* [for-next 1/5] RDMA/bnxt_re: Enable RoCE on virtual functions
[not found] ` <1515152434-13105-1-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
@ 2018-01-05 11:40 ` Devesh Sharma
[not found] ` <1515152434-13105-2-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
2018-01-05 11:40 ` [for-next 2/5] RDMA/bnxt_re: Add support for query firmware version Devesh Sharma
` (3 subsequent siblings)
4 siblings, 1 reply; 14+ messages in thread
From: Devesh Sharma @ 2018-01-05 11:40 UTC (permalink / raw)
To: dledford-H+wXaHxf7aLQT0dZR+AlfA, jgg-VPRAkNaXOzVWk0Htik3J/w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Selvin Xavier, Devesh Sharma
From: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
RoCE can be used by virtual functions (VFs) as well. Adding
code changes to allow resource reservation, initialization
and avail the resources to the RDMA applications running on
those VFs.
Currently, fifty percent of the total available resources
are reserved for PF and remaining are equally divided among
active VFs.
Signed-off-by: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Devesh Sharma <devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/bnxt_re/bnxt_re.h | 15 ++++
drivers/infiniband/hw/bnxt_re/ib_verbs.c | 17 ++--
drivers/infiniband/hw/bnxt_re/main.c | 135 ++++++++++++++++++++++++-------
drivers/infiniband/hw/bnxt_re/qplib_sp.c | 37 ++++++++-
drivers/infiniband/hw/bnxt_re/qplib_sp.h | 5 +-
drivers/infiniband/hw/bnxt_re/roce_hsi.h | 10 +++
6 files changed, 182 insertions(+), 37 deletions(-)
diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index ecbac91..b604277 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -57,6 +57,19 @@
#define BNXT_RE_MAX_MRW_COUNT (64 * 1024)
#define BNXT_RE_MAX_SRQC_COUNT (64 * 1024)
#define BNXT_RE_MAX_CQ_COUNT (64 * 1024)
+#define BNXT_RE_MAX_MRW_COUNT_64K (64 * 1024)
+#define BNXT_RE_MAX_MRW_COUNT_256K (256 * 1024)
+
+/* Number of MRs to reserve for PF, leaving remainder for VFs */
+#define BNXT_RE_RESVD_MR_FOR_PF (32 * 1024)
+#define BNXT_RE_MAX_GID_PER_VF 128
+
+/*
+ * Percentage of resources of each type reserved for PF.
+ * Remaining resources are divided equally among VFs.
+ * [0, 100]
+ */
+#define BNXT_RE_PCT_RSVD_FOR_PF 50
#define BNXT_RE_UD_QP_HW_STALL 0x400000
@@ -145,6 +158,8 @@ struct bnxt_re_dev {
struct bnxt_re_ah *sqp_ah;
struct bnxt_re_sqp_entries sqp_tbl[1024];
atomic_t nq_alloc_cnt;
+ u32 is_virtfn;
+ u32 num_vfs;
};
#define to_bnxt_re_dev(ptr, member) \
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 2032db7..6dd2fe1 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -2295,10 +2295,14 @@ int bnxt_re_post_recv(struct ib_qp *ib_qp, struct ib_recv_wr *wr,
/* Completion Queues */
int bnxt_re_destroy_cq(struct ib_cq *ib_cq)
{
- struct bnxt_re_cq *cq = container_of(ib_cq, struct bnxt_re_cq, ib_cq);
- struct bnxt_re_dev *rdev = cq->rdev;
int rc;
- struct bnxt_qplib_nq *nq = cq->qplib_cq.nq;
+ struct bnxt_re_cq *cq;
+ struct bnxt_qplib_nq *nq;
+ struct bnxt_re_dev *rdev;
+
+ cq = container_of(ib_cq, struct bnxt_re_cq, ib_cq);
+ rdev = cq->rdev;
+ nq = cq->qplib_cq.nq;
rc = bnxt_qplib_destroy_cq(&rdev->qplib_res, &cq->qplib_cq);
if (rc) {
@@ -2308,12 +2312,11 @@ int bnxt_re_destroy_cq(struct ib_cq *ib_cq)
if (!IS_ERR_OR_NULL(cq->umem))
ib_umem_release(cq->umem);
- if (cq) {
- kfree(cq->cql);
- kfree(cq);
- }
atomic_dec(&rdev->cq_count);
nq->budget--;
+ kfree(cq->cql);
+ kfree(cq);
+
return 0;
}
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index aafc19a..d825df0 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -80,6 +80,79 @@
static struct workqueue_struct *bnxt_re_wq;
static void bnxt_re_ib_unreg(struct bnxt_re_dev *rdev, bool lock_wait);
+/* SR-IOV helper functions */
+
+static void bnxt_re_get_sriov_func_type(struct bnxt_re_dev *rdev)
+{
+ struct bnxt *bp;
+
+ bp = netdev_priv(rdev->en_dev->net);
+ if (BNXT_VF(bp))
+ rdev->is_virtfn = 1;
+}
+
+/* Set the maximum number of each resource that the driver actually wants
+ * to allocate. This may be up to the maximum number the firmware has
+ * reserved for the function. The driver may choose to allocate fewer
+ * resources than the firmware maximum.
+ */
+static void bnxt_re_set_resource_limits(struct bnxt_re_dev *rdev)
+{
+ u32 vf_qps = 0, vf_srqs = 0, vf_cqs = 0, vf_mrws = 0, vf_gids = 0;
+ u32 i;
+ u32 vf_pct;
+ u32 num_vfs;
+ struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
+
+ rdev->qplib_ctx.qpc_count = min_t(u32, BNXT_RE_MAX_QPC_COUNT,
+ dev_attr->max_qp);
+
+ rdev->qplib_ctx.mrw_count = BNXT_RE_MAX_MRW_COUNT_256K;
+ /* Use max_mr from fw since max_mrw does not get set */
+ rdev->qplib_ctx.mrw_count = min_t(u32, rdev->qplib_ctx.mrw_count,
+ dev_attr->max_mr);
+ rdev->qplib_ctx.srqc_count = min_t(u32, BNXT_RE_MAX_SRQC_COUNT,
+ dev_attr->max_srq);
+ rdev->qplib_ctx.cq_count = min_t(u32, BNXT_RE_MAX_CQ_COUNT,
+ dev_attr->max_cq);
+
+ for (i = 0; i < MAX_TQM_ALLOC_REQ; i++)
+ rdev->qplib_ctx.tqm_count[i] =
+ rdev->dev_attr.tqm_alloc_reqs[i];
+
+ if (rdev->num_vfs) {
+ /*
+ * Reserve a set of resources for the PF. Divide the remaining
+ * resources among the VFs
+ */
+ vf_pct = 100 - BNXT_RE_PCT_RSVD_FOR_PF;
+ num_vfs = 100 * rdev->num_vfs;
+ vf_qps = (rdev->qplib_ctx.qpc_count * vf_pct) / num_vfs;
+ vf_srqs = (rdev->qplib_ctx.srqc_count * vf_pct) / num_vfs;
+ vf_cqs = (rdev->qplib_ctx.cq_count * vf_pct) / num_vfs;
+ /*
+ * The driver allows many more MRs than other resources. If the
+ * firmware does also, then reserve a fixed amount for the PF
+ * and divide the rest among VFs. VFs may use many MRs for NFS
+ * mounts, ISER, NVME applications, etc. If the firmware
+ * severely restricts the number of MRs, then let PF have
+ * half and divide the rest among VFs, as for the other
+ * resource types.
