From: Raja Zidane <rzidane@nvidia.com>
To: <dev@dpdk.org>
Cc: Matan Azrad <matan@nvidia.com>, <stable@dpdk.org>
Subject: [dpdk-dev] [PATCH V2 3/4] crypto/mlx5: fix the queue size configuration
Date: Mon, 8 Nov 2021 13:09:20 +0000 [thread overview]
Message-ID: <20211108130921.19143-4-rzidane@nvidia.com> (raw)
In-Reply-To: <20211108130921.19143-1-rzidane@nvidia.com>
The DevX interface for QP creation expects the number of WQEBBs.
Wrongly, the number of descriptors was provided to the QP creation.
In addition, the QP size must be a power of 2 what was not guaranteed.
Provide the number of WQEBBs to the QP creation API.
Round up the SQ size to a power of 2.
Rename (sq/rq)_size to num_of_(send/receive)_wqes.
Fixes: 6152534e211e ("crypto/mlx5: support queue pairs operations")
Cc: stable@dpdk.org
Signed-off-by: Raja Zidane <rzidane@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
---
drivers/common/mlx5/mlx5_devx_cmds.c | 14 +--
drivers/common/mlx5/mlx5_devx_cmds.h | 5 +-
drivers/compress/mlx5/mlx5_compress.c | 4 +-
drivers/crypto/mlx5/mlx5_crypto.c | 120 +++++++++++++++++++-----
drivers/crypto/mlx5/mlx5_crypto.h | 7 ++
drivers/regex/mlx5/mlx5_regex_control.c | 4 +-
drivers/vdpa/mlx5/mlx5_vdpa_event.c | 4 +-
7 files changed, 120 insertions(+), 38 deletions(-)
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c b/drivers/common/mlx5/mlx5_devx_cmds.c
index cecbf541f6..e52b995ee3 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.c
+++ b/drivers/common/mlx5/mlx5_devx_cmds.c
@@ -832,6 +832,7 @@ mlx5_devx_cmd_query_hca_attr(void *ctx,
MLX5_HCA_CAP_OPMOD_GET_CUR);
if (!hcattr)
return rc;
+ attr->max_wqe_sz_sq = MLX5_GET(cmd_hca_cap, hcattr, max_wqe_sz_sq);
attr->flow_counter_bulk_alloc_bitmap =
MLX5_GET(cmd_hca_cap, hcattr, flow_counter_bulk_alloc);
attr->flow_counters_dump = MLX5_GET(cmd_hca_cap, hcattr,
@@ -2153,21 +2154,22 @@ mlx5_devx_cmd_create_qp(void *ctx,
if (attr->log_page_size > MLX5_ADAPTER_PAGE_SHIFT)
MLX5_SET(qpc, qpc, log_page_size,
attr->log_page_size - MLX5_ADAPTER_PAGE_SHIFT);
- if (attr->sq_size) {
- MLX5_ASSERT(RTE_IS_POWER_OF_2(attr->sq_size));
+ if (attr->num_of_send_wqbbs) {
+ MLX5_ASSERT(RTE_IS_POWER_OF_2(attr->num_of_send_wqbbs));
MLX5_SET(qpc, qpc, cqn_snd, attr->cqn);
MLX5_SET(qpc, qpc, log_sq_size,
- rte_log2_u32(attr->sq_size));
+ rte_log2_u32(attr->num_of_send_wqbbs));
} else {
MLX5_SET(qpc, qpc, no_sq, 1);
}
- if (attr->rq_size) {
- MLX5_ASSERT(RTE_IS_POWER_OF_2(attr->rq_size));
