#include "mlx5_rxtx.h"
#include "mlx5_common_os.h"
+static_assert(sizeof(struct mlx5_cqe_ts) == sizeof(rte_int128_t),
+ "Wrong timestamp CQE part size");
+
static const char * const mlx5_txpp_stat_names[] = {
- "txpp_err_miss_int", /* Missed service interrupt. */
- "txpp_err_rearm_queue", /* Rearm Queue errors. */
- "txpp_err_clock_queue", /* Clock Queue errors. */
- "txpp_err_ts_past", /* Timestamp in the past. */
- "txpp_err_ts_future", /* Timestamp in the distant future. */
- "txpp_jitter", /* Timestamp jitter (one Clock Queue completion). */
- "txpp_wander", /* Timestamp jitter (half of Clock Queue completions). */
- "txpp_sync_lost", /* Scheduling synchronization lost. */
+ "tx_pp_missed_interrupt_errors", /* Missed service interrupt. */
+ "tx_pp_rearm_queue_errors", /* Rearm Queue errors. */
+ "tx_pp_clock_queue_errors", /* Clock Queue errors. */
+ "tx_pp_timestamp_past_errors", /* Timestamp in the past. */
+ "tx_pp_timestamp_future_errors", /* Timestamp in the distant future. */
+ "tx_pp_jitter", /* Timestamp jitter (one Clock Queue completion). */
+ "tx_pp_wander", /* Timestamp wander (half of Clock Queue CQEs). */
+ "tx_pp_sync_lost", /* Scheduling synchronization lost. */
};
/* Destroy Event Queue Notification Channel. */
static void
-mlx5_txpp_destroy_eqn(struct mlx5_dev_ctx_shared *sh)
+mlx5_txpp_destroy_event_channel(struct mlx5_dev_ctx_shared *sh)
{
if (sh->txpp.echan) {
- mlx5_glue->devx_destroy_event_channel(sh->txpp.echan);
+ mlx5_os_devx_destroy_event_channel(sh->txpp.echan);
sh->txpp.echan = NULL;
}
- sh->txpp.eqn = 0;
}
/* Create Event Queue Notification Channel. */
static int
-mlx5_txpp_create_eqn(struct mlx5_dev_ctx_shared *sh)
+mlx5_txpp_create_event_channel(struct mlx5_dev_ctx_shared *sh)
{
- uint32_t lcore;
-
MLX5_ASSERT(!sh->txpp.echan);
- lcore = (uint32_t)rte_lcore_to_cpu_id(-1);
- if (mlx5_glue->devx_query_eqn(sh->ctx, lcore, &sh->txpp.eqn)) {
- rte_errno = errno;
- DRV_LOG(ERR, "Failed to query EQ number %d.", rte_errno);
- sh->txpp.eqn = 0;
- return -rte_errno;
- }
- sh->txpp.echan = mlx5_glue->devx_create_event_channel(sh->ctx,
+ sh->txpp.echan = mlx5_os_devx_create_event_channel(sh->ctx,
MLX5DV_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA);
if (!sh->txpp.echan) {
- sh->txpp.eqn = 0;
rte_errno = errno;
- DRV_LOG(ERR, "Failed to create event channel %d.",
- rte_errno);
+ DRV_LOG(ERR, "Failed to create event channel %d.", rte_errno);
return -rte_errno;
}
return 0;
static void
mlx5_txpp_free_pp_index(struct mlx5_dev_ctx_shared *sh)
{
+#ifdef HAVE_MLX5DV_PP_ALLOC
if (sh->txpp.pp) {
mlx5_glue->dv_free_pp(sh->txpp.pp);
sh->txpp.pp = NULL;
sh->txpp.pp_id = 0;
}
+#else
+ RTE_SET_USED(sh);
+ DRV_LOG(ERR, "Freeing pacing index is not supported.");
+#endif
}
/* Allocate Packet Pacing index from kernel via mlx5dv call. */
if (wq->sq)
claim_zero(mlx5_devx_cmd_destroy(wq->sq));
if (wq->sq_umem)
- claim_zero(mlx5_glue->devx_umem_dereg(wq->sq_umem));
+ claim_zero(mlx5_os_umem_dereg(wq->sq_umem));
if (wq->sq_buf)
mlx5_free((void *)(uintptr_t)wq->sq_buf);
if (wq->cq)
claim_zero(mlx5_devx_cmd_destroy(wq->cq));
if (wq->cq_umem)
- claim_zero(mlx5_glue->devx_umem_dereg(wq->cq_umem));
+ claim_zero(mlx5_os_umem_dereg(wq->cq_umem));
if (wq->cq_buf)
mlx5_free((void *)(uintptr_t)wq->cq_buf);
memset(wq, 0, sizeof(*wq));
return -ENOMEM;
}
/* Register allocated buffer in user space with DevX. */
- wq->cq_umem = mlx5_glue->devx_umem_reg(sh->ctx,
+ wq->cq_umem = mlx5_os_umem_reg(sh->ctx,
(void *)(uintptr_t)wq->cq_buf,
umem_size,
IBV_ACCESS_LOCAL_WRITE);
cq_attr.cqe_size = (sizeof(struct mlx5_cqe) == 128) ?
