mlx5 updates 2020-09-15
Subsystems (5)
From: Saeed Mahameed <[email protected]>
Hi Dave & Jakub,
This series adds some misc updates to mlx5 driver.
For more information please see tag log below.
Please pull and let me know if there is any problem.
Thanks,
Saeed.
---
The following changes since commit ed6d9b0228132fee03314b276d946fc3f0cc9e4f:
ionic: fix up debugfs after queue swap (2020-09-14 16:55:54 -0700)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-updates-2020-09-15
for you to fetch changes up to b7cf0806e8783e38f881cae3c56f0869e70b8da2:
net/mlx5e: Add CQE compression support for multi-strides packets (2020-09-15 11:59:53 -0700)
----------------------------------------------------------------
mlx5-updates-2020-09-15
Various updates to mlx5 driver,
1) Eli adds support for TC trap action.
2) Eran, minor improvements to clock.c code structure
3) Better handling of error reporting in LAG from Jianbo
4) IPv6 traffic class (DSCP) header rewrite support from Maor
5) Ofer Levi adds support for CQE compression of multi-strides packets
6) Vu, Enables use of vport meta data by default.
7) Some minor code cleanup
----------------------------------------------------------------
Dan Carpenter (1):
net/mlx5: remove erroneous fallthrough
Eli Cohen (1):
net/mlx5e: Add support for tc trap
Eran Ben Elisha (4):
net/mlx5: Always use container_of to find mdev pointer from clock struct net/mlx5: Rename ptp clock info
net/mlx5: Release clock lock before scheduling a PPS work net/mlx5: Don't call timecounter cyc2time directly from 1PPS flow
Jianbo Liu (3):
net/mlx5e: Return a valid errno if can't get lag device index net/mlx5e: Add LAG warning for unsupported tx type net/mlx5e: Add LAG warning if bond slave is not lag master
Maor Dickman (1):
net/mlx5e: Add IPv6 traffic class (DSCP) header rewrite support
Moshe Tal (1):
net/mlx5: Fix uninitialized variable warning
Ofer Levi (1):
net/mlx5e: Add CQE compression support for multi-strides packets
Vu Pham (4):
net/mlx5: E-Switch, Check and enable metadata support flag before using net/mlx5: E-Switch, Dedicated metadata for uplink vport net/mlx5: E-Switch, Setup all vports' metadata to support peer miss rule net/mlx5: E-Switch, Use vport metadata matching by default
drivers/net/ethernet/mellanox/mlx5/core/en.h | 1 + .../net/ethernet/mellanox/mlx5/core/en/health.c | 2 +- .../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 1 - drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 12 ++- drivers/net/ethernet/mellanox/mlx5/core/en_rx.c | 11 ++- drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 11 +++ .../ethernet/mellanox/mlx5/core/eswitch_offloads.c | 85 +++++++++++++--------- drivers/net/ethernet/mellanox/mlx5/core/lag.c | 66 ++++++++++++----- drivers/net/ethernet/mellanox/mlx5/core/lag.h | 7 ++ drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c | 9 ++- .../net/ethernet/mellanox/mlx5/core/lib/clock.c | 61 +++++++++------- include/linux/mlx5/device.h | 3 +- include/linux/mlx5/driver.h | 1 - 13 files changed, 181 insertions(+), 89 deletions(-)
PATCH 3/16 net/mlx5: Always use container_of to find mdev pointer from clock struct
From: Eran Ben Elisha <[email protected]>
Clock struct is part of struct mlx5_core_dev. Code was inconsistent, on some cases used container_of and on another used clock->mdev.
Align code to use container_of amd remove clock->mdev pointer. While here, fix reverse xmas tree coding style.
--- .../ethernet/mellanox/mlx5/core/lib/clock.c | 52 +++++++++++-------- include/linux/mlx5/driver.h | 1 - 2 files changed, 29 insertions(+), 24 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +29 -23
@@ -150,28 +150,30 @@ static void mlx5_pps_out(struct work_struct *work)static void mlx5_timestamp_overflow (struct work_struct *work) {struct delayed_work *dwork =to_delayed_work (work); - struct mlx5_clock *clock = container_of(dwork, struct mlx5_clock, - overflow_work); +struct mlx5_core_dev *mdev; +struct mlx5_clock *clock;unsigned long flags; + clock = container_of(dwork, struct mlx5_clock, overflow_work); + mdev = container_of(clock, struct mlx5_core_dev, clock);write_seqlock_irqsave (&clock->lock, flags);timecounter_read (&clock->tc); -mlx5_update_clock_info_page (clock->mdev); +mlx5_update_clock_info_page (mdev);write_sequnlock_irqrestore (&clock->lock, flags);schedule_delayed_work (&clock->overflow_work, clock->overflow_period); } -static int mlx5_ptp_settime (struct ptp_clock_info *ptp, -const struct timespec64 *ts) +static int mlx5_ptp_settime (struct ptp_clock_info *ptp,const struct timespec64 *ts) { - struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, - ptp_info); + struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info);u64 ns =timespec64_to_ns (ts); +struct mlx5_core_dev *mdev;unsigned long flags; + mdev = container_of(clock, struct mlx5_core_dev, clock);write_seqlock_irqsave (&clock->lock, flags);timecounter_init (&clock->tc, &clock->cycles, ns); -mlx5_update_clock_info_page (clock->mdev); +mlx5_update_clock_info_page (mdev);write_sequnlock_irqrestore (&clock->lock, flags);return 0 ;