+ if (attr->num_of_receive_wqes) {
+ MLX5_ASSERT(RTE_IS_POWER_OF_2(
+ attr->num_of_receive_wqes));
MLX5_SET(qpc, qpc, cqn_rcv, attr->cqn);
MLX5_SET(qpc, qpc, log_rq_stride, attr->log_rq_stride -
MLX5_LOG_RQ_STRIDE_SHIFT);
MLX5_SET(qpc, qpc, log_rq_size,
- rte_log2_u32(attr->rq_size));
+ rte_log2_u32(attr->num_of_receive_wqes));
MLX5_SET(qpc, qpc, rq_type, MLX5_NON_ZERO_RQ);
} else {
MLX5_SET(qpc, qpc, rq_type, MLX5_ZERO_LEN_RQ);
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h b/drivers/common/mlx5/mlx5_devx_cmds.h
index 447f76f1f9..d7f71646a3 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.h
+++ b/drivers/common/mlx5/mlx5_devx_cmds.h
@@ -251,6 +251,7 @@ struct mlx5_hca_attr {
uint32_t log_max_mmo_decompress:5;
uint32_t umr_modify_entity_size_disabled:1;
uint32_t umr_indirect_mkey_disabled:1;
+ uint16_t max_wqe_sz_sq;
};
/* LAG Context. */
@@ -477,9 +478,9 @@ struct mlx5_devx_qp_attr {
uint32_t uar_index:24;
uint32_t cqn:24;
uint32_t log_page_size:5;
- uint32_t rq_size:17; /* Must be power of 2. */
+ uint32_t num_of_receive_wqes:17; /* Must be power of 2. */
uint32_t log_rq_stride:3;
- uint32_t sq_size:17; /* Must be power of 2. */
+ uint32_t num_of_send_wqbbs:17; /* Must be power of 2. */
uint32_t ts_format:2;
uint32_t dbr_umem_valid:1;
uint32_t dbr_umem_id;
diff --git a/drivers/compress/mlx5/mlx5_compress.c b/drivers/compress/mlx5/mlx5_compress.c
index d5511aebdf..7813af38e6 100644
--- a/drivers/compress/mlx5/mlx5_compress.c
+++ b/drivers/compress/mlx5/mlx5_compress.c
@@ -244,8 +244,8 @@ mlx5_compress_qp_setup(struct rte_compressdev *dev, uint16_t qp_id,
qp_attr.cqn = qp->cq.cq->id;
qp_attr.ts_format =
mlx5_ts_format_conv(priv->cdev->config.hca_attr.qp_ts_format);
- qp_attr.rq_size = 0;
- qp_attr.sq_size = RTE_BIT32(log_ops_n);
+ qp_attr.num_of_receive_wqes = 0;
+ qp_attr.num_of_send_wqbbs = RTE_BIT32(log_ops_n);
qp_attr.mmo = priv->mmo_decomp_qp && priv->mmo_comp_qp
&& priv->mmo_dma_qp;
ret = mlx5_devx_qp_create(priv->cdev->ctx, &qp->qp, log_ops_n, &qp_attr,
diff --git a/drivers/crypto/mlx5/mlx5_crypto.c b/drivers/crypto/mlx5/mlx5_crypto.c
index 1d0f1f3cfc..9fdbee9be1 100644
--- a/drivers/crypto/mlx5/mlx5_crypto.c
+++ b/drivers/crypto/mlx5/mlx5_crypto.c
@@ -545,7 +545,7 @@ mlx5_crypto_qp_init(struct mlx5_crypto_priv *priv, struct mlx5_crypto_qp *qp)
ucseg->if_cf_toe_cq_res = RTE_BE32(1u << MLX5_UMRC_IF_OFFSET);
ucseg->mkey_mask = RTE_BE64(1u << 0); /* Mkey length bit. */