MLX5_CQE_SIZE_128B : MLX5_CQE_SIZE_64B;
cq_attr.uar_page_id = mlx5_os_get_devx_uar_page_id(sh->tx_uar);
- cq_attr.eqn = sh->txpp.eqn;
+ cq_attr.eqn = sh->eqn;
cq_attr.q_umem_valid = 1;
cq_attr.q_umem_offset = 0;
cq_attr.q_umem_id = mlx5_os_get_umem_id(wq->cq_umem);
goto error;
}
/* Register allocated buffer in user space with DevX. */
- wq->sq_umem = mlx5_glue->devx_umem_reg(sh->ctx,
+ wq->sq_umem = mlx5_os_umem_reg(sh->ctx,
(void *)(uintptr_t)wq->sq_buf,
umem_size,
IBV_ACCESS_LOCAL_WRITE);
return -ENOMEM;
}
/* Register allocated buffer in user space with DevX. */
- wq->cq_umem = mlx5_glue->devx_umem_reg(sh->ctx,
+ wq->cq_umem = mlx5_os_umem_reg(sh->ctx,
(void *)(uintptr_t)wq->cq_buf,
umem_size,
IBV_ACCESS_LOCAL_WRITE);
cq_attr.use_first_only = 1;
cq_attr.overrun_ignore = 1;
cq_attr.uar_page_id = mlx5_os_get_devx_uar_page_id(sh->tx_uar);
- cq_attr.eqn = sh->txpp.eqn;
+ cq_attr.eqn = sh->eqn;
cq_attr.q_umem_valid = 1;
cq_attr.q_umem_offset = 0;
cq_attr.q_umem_id = mlx5_os_get_umem_id(wq->cq_umem);
goto error;
}
/* Register allocated buffer in user space with DevX. */
- wq->sq_umem = mlx5_glue->devx_umem_reg(sh->ctx,
+ wq->sq_umem = mlx5_os_umem_reg(sh->ctx,
(void *)(uintptr_t)wq->sq_buf,
umem_size,
IBV_ACCESS_LOCAL_WRITE);
aq->arm_sn++;
}
+#if defined(RTE_ARCH_X86_64)
+static inline int
+mlx5_atomic128_compare_exchange(rte_int128_t *dst,
+ rte_int128_t *exp,
+ const rte_int128_t *src)
+{
+ uint8_t res;
+
+ asm volatile (MPLOCKED
+ "cmpxchg16b %[dst];"
+ " sete %[res]"
+ : [dst] "=m" (dst->val[0]),
+ "=a" (exp->val[0]),
+ "=d" (exp->val[1]),
+ [res] "=r" (res)
+ : "b" (src->val[0]),
+ "c" (src->val[1]),
+ "a" (exp->val[0]),
+ "d" (exp->val[1]),
+ "m" (dst->val[0])
+ : "memory");
+
+ return res;
+}
+#endif
+
static inline void
mlx5_atomic_read_cqe(rte_int128_t *from, rte_int128_t *ts)
{
* update by hardware with soecified rate. We have to
* read timestump and WQE completion index atomically.
*/
-#if defined(RTE_ARCH_X86_64) || defined(RTE_ARCH_ARM64)
+#if defined(RTE_ARCH_X86_64)
rte_int128_t src;
memset(&src, 0, sizeof(src));
*ts = src;
/* if (*from == *ts) *from = *src else *ts = *from; */
- rte_atomic128_cmp_exchange(from, ts, &src, 0,
- __ATOMIC_RELAXED, __ATOMIC_RELAXED);
+ mlx5_atomic128_compare_exchange(from, ts, &src);
#else
- rte_atomic64_t *cqe = (rte_atomic64_t *)from;
+ uint64_t *cqe = (uint64_t *)from;
- /* Power architecture does not support 16B compare-and-swap. */
+ /*
+ * Power architecture does not support 16B compare-and-swap.