+ */
+ if (rdev->qplib_ctx.mrw_count < BNXT_RE_MAX_MRW_COUNT_64K)
+ vf_mrws = rdev->qplib_ctx.mrw_count * vf_pct / num_vfs;
+ else
+ vf_mrws = (rdev->qplib_ctx.mrw_count -
+ BNXT_RE_RESVD_MR_FOR_PF) / rdev->num_vfs;
+ vf_gids = BNXT_RE_MAX_GID_PER_VF;
+ }
+ rdev->qplib_ctx.vf_res.max_mrw_per_vf = vf_mrws;
+ rdev->qplib_ctx.vf_res.max_gid_per_vf = vf_gids;
+ rdev->qplib_ctx.vf_res.max_qp_per_vf = vf_qps;
+ rdev->qplib_ctx.vf_res.max_srq_per_vf = vf_srqs;
+ rdev->qplib_ctx.vf_res.max_cq_per_vf = vf_cqs;
+}
+
/* for handling bnxt_en callbacks later */
static void bnxt_re_stop(void *p)
{
@@ -91,6 +164,15 @@ static void bnxt_re_start(void *p)
static void bnxt_re_sriov_config(void *p, int num_vfs)
{
+ struct bnxt_re_dev *rdev = p;
+
+ if (!rdev)
+ return;
+
+ rdev->num_vfs = num_vfs;
+ bnxt_re_set_resource_limits(rdev);
+ bnxt_qplib_set_func_resources(&rdev->qplib_res, &rdev->rcfw,
+ &rdev->qplib_ctx);
}
static void bnxt_re_shutdown(void *p)
@@ -734,7 +816,8 @@ static int bnxt_re_alloc_res(struct bnxt_re_dev *rdev)
/* Configure and allocate resources for qplib */
rdev->qplib_res.rcfw = &rdev->rcfw;
- rc = bnxt_qplib_get_dev_attr(&rdev->rcfw, &rdev->dev_attr);
+ rc = bnxt_qplib_get_dev_attr(&rdev->rcfw, &rdev->dev_attr,
+ rdev->is_virtfn);
if (rc)
goto fail;
@@ -1035,19 +1118,6 @@ static void bnxt_re_ib_unreg(struct bnxt_re_dev *rdev, bool lock_wait)
}
}
-static void bnxt_re_set_resource_limits(struct bnxt_re_dev *rdev)
-{
- u32 i;
-
- rdev->qplib_ctx.qpc_count = BNXT_RE_MAX_QPC_COUNT;
- rdev->qplib_ctx.mrw_count = BNXT_RE_MAX_MRW_COUNT;
- rdev->qplib_ctx.srqc_count = BNXT_RE_MAX_SRQC_COUNT;
- rdev->qplib_ctx.cq_count = BNXT_RE_MAX_CQ_COUNT;
- for (i = 0; i < MAX_TQM_ALLOC_REQ; i++)
- rdev->qplib_ctx.tqm_count[i] =
- rdev->dev_attr.tqm_alloc_reqs[i];
-}
-
/* worker thread for polling periodic events. Now used for QoS programming*/
static void bnxt_re_worker(struct work_struct *work)
{
@@ -1070,6 +1140,9 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
}
set_bit(BNXT_RE_FLAG_NETDEV_REGISTERED, &rdev->flags);
+ /* Check whether VF or PF */
+ bnxt_re_get_sriov_func_type(rdev);
+
rc = bnxt_re_request_msix(rdev);
if (rc) {
pr_err("Failed to get MSI-X vectors: %#x\n", rc);
@@ -1101,16 +1174,18 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
(rdev->en_dev->pdev, &rdev->rcfw,
rdev->msix_entries[BNXT_RE_AEQ_IDX].vector,
rdev->msix_entries[BNXT_RE_AEQ_IDX].db_offset,
- 0, &bnxt_re_aeq_handler);
+ rdev->is_virtfn, &bnxt_re_aeq_handler);
if (rc) {
pr_err("Failed to enable RCFW channel: %#x\n", rc);
goto free_ring;
}
- rc = bnxt_qplib_get_dev_attr(&rdev->rcfw, &rdev->dev_attr);
+ rc = bnxt_qplib_get_dev_attr(&rdev->rcfw, &rdev->dev_attr,
+ rdev->is_virtfn);
if (rc)
goto disable_rcfw;
- bnxt_re_set_resource_limits(rdev);
+ if (!rdev->is_virtfn)
+ bnxt_re_set_resource_limits(rdev);
rc = bnxt_qplib_alloc_ctx(rdev->en_dev->pdev, &rdev->qplib_ctx, 0);
if (rc) {
@@ -1125,7 +1200,8 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
goto free_ctx;
}
- rc = bnxt_qplib_init_rcfw(&rdev->rcfw, &rdev->qplib_ctx, 0);
+ rc = bnxt_qplib_init_rcfw(&rdev->rcfw, &rdev->qplib_ctx,
+ rdev->is_virtfn);
if (rc) {
pr_err("Failed to initialize RCFW: %#x\n", rc);
goto free_sctx;
@@ -1144,13 +1220,15 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
goto fail;
}
- rc = bnxt_re_setup_qos(rdev);
- if (rc)
- pr_info("RoCE priority not yet configured\n");
+ if (!rdev->is_virtfn) {
+ rc = bnxt_re_setup_qos(rdev);
+ if (rc)
+ pr_info("RoCE priority not yet configured\n");
- INIT_DELAYED_WORK(&rdev->worker, bnxt_re_worker);
- set_bit(BNXT_RE_FLAG_QOS_WORK_REG, &rdev->flags);
- schedule_delayed_work(&rdev->worker, msecs_to_jiffies(30000));
+ INIT_DELAYED_WORK(&rdev->worker, bnxt_re_worker);
+ set_bit(BNXT_RE_FLAG_QOS_WORK_REG, &rdev->flags);
+ schedule_delayed_work(&rdev->worker, msecs_to_jiffies(30000));
+ }
/* Register ib dev */
rc = bnxt_re_register_ib(rdev);
@@ -1400,7 +1478,7 @@ static int __init bnxt_re_mod_init(void)
static void __exit bnxt_re_mod_exit(void)
{
- struct bnxt_re_dev *rdev;
+ struct bnxt_re_dev *rdev, *next;
LIST_HEAD(to_be_deleted);
mutex_lock(&bnxt_re_dev_lock);
@@ -1408,8 +1486,11 @@ static void __exit bnxt_re_mod_exit(void)
if (!list_empty(&bnxt_re_dev_list))
list_splice_init(&bnxt_re_dev_list, &to_be_deleted);
mutex_unlock(&bnxt_re_dev_lock);
-
- list_for_each_entry(rdev, &to_be_deleted, list) {
+ /*
+ * Cleanup the devices in reverse order so that the VF device
+ * cleanup is done before PF cleanup
+ */
+ list_for_each_entry_safe_reverse(rdev, next, &to_be_deleted, list) {
dev_info(rdev_to_dev(rdev), "Unregistering Device");
bnxt_re_dev_stop(rdev);