@@ -180,13 +182,12 @@ static int mlx5_ptp_settime(struct ptp_clock_info *ptp,static int mlx5_ptp_gettimex (struct ptp_clock_info *ptp,struct timespec64 *ts,struct ptp_system_timestamp *sts) { - struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, - ptp_info); - struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev, - clock); + struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info); +struct mlx5_core_dev *mdev;unsigned long flags;u64 cycles, ns; + mdev = container_of(clock, struct mlx5_core_dev, clock);write_seqlock_irqsave (&clock->lock, flags); cycles =mlx5_read_internal_timer (mdev, sts); ns =timecounter_cyc2time (&clock->tc, cycles);
@@ -199,13 +200,14 @@ static int mlx5_ptp_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,static int mlx5_ptp_adjtime (struct ptp_clock_info *ptp,s64 delta) { - struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, - ptp_info); + struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info); +struct mlx5_core_dev *mdev;unsigned long flags; + mdev = container_of(clock, struct mlx5_core_dev, clock);write_seqlock_irqsave (&clock->lock, flags);timecounter_adjtime (&clock->tc, delta); -mlx5_update_clock_info_page (clock->mdev); +mlx5_update_clock_info_page (mdev);write_sequnlock_irqrestore (&clock->lock, flags);return 0 ;
@@ -213,12 +215,13 @@ static int mlx5_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)static int mlx5_ptp_adjfreq (struct ptp_clock_info *ptp,s32 delta) { -u64 adj; -u32 diff; + struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, ptp_info); +struct mlx5_core_dev *mdev;unsigned long flags;int neg_adj =0 ; - struct mlx5_clock *clock = container_of(ptp, struct mlx5_clock, - ptp_info); +u32 diff; +u64 adj; +if (delta <0 ) { neg_adj =1 ;
@@ -229,11 +232,12 @@ static int mlx5_ptp_adjfreq(struct ptp_clock_info *ptp, s32 delta) adj *= delta; diff =div_u64 (adj,1000000000ULL ); + mdev = container_of(clock, struct mlx5_core_dev, clock);write_seqlock_irqsave (&clock->lock, flags);timecounter_read (&clock->tc); clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff : clock->nominal_c_mult + diff; -mlx5_update_clock_info_page (clock->mdev); +mlx5_update_clock_info_page (mdev);write_sequnlock_irqrestore (&clock->lock, flags);return 0 ;
@@ -465,7 +469,8 @@ static int mlx5_query_mtpps_pin_mode(struct mlx5_core_dev *mdev, u8 pin,static int mlx5_get_pps_pin_mode (struct mlx5_clock *clock,u8 pin) { -struct mlx5_core_dev *mdev = clock->mdev; + struct mlx5_core_dev *mdev = container_of(clock, struct mlx5_core_dev, clock); +u32 out[MLX5_ST_SZ_DW (mtpps_reg)] = {};u8 mode;int err;
@@ -538,15 +543,17 @@ static int mlx5_pps_event(struct notifier_block *nb, unsigned long type, void *data) { struct mlx5_clock *clock = mlx5_nb_cof(nb, struct mlx5_clock, pps_nb); -struct mlx5_core_dev *mdev = clock->mdev;struct ptp_clock_event ptp_event;u64 cycles_now, cycles_delta;u64 nsec_now, nsec_delta, ns;struct mlx5_eqe *eqe = data;int pin = eqe->data.pps.pin; +struct mlx5_core_dev *mdev;struct timespec64 ts;unsigned long flags; + mdev = container_of(clock, struct mlx5_core_dev, clock); +switch (clock->ptp_info.pin_config[pin].func) {case PTP_PF_EXTTS : ptp_event.index = pin;
@@ -605,7 +612,6 @@ void mlx5_init_clock(struct mlx5_core_dev *mdev) clock->cycles.shift); clock->nominal_c_mult = clock->cycles.mult; clock->cycles.mask =CLOCKSOURCE_MASK (41 ); - clock->mdev = mdev;timecounter_init (&clock->tc, &clock->cycles,ktime_to_ns (ktime_get_real ()));
include/linux/mlx5/driver.h +0 -1
@@ -643,7 +643,6 @@ struct mlx5_pps { };struct mlx5_clock { -struct mlx5_core_dev *mdev;struct mlx5_nb pps_nb;seqlock_t lock;struct cyclecounter cycles;
Signature · 1 line
2.26.2
PATCH 15/16 net/mlx5e: Add IPv6 traffic class (DSCP) header rewrite support
In reply to [email protected]From: Maor Dickman <[email protected]>
Add support for rewriting of IPV6 DSCP part of traffic class field. Next commands, for example, can be used to offload rewrite action:
OVS:
$ ovs-ofctl add-flow ovs-sriov "tcpv6, in_port=REP, \
actions=mod_nw_tos:68, output:NIC"iproute2:
$ tc filter add dev REP ingress protocol ipv6 prio 1 flower skip_sw \
ip_proto tcp \
action pedit ex munge ip6 traffic_class set 68 retain 0xfc pipe \
action mirred egress redirect dev NIC--- drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 1 + 1 file changed, 1 insertion(+)
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +1 -0
@@ -2615,6 +2615,7 @@ static struct mlx5_fields fields[] = {OFFLOAD (DIPV6_31_0 ,32 ,U32_MAX , ip6.daddr.s6_addr32[3 ],0 , dst_ipv4_dst_ipv6.ipv6_layout.ipv6[12 ]),OFFLOAD (IPV6_HOPLIMIT ,8 ,U8_MAX , ip6.hop_limit,0 , ttl_hoplimit), +OFFLOAD (IP_DSCP ,16 ,0xc00f , ip6,0 , ip_dscp),OFFLOAD (TCP_SPORT ,16 ,U16_MAX , tcp.source,0 , tcp_sport), OFFLOAD(TCP_DPORT, 16, U16_MAX, tcp.dest, 0, tcp_dport),
Signature · 1 line
2.26.2
PATCH 16/16 net/mlx5e: Add CQE compression support for multi-strides packets
In reply to [email protected]From: Ofer Levi <[email protected]>
Add CQE compression support for completions of packets that span multiple strides in a Striding RQ, per the HW capability. In our memory model, we use small strides (256B as of today) for the non-linear SKB mode. This feature allows CQE compression to work also for multiple strides packets. In this case decompressing the mini CQE array will use stride index provided by HW as part of the mini CQE. Before this feature, compression was possible only for single-strided packets, i.e. for packets of size up to 256 bytes when in non-linear mode, and the index was maintained by SW.