ucseg->ko_to_bs = rte_cpu_to_be_32
- ((RTE_ALIGN(priv->max_segs_num, 4u) <<
+ ((MLX5_CRYPTO_KLM_SEGS_NUM(priv->umr_wqe_size) <<
MLX5_UMRC_KO_OFFSET) | (4 << MLX5_UMRC_TO_BS_OFFSET));
bsf->keytag = priv->keytag;
/* Init RDMA WRITE WQE. */
@@ -569,7 +569,7 @@ mlx5_crypto_indirect_mkeys_prepare(struct mlx5_crypto_priv *priv,
.umr_en = 1,
.crypto_en = 1,
.set_remote_rw = 1,
- .klm_num = RTE_ALIGN(priv->max_segs_num, 4),
+ .klm_num = MLX5_CRYPTO_KLM_SEGS_NUM(priv->umr_wqe_size),
};
for (umr = (struct mlx5_umr_wqe *)qp->qp_obj.umem_buf, i = 0;
@@ -597,6 +597,7 @@ mlx5_crypto_queue_pair_setup(struct rte_cryptodev *dev, uint16_t qp_id,
uint16_t log_nb_desc = rte_log2_u32(qp_conf->nb_descriptors);
uint32_t ret;
uint32_t alloc_size = sizeof(*qp);
+ uint32_t log_wqbb_n;
struct mlx5_devx_cq_attr cq_attr = {
.uar_page_id = mlx5_os_get_devx_uar_page_id(priv->uar.obj),
};
@@ -619,14 +620,16 @@ mlx5_crypto_queue_pair_setup(struct rte_cryptodev *dev, uint16_t qp_id,
DRV_LOG(ERR, "Failed to create CQ.");
goto error;
}
+ log_wqbb_n = rte_log2_u32(RTE_BIT32(log_nb_desc) *
+ (priv->wqe_set_size / MLX5_SEND_WQE_BB));
attr.pd = priv->cdev->pdn;
attr.uar_index = mlx5_os_get_devx_uar_page_id(priv->uar.obj);
attr.cqn = qp->cq_obj.cq->id;
- attr.rq_size = 0;
- attr.sq_size = RTE_BIT32(log_nb_desc);
+ attr.num_of_receive_wqes = 0;
+ attr.num_of_send_wqbbs = RTE_BIT32(log_wqbb_n);
attr.ts_format =
mlx5_ts_format_conv(priv->cdev->config.hca_attr.qp_ts_format);
- ret = mlx5_devx_qp_create(priv->cdev->ctx, &qp->qp_obj, log_nb_desc,
+ ret = mlx5_devx_qp_create(priv->cdev->ctx, &qp->qp_obj, log_wqbb_n,
&attr, socket_id);
if (ret) {
DRV_LOG(ERR, "Failed to create QP.");
@@ -747,10 +750,8 @@ mlx5_crypto_args_check_handler(const char *key, const char *val, void *opaque)
return -errno;
}
if (strcmp(key, "max_segs_num") == 0) {
- if (!tmp || tmp > MLX5_CRYPTO_MAX_SEGS) {
- DRV_LOG(WARNING, "Invalid max_segs_num: %d, should"
- " be less than %d.",
- (uint32_t)tmp, MLX5_CRYPTO_MAX_SEGS);
+ if (!tmp) {
+ DRV_LOG(ERR, "max_segs_num must be greater than 0.");
rte_errno = EINVAL;
return -rte_errno;
}
@@ -809,6 +810,81 @@ mlx5_crypto_parse_devargs(struct rte_devargs *devargs,
return 0;
}
+/*
+ * Calculate UMR WQE size and RDMA Write WQE size with the
+ * following limitations:
+ * - Each WQE size is multiple of 64.
+ * - The summarize of both UMR WQE and RDMA_W WQE is a power of 2.
+ * - The number of entries in the UMR WQE's KLM list is multiple of 4.