bnxt_re_ib_unreg(rdev, true);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
index 9543ce5..f5450b7 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
@@ -65,7 +65,7 @@ static bool bnxt_qplib_is_atomic_cap(struct bnxt_qplib_rcfw *rcfw)
}
int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
- struct bnxt_qplib_dev_attr *attr)
+ struct bnxt_qplib_dev_attr *attr, bool vf)
{
struct cmdq_query_func req;
struct creq_query_func_resp resp;
@@ -95,7 +95,8 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
/* Extract the context from the side buffer */
attr->max_qp = le32_to_cpu(sb->max_qp);
/* max_qp value reported by FW for PF doesn't include the QP1 for PF */
- attr->max_qp += 1;
+ if (!vf)
+ attr->max_qp += 1;
attr->max_qp_rd_atom =
sb->max_qp_rd_atom > BNXT_QPLIB_MAX_OUT_RD_ATOM ?
BNXT_QPLIB_MAX_OUT_RD_ATOM : sb->max_qp_rd_atom;
@@ -150,6 +151,38 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
return rc;
}
+int bnxt_qplib_set_func_resources(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_rcfw *rcfw,
+ struct bnxt_qplib_ctx *ctx)
+{
+ struct cmdq_set_func_resources req;
+ struct creq_set_func_resources_resp resp;
+ u16 cmd_flags = 0;
+ int rc = 0;
+
+ RCFW_CMD_PREP(req, SET_FUNC_RESOURCES, cmd_flags);
+
+ req.number_of_qp = cpu_to_le32(ctx->qpc_count);
+ req.number_of_mrw = cpu_to_le32(ctx->mrw_count);
+ req.number_of_srq = cpu_to_le32(ctx->srqc_count);
+ req.number_of_cq = cpu_to_le32(ctx->cq_count);
+
+ req.max_qp_per_vf = cpu_to_le32(ctx->vf_res.max_qp_per_vf);
+ req.max_mrw_per_vf = cpu_to_le32(ctx->vf_res.max_mrw_per_vf);
+ req.max_srq_per_vf = cpu_to_le32(ctx->vf_res.max_srq_per_vf);
+ req.max_cq_per_vf = cpu_to_le32(ctx->vf_res.max_cq_per_vf);
+ req.max_gid_per_vf = cpu_to_le32(ctx->vf_res.max_gid_per_vf);
+
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req,
+ (void *)&resp,
+ NULL, 0);
+ if (rc) {
+ dev_err(&res->pdev->dev,
+ "QPLIB: Failed to set function resources");
+ }
+ return rc;
+}
+
/* SGID */
int bnxt_qplib_get_sgid(struct bnxt_qplib_res *res,
struct bnxt_qplib_sgid_tbl *sgid_tbl, int index,
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
index 1132258..51bb784 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
@@ -147,7 +147,10 @@ int bnxt_qplib_add_pkey(struct bnxt_qplib_res *res,
struct bnxt_qplib_pkey_tbl *pkey_tbl, u16 *pkey,
bool update);
int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
- struct bnxt_qplib_dev_attr *attr);
+ struct bnxt_qplib_dev_attr *attr, bool vf);
+int bnxt_qplib_set_func_resources(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_rcfw *rcfw,
+ struct bnxt_qplib_ctx *ctx);
int bnxt_qplib_create_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah);
int bnxt_qplib_destroy_ah(struct bnxt_qplib_res *res, struct bnxt_qplib_ah *ah);
int bnxt_qplib_alloc_mrw(struct bnxt_qplib_res *res,
diff --git a/drivers/infiniband/hw/bnxt_re/roce_hsi.h b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
index c3cba606..f429fdb 100644
--- a/drivers/infiniband/hw/bnxt_re/roce_hsi.h
+++ b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
@@ -1799,6 +1799,16 @@ struct cmdq_set_func_resources {
u8 resp_size;
u8 reserved8;
__le64 resp_addr;
+ __le32 number_of_qp;
+ __le32 number_of_mrw;
+ __le32 number_of_srq;
+ __le32 number_of_cq;
+ __le32 max_qp_per_vf;
+ __le32 max_mrw_per_vf;
+ __le32 max_srq_per_vf;
+ __le32 max_cq_per_vf;
+ __le32 max_gid_per_vf;
+ __le32 stat_ctx_id;
};
/* Read hardware resource context command (24 bytes) */
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 14+ messages in thread* [for-next 2/5] RDMA/bnxt_re: Add support for query firmware version
[not found] ` <1515152434-13105-1-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
2018-01-05 11:40 ` [for-next 1/5] RDMA/bnxt_re: Enable RoCE on virtual functions Devesh Sharma
@ 2018-01-05 11:40 ` Devesh Sharma
[not found] ` <1515152434-13105-3-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
2018-01-05 11:40 ` [for-next 3/5] RDMA/bnxt_re: Add support for MRs with Huge pages Devesh Sharma
` (2 subsequent siblings)
4 siblings, 1 reply; 14+ messages in thread
From: Devesh Sharma @ 2018-01-05 11:40 UTC (permalink / raw)
To: dledford-H+wXaHxf7aLQT0dZR+AlfA, jgg-VPRAkNaXOzVWk0Htik3J/w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Selvin Xavier, Devesh Sharma
From: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
The device now reports firmware version thus, removing
the hard coded values of the FW version string and
redundant fw_rev hook from sysfs. Adding code to query
firmware version from underlying device and report it
through the kernel verb to get firmware version string.