This feature is supported for ConnectX-5 and above.
Feature performance test:
This was whitebox-tested, we reduced the PCI speed from 125Gb/s to 62.5Gb/s to overload pci and manipulated mlx5 driver to drop incoming packets before building the SKB to achieve low cpu utilization. Outcome is low cpu utilization and bottleneck on pci only.
Test setup:
Server: Intel(R) Xeon(R) Silver 4108 CPU @ 1.80GHz server, 32 cores
NIC: ConnectX-6 DX.
Sender side generates 300 byte packets at full pci bandwidth. Receiver side configuration:
Single channel, one cpu processing with one ring allocated. Cpu utilization is ~20% while pci bandwidth is fully utilized.
For the generated traffic and interface MTU of 4500B (to activate the non-linear SKB mode), packet rate improvement is about 19% from ~17.6Mpps to ~21Mpps.
Without this feature, counters show no CQE compression blocks for this setup, while with the feature, counters show ~20.7Mpps compressed CQEs in ~500K compression blocks.
--- drivers/net/ethernet/mellanox/mlx5/core/en.h | 1 + drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 12 +++++++++++- drivers/net/ethernet/mellanox/mlx5/core/en_rx.c | 11 ++++++++++- include/linux/mlx5/device.h | 3 ++- 4 files changed, 24 insertions(+), 3 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en.h +1 -0
@@ -265,6 +265,7 @@ enum {MLX5E_RQ_STATE_NO_CSUM_COMPLETE ,MLX5E_RQ_STATE_CSUM_FULL ,/* cqe_csum_full hw bit is set */ MLX5E_RQ_STATE_FPGA_TLS ,/* FPGA TLS enabled */ +MLX5E_RQ_STATE_MINI_CQE_HW_STRIDX /* set when mini_cqe_resp_stride_index cap is used */ }; struct mlx5e_cq {
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +11 -1
@@ -848,6 +848,13 @@ int mlx5e_open_rq(struct mlx5e_channel *c, struct mlx5e_params *params,if (MLX5E_GET_PFLAG (params,MLX5E_PFLAG_RX_NO_CSUM_COMPLETE ) || c->xdp)__set_bit (MLX5E_RQ_STATE_NO_CSUM_COMPLETE , &c->rq.state); +/* For CQE compression on striding RQ, use stride index provided by +* HW if capability is supported. +*/ +if (MLX5E_GET_PFLAG (params,MLX5E_PFLAG_RX_STRIDING_RQ ) && +MLX5_CAP_GEN (c->mdev, mini_cqe_resp_stride_index)) +__set_bit (MLX5E_RQ_STATE_MINI_CQE_HW_STRIDX , &c->rq.state); +return 0 ; err_destroy_rq:
@@ -2182,6 +2189,7 @@ void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv, struct mlx5e_cq_param *param) {struct mlx5_core_dev *mdev = priv->mdev; +bool hw_stridx = false;void *cqc = param->cqc;u8 log_cq_size;
@@ -2189,6 +2197,7 @@ void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ : log_cq_size =mlx5e_mpwqe_get_log_rq_size (params, xsk) +mlx5e_mpwqe_get_log_num_strides (mdev, params, xsk); + hw_stridx =MLX5_CAP_GEN (mdev, mini_cqe_resp_stride_index);break ;default :/* MLX5_WQ_TYPE_CYCLIC */ log_cq_size = params->log_rq_mtu_frames;
@@ -2196,7 +2205,8 @@ void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,MLX5_SET (cqc, cqc, log_cq_size, log_cq_size);if (MLX5E_GET_PFLAG (params,MLX5E_PFLAG_RX_CQE_COMPRESS )) { -MLX5_SET (cqc, cqc, mini_cqe_res_format,MLX5_CQE_FORMAT_CSUM ); +MLX5_SET (cqc, cqc, mini_cqe_res_format, hw_stridx ? +MLX5_CQE_FORMAT_CSUM_STRIDX :MLX5_CQE_FORMAT_CSUM );MLX5_SET (cqc, cqc, cqe_comp_en,1 ); }
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +10 -1
@@ -137,8 +137,17 @@ static inline void mlx5e_decompress_cqe(struct mlx5e_rq *rq, title->check_sum = mini_cqe->checksum; title->op_own &=0xf0 ; title->op_own |=0x01 & (cqcc >> wq->fbc.log_sz); - title->wqe_counter =cpu_to_be16 (cqd->wqe_counter); +/* state bit set implies linked-list striding RQ wq type and +* HW stride index capability supported +*/ +if (test_bit (MLX5E_RQ_STATE_MINI_CQE_HW_STRIDX , &rq->state)) { + title->wqe_counter = mini_cqe->stridx; +return ; + } + +/* HW stride index capability not supported */ + title->wqe_counter =cpu_to_be16 (cqd->wqe_counter);if (rq->wq_type ==MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ ) cqd->wqe_counter +=mpwrq_get_cqe_consumed_strides (title);else
include/linux/mlx5/device.h +2 -1
@@ -816,7 +816,7 @@ struct mlx5_mini_cqe8 {__be32 rx_hash_result;struct {__be16 checksum; -__be16 rsvd; +__be16 stridx; };struct {__be16 wqe_counter;
@@ -836,6 +836,7 @@ enum {enum {MLX5_CQE_FORMAT_CSUM =0x1 , +MLX5_CQE_FORMAT_CSUM_STRIDX =0x3 , };#define MLX5_MINI_CQE_ARRAY_SIZE 8
Signature · 1 line
2.26.2
PATCH 14/16 net/mlx5e: Add support for tc trap
In reply to [email protected]From: Eli Cohen <[email protected]>
Support tc trap such that packets can explicitly be forwarded to slow path if they match a specific rule.