+ */
+static void
+mlx5_crypto_get_wqe_sizes(uint32_t segs_num, uint32_t *umr_size,
+ uint32_t *rdmaw_size)
+{
+ uint32_t diff, wqe_set_size;
+
+ *umr_size = MLX5_CRYPTO_UMR_WQE_STATIC_SIZE +
+ RTE_ALIGN(segs_num, 4) *
+ sizeof(struct mlx5_wqe_dseg);
+ /* Make sure UMR WQE size is multiple of WQBB. */
+ *umr_size = RTE_ALIGN(*umr_size, MLX5_SEND_WQE_BB);
+ *rdmaw_size = sizeof(struct mlx5_rdma_write_wqe) +
+ sizeof(struct mlx5_wqe_dseg) *
+ (segs_num <= 2 ? 2 : 2 +
+ RTE_ALIGN(segs_num - 2, 4));
+ /* Make sure RDMA_WRITE WQE size is multiple of WQBB. */
+ *rdmaw_size = RTE_ALIGN(*rdmaw_size, MLX5_SEND_WQE_BB);
+ wqe_set_size = *rdmaw_size + *umr_size;
+ diff = rte_align32pow2(wqe_set_size) - wqe_set_size;
+ /* Make sure wqe_set size is power of 2. */
+ if (diff)
+ *umr_size += diff;
+}
+
+static uint8_t
+mlx5_crypto_max_segs_num(uint16_t max_wqe_size)
+{
+ int klms_sizes = max_wqe_size - MLX5_CRYPTO_UMR_WQE_STATIC_SIZE;
+ uint32_t max_segs_cap = RTE_ALIGN_FLOOR(klms_sizes, MLX5_SEND_WQE_BB) /
+ sizeof(struct mlx5_wqe_dseg);
+
+ MLX5_ASSERT(klms_sizes >= MLX5_SEND_WQE_BB);
+ while (max_segs_cap) {
+ uint32_t umr_wqe_size, rdmw_wqe_size;
+
+ mlx5_crypto_get_wqe_sizes(max_segs_cap, &umr_wqe_size,
+ &rdmw_wqe_size);
+ if (umr_wqe_size <= max_wqe_size &&
+ rdmw_wqe_size <= max_wqe_size)
+ break;
+ max_segs_cap -= 4;
+ }
+ return max_segs_cap;
+}
+
+static int
+mlx5_crypto_configure_wqe_size(struct mlx5_crypto_priv *priv,
+ uint16_t max_wqe_size, uint32_t max_segs_num)
+{
+ uint32_t rdmw_wqe_size, umr_wqe_size;
+
+ mlx5_crypto_get_wqe_sizes(max_segs_num, &umr_wqe_size,
+ &rdmw_wqe_size);
+ priv->wqe_set_size = rdmw_wqe_size + umr_wqe_size;
+ if (umr_wqe_size > max_wqe_size ||
+ rdmw_wqe_size > max_wqe_size) {
+ DRV_LOG(ERR, "Invalid max_segs_num: %u. should be %u or lower.",
+ max_segs_num,
+ mlx5_crypto_max_segs_num(max_wqe_size));
+ rte_errno = EINVAL;
+ return -EINVAL;
+ }
+ priv->umr_wqe_size = (uint16_t)umr_wqe_size;
+ priv->umr_wqe_stride = priv->umr_wqe_size / MLX5_SEND_WQE_BB;
+ priv->max_rdmar_ds = rdmw_wqe_size / sizeof(struct mlx5_wqe_dseg);
+ return 0;
+}
+
static int
mlx5_crypto_dev_probe(struct mlx5_common_device *cdev)
{
@@ -824,7 +900,6 @@ mlx5_crypto_dev_probe(struct mlx5_common_device *cdev)
RTE_CRYPTODEV_PMD_DEFAULT_MAX_NB_QUEUE_PAIRS,
};
const char *ibdev_name = mlx5_os_get_ctx_device_name(cdev->ctx);
- uint16_t rdmw_wqe_size;
int ret;
if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
@@ -873,20 +948,17 @@ mlx5_crypto_dev_probe(struct mlx5_common_device *cdev)
}
priv->login_obj = login;
priv->keytag = rte_cpu_to_be_64(devarg_prms.keytag);
- priv->max_segs_num = devarg_prms.max_segs_num;