Signed-off-by: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Devesh Sharma <devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/bnxt_re/ib_verbs.c | 14 ++++++++++++--
drivers/infiniband/hw/bnxt_re/ib_verbs.h | 1 +
drivers/infiniband/hw/bnxt_re/main.c | 11 +----------
drivers/infiniband/hw/bnxt_re/qplib_rcfw.c | 3 ++-
drivers/infiniband/hw/bnxt_re/qplib_sp.c | 25 ++++++++++++++++++++++++-
drivers/infiniband/hw/bnxt_re/qplib_sp.h | 3 ++-
6 files changed, 42 insertions(+), 15 deletions(-)
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 6dd2fe1..8a80e95 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -141,8 +141,9 @@ int bnxt_re_query_device(struct ib_device *ibdev,
struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
memset(ib_attr, 0, sizeof(*ib_attr));
-
- ib_attr->fw_ver = (u64)(unsigned long)(dev_attr->fw_ver);
+ memcpy(&ib_attr->fw_ver, dev_attr->fw_ver,
+ min(sizeof(dev_attr->fw_ver),
+ sizeof(ib_attr->fw_ver)));
bnxt_qplib_get_guid(rdev->netdev->dev_addr,
(u8 *)&ib_attr->sys_image_guid);
ib_attr->max_mr_size = BNXT_RE_MAX_MR_SIZE;
@@ -281,6 +282,15 @@ int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num,
return 0;
}
+void bnxt_re_query_fw_str(struct ib_device *ibdev, char *str)
+{
+ struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
+
+ snprintf(str, IB_FW_VERSION_NAME_MAX, "%d.%d.%d.%d",
+ rdev->dev_attr.fw_ver[0], rdev->dev_attr.fw_ver[1],
+ rdev->dev_attr.fw_ver[2], rdev->dev_attr.fw_ver[3]);
+}
+
int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num,
u16 index, u16 *pkey)
{
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index 1df11ed2..66dd8d2 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -143,6 +143,7 @@ int bnxt_re_query_port(struct ib_device *ibdev, u8 port_num,
struct ib_port_attr *port_attr);
int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num,
struct ib_port_immutable *immutable);
+void bnxt_re_query_fw_str(struct ib_device *ibdev, char *str);
int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num,
u16 index, u16 *pkey);
int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index d825df0..38a5d4b 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -572,6 +572,7 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
ibdev->query_port = bnxt_re_query_port;
ibdev->get_port_immutable = bnxt_re_get_port_immutable;
+ ibdev->get_dev_fw_str = bnxt_re_query_fw_str;
ibdev->query_pkey = bnxt_re_query_pkey;
ibdev->query_gid = bnxt_re_query_gid;
ibdev->get_netdev = bnxt_re_get_netdev;
@@ -623,14 +624,6 @@ static ssize_t show_rev(struct device *device, struct device_attribute *attr,
return scnprintf(buf, PAGE_SIZE, "0x%x\n", rdev->en_dev->pdev->vendor);
}
-static ssize_t show_fw_ver(struct device *device, struct device_attribute *attr,
- char *buf)
-{
- struct bnxt_re_dev *rdev = to_bnxt_re_dev(device, ibdev.dev);
-
- return scnprintf(buf, PAGE_SIZE, "%s\n", rdev->dev_attr.fw_ver);
-}
-
static ssize_t show_hca(struct device *device, struct device_attribute *attr,
char *buf)
{
@@ -640,12 +633,10 @@ static ssize_t show_hca(struct device *device, struct device_attribute *attr,
}
static DEVICE_ATTR(hw_rev, 0444, show_rev, NULL);
-static DEVICE_ATTR(fw_rev, 0444, show_fw_ver, NULL);
static DEVICE_ATTR(hca_type, 0444, show_hca, NULL);
static struct device_attribute *bnxt_re_attributes[] = {
&dev_attr_hw_rev,
- &dev_attr_fw_rev,
&dev_attr_hca_type
};
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
index bb5574a..6a3633a 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
@@ -93,7 +93,8 @@ static int __send_message(struct bnxt_qplib_rcfw *rcfw, struct cmdq_base *req,
opcode = req->opcode;
if (!test_bit(FIRMWARE_INITIALIZED_FLAG, &rcfw->flags) &&
(opcode != CMDQ_BASE_OPCODE_QUERY_FUNC &&
- opcode != CMDQ_BASE_OPCODE_INITIALIZE_FW)) {
+ opcode != CMDQ_BASE_OPCODE_INITIALIZE_FW &&
+ opcode != CMDQ_BASE_OPCODE_QUERY_VERSION)) {
dev_err(&rcfw->pdev->dev,
"QPLIB: RCFW not initialized, reject opcode 0x%x",
opcode);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
index f5450b7..60377a3 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
@@ -64,6 +64,29 @@ static bool bnxt_qplib_is_atomic_cap(struct bnxt_qplib_rcfw *rcfw)
return !!(pcie_ctl2 & PCI_EXP_DEVCTL2_ATOMIC_REQ);
}
+static void bnxt_qplib_query_version(struct bnxt_qplib_rcfw *rcfw,
+ char *fw_ver)
+{
+ struct cmdq_query_version req;
+ struct creq_query_version_resp resp;
+ u16 cmd_flags = 0;
+ int rc = 0;
+
+ RCFW_CMD_PREP(req, QUERY_VERSION, cmd_flags);
+
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req,
+ (void *)&resp, NULL, 0);
+ if (rc) {
+ dev_err(&rcfw->pdev->dev, "QPLIB: Failed to query version");
+ return;
+ }
+
+ fw_ver[0] = resp.fw_maj;
+ fw_ver[1] = resp.fw_minor;
+ fw_ver[2] = resp.fw_bld;
+ fw_ver[3] = resp.fw_rsvd;
+}
+
int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
struct bnxt_qplib_dev_attr *attr, bool vf)
{
@@ -134,7 +157,7 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
attr->l2_db_size = (sb->l2_db_space_size + 1) * PAGE_SIZE;
attr->max_sgid = le32_to_cpu(sb->max_gid);
- strlcpy(attr->fw_ver, "20.6.28.0", sizeof(attr->fw_ver));
+ bnxt_qplib_query_version(rcfw, attr->fw_ver);
for (i = 0; i < MAX_TQM_ALLOC_REQ / 4; i++) {
temp = le32_to_cpu(sb->tqm_alloc_reqs[i]);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