In the example below, we want packets with src IP equals 7.7.7.8 to be forwarded to software, in which case it will get to the appropriate representor net device.
$ tc filter add dev eth1 protocol ip prio 1 root flower skip_sw \
src_ip 7.7.7.8 action trap--- drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +10 -0
@@ -3943,6 +3943,16 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, action |=MLX5_FLOW_CONTEXT_ACTION_DROP |MLX5_FLOW_CONTEXT_ACTION_COUNT ;break ; +case FLOW_ACTION_TRAP : +if (!flow_offload_has_one_action (flow_action)) { +NL_SET_ERR_MSG_MOD (extack, +"action trap is supported as a sole action only" ); +return -EOPNOTSUPP ; + } + action |= (MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | +MLX5_FLOW_CONTEXT_ACTION_COUNT ); + attr->flags |=MLX5_ESW_ATTR_FLAG_SLOW_PATH ; +break ;case FLOW_ACTION_MPLS_PUSH : if (!MLX5_CAP_ESW_FLOWTABLE_FDB(priv->mdev, reformat_l2_to_l3_tunnel) ||
Signature · 1 line
2.26.2
PATCH 8/16 net/mlx5e: Add LAG warning for unsupported tx type
In reply to [email protected]From: Jianbo Liu <[email protected]>
If bond tx type is not active-backup or hash, LAG offload can't be enabled. When CHANGEUPPER event is received, and both PFs (and only them) under the same lag master are about to be enslaved, a warning is returned for user to know the offload failure, otherwise PFs are enslaved silently without LAG offload activated.
--- drivers/net/ethernet/mellanox/mlx5/core/lag.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lag.c +11 -6
@@ -355,7 +355,7 @@ static int mlx5_handle_changeupper_event(struct mlx5_lag *ldev, {struct net_device *upper = info->upper_dev, *ndev_tmp;struct netdev_lag_upper_info *lag_upper_info =NULL ; -bool is_bonded; +bool is_bonded, is_in_lag, mode_supported;int bond_status =0 ;int num_slaves =0 ;int idx;
@@ -391,13 +391,18 @@ static int mlx5_handle_changeupper_event(struct mlx5_lag *ldev,/* Determine bonding status: * A device is considered bonded if both its physical ports are slaves * of the same lag master, and only them. -* Lag mode must be activebackup or hash. */ - is_bonded = (num_slaves ==MLX5_MAX_PORTS ) && - (bond_status ==0x3 ) && - ((tracker->tx_type ==NETDEV_LAG_TX_TYPE_ACTIVEBACKUP ) || - (tracker->tx_type ==NETDEV_LAG_TX_TYPE_HASH )); + is_in_lag = num_slaves ==MLX5_MAX_PORTS && bond_status ==0x3 ; +/* Lag mode must be activebackup or hash. */ + mode_supported = tracker->tx_type ==NETDEV_LAG_TX_TYPE_ACTIVEBACKUP || + tracker->tx_type ==NETDEV_LAG_TX_TYPE_HASH ; + +if (is_in_lag && !mode_supported) +NL_SET_ERR_MSG_MOD (info->info.extack, +"Can't activate LAG offload, TX type isn't supported" ); + + is_bonded = is_in_lag && mode_supported;if (tracker->is_bonded != is_bonded) { tracker->is_bonded = is_bonded;return 1 ;
Signature · 1 line
2.26.2
PATCH 7/16 net/mlx5e: Return a valid errno if can't get lag device index
In reply to [email protected]From: Jianbo Liu <[email protected]>
Change the return value to -ENOENT, to make it more meaningful.
--- drivers/net/ethernet/mellanox/mlx5/core/lag.c | 6 +++--- drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c | 7 ++++++- 2 files changed, 9 insertions(+), 4 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lag.c +3 -3
@@ -102,7 +102,7 @@ int mlx5_lag_dev_get_netdev_idx(struct mlx5_lag *ldev,if (ldev->pf[i].netdev == ndev)return i; -return -1 ; +return -ENOENT ; }static bool __mlx5_lag_is_roce (struct mlx5_lag *ldev)
@@ -374,7 +374,7 @@ static int mlx5_handle_changeupper_event(struct mlx5_lag *ldev,rcu_read_lock ();for_each_netdev_in_bond_rcu (upper, ndev_tmp) { idx =mlx5_lag_dev_get_netdev_idx (ldev, ndev_tmp); -if (idx >-1 ) +if (idx >=0 ) bond_status |= (1 << idx); num_slaves++;
@@ -418,7 +418,7 @@ static int mlx5_handle_changelowerstate_event(struct mlx5_lag *ldev,return 0 ; idx =mlx5_lag_dev_get_netdev_idx (ldev, ndev); -if (idx ==-1 ) +if (idx <0 )return 0 ;/* This information is used to determine virtual to physical
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c +6 -1
@@ -131,7 +131,12 @@ static void mlx5_lag_fib_route_event(struct mlx5_lag *ldev,struct net_device *nh_dev = nh->fib_nh_dev;int i =mlx5_lag_dev_get_netdev_idx (ldev, nh_dev); -mlx5_lag_set_port_affinity (ldev, ++i); +if (i <0 ) + i =MLX5_LAG_NORMAL_AFFINITY ; +else + ++i; + +mlx5_lag_set_port_affinity (ldev, i); }return ; }
Signature · 1 line
2.26.2
PATCH 5/16 net/mlx5: Release clock lock before scheduling a PPS work
In reply to [email protected]From: Eran Ben Elisha <[email protected]>
Holding the clock lock is not required when scheduling a PPS work.
--- drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +1 -1
@@ -581,8 +581,8 @@ static int mlx5_pps_event(struct notifier_block *nb, cycles_delta =div64_u64 (nsec_delta << clock->cycles.shift, clock->cycles.mult); clock->pps_info.start[pin] = cycles_now + cycles_delta; -schedule_work (&clock->pps_info.out_work);write_sequnlock_irqrestore (&clock->lock, flags); +schedule_work (&clock->pps_info.out_work);break ;default : mlx5_core_err(mdev, " Unhandled clock PPS event, func %d\n",
Signature · 1 line
2.26.2
PATCH 13/16 net/mlx5: E-Switch, Use vport metadata matching by default
In reply to [email protected]From: Vu Pham <[email protected]>
Multiple features use metadata matching such as bond vport in live migration, multi-port RoCE mode, stacked devices; hence, enable vport metadata matching by default.
--- .../mellanox/mlx5/core/eswitch_offloads.c | 16 ++-------------- 1 file changed, 2 insertions(+), 14 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c +2 -14
@@ -1864,18 +1864,6 @@ esw_check_vport_match_metadata_supported(const struct mlx5_eswitch *esw)return true; } -static bool -esw_check_vport_match_metadata_mandatory (const struct mlx5_eswitch *esw) -{ -return mlx5_core_mp_enabled (esw->dev); -} - -static bool esw_use_vport_metadata (const struct mlx5_eswitch *esw) -{ -return esw_check_vport_match_metadata_mandatory (esw) && -esw_check_vport_match_metadata_supported (esw); -} -u32 mlx5_esw_match_metadata_alloc (struct mlx5_eswitch *esw) {u32 num_vports =GENMASK (ESW_VPORT_BITS -1 ,0 ) -1 ;
@@ -2159,9 +2147,9 @@ int esw_offloads_enable(struct mlx5_eswitch *esw) err =mlx5_esw_host_number_init (esw);if (err) - goto err_vport_metadata; + goto err_metadata; -if (esw_use_vport_metadata (esw)) +if (esw_check_vport_match_metadata_supported (esw)) esw->flags |=MLX5_ESWITCH_VPORT_MATCH_METADATA ; err =esw_offloads_metadata_init (esw);
Signature · 1 line
2.26.2
PATCH 12/16 net/mlx5: E-Switch, Setup all vports' metadata to support peer miss rule
In reply to [email protected]From: Vu Pham <[email protected]>
In merged eswitch configuration, peer miss rule is setup for all vports. If metadata is enabled, peer miss rule with metadata matching will be configured instead of source port matching; however, some vports that have not yet been enabled don't have default_metadata setup and their default_metadata will be zero.
Hence, setup/cleanup default metadata for all vports when eswitch moves in/out of offloads mode.
--- .../mellanox/mlx5/core/eswitch_offloads.c | 51 +++++++++++++++---- 1 file changed, 42 insertions(+), 9 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c +42 -9
@@ -1923,19 +1923,49 @@ static void esw_offloads_vport_metadata_cleanup(struct mlx5_eswitch *esw,mlx5_esw_match_metadata_free (esw, vport->default_metadata); } +static void esw_offloads_metadata_uninit (struct mlx5_eswitch *esw) +{ +struct mlx5_vport *vport; +int i; + +if (!mlx5_eswitch_vport_match_metadata_enabled (esw)) +return ; + + mlx5_esw_for_all_vports_reverse(esw, i, vport) +esw_offloads_vport_metadata_cleanup (esw, vport); +} + +static int esw_offloads_metadata_init (struct mlx5_eswitch *esw) +{ +struct mlx5_vport *vport; +int err; +int i; + +if (!mlx5_eswitch_vport_match_metadata_enabled (esw)) +return 0 ; + + mlx5_esw_for_all_vports(esw, i, vport) { + err =esw_offloads_vport_metadata_setup (esw, vport); +if (err) + goto metadata_err; + } + +return 0 ; + +metadata_err: +esw_offloads_metadata_uninit (esw); +return err; +} +int esw_vport_create_offloads_acl_tables (struct mlx5_eswitch *esw,struct mlx5_vport *vport) {int err; - err =esw_offloads_vport_metadata_setup (esw, vport); -if (err) - goto metadata_err; - err =esw_acl_ingress_ofld_setup (esw, vport);if (err) - goto ingress_err; +return err;if (mlx5_eswitch_is_vf_vport (esw, vport->vport)) { err =esw_acl_egress_ofld_setup (esw, vport);
@@ -1947,9 +1977,6 @@ esw_vport_create_offloads_acl_tables(struct mlx5_eswitch *esw, egress_err:esw_acl_ingress_ofld_cleanup (esw, vport); -ingress_err: -esw_offloads_vport_metadata_cleanup (esw, vport); -metadata_err:return err; }
@@ -1959,7 +1986,6 @@ esw_vport_destroy_offloads_acl_tables(struct mlx5_eswitch *esw, {esw_acl_egress_ofld_cleanup (vport);esw_acl_ingress_ofld_cleanup (esw, vport); -esw_offloads_vport_metadata_cleanup (esw, vport); }static int esw_create_uplink_offloads_acl_tables (struct mlx5_eswitch *esw)
@@ -2138,6 +2164,10 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)if (esw_use_vport_metadata (esw)) esw->flags |=MLX5_ESWITCH_VPORT_MATCH_METADATA ; + err =esw_offloads_metadata_init (esw); +if (err) + goto err_metadata; + err =esw_set_passing_vport_metadata (esw, true);if (err) goto err_vport_metadata;
@@ -2170,6 +2200,8 @@ int esw_offloads_enable(struct mlx5_eswitch *esw) err_steering_init:esw_set_passing_vport_metadata (esw, false); err_vport_metadata: +esw_offloads_metadata_uninit (esw); +err_metadata: esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ;mlx5_rdma_disable_roce (esw->dev);mutex_destroy (&esw->offloads.termtbl_mutex);
@@ -2204,6 +2236,7 @@ void esw_offloads_disable(struct mlx5_eswitch *esw)esw_offloads_unload_rep (esw,MLX5_VPORT_UPLINK );esw_set_passing_vport_metadata (esw, false);esw_offloads_steering_cleanup (esw); +esw_offloads_metadata_uninit (esw); esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ;mlx5_rdma_disable_roce (esw->dev);mutex_destroy (&esw->offloads.termtbl_mutex);
Signature · 1 line
2.26.2
PATCH 10/16 net/mlx5: E-Switch, Check and enable metadata support flag before using
In reply to [email protected]From: Vu Pham <[email protected]>
Check E-Switch capabilities and enable metadata support flag before using it to setup other features that need metadata.