- priv->umr_wqe_size = sizeof(struct mlx5_wqe_umr_bsf_seg) +
- sizeof(struct mlx5_wqe_cseg) +
- sizeof(struct mlx5_wqe_umr_cseg) +
- sizeof(struct mlx5_wqe_mkey_cseg) +
- RTE_ALIGN(priv->max_segs_num, 4) *
- sizeof(struct mlx5_wqe_dseg);
- rdmw_wqe_size = sizeof(struct mlx5_rdma_write_wqe) +
- sizeof(struct mlx5_wqe_dseg) *
- (priv->max_segs_num <= 2 ? 2 : 2 +
- RTE_ALIGN(priv->max_segs_num - 2, 4));
- priv->wqe_set_size = priv->umr_wqe_size + rdmw_wqe_size;
- priv->umr_wqe_stride = priv->umr_wqe_size / MLX5_SEND_WQE_BB;
- priv->max_rdmar_ds = rdmw_wqe_size / sizeof(struct mlx5_wqe_dseg);
+ ret = mlx5_crypto_configure_wqe_size(priv,
+ cdev->config.hca_attr.max_wqe_sz_sq, devarg_prms.max_segs_num);
+ if (ret) {
+ mlx5_devx_uar_release(&priv->uar);
+ rte_cryptodev_pmd_destroy(priv->crypto_dev);
+ return -1;
+ }
+ DRV_LOG(INFO, "Max number of segments: %u.",
+ (unsigned int)RTE_MIN(
+ MLX5_CRYPTO_KLM_SEGS_NUM(priv->umr_wqe_size),
+ (uint16_t)(priv->max_rdmar_ds - 2)));
pthread_mutex_lock(&priv_list_lock);
TAILQ_INSERT_TAIL(&mlx5_crypto_priv_list, priv, next);
pthread_mutex_unlock(&priv_list_lock);
diff --git a/drivers/crypto/mlx5/mlx5_crypto.h b/drivers/crypto/mlx5/mlx5_crypto.h
index 135cd78212..f04b3d8c20 100644
--- a/drivers/crypto/mlx5/mlx5_crypto.h
+++ b/drivers/crypto/mlx5/mlx5_crypto.h
@@ -16,6 +16,13 @@
#define MLX5_CRYPTO_DEK_HTABLE_SZ (1 << 11)
#define MLX5_CRYPTO_KEY_LENGTH 80
+#define MLX5_CRYPTO_UMR_WQE_STATIC_SIZE (sizeof(struct mlx5_wqe_cseg) +\
+ sizeof(struct mlx5_wqe_umr_cseg) +\
+ sizeof(struct mlx5_wqe_mkey_cseg) +\
+ sizeof(struct mlx5_wqe_umr_bsf_seg))
+#define MLX5_CRYPTO_KLM_SEGS_NUM(umr_wqe_sz) ((umr_wqe_sz -\
+ MLX5_CRYPTO_UMR_WQE_STATIC_SIZE) /\
+ MLX5_WSEG_SIZE)
struct mlx5_crypto_priv {
TAILQ_ENTRY(mlx5_crypto_priv) next;
diff --git a/drivers/regex/mlx5/mlx5_regex_control.c b/drivers/regex/mlx5/mlx5_regex_control.c
index d184b1a921..46e400a93f 100644
--- a/drivers/regex/mlx5/mlx5_regex_control.c
+++ b/drivers/regex/mlx5/mlx5_regex_control.c
@@ -149,8 +149,8 @@ regex_ctrl_create_hw_qp(struct mlx5_regex_priv *priv, struct mlx5_regex_qp *qp,
qp_obj->qpn = q_ind;
qp_obj->ci = 0;
qp_obj->pi = 0;
- attr.rq_size = 0;
- attr.sq_size = RTE_BIT32(MLX5_REGEX_WQE_LOG_NUM(priv->has_umr,
+ attr.num_of_receive_wqes = 0;
+ attr.num_of_send_wqbbs = RTE_BIT32(MLX5_REGEX_WQE_LOG_NUM(priv->has_umr,
log_nb_desc));
attr.mmo = priv->mmo_regex_qp_cap;
ret = mlx5_devx_qp_create(priv->cdev->ctx, &qp_obj->qp_obj,
diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c b/drivers/vdpa/mlx5/mlx5_vdpa_event.c
index 9cc71714a2..657c39dae1 100644
--- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c