index 51bb784..0828bb1 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
@@ -45,7 +45,8 @@
#define PCI_EXP_DEVCTL2_ATOMIC_REQ 0x0040
struct bnxt_qplib_dev_attr {
- char fw_ver[32];
+#define FW_VER_ARR_LEN 4
+ u8 fw_ver[FW_VER_ARR_LEN];
u16 max_sgid;
u16 max_mrw;
u32 max_qp;
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 14+ messages in thread* [for-next 3/5] RDMA/bnxt_re: Add support for MRs with Huge pages
[not found] ` <1515152434-13105-1-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
2018-01-05 11:40 ` [for-next 1/5] RDMA/bnxt_re: Enable RoCE on virtual functions Devesh Sharma
2018-01-05 11:40 ` [for-next 2/5] RDMA/bnxt_re: Add support for query firmware version Devesh Sharma
@ 2018-01-05 11:40 ` Devesh Sharma
2018-01-05 11:40 ` [for-next 4/5] RDMA/bnxt_re: expose detailed stats retrieved from HW Devesh Sharma
2018-01-05 11:40 ` [for-next 5/5] RDMA/bnxt_re: Add SRQ support for Broadcom adapters Devesh Sharma
4 siblings, 0 replies; 14+ messages in thread
From: Devesh Sharma @ 2018-01-05 11:40 UTC (permalink / raw)
To: dledford-H+wXaHxf7aLQT0dZR+AlfA, jgg-VPRAkNaXOzVWk0Htik3J/w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Somnath Kotur, Selvin Xavier,
Devesh Sharma
From: Somnath Kotur <somnath.kotur-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Depending on the OS page-table configurations, applications
may request MRs which has page size alignment other than 4K
Underlying provider driver needs to adjust its PBL boundaries
according to the incoming page boundaries in the PA list.
Adding a capability to register MRs having pages-sizes other
than 4K (Hugepages).
Signed-off-by: Somnath Kotur <somnath.kotur-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Devesh Sharma <devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/bnxt_re/bnxt_re.h | 26 ++++---
drivers/infiniband/hw/bnxt_re/ib_verbs.c | 120 +++++++++++++++++++++----------
drivers/infiniband/hw/bnxt_re/qplib_sp.c | 12 +++-
drivers/infiniband/hw/bnxt_re/qplib_sp.h | 2 +-
drivers/infiniband/hw/bnxt_re/roce_hsi.h | 28 +++++++-
5 files changed, 137 insertions(+), 51 deletions(-)
diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index b604277..085ca00 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -43,15 +43,23 @@
#define ROCE_DRV_MODULE_VERSION "1.0.0"
#define BNXT_RE_DESC "Broadcom NetXtreme-C/E RoCE Driver"
-
-#define BNXT_RE_PAGE_SIZE_4K BIT(12)
-#define BNXT_RE_PAGE_SIZE_8K BIT(13)
-#define BNXT_RE_PAGE_SIZE_64K BIT(16)
-#define BNXT_RE_PAGE_SIZE_2M BIT(21)
-#define BNXT_RE_PAGE_SIZE_8M BIT(23)
-#define BNXT_RE_PAGE_SIZE_1G BIT(30)
-
-#define BNXT_RE_MAX_MR_SIZE BIT(30)
+#define BNXT_RE_PAGE_SHIFT_4K (12)
+#define BNXT_RE_PAGE_SHIFT_8K (13)
+#define BNXT_RE_PAGE_SHIFT_64K (16)
+#define BNXT_RE_PAGE_SHIFT_2M (21)
+#define BNXT_RE_PAGE_SHIFT_8M (23)
+#define BNXT_RE_PAGE_SHIFT_1G (30)
+
+#define BNXT_RE_PAGE_SIZE_4K BIT(BNXT_RE_PAGE_SHIFT_4K)
+#define BNXT_RE_PAGE_SIZE_8K BIT(BNXT_RE_PAGE_SHIFT_8K)
+#define BNXT_RE_PAGE_SIZE_64K BIT(BNXT_RE_PAGE_SHIFT_64K)
+#define BNXT_RE_PAGE_SIZE_2M BIT(BNXT_RE_PAGE_SHIFT_2M)
+#define BNXT_RE_PAGE_SIZE_8M BIT(BNXT_RE_PAGE_SHIFT_8M)
+#define BNXT_RE_PAGE_SIZE_1G BIT(BNXT_RE_PAGE_SHIFT_1G)
+
+#define BNXT_RE_MAX_MR_SIZE_LOW BIT(BNXT_RE_PAGE_SHIFT_1G)
+#define BNXT_RE_MAX_MR_SIZE_HIGH BIT(39)
+#define BNXT_RE_MAX_MR_SIZE BNXT_RE_MAX_MR_SIZE_HIGH
#define BNXT_RE_MAX_QPC_COUNT (64 * 1024)
#define BNXT_RE_MAX_MRW_COUNT (64 * 1024)
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 8a80e95..c3135f6 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -147,7 +147,7 @@ int bnxt_re_query_device(struct ib_device *ibdev,
bnxt_qplib_get_guid(rdev->netdev->dev_addr,
(u8 *)&ib_attr->sys_image_guid);
ib_attr->max_mr_size = BNXT_RE_MAX_MR_SIZE;
- ib_attr->page_size_cap = BNXT_RE_PAGE_SIZE_4K;
+ ib_attr->page_size_cap = BNXT_RE_PAGE_SIZE_4K | BNXT_RE_PAGE_SIZE_2M;
ib_attr->vendor_id = rdev->en_dev->pdev->vendor;
ib_attr->vendor_part_id = rdev->en_dev->pdev->device;
@@ -248,8 +248,7 @@ int bnxt_re_query_port(struct ib_device *ibdev, u8 port_num,
IB_PORT_VENDOR_CLASS_SUP |
IB_PORT_IP_BASED_GIDS;
- /* Max MSG size set to 2G for now */
- port_attr->max_msg_sz = 0x80000000;
+ port_attr->max_msg_sz = (u32)BNXT_RE_MAX_MR_SIZE_LOW;
port_attr->bad_pkey_cntr = 0;
port_attr->qkey_viol_cntr = 0;
port_attr->pkey_tbl_len = dev_attr->max_pkey;
@@ -542,7 +541,7 @@ static int bnxt_re_create_fence_mr(struct bnxt_re_pd *pd)
mr->qplib_mr.total_size = BNXT_RE_FENCE_BYTES;
pbl_tbl = dma_addr;
rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, &pbl_tbl,
- BNXT_RE_FENCE_PBL_SIZE, false);
+ BNXT_RE_FENCE_PBL_SIZE, false, PAGE_SIZE);
if (rc) {
dev_err(rdev_to_dev(rdev), "Failed to register fence-MR\n");
goto fail;
@@ -3091,7 +3090,8 @@ struct ib_mr *bnxt_re_get_dma_mr(struct ib_pd *ib_pd, int mr_access_flags)
mr->qplib_mr.hwq.level = PBL_LVL_MAX;
mr->qplib_mr.total_size = -1; /* Infinte length */
- rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, &pbl, 0, false);
+ rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, &pbl, 0, false,
+ PAGE_SIZE);
if (rc)
goto fail_mr;
@@ -3261,6 +3261,46 @@ int bnxt_re_dealloc_mw(struct ib_mw *ib_mw)
return rc;
}
+static int bnxt_re_page_size_ok(int page_shift)
+{
+ switch (page_shift) {