--- .../mellanox/mlx5/core/eswitch_offloads.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c +6 -9
@@ -1968,16 +1968,9 @@ esw_vport_destroy_offloads_acl_tables(struct mlx5_eswitch *esw,static int esw_create_uplink_offloads_acl_tables (struct mlx5_eswitch *esw) {struct mlx5_vport *vport; -int err; - -if (esw_use_vport_metadata (esw)) - esw->flags |=MLX5_ESWITCH_VPORT_MATCH_METADATA ; vport =mlx5_eswitch_get_vport (esw,MLX5_VPORT_UPLINK ); - err =esw_vport_create_offloads_acl_tables (esw, vport); -if (err) - esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ; -return err; +return esw_vport_create_offloads_acl_tables (esw, vport); }static void esw_destroy_uplink_offloads_acl_tables (struct mlx5_eswitch *esw)
@@ -1986,7 +1979,6 @@ static void esw_destroy_uplink_offloads_acl_tables(struct mlx5_eswitch *esw) vport =mlx5_eswitch_get_vport (esw,MLX5_VPORT_UPLINK );esw_vport_destroy_offloads_acl_tables (esw, vport); - esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ; }static int esw_offloads_steering_init (struct mlx5_eswitch *esw)
@@ -2146,6 +2138,9 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)if (err) goto err_vport_metadata; +if (esw_use_vport_metadata (esw)) + esw->flags |=MLX5_ESWITCH_VPORT_MATCH_METADATA ; + err =esw_set_passing_vport_metadata (esw, true);if (err) goto err_vport_metadata;
@@ -2178,6 +2173,7 @@ int esw_offloads_enable(struct mlx5_eswitch *esw) err_steering_init:esw_set_passing_vport_metadata (esw, false); err_vport_metadata: + esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ;mlx5_rdma_disable_roce (esw->dev);mutex_destroy (&esw->offloads.termtbl_mutex);return err;
@@ -2211,6 +2207,7 @@ void esw_offloads_disable(struct mlx5_eswitch *esw)esw_offloads_unload_rep (esw,MLX5_VPORT_UPLINK );esw_set_passing_vport_metadata (esw, false);esw_offloads_steering_cleanup (esw); + esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA ;mlx5_rdma_disable_roce (esw->dev);mutex_destroy (&esw->offloads.termtbl_mutex); esw->offloads.encap =DEVLINK_ESWITCH_ENCAP_MODE_NONE ;
Signature · 1 line
2.26.2
PATCH 9/16 net/mlx5e: Add LAG warning if bond slave is not lag master
In reply to [email protected]From: Jianbo Liu <[email protected]>
LAG offload can't be enabled if the enslaved PF is not lag master, which is indicated by HCA capabilities bit. It is cleared if more than 64 VFs are configured for this PF.
Previously, a data structure is created to store lag info, including PFs to be enslaved, then a handler is registered for netdev notifier. However, this initialization is skipped if PF is not lag master. So PF can't handle CHANGEUPPER event from upper bond device. Even worse, PF is enslaved silently, and LAG offload is not activated.
Fix this by registering netdev notifier for PFs which are not lag masters. When CHANGEUPPER event is received, and both physical ports (and only them) on the same NIC are about to be enslaved, a warning is returned for user to know it.
