Add support for report and recovery from error on completion on ICOSQ.
Deactivate RQ and flush, then deactivate ICOSQ. Set the queue back to
ready state (firmware) and reset the ICOSQ and the RQ (software
resources). Finally, activate the ICOSQ and the RQ.
Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
/* control path */
struct mlx5_wq_ctrl wq_ctrl;
struct mlx5e_channel *channel;
+
+ struct work_struct recover_work;
} ____cacheline_aligned_in_smp;
struct mlx5e_wqe_frag_info {
void mlx5e_set_rq_type(struct mlx5_core_dev *mdev, struct mlx5e_params *params);
void mlx5e_init_rq_type_params(struct mlx5_core_dev *mdev,
struct mlx5e_params *params);
+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_activate_icosq(struct mlx5e_icosq *icosq);
+void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq);
int mlx5e_modify_sq(struct mlx5_core_dev *mdev, u32 sqn,
struct mlx5e_modify_sq_param *p);
int mlx5e_reporter_rx_create(struct mlx5e_priv *priv);
void mlx5e_reporter_rx_destroy(struct mlx5e_priv *priv);
+void mlx5e_reporter_icosq_cqe_err(struct mlx5e_icosq *icosq);
#define MLX5E_REPORTER_PER_Q_MAX_LEN 256
return err;
}
+static int mlx5e_wait_for_icosq_flush(struct mlx5e_icosq *icosq)
+{
+ unsigned long exp_time = jiffies + msecs_to_jiffies(2000);
+
+ while (time_before(jiffies, exp_time)) {
+ if (icosq->cc == icosq->pc)
+ return 0;
+
+ msleep(20);
+ }
+
+ netdev_err(icosq->channel->netdev,
+ "Wait for ICOSQ 0x%x flush timeout (cc = 0x%x, pc = 0x%x)\n",
+ icosq->sqn, icosq->cc, icosq->pc);
+
+ return -ETIMEDOUT;
+}
+
+static void mlx5e_reset_icosq_cc_pc(struct mlx5e_icosq *icosq)
+{
+ WARN_ONCE(icosq->cc != icosq->pc, "ICOSQ 0x%x: cc (0x%x) != pc (0x%x)\n",
+ icosq->sqn, icosq->cc, icosq->pc);
+ icosq->cc = 0;
+ icosq->pc = 0;
+}
+
+static int mlx5e_rx_reporter_err_icosq_cqe_recover(void *ctx)
+{
+ struct mlx5_core_dev *mdev;
+ struct mlx5e_icosq *icosq;
+ struct net_device *dev;
+ struct mlx5e_rq *rq;
+ u8 state;
+ int err;
+
+ icosq = ctx;
+ rq = &icosq->channel->rq;
+ mdev = icosq->channel->mdev;
+ dev = icosq->channel->netdev;
+ err = mlx5_core_query_sq_state(mdev, icosq->sqn, &state);
+ if (err) {
+ netdev_err(dev, "Failed to query ICOSQ 0x%x state. err = %d\n",
+ icosq->sqn, err);
+ goto out;
+ }
+
+ if (state != MLX5_SQC_STATE_ERR)
+ goto out;
+
+ mlx5e_deactivate_rq(rq);
+ err = mlx5e_wait_for_icosq_flush(icosq);
+ if (err)
+ goto out;
+
+ mlx5e_deactivate_icosq(icosq);
+
+ /* At this point, both the rq and the icosq are disabled */
+
+ err = mlx5e_health_sq_to_ready(icosq->channel, icosq->sqn);
+ if (err)
+ goto out;
+
+ mlx5e_reset_icosq_cc_pc(icosq);
+ mlx5e_free_rx_descs(rq);
+ clear_bit(MLX5E_SQ_STATE_RECOVERING, &icosq->state);
+ mlx5e_activate_icosq(icosq);
+ mlx5e_activate_rq(rq);
+
+ rq->stats->recover++;
+ return 0;
+out:
+ clear_bit(MLX5E_SQ_STATE_RECOVERING, &icosq->state);
+ return err;
+}
+
+void mlx5e_reporter_icosq_cqe_err(struct mlx5e_icosq *icosq)
+{
+ struct mlx5e_priv *priv = icosq->channel->priv;
+ char err_str[MLX5E_REPORTER_PER_Q_MAX_LEN];
+ struct mlx5e_err_ctx err_ctx = {};
+
+ err_ctx.ctx = icosq;
+ err_ctx.recover = mlx5e_rx_reporter_err_icosq_cqe_recover;
+ sprintf(err_str, "ERR CQE on ICOSQ: 0x%x", icosq->sqn);
+
+ mlx5e_health_report(priv, priv->rx_reporter, err_str, &err_ctx);
+}
+
+static int mlx5e_rx_reporter_recover_from_ctx(struct mlx5e_err_ctx *err_ctx)
+{
+ return err_ctx->recover(err_ctx->ctx);
+}
+
+static int mlx5e_rx_reporter_recover(struct devlink_health_reporter *reporter,
+ void *context)
+{
+ struct mlx5e_priv *priv = devlink_health_reporter_priv(reporter);
+ struct mlx5e_err_ctx *err_ctx = context;
+
+ return err_ctx ? mlx5e_rx_reporter_recover_from_ctx(err_ctx) :
+ mlx5e_health_recover_channels(priv);
+}
+