+ * ARM implements it in software, code below is more relevant.
+ */
for (;;) {
- int64_t tm, op;
- int64_t *ps;
+ uint64_t tm, op;
+ uint64_t *ps;
rte_compiler_barrier();
- tm = rte_atomic64_read(cqe + 0);
- op = rte_atomic64_read(cqe + 1);
+ tm = __atomic_load_n(cqe + 0, __ATOMIC_RELAXED);
+ op = __atomic_load_n(cqe + 1, __ATOMIC_RELAXED);
rte_compiler_barrier();
- if (tm != rte_atomic64_read(cqe + 0))
+ if (tm != __atomic_load_n(cqe + 0, __ATOMIC_RELAXED))
continue;
- if (op != rte_atomic64_read(cqe + 1))
+ if (op != __atomic_load_n(cqe + 1, __ATOMIC_RELAXED))
continue;
- ps = (int64_t *)ts;
+ ps = (uint64_t *)ts;
ps[0] = tm;
ps[1] = op;
return;
ci = ci << (64 - MLX5_CQ_INDEX_WIDTH);
ci |= (ts << MLX5_CQ_INDEX_WIDTH) >> MLX5_CQ_INDEX_WIDTH;
rte_compiler_barrier();
- rte_atomic64_set(&sh->txpp.ts.ts, ts);
- rte_atomic64_set(&sh->txpp.ts.ci_ts, ci);
+ __atomic_store_n(&sh->txpp.ts.ts, ts, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.ts.ci_ts, ci, __ATOMIC_RELAXED);
rte_wmb();
}
uint64_t ts;
uint16_t ci;
- static_assert(sizeof(struct mlx5_cqe_ts) == sizeof(rte_int128_t),
- "Wrong timestamp CQE part size");
mlx5_atomic_read_cqe((rte_int128_t *)&cqe->timestamp, &to.u128);
if (to.cts.op_own >> 4) {
DRV_LOG(DEBUG, "Clock Queue error sync lost.");
- rte_atomic32_inc(&sh->txpp.err_clock_queue);
+ __atomic_fetch_add(&sh->txpp.err_clock_queue,
+ 1, __ATOMIC_RELAXED);
sh->txpp.sync_lost = 1;
return;
}
sh->txpp.ts_p = 0;
sh->txpp.ts_n = 0;
for (wait = 0; wait < MLX5_TXPP_WAIT_INIT_TS; wait++) {
- struct timespec onems;
-
mlx5_txpp_update_timestamp(sh);
if (wq->sq_ci)
return;
/* Wait one millisecond and try again. */
- onems.tv_sec = 0;
- onems.tv_nsec = NS_PER_S / MS_PER_S;
- nanosleep(&onems, 0);
+ rte_delay_us_sleep(US_PER_S / MS_PER_S);
}
DRV_LOG(ERR, "Unable to initialize timestamp.");
sh->txpp.sync_lost = 1;
if (!sh->txpp.clock_queue.sq_ci && !sh->txpp.ts_n)
return;
MLX5_ASSERT(sh->txpp.ts_p < MLX5_TXPP_REARM_SQ_SIZE);
- sh->txpp.tsa[sh->txpp.ts_p] = sh->txpp.ts;
+ __atomic_store_n(&sh->txpp.tsa[sh->txpp.ts_p].ts,
+ sh->txpp.ts.ts, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.tsa[sh->txpp.ts_p].ci_ts,
+ sh->txpp.ts.ci_ts, __ATOMIC_RELAXED);
if (++sh->txpp.ts_p >= MLX5_TXPP_REARM_SQ_SIZE)
sh->txpp.ts_p = 0;
if (sh->txpp.ts_n < MLX5_TXPP_REARM_SQ_SIZE)
/* Check whether we have missed interrupts. */
if (cq_ci - wq->cq_ci != 1) {
DRV_LOG(DEBUG, "Rearm Queue missed interrupt.");
- rte_atomic32_inc(&sh->txpp.err_miss_int);
+ __atomic_fetch_add(&sh->txpp.err_miss_int,
+ 1, __ATOMIC_RELAXED);
/* Check sync lost on wqe index. */
if (cq_ci - wq->cq_ci >=
(((1UL << MLX5_WQ_INDEX_WIDTH) /
/* Fire new requests to Rearm Queue. */
if (error) {
DRV_LOG(DEBUG, "Rearm Queue error sync lost.");