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4K:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_8K:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_64K:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_2M:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_256K:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1M:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4M:
+ case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1G:
+ return 1;
+ default:
+ return 0;
+ }
+}
+
+static int fill_umem_pbl_tbl(struct ib_umem *umem, u64 *pbl_tbl_orig,
+ int page_shift)
+{
+ u64 *pbl_tbl = pbl_tbl_orig;
+ u64 paddr;
+ u64 page_mask = (1ULL << page_shift) - 1;
+ int i, pages;
+ struct scatterlist *sg;
+ int entry;
+
+ for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+ pages = sg_dma_len(sg) >> PAGE_SHIFT;
+ for (i = 0; i < pages; i++) {
+ paddr = sg_dma_address(sg) + (i << PAGE_SHIFT);
+ if (pbl_tbl == pbl_tbl_orig)
+ *pbl_tbl++ = paddr & ~page_mask;
+ else if ((paddr & page_mask) == 0)
+ *pbl_tbl++ = paddr;
+ }
+ }
+ return pbl_tbl - pbl_tbl_orig;
+}
+
/* uverbs */
struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
u64 virt_addr, int mr_access_flags,
@@ -3270,10 +3310,8 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
struct bnxt_re_dev *rdev = pd->rdev;
struct bnxt_re_mr *mr;
struct ib_umem *umem;
- u64 *pbl_tbl, *pbl_tbl_orig;
- int i, umem_pgs, pages, rc;
- struct scatterlist *sg;
- int entry;
+ u64 *pbl_tbl = NULL;
+ int umem_pgs, page_shift, rc;
if (length > BNXT_RE_MAX_MR_SIZE) {
dev_err(rdev_to_dev(rdev), "MR Size: %lld > Max supported:%ld\n",
@@ -3290,64 +3328,70 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
mr->qplib_mr.flags = __from_ib_access_flags(mr_access_flags);
mr->qplib_mr.type = CMDQ_ALLOCATE_MRW_MRW_FLAGS_MR;
+ rc = bnxt_qplib_alloc_mrw(&rdev->qplib_res, &mr->qplib_mr);
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "Failed to allocate MR");
+ goto free_mr;
+ }
+ /* The fixed portion of the rkey is the same as the lkey */
+ mr->ib_mr.rkey = mr->qplib_mr.rkey;
+
umem = ib_umem_get(ib_pd->uobject->context, start, length,
mr_access_flags, 0);
if (IS_ERR(umem)) {
dev_err(rdev_to_dev(rdev), "Failed to get umem");
rc = -EFAULT;
- goto free_mr;
+ goto free_mrw;
}
mr->ib_umem = umem;
- rc = bnxt_qplib_alloc_mrw(&rdev->qplib_res, &mr->qplib_mr);
- if (rc) {
- dev_err(rdev_to_dev(rdev), "Failed to allocate MR");
- goto release_umem;
- }
- /* The fixed portion of the rkey is the same as the lkey */
- mr->ib_mr.rkey = mr->qplib_mr.rkey;
-
mr->qplib_mr.va = virt_addr;
umem_pgs = ib_umem_page_count(umem);
if (!umem_pgs) {
dev_err(rdev_to_dev(rdev), "umem is invalid!");
rc = -EINVAL;
- goto free_mrw;
+ goto free_umem;
}
mr->qplib_mr.total_size = length;
pbl_tbl = kcalloc(umem_pgs, sizeof(u64 *), GFP_KERNEL);
if (!pbl_tbl) {
- rc = -EINVAL;
+ rc = -ENOMEM;
goto free_mrw;
}
- pbl_tbl_orig = pbl_tbl;
- if (umem->hugetlb) {
- dev_err(rdev_to_dev(rdev), "umem hugetlb not supported!");
+ page_shift = umem->page_shift;
+
+ if (!bnxt_re_page_size_ok(page_shift)) {
+ dev_err(rdev_to_dev(rdev), "umem page size unsupported!");
+ kfree(pbl_tbl);
rc = -EFAULT;
- goto fail;
+ goto free_umem;
}
- if (umem->page_shift != PAGE_SHIFT) {
- dev_err(rdev_to_dev(rdev), "umem page shift unsupported!");
- rc = -EFAULT;
+ if (!umem->hugetlb && length > BNXT_RE_MAX_MR_SIZE_LOW) {
+ dev_err(rdev_to_dev(rdev), "Requested MR Sz:%llu Max sup:%llu",
+ length, (u64)BNXT_RE_MAX_MR_SIZE_LOW);
+ kfree(pbl_tbl);
+ rc = -EINVAL;
goto fail;
}
- /* Map umem buf ptrs to the PBL */
- for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
- pages = sg_dma_len(sg) >> umem->page_shift;
- for (i = 0; i < pages; i++, pbl_tbl++)
- *pbl_tbl = sg_dma_address(sg) + (i << umem->page_shift);
+ if (umem->hugetlb && length > BNXT_RE_PAGE_SIZE_2M) {
+ page_shift = BNXT_RE_PAGE_SHIFT_2M;
+ dev_warn(rdev_to_dev(rdev), "umem hugetlb set page_size %x",
+ 1 << page_shift);
}
- rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, pbl_tbl_orig,
- umem_pgs, false);
+
+ /* Map umem buf ptrs to the PBL */
+ umem_pgs = fill_umem_pbl_tbl(umem, pbl_tbl, page_shift);
+ rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, pbl_tbl,
+ umem_pgs, false, 1 << page_shift);
if (rc) {
dev_err(rdev_to_dev(rdev), "Failed to register user MR");
goto fail;
}
- kfree(pbl_tbl_orig);
+ kfree(pbl_tbl);
mr->ib_mr.lkey = mr->qplib_mr.lkey;
mr->ib_mr.rkey = mr->qplib_mr.lkey;
@@ -3355,11 +3399,11 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
return &mr->ib_mr;
fail:
- kfree(pbl_tbl_orig);
+ kfree(pbl_tbl);
+free_umem:
+ ib_umem_release(umem);
free_mrw:
bnxt_qplib_free_mrw(&rdev->qplib_res, &mr->qplib_mr);
-release_umem:
- ib_umem_release(umem);
free_mr:
kfree(mr);
return ERR_PTR(rc);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
index 60377a3..b805495 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
@@ -660,7 +660,7 @@ int bnxt_qplib_dereg_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mrw,
}
int bnxt_qplib_reg_mr(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr,
- u64 *pbl_tbl, int num_pbls, bool block)
+ u64 *pbl_tbl, int num_pbls, bool block, u32 buf_pg_size)
{
struct bnxt_qplib_rcfw *rcfw = res->rcfw;
struct cmdq_register_mr req;
@@ -671,6 +671,9 @@ int bnxt_qplib_reg_mr(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr,
u32 pg_size;
if (num_pbls) {
+ /* Allocate memory for the non-leaf pages to store buf ptrs.