static int mlx5e_rx_reporter_build_diagnose_output(struct mlx5e_rq *rq,
struct devlink_fmsg *fmsg)
{
static const struct devlink_health_reporter_ops mlx5_rx_reporter_ops = {
.name = "rx",
+ .recover = mlx5e_rx_reporter_recover,
.diagnose = mlx5e_rx_reporter_diagnose,
};
+#define MLX5E_REPORTER_RX_GRACEFUL_PERIOD 500
+
int mlx5e_reporter_rx_create(struct mlx5e_priv *priv)
{
struct devlink *devlink = priv_to_devlink(priv->mdev);
reporter = devlink_health_reporter_create(devlink,
&mlx5_rx_reporter_ops,
- 0, false, priv);
+ MLX5E_REPORTER_RX_GRACEFUL_PERIOD,
+ true, priv);
if (IS_ERR(reporter)) {
netdev_warn(priv->netdev, "Failed to create rx reporter, err = %ld\n",
PTR_ERR(reporter));
return err;
}
-static int mlx5e_modify_rq_state(struct mlx5e_rq *rq, int curr_state,
- int next_state)
+int mlx5e_modify_rq_state(struct mlx5e_rq *rq, int curr_state, int next_state)
{
struct mlx5_core_dev *mdev = rq->mdev;
return -ETIMEDOUT;
}
-static void mlx5e_free_rx_descs(struct mlx5e_rq *rq)
+void mlx5e_free_rx_descs(struct mlx5e_rq *rq)
{
__be16 wqe_ix_be;
u16 wqe_ix;
return err;
}
-static void mlx5e_activate_rq(struct mlx5e_rq *rq)
+void mlx5e_activate_rq(struct mlx5e_rq *rq)
{
set_bit(MLX5E_RQ_STATE_ENABLED, &rq->state);
mlx5e_trigger_irq(&rq->channel->icosq);
void mlx5e_close_rq(struct mlx5e_rq *rq)
{
cancel_work_sync(&rq->dim.work);
+ cancel_work_sync(&rq->channel->icosq.recover_work);
mlx5e_destroy_rq(rq);
mlx5e_free_rx_descs(rq);
mlx5e_free_rq(rq);
return 0;
}
+static void mlx5e_icosq_err_cqe_work(struct work_struct *recover_work)
+{
+ struct mlx5e_icosq *sq = container_of(recover_work, struct mlx5e_icosq,
+ recover_work);
+
+ mlx5e_reporter_icosq_cqe_err(sq);
+}
+
static int mlx5e_alloc_icosq(struct mlx5e_channel *c,
struct mlx5e_sq_param *param,
struct mlx5e_icosq *sq)
if (err)
goto err_sq_wq_destroy;
+ INIT_WORK(&sq->recover_work, mlx5e_icosq_err_cqe_work);
+
return 0;
err_sq_wq_destroy:
return err;
}
-static void mlx5e_activate_icosq(struct mlx5e_icosq *icosq)
+void mlx5e_activate_icosq(struct mlx5e_icosq *icosq)
{
set_bit(MLX5E_SQ_STATE_ENABLED, &icosq->state);
}
-static void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq)
+void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq)
{
struct mlx5e_channel *c = icosq->channel;
if (unlikely(get_cqe_opcode(cqe) != MLX5_CQE_REQ)) {
netdev_WARN_ONCE(cq->channel->netdev,
"Bad OP in ICOSQ CQE: 0x%x\n", get_cqe_opcode(cqe));
+ if (!test_and_set_bit(MLX5E_SQ_STATE_RECOVERING, &sq->state))
+ queue_work(cq->channel->priv->wq, &sq->recover_work);
break;
}
do {
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_cache_waive) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_congst_umr) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_arfs_err) },
+ { MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_recover) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, ch_events) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, ch_poll) },
{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, ch_arm) },
s->rx_cache_waive += rq_stats->cache_waive;
s->rx_congst_umr += rq_stats->congst_umr;
s->rx_arfs_err += rq_stats->arfs_err;
+ s->rx_recover += rq_stats->recover;
s->ch_events += ch_stats->events;
s->ch_poll += ch_stats->poll;
s->ch_arm += ch_stats->arm;
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, cache_waive) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, congst_umr) },
{ MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, arfs_err) },
+ { MLX5E_DECLARE_RX_STAT(struct mlx5e_rq_stats, recover) },
};
static const struct counter_desc sq_stats_desc[] = {
u64 rx_cache_waive;
u64 rx_congst_umr;
u64 rx_arfs_err;
+ u64 rx_recover;
u64 ch_events;
u64 ch_poll;
u64 ch_arm;
u64 cache_waive;
u64 congst_umr;
u64 arfs_err;
+ u64 recover;
};
struct mlx5e_sq_stats {