--- drivers/net/ethernet/mellanox/mlx5/core/lag.c | 43 ++++++++++++++----- drivers/net/ethernet/mellanox/mlx5/core/lag.h | 7 +++ .../net/ethernet/mellanox/mlx5/core/lag_mp.c | 2 +- 3 files changed, 41 insertions(+), 11 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lag.c +33 -10
@@ -271,7 +271,7 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)bool do_bond, roce_lag;int err; -if (!dev0 || !dev1) +if (!mlx5_lag_is_ready (ldev))return ;spin_lock (&lag_lock);
@@ -394,6 +394,12 @@ static int mlx5_handle_changeupper_event(struct mlx5_lag *ldev,*/ is_in_lag = num_slaves ==MLX5_MAX_PORTS && bond_status ==0x3 ; +if (!mlx5_lag_is_ready (ldev) && is_in_lag) { +NL_SET_ERR_MSG_MOD (info->info.extack, +"Can't activate LAG offload, PF is configured with more than 64 VFs" ); +return 0 ; + } +/* Lag mode must be activebackup or hash. */ mode_supported = tracker->tx_type ==NETDEV_LAG_TX_TYPE_ACTIVEBACKUP || tracker->tx_type ==NETDEV_LAG_TX_TYPE_HASH ;
@@ -450,6 +456,10 @@ static int mlx5_lag_netdev_event(struct notifier_block *this,return NOTIFY_DONE ; ldev = container_of(this, struct mlx5_lag, nb); + +if (!mlx5_lag_is_ready (ldev) && event ==NETDEV_CHANGELOWERSTATE ) +return NOTIFY_DONE ; + tracker = ldev->tracker;switch (event) {
@@ -498,14 +508,14 @@ static void mlx5_lag_dev_free(struct mlx5_lag *ldev)kfree (ldev); } -static void mlx5_lag_dev_add_pf (struct mlx5_lag *ldev, -struct mlx5_core_dev *dev, -struct net_device *netdev) +static int mlx5_lag_dev_add_pf (struct mlx5_lag *ldev, +struct mlx5_core_dev *dev, +struct net_device *netdev) {unsigned int fn =PCI_FUNC (dev->pdev->devfn);if (fn >=MLX5_MAX_PORTS ) -return ; +return -EPERM ;spin_lock (&lag_lock); ldev->pf[fn].dev = dev;
@@ -516,6 +526,8 @@ static void mlx5_lag_dev_add_pf(struct mlx5_lag *ldev, dev->priv.lag = ldev;spin_unlock (&lag_lock); + +return fn; } static void mlx5_lag_dev_remove_pf(struct mlx5_lag *ldev,
@@ -542,11 +554,9 @@ void mlx5_lag_add(struct mlx5_core_dev *dev, struct net_device *netdev) {struct mlx5_lag *ldev =NULL ;struct mlx5_core_dev *tmp_dev; -int err; +int i, err; -if (!MLX5_CAP_GEN (dev, vport_group_manager) || - !MLX5_CAP_GEN (dev, lag_master) || - (MLX5_CAP_GEN (dev, num_lag_ports) !=MLX5_MAX_PORTS )) +if (!MLX5_CAP_GEN (dev, vport_group_manager))return ; tmp_dev =mlx5_get_next_phys_dev (dev);
@@ -561,7 +571,18 @@ void mlx5_lag_add(struct mlx5_core_dev *dev, struct net_device *netdev) } } -mlx5_lag_dev_add_pf (ldev, dev, netdev); +if (mlx5_lag_dev_add_pf (ldev, dev, netdev) <0 ) +return ; + +for (i =0 ; i <MLX5_MAX_PORTS ; i++) { + tmp_dev = ldev->pf[i].dev; +if (!tmp_dev || !MLX5_CAP_GEN (tmp_dev, lag_master) || +MLX5_CAP_GEN (tmp_dev, num_lag_ports) !=MLX5_MAX_PORTS ) +break ; + } + +if (i >=MLX5_MAX_PORTS ) + ldev->flags |=MLX5_LAG_FLAG_READY ;if (!ldev->nb.notifier_call) { ldev->nb.notifier_call = mlx5_lag_netdev_event;
@@ -592,6 +613,8 @@ void mlx5_lag_remove(struct mlx5_core_dev *dev)mlx5_lag_dev_remove_pf (ldev, dev); + ldev->flags &= ~MLX5_LAG_FLAG_READY ; +for (i =0 ; i <MLX5_MAX_PORTS ; i++)if (ldev->pf[i].dev)break ;
drivers/net/ethernet/mellanox/mlx5/core/lag.h +7 -0
@@ -16,6 +16,7 @@ enum {MLX5_LAG_FLAG_ROCE =1 <<0 ,MLX5_LAG_FLAG_SRIOV =1 <<1 ,MLX5_LAG_FLAG_MULTIPATH =1 <<2 , + MLX5_LAG_FLAG_READY = 1 << 3, };#define MLX5_LAG_MODE_FLAGS (MLX5_LAG_FLAG_ROCE | MLX5_LAG_FLAG_SRIOV |\
@@ -59,6 +60,12 @@ __mlx5_lag_is_active(struct mlx5_lag *ldev)return !!(ldev->flags &MLX5_LAG_MODE_FLAGS ); } +static inline bool +mlx5_lag_is_ready (struct mlx5_lag *ldev) +{ +return ldev->flags &MLX5_LAG_FLAG_READY ; +} +void mlx5_modify_lag (struct mlx5_lag *ldev,struct lag_tracker *tracker); int mlx5_activate_lag(struct mlx5_lag *ldev,
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c +1 -1
@@ -11,7 +11,7 @@static bool mlx5_lag_multipath_check_prereq (struct mlx5_lag *ldev) { -if (!ldev->pf[MLX5_LAG_P1 ].dev || !ldev->pf[MLX5_LAG_P2 ].dev) +if (!mlx5_lag_is_ready (ldev))return false; return mlx5_esw_multipath_prereq(ldev->pf[MLX5_LAG_P1].dev,
Signature · 1 line
2.26.2
PATCH 6/16 net/mlx5: Don't call timecounter cyc2time directly from 1PPS flow
In reply to [email protected]From: Eran Ben Elisha <[email protected]>
Before calling timecounter_cyc2time(), clock->lock must be taken. Use mlx5_timecounter_cyc2time instead which guarantees a safe access.
--- drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +3 -2
@@ -557,8 +557,9 @@ static int mlx5_pps_event(struct notifier_block *nb,switch (clock->ptp_info.pin_config[pin].func) {case PTP_PF_EXTTS : ptp_event.index = pin; - ptp_event.timestamp =timecounter_cyc2time (&clock->tc, -be64_to_cpu (eqe->data.pps.time_stamp)); + ptp_event.timestamp = +mlx5_timecounter_cyc2time (clock, +be64_to_cpu (eqe->data.pps.time_stamp));if (clock->pps_info.enabled) { ptp_event.type =PTP_CLOCK_PPSUSR ; ptp_event.pps_times.ts_real =
Signature · 1 line
2.26.2
PATCH 11/16 net/mlx5: E-Switch, Dedicated metadata for uplink vport
In reply to [email protected]From: Vu Pham <[email protected]>
Uplink vport must have a dedicated metadata with vhca_id being part of the metadata.
--- drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c +1 -4
@@ -1908,9 +1908,6 @@ void mlx5_esw_match_metadata_free(struct mlx5_eswitch *esw, u32 metadata)static int esw_offloads_vport_metadata_setup (struct mlx5_eswitch *esw,struct mlx5_vport *vport) { -if (vport->vport ==MLX5_VPORT_UPLINK ) -return 0 ; - vport->default_metadata =mlx5_esw_match_metadata_alloc (esw); vport->metadata = vport->default_metadata;return vport->metadata ?0 : -ENOSPC ;
@@ -1919,7 +1916,7 @@ static int esw_offloads_vport_metadata_setup(struct mlx5_eswitch *esw,static void esw_offloads_vport_metadata_cleanup (struct mlx5_eswitch *esw,struct mlx5_vport *vport) { -if (vport->vport ==MLX5_VPORT_UPLINK || !vport->default_metadata) +if (!vport->default_metadata)return ;WARN_ON (vport->metadata != vport->default_metadata);
Signature · 1 line
2.26.2
PATCH 2/16 net/mlx5: remove erroneous fallthrough
In reply to [email protected]From: Dan Carpenter <[email protected]>
This isn't a fall through because it was after a return statement. The fall through annotation leads to a Smatch warning:
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c:246 mlx5e_ethtool_get_sset_count() warn: ignoring unreachable code.
--- drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c | 1 - 1 file changed, 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +0 -1
@@ -243,7 +243,6 @@ int mlx5e_ethtool_get_sset_count(struct mlx5e_priv *priv, int sset)return MLX5E_NUM_PFLAGS ;case ETH_SS_TEST :return mlx5e_self_test_num (priv); - fallthrough;default :return -EOPNOTSUPP ; }
Signature · 1 line
2.26.2
PATCH 1/16 net/mlx5: Fix uninitialized variable warning
In reply to [email protected]From: Moshe Tal <[email protected]>
Add variable initialization to eliminate the warning "variable may be used uninitialized".
--- drivers/net/ethernet/mellanox/mlx5/core/en/health.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/en/health.c +1 -1
@@ -242,8 +242,8 @@ static int mlx5e_health_rsc_fmsg_binary(struct devlink_fmsg *fmsg, {u32 data_size; +int err =0 ;u32 offset; -int err;for (offset =0 ; offset < value_len; offset += data_size) { data_size = value_len - offset;
Signature · 1 line
2.26.2
PATCH 4/16 net/mlx5: Rename ptp clock info
In reply to [email protected]From: Eran Ben Elisha <[email protected]>
Fix a typo in ptp_clock_info naming: mlx5_p2p -> mlx5_ptp.
--- drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +1 -1
@@ -441,7 +441,7 @@ static int mlx5_ptp_verify(struct ptp_clock_info *ptp, unsigned int pin,static const struct ptp_clock_info mlx5_ptp_clock_info = { .owner =THIS_MODULE , - .name ="mlx5_p2p" , + .name ="mlx5_ptp" , .max_adj =100000000 , .n_alarm =0 , .n_ext_ts =0 ,
Signature · 1 line
2.26.2
Quoted 9 lines · On Tue, 15 Sep 2020 13:25:17 -0700 [email protected] wrote:
From: Saeed Mahameed <[email protected]>
Hi Dave & Jakub,
This series adds some misc updates to mlx5 driver.
For more information please see tag log below.
Please pull and let me know if there is any problem.
I don't really know what the metadata stuff is for but nothing looks alarming so:
That said some of the patches look like they should rather go to net. e.g. 4, 6, 11.. why not?
David Miller
maintainerFrom: [email protected]
Date: Tue, 15 Sep 2020 13:25:17 -0700
Quoted 9 lines
From: Saeed Mahameed <[email protected]>
Hi Dave & Jakub,
This series adds some misc updates to mlx5 driver.
For more information please see tag log below.
Please pull and let me know if there is any problem.
Pulled, thank you.
Quoted 5 lines · Jakub Kicinski
Quoted 4 lines · On Tue, 15 Sep 2020 13:25:17 -0700 [email protected] wrote:
From: Saeed Mahameed <[email protected]>
Hi Dave & Jakub,
Quoted 8 lines
Quoted 5 lines
This series adds some misc updates to mlx5 driver.
For more information please see tag log below.
Please pull and let me know if there is any problem.
I don't really know what the metadata stuff is for but nothing looks alarming so:
In simple words: the NIC steering pipeline knows to look into the packet fields in addition to that it can also provide and match on "metadata" such as, origin function, dest function/port if already determined, and some other well known hw registers..
Quoted 5 lines
Reviewed-by Jakub Kicinski <[email protected]>That said some of the patches look like they should rather go to net. e.g. 4, 6, 11.. why not?
nothing critical really.