64(hca_vport_context, ctx, node_guid, req->node_guid); + MLX5_SET(hca_vport_context, ctx, cap_mask1, req->cap_mask1); + MLX5_SET(hca_vport_context, ctx, cap_mask1_field_select, + req->cap_mask1_perm); err = mlx5_cmd_exec(dev, in, in_sz, out, sizeof(out)); ex: kfree(in); -- cgit v1.2.3 From 1de0306c3a05d305e45b1f1fabe2f4e94222eb6b Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Mon, 9 Mar 2020 09:44:18 +0200 Subject: net/mlx5e: Enhance ICOSQ WQE info fields Add number of WQEBBs (WQE's Basic Block) to WQE info struct. Set the number of WQEBBs on WQE post, and increment the consumer counter (cc) on completion. In case of error completions, the cc was mistakenly not incremented, keeping a gap between cc and pc (producer counter). This failed the recovery flow on the ICOSQ from a CQE error which timed-out waiting for the cc and pc to meet. Fixes: be5323c8379f ("net/mlx5e: Report and recover from CQE error on ICOSQ") Signed-off-by: Aya Levin Reviewed-by: Tariq Toukan Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en.h | 1 + drivers/net/ethernet/mellanox/mlx5/core/en_rx.c | 11 +++++------ drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 1 + 3 files changed, 7 insertions(+), 6 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h index 220ef9f06f84..571ac5976549 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h @@ -371,6 +371,7 @@ enum { struct mlx5e_sq_wqe_info { u8 opcode; + u8 num_wqebbs; /* Auxiliary data for different opcodes. */ union { diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c index 1c3ab69cbd96..312d4692425b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c @@ -477,6 +477,7 @@ static inline void mlx5e_fill_icosq_frag_edge(struct mlx5e_icosq *sq, /* fill sq frag edge with nops to avoid wqe wrapping two pages */ for (; wi < edge_wi; wi++) { wi->opcode = MLX5_OPCODE_NOP; + wi->num_wqebbs = 1; mlx5e_post_nop(wq, sq->sqn, &sq->pc); } } @@ -525,6 +526,7 @@ static int mlx5e_alloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix) umr_wqe->uctrl.xlt_offset = cpu_to_be16(xlt_offset); sq->db.ico_wqe[pi].opcode = MLX5_OPCODE_UMR; + sq->db.ico_wqe[pi].num_wqebbs = MLX5E_UMR_WQEBBS; sq->db.ico_wqe[pi].umr.rq = rq; sq->pc += MLX5E_UMR_WQEBBS; @@ -621,6 +623,7 @@ void mlx5e_poll_ico_cq(struct mlx5e_cq *cq) ci = mlx5_wq_cyc_ctr2ix(&sq->wq, sqcc); wi = &sq->db.ico_wqe[ci]; + sqcc += wi->num_wqebbs; if (last_wqe && unlikely(get_cqe_opcode(cqe) != MLX5_CQE_REQ)) { netdev_WARN_ONCE(cq->channel->netdev, @@ -631,16 +634,12 @@ void mlx5e_poll_ico_cq(struct mlx5e_cq *cq) break; } - if (likely(wi->opcode == MLX5_OPCODE_UMR)) { - sqcc += MLX5E_UMR_WQEBBS; + if (likely(wi->opcode == MLX5_OPCODE_UMR)) wi->umr.rq->mpwqe.umr_completed++; - } else if (likely(wi->opcode == MLX5_OPCODE_NOP)) { - sqcc++; - } else { + else if (unlikely(wi->opcode != MLX5_OPCODE_NOP)) netdev_WARN_ONCE(cq->channel->netdev, "Bad OPCODE in ICOSQ WQE info: 0x%x\n", wi->opcode); - } } while (!last_wqe); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c index 257a7c9f7a14..800d34ed8a96 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c @@ -78,6 +78,7 @@ void mlx5e_trigger_irq(struct mlx5e_icosq *sq) u16 pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc); sq->db.ico_wqe[pi].opcode = MLX5_OPCODE_NOP; + sq->db.ico_wqe[pi].num_wqebbs = 1; nopwqe = mlx5e_post_nop(wq, sq->sqn, &sq->pc); mlx5e_notify_hw(wq, sq->pc, sq->uar_map, &nopwqe->ctrl); } -- cgit v1.2.3 From 39369fd536d485a99a59d8e357c0d4d3ce19a3b8 Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Thu, 12 Mar 2020 12:35:32 +0200 Subject: net/mlx5e: Fix missing reset of SW metadata in Striding RQ reset When resetting the RQ (moving RQ state from RST to RDY), the driver resets the WQ's SW metadata. In striding RQ mode, we maintain a field that reflects the actual expected WQ head (including in progress WQEs posted to the ICOSQ). It was mistakenly not reset together with the WQ. Fix this here. Fixes: 8276ea1353a4 ("net/mlx5e: Report and recover from CQE with error on RQ") Signed-off-by: Aya Levin Reviewed-by: Tariq Toukan Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h index a226277b0980..f07b1399744e 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h @@ -181,10 +181,12 @@ mlx5e_tx_dma_unmap(struct device *pdev, struct mlx5e_sq_dma *dma) static inline void mlx5e_rqwq_reset(struct mlx5e_rq *rq) { - if (rq->wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) + if (rq->wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) { mlx5_wq_ll_reset(&rq->mpwqe.wq); - else + rq->mpwqe.actual_wq_head = 0; + } else { mlx5_wq_cyc_reset(&rq->wqe.wq); + } } /* SW parser related functions */ -- cgit v1.2.3 From e239c6d686e1c37fb2ab143162dfb57471a8643f Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Mon, 