+ * Non-leaf pages always uses system PAGE_SIZE
+ */
pg_ptrs = roundup_pow_of_two(num_pbls);
pages = pg_ptrs >> MAX_PBL_LVL_1_PGS_SHIFT;
if (!pages)
@@ -688,6 +691,7 @@ int bnxt_qplib_reg_mr(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr,
bnxt_qplib_free_hwq(res->pdev, &mr->hwq);
mr->hwq.max_elements = pages;
+ /* Use system PAGE_SIZE */
rc = bnxt_qplib_alloc_init_hwq(res->pdev, &mr->hwq, NULL, 0,
&mr->hwq.max_elements,
PAGE_SIZE, 0, PAGE_SIZE,
@@ -708,18 +712,22 @@ int bnxt_qplib_reg_mr(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr,
/* Configure the request */
if (mr->hwq.level == PBL_LVL_MAX) {
+ /* No PBL provided, just use system PAGE_SIZE */
level = 0;
req.pbl = 0;
pg_size = PAGE_SIZE;
} else {
level = mr->hwq.level + 1;
req.pbl = cpu_to_le64(mr->hwq.pbl[PBL_LVL_0].pg_map_arr[0]);
- pg_size = mr->hwq.pbl[PBL_LVL_0].pg_size;
}
+ pg_size = buf_pg_size ? buf_pg_size : PAGE_SIZE;
req.log2_pg_size_lvl = (level << CMDQ_REGISTER_MR_LVL_SFT) |
((ilog2(pg_size) <<
CMDQ_REGISTER_MR_LOG2_PG_SIZE_SFT) &
CMDQ_REGISTER_MR_LOG2_PG_SIZE_MASK);
+ req.log2_pbl_pg_size = cpu_to_le16(((ilog2(PAGE_SIZE) <<
+ CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_SFT) &
+ CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_MASK));
req.access = (mr->flags & 0xFFFF);
req.va = cpu_to_le64(mr->va);
req.key = cpu_to_le32(mr->lkey);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
index 0828bb1..074e5e3 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
@@ -159,7 +159,7 @@ int bnxt_qplib_alloc_mrw(struct bnxt_qplib_res *res,
int bnxt_qplib_dereg_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mrw,
bool block);
int bnxt_qplib_reg_mr(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr,
- u64 *pbl_tbl, int num_pbls, bool block);
+ u64 *pbl_tbl, int num_pbls, bool block, u32 buf_pg_size);
int bnxt_qplib_free_mrw(struct bnxt_qplib_res *res, struct bnxt_qplib_mrw *mr);
int bnxt_qplib_alloc_fast_reg_mr(struct bnxt_qplib_res *res,
struct bnxt_qplib_mrw *mr, int max);
diff --git a/drivers/infiniband/hw/bnxt_re/roce_hsi.h b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
index f429fdb..5cd31de 100644
--- a/drivers/infiniband/hw/bnxt_re/roce_hsi.h
+++ b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
@@ -1383,8 +1383,20 @@ struct cmdq_register_mr {
#define CMDQ_REGISTER_MR_LVL_LVL_0 0x0UL
#define CMDQ_REGISTER_MR_LVL_LVL_1 0x1UL
#define CMDQ_REGISTER_MR_LVL_LVL_2 0x2UL
+ #define CMDQ_REGISTER_MR_LVL_LAST CMDQ_REGISTER_MR_LVL_LVL_2
#define CMDQ_REGISTER_MR_LOG2_PG_SIZE_MASK 0x7cUL
#define CMDQ_REGISTER_MR_LOG2_PG_SIZE_SFT 2
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_4K (0xcUL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_8K (0xdUL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_64K (0x10UL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_256K (0x12UL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_1M (0x14UL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_2M (0x15UL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_4M (0x16UL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_1G (0x1eUL << 2)
+ #define CMDQ_REGISTER_MR_LOG2_PG_SIZE_LAST \
+ CMDQ_REGISTER_MR_LOG2_PG_SIZE_PG_1G
+ #define CMDQ_REGISTER_MR_UNUSED1 0x80UL
u8 access;
#define CMDQ_REGISTER_MR_ACCESS_LOCAL_WRITE 0x1UL
#define CMDQ_REGISTER_MR_ACCESS_REMOTE_READ 0x2UL
@@ -1392,7 +1404,21 @@ struct cmdq_register_mr {
#define CMDQ_REGISTER_MR_ACCESS_REMOTE_ATOMIC 0x8UL
#define CMDQ_REGISTER_MR_ACCESS_MW_BIND 0x10UL
#define CMDQ_REGISTER_MR_ACCESS_ZERO_BASED 0x20UL
- __le16 unused_1;
+ __le16 log2_pbl_pg_size;
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_MASK 0x1fUL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_SFT 0
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4K 0xcUL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_8K 0xdUL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_64K 0x10UL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_256K 0x12UL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1M 0x14UL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_2M 0x15UL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4M 0x16UL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1G 0x1eUL
+ #define CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_LAST \
+ CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1G
+ #define CMDQ_REGISTER_MR_UNUSED11_MASK 0xffe0UL
+ #define CMDQ_REGISTER_MR_UNUSED11_SFT 5
__le32 key;
__le64 pbl;
__le64 va;
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 14+ messages in thread* [for-next 4/5] RDMA/bnxt_re: expose detailed stats retrieved from HW
[not found] ` <1515152434-13105-1-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
` (2 preceding siblings ...)