- rte_atomic32_inc(&sh->txpp.err_rearm_queue);
+ __atomic_fetch_add(&sh->txpp.err_rearm_queue,
+ 1, __ATOMIC_RELAXED);
sh->txpp.sync_lost = 1;
}
}
int ret;
int fd;
- rte_atomic32_set(&sh->txpp.err_miss_int, 0);
- rte_atomic32_set(&sh->txpp.err_rearm_queue, 0);
- rte_atomic32_set(&sh->txpp.err_clock_queue, 0);
- rte_atomic32_set(&sh->txpp.err_ts_past, 0);
- rte_atomic32_set(&sh->txpp.err_ts_future, 0);
+ sh->txpp.err_miss_int = 0;
+ sh->txpp.err_rearm_queue = 0;
+ sh->txpp.err_clock_queue = 0;
+ sh->txpp.err_ts_past = 0;
+ sh->txpp.err_ts_future = 0;
/* Attach interrupt handler to process Rearm Queue completions. */
fd = mlx5_os_get_devx_channel_fd(sh->txpp.echan);
ret = mlx5_os_set_nonblock_channel_fd(fd);
return -rte_errno;
}
/* Subscribe CQ event to the event channel controlled by the driver. */
- ret = mlx5_glue->devx_subscribe_devx_event(sh->txpp.echan,
+ ret = mlx5_os_devx_subscribe_devx_event(sh->txpp.echan,
sh->txpp.rearm_queue.cq->obj,
sizeof(event_nums),
event_nums, 0);
sh->txpp.test = !!(tx_pp < 0);
sh->txpp.skew = priv->config.tx_skew;
sh->txpp.freq = priv->config.hca_attr.dev_freq_khz;
- ret = mlx5_txpp_create_eqn(sh);
+ ret = mlx5_txpp_create_event_channel(sh);
if (ret)
goto exit;
ret = mlx5_txpp_alloc_pp_index(sh);
mlx5_txpp_destroy_rearm_queue(sh);
mlx5_txpp_destroy_clock_queue(sh);
mlx5_txpp_free_pp_index(sh);
- mlx5_txpp_destroy_eqn(sh);
+ mlx5_txpp_destroy_event_channel(sh);
sh->txpp.tick = 0;
sh->txpp.test = 0;
sh->txpp.skew = 0;
mlx5_txpp_destroy_rearm_queue(sh);
mlx5_txpp_destroy_clock_queue(sh);
mlx5_txpp_free_pp_index(sh);
- mlx5_txpp_destroy_eqn(sh);
+ mlx5_txpp_destroy_event_channel(sh);
sh->txpp.tick = 0;
sh->txpp.test = 0;
sh->txpp.skew = 0;
mlx5_atomic_read_cqe((rte_int128_t *)&cqe->timestamp, &to.u128);
if (to.cts.op_own >> 4) {
DRV_LOG(DEBUG, "Clock Queue error sync lost.");
- rte_atomic32_inc(&sh->txpp.err_clock_queue);
+ __atomic_fetch_add(&sh->txpp.err_clock_queue,
+ 1, __ATOMIC_RELAXED);
sh->txpp.sync_lost = 1;
return -EIO;
}
struct mlx5_priv *priv = dev->data->dev_private;
struct mlx5_dev_ctx_shared *sh = priv->sh;
- rte_atomic32_set(&sh->txpp.err_miss_int, 0);
- rte_atomic32_set(&sh->txpp.err_rearm_queue, 0);
- rte_atomic32_set(&sh->txpp.err_clock_queue, 0);
- rte_atomic32_set(&sh->txpp.err_ts_past, 0);
- rte_atomic32_set(&sh->txpp.err_ts_future, 0);
+ __atomic_store_n(&sh->txpp.err_miss_int, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.err_rearm_queue, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.err_clock_queue, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.err_ts_past, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&sh->txpp.err_ts_future, 0, __ATOMIC_RELAXED);
return 0;
}
struct mlx5_txpp_ts *tsa, uint16_t idx)