16 Mar 2020 16:53:10 +0200 Subject: net/mlx5e: Fix ICOSQ recovery flow with Striding RQ In striding RQ mode, the buffers of an RX WQE are first prepared and posted to the HW using a UMR WQEs via the ICOSQ. We maintain the state of these in-progress WQEs in the RQ SW struct. In the flow of ICOSQ recovery, the corresponding RQ is not in error state, hence: - The buffers of the in-progress WQEs must be released and the RQ metadata should reflect it. - Existing RX WQEs in the RQ should not be affected. For this, wrap the dealloc of the in-progress WQEs in a function, and use it in the ICOSQ recovery flow instead of mlx5e_free_rx_descs(). Fixes: be5323c8379f ("net/mlx5e: Report and recover from CQE error on ICOSQ") Signed-off-by: Aya Levin Reviewed-by: Tariq Toukan Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en.h | 1 + .../ethernet/mellanox/mlx5/core/en/reporter_rx.c | 2 +- drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 31 +++++++++++++++++----- 3 files changed, 26 insertions(+), 8 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h index 571ac5976549..c9606b8ab6ef 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h @@ -1060,6 +1060,7 @@ int mlx5e_modify_rq_state(struct mlx5e_rq *rq, int curr_state, int next_state); void mlx5e_activate_rq(struct mlx5e_rq *rq); void mlx5e_deactivate_rq(struct mlx5e_rq *rq); void mlx5e_free_rx_descs(struct mlx5e_rq *rq); +void mlx5e_free_rx_in_progress_descs(struct mlx5e_rq *rq); void mlx5e_activate_icosq(struct mlx5e_icosq *icosq); void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c index 6c72b592315b..a01e2de2488f 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c @@ -90,7 +90,7 @@ static int mlx5e_rx_reporter_err_icosq_cqe_recover(void *ctx) goto out; mlx5e_reset_icosq_cc_pc(icosq); - mlx5e_free_rx_descs(rq); + mlx5e_free_rx_in_progress_descs(rq); clear_bit(MLX5E_SQ_STATE_RECOVERING, &icosq->state); mlx5e_activate_icosq(icosq); mlx5e_activate_rq(rq); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index 21de4764d4c0..4ef3dc79f73c 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -813,6 +813,29 @@ int mlx5e_wait_for_min_rx_wqes(struct mlx5e_rq *rq, int wait_time) return -ETIMEDOUT; } +void mlx5e_free_rx_in_progress_descs(struct mlx5e_rq *rq) +{ + struct mlx5_wq_ll *wq; + u16 head; + int i; + + if (rq->wq_type != MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) + return; + + wq = &rq->mpwqe.wq; + head = wq->head; + + /* Outstanding UMR WQEs (in progress) start at wq->head */ + for (i = 0; i < rq->mpwqe.umr_in_progress; i++) { + rq->dealloc_wqe(rq, head); + head = mlx5_wq_ll_get_wqe_next_ix(wq, head); + } + + rq->mpwqe.actual_wq_head = wq->head; + rq->mpwqe.umr_in_progress = 0; + rq->mpwqe.umr_completed = 0; +} + void mlx5e_free_rx_descs(struct mlx5e_rq *rq) { __be16 wqe_ix_be; @@ -820,14 +843,8 @@ void mlx5e_free_rx_descs(struct mlx5e_rq *rq) if (rq->wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) { struct mlx5_wq_ll *wq = &rq->mpwqe.wq; - u16 head = wq->head; - int i; - /* Outstanding UMR WQEs (in progress) start at wq->head */ - for (i = 0; i < rq->mpwqe.umr_in_progress; i++) { - rq->dealloc_wqe(rq, head); - head = mlx5_wq_ll_get_wqe_next_ix(wq, head); - } + mlx5e_free_rx_in_progress_descs(rq); while (!mlx5_wq_ll_is_empty(wq)) { struct mlx5e_rx_wqe_ll *wqe; -- cgit v1.2.3 From 187a9830c921d92c4a9a8e2921ecc4b35a97532c Mon Sep 17 00:00:00 2001 From: Aya Levin Date: Thu, 19 Mar 2020 13:25:17 +0200 Subject: net/mlx5e: Do not recover from a non-fatal syndrome For non-fatal syndromes like LOCAL_LENGTH_ERR, recovery shouldn't be triggered. In these scenarios, the RQ is not actually in ERR state. This misleads the recovery flow which assumes that the RQ is really in error state and no more completions arrive, causing crashes on bad page state. Fixes: 8276ea1353a4 ("net/mlx5e: Report and recover from CQE with error on RQ") Signed-off-by: Aya Levin Reviewed-by: Tariq Toukan Signed-off-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en/health.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/net/ethernet/mellanox') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/health.h b/drivers/net/ethernet/mellanox/mlx5/core/en/health.h index d3693fa547ac..e54f70d9af22 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/health.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/health.h @@ -10,8 +10,7 @@ static inline bool cqe_syndrome_needs_recover(u8 syndrome) { - return syndrome == MLX5_CQE_SYNDROME_LOCAL_LENGTH_ERR || - syndrome == MLX5_CQE_SYNDROME_LOCAL_QP_OP_ERR || + return syndrome == MLX5_CQE_SYNDROME_LOCAL_QP_OP_ERR || syndrome == MLX5_CQE_SYNDROME_LOCAL_PROT_ERR || syndrome == MLX5_CQE_SYNDROME_WR_FLUSH_ERR; } -- cgit v1.2.3