+++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c
@@ -589,9 +589,9 @@ mlx5_vdpa_event_qp_create(struct mlx5_vdpa_priv *priv, uint16_t desc_n,
}
attr.uar_index = mlx5_os_get_devx_uar_page_id(priv->uar.obj);
attr.cqn = eqp->cq.cq_obj.cq->id;
- attr.rq_size = RTE_BIT32(log_desc_n);
+ attr.num_of_receive_wqes = RTE_BIT32(log_desc_n);
attr.log_rq_stride = rte_log2_u32(MLX5_WSEG_SIZE);
- attr.sq_size = 0; /* No need SQ. */
+ attr.num_of_send_wqbbs = 0; /* No need SQ. */
attr.ts_format =
mlx5_ts_format_conv(priv->cdev->config.hca_attr.qp_ts_format);
ret = mlx5_devx_qp_create(priv->cdev->ctx, &(eqp->sw_qp), log_desc_n,
--
2.17.1
next prev parent reply other threads:[~2021-11-08 13:10 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-11-04 12:49 [dpdk-dev] [PATCH 0/4] fixes to queue size config Raja Zidane
2021-11-04 12:49 ` [dpdk-dev] [PATCH 1/4] common/mlx5: fix overflows in DevX queues size calculations Raja Zidane
2021-11-04 12:49 ` [dpdk-dev] [PATCH 2/4] crypto/mlx5: fix driver destroy before the configuration Raja Zidane
2021-11-04 12:49 ` [dpdk-dev] [PATCH 3/4] crypto/mlx5: fix the queue size configuration Raja Zidane
2021-11-04 18:59 ` Tal Shnaiderman
2021-11-04 12:49 ` [dpdk-dev] [PATCH 4/4] common/mlx5: fix RQ size configuration in QP create Raja Zidane
2021-11-08 9:02 ` [dpdk-dev] [PATCH 0/4] fixes to queue size config Thomas Monjalon
2021-11-08 12:33 ` [dpdk-dev] [PATCH V2 " Raja Zidane
2021-11-08 12:33 ` [dpdk-dev] [PATCH V2 1/4] common/mlx5: fix overflows in DevX queues size calculations Raja Zidane
2021-11-08 12:33 ` [dpdk-dev] [PATCH V2 2/4] crypto/mlx5: fix driver destroy before the configuration Raja Zidane
2021-11-08 12:33 ` [dpdk-dev] [PATCH V2 3/4] crypto/mlx5: fix the queue size configuration Raja Zidane
2021-11-08 12:33 ` [dpdk-dev] [PATCH V2 4/4] common/mlx5: fix RQ size configuration in QP create Raja Zidane
2021-11-08 13:09 ` [dpdk-dev] [PATCH V2 0/4] fixes to queue size config Raja Zidane
2021-11-08 13:09 ` [dpdk-dev] [PATCH V2 1/4] common/mlx5: fix overflows in DevX queues size calculations Raja Zidane
2021-11-08 13:09 ` [dpdk-dev] [PATCH V2 2/4] crypto/mlx5: fix driver destroy before the configuration Raja Zidane
2021-11-08 13:09 ` Raja Zidane [this message]
2021-11-08 13:09 ` [dpdk-dev] [PATCH V2 4/4] common/mlx5: fix RQ size configuration in QP create Raja Zidane
2021-11-08 18:44 ` [dpdk-dev] [PATCH V2 0/4] fixes to queue size config Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20211108130921.19143-4-rzidane@nvidia.com \
--to=rzidane@nvidia.com \
--cc=dev@dpdk.org \
--cc=matan@nvidia.com \
--cc=stable@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).