{
do {
- int64_t ts, ci;
+ uint64_t ts, ci;
- ts = rte_atomic64_read(&txpp->tsa[idx].ts);
- ci = rte_atomic64_read(&txpp->tsa[idx].ci_ts);
+ ts = __atomic_load_n(&txpp->tsa[idx].ts, __ATOMIC_RELAXED);
+ ci = __atomic_load_n(&txpp->tsa[idx].ci_ts, __ATOMIC_RELAXED);
rte_compiler_barrier();
if ((ci ^ ts) << MLX5_CQ_INDEX_WIDTH != 0)
continue;
- if (rte_atomic64_read(&txpp->tsa[idx].ts) != ts)
+ if (__atomic_load_n(&txpp->tsa[idx].ts,
+ __ATOMIC_RELAXED) != ts)
continue;
- if (rte_atomic64_read(&txpp->tsa[idx].ci_ts) != ci)
+ if (__atomic_load_n(&txpp->tsa[idx].ci_ts,
+ __ATOMIC_RELAXED) != ci)
continue;
- rte_atomic64_set(&tsa->ts, ts);
- rte_atomic64_set(&tsa->ci_ts, ci);
+ tsa->ts = ts;
+ tsa->ci_ts = ci;
return;
} while (true);
}
rte_compiler_barrier();
} while (ts_p != txpp->ts_p);
/* We have two neighbor reports, calculate the jitter. */
- dts = rte_atomic64_read(&tsa1.ts) - rte_atomic64_read(&tsa0.ts);
- dci = (rte_atomic64_read(&tsa1.ci_ts) >> (64 - MLX5_CQ_INDEX_WIDTH)) -
- (rte_atomic64_read(&tsa0.ci_ts) >> (64 - MLX5_CQ_INDEX_WIDTH));
+ dts = tsa1.ts - tsa0.ts;
+ dci = (tsa1.ci_ts >> (64 - MLX5_CQ_INDEX_WIDTH)) -
+ (tsa0.ci_ts >> (64 - MLX5_CQ_INDEX_WIDTH));
if (dci < 0)
dci += 1 << MLX5_CQ_INDEX_WIDTH;
dci *= txpp->tick;
rte_compiler_barrier();
} while (ts_p != txpp->ts_p);
/* We have two neighbor reports, calculate the jitter. */
- dts = rte_atomic64_read(&tsa1.ts) - rte_atomic64_read(&tsa0.ts);
- dci = (rte_atomic64_read(&tsa1.ci_ts) >> (64 - MLX5_CQ_INDEX_WIDTH)) -
- (rte_atomic64_read(&tsa0.ci_ts) >> (64 - MLX5_CQ_INDEX_WIDTH));
+ dts = tsa1.ts - tsa0.ts;
+ dci = (tsa1.ci_ts >> (64 - MLX5_CQ_INDEX_WIDTH)) -
+ (tsa0.ci_ts >> (64 - MLX5_CQ_INDEX_WIDTH));
dci += 1 << MLX5_CQ_INDEX_WIDTH;
dci *= txpp->tick;
return (dts > dci) ? dts - dci : dci - dts;
for (i = 0; i < n_txpp; ++i)
stats[n_used + i].id = n_used + i;
stats[n_used + 0].value =
- rte_atomic32_read(&sh->txpp.err_miss_int);
+ __atomic_load_n(&sh->txpp.err_miss_int,
+ __ATOMIC_RELAXED);
stats[n_used + 1].value =
- rte_atomic32_read(&sh->txpp.err_rearm_queue);
+ __atomic_load_n(&sh->txpp.err_rearm_queue,
+ __ATOMIC_RELAXED);
stats[n_used + 2].value =
- rte_atomic32_read(&sh->txpp.err_clock_queue);
+ __atomic_load_n(&sh->txpp.err_clock_queue,
+ __ATOMIC_RELAXED);
stats[n_used + 3].value =
- rte_atomic32_read(&sh->txpp.err_ts_past);
+ __atomic_load_n(&sh->txpp.err_ts_past,
+ __ATOMIC_RELAXED);
stats[n_used + 4].value =
- rte_atomic32_read(&sh->txpp.err_ts_future);
+ __atomic_load_n(&sh->txpp.err_ts_future,
+ __ATOMIC_RELAXED);
stats[n_used + 5].value = mlx5_txpp_xstats_jitter(&sh->txpp);
stats[n_used + 6].value = mlx5_txpp_xstats_wander(&sh->txpp);
stats[n_used + 7].value = sh->txpp.sync_lost;