2018-01-05 11:40 ` [for-next 3/5] RDMA/bnxt_re: Add support for MRs with Huge pages Devesh Sharma
@ 2018-01-05 11:40 ` Devesh Sharma
2018-01-05 11:40 ` [for-next 5/5] RDMA/bnxt_re: Add SRQ support for Broadcom adapters Devesh Sharma
4 siblings, 0 replies; 14+ messages in thread
From: Devesh Sharma @ 2018-01-05 11:40 UTC (permalink / raw)
To: dledford-H+wXaHxf7aLQT0dZR+AlfA, jgg-VPRAkNaXOzVWk0Htik3J/w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Selvin Xavier, Devesh Sharma
From: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Broadcom's adapter supports more granular statistics
to allow better understanding about the state of the
chip when data traffic is flowing.
Exposing the detailed stats to the consumer through
the standard hook available in the kverbs interface.
In order to retrieve all the information, driver
implements a firmware command.
Signed-off-by: Selvin Xavier <selvin.xavier-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
Signed-off-by: Devesh Sharma <devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/bnxt_re/bnxt_re.h | 2 +
drivers/infiniband/hw/bnxt_re/hw_counters.c | 145 ++++++++++++++++++++++++++--
drivers/infiniband/hw/bnxt_re/hw_counters.h | 39 ++++++++
drivers/infiniband/hw/bnxt_re/main.c | 1 +
drivers/infiniband/hw/bnxt_re/qplib_sp.c | 70 ++++++++++++++
drivers/infiniband/hw/bnxt_re/qplib_sp.h | 81 ++++++++++++++++
drivers/infiniband/hw/bnxt_re/roce_hsi.h | 89 +++++++++++++++++
7 files changed, 417 insertions(+), 10 deletions(-)
diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index 085ca00..ca32057 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -121,6 +121,7 @@ struct bnxt_re_dev {
#define BNXT_RE_FLAG_RCFW_CHANNEL_EN 4
#define BNXT_RE_FLAG_QOS_WORK_REG 5
#define BNXT_RE_FLAG_TASK_IN_PROG 6
+#define BNXT_RE_FLAG_ISSUE_ROCE_STATS 29
struct net_device *netdev;
unsigned int version, major, minor;
struct bnxt_en_dev *en_dev;
@@ -168,6 +169,7 @@ struct bnxt_re_dev {
atomic_t nq_alloc_cnt;
u32 is_virtfn;
u32 num_vfs;
+ struct bnxt_qplib_roce_stats stats;
};
#define to_bnxt_re_dev(ptr, member) \
diff --git a/drivers/infiniband/hw/bnxt_re/hw_counters.c b/drivers/infiniband/hw/bnxt_re/hw_counters.c
index 7b28219..77416bc 100644
--- a/drivers/infiniband/hw/bnxt_re/hw_counters.c
+++ b/drivers/infiniband/hw/bnxt_re/hw_counters.c
@@ -58,16 +58,55 @@
#include "hw_counters.h"
static const char * const bnxt_re_stat_name[] = {
- [BNXT_RE_ACTIVE_QP] = "active_qps",
- [BNXT_RE_ACTIVE_SRQ] = "active_srqs",
- [BNXT_RE_ACTIVE_CQ] = "active_cqs",
- [BNXT_RE_ACTIVE_MR] = "active_mrs",
- [BNXT_RE_ACTIVE_MW] = "active_mws",
- [BNXT_RE_RX_PKTS] = "rx_pkts",
- [BNXT_RE_RX_BYTES] = "rx_bytes",
- [BNXT_RE_TX_PKTS] = "tx_pkts",
- [BNXT_RE_TX_BYTES] = "tx_bytes",
- [BNXT_RE_RECOVERABLE_ERRORS] = "recoverable_errors"
+ [BNXT_RE_ACTIVE_QP] = "active_qps",
+ [BNXT_RE_ACTIVE_SRQ] = "active_srqs",
+ [BNXT_RE_ACTIVE_CQ] = "active_cqs",
+ [BNXT_RE_ACTIVE_MR] = "active_mrs",
+ [BNXT_RE_ACTIVE_MW] = "active_mws",
+ [BNXT_RE_RX_PKTS] = "rx_pkts",
+ [BNXT_RE_RX_BYTES] = "rx_bytes",
+ [BNXT_RE_TX_PKTS] = "tx_pkts",
+ [BNXT_RE_TX_BYTES] = "tx_bytes",
+ [BNXT_RE_RECOVERABLE_ERRORS] = "recoverable_errors",
+ [BNXT_RE_TO_RETRANSMITS] = "to_retransmits",
+ [BNXT_RE_SEQ_ERR_NAKS_RCVD] = "seq_err_naks_rcvd",
+ [BNXT_RE_MAX_RETRY_EXCEEDED] = "max_retry_exceeded",
+ [BNXT_RE_RNR_NAKS_RCVD] = "rnr_naks_rcvd",
+ [BNXT_RE_MISSING_RESP] = "missin_resp",
+ [BNXT_RE_UNRECOVERABLE_ERR] = "unrecoverable_err",
+ [BNXT_RE_BAD_RESP_ERR] = "bad_resp_err",
+ [BNXT_RE_LOCAL_QP_OP_ERR] = "local_qp_op_err",
+ [BNXT_RE_LOCAL_PROTECTION_ERR] = "local_protection_err",
+ [BNXT_RE_MEM_MGMT_OP_ERR] = "mem_mgmt_op_err",
+ [BNXT_RE_REMOTE_INVALID_REQ_ERR] = "remote_invalid_req_err",
+ [BNXT_RE_REMOTE_ACCESS_ERR] = "remote_access_err",
+ [BNXT_RE_REMOTE_OP_ERR] = "remote_op_err",
+ [BNXT_RE_DUP_REQ] = "dup_req",
+ [BNXT_RE_RES_EXCEED_MAX] = "res_exceed_max",
+ [BNXT_RE_RES_LENGTH_MISMATCH] = "res_length_mismatch",
+ [BNXT_RE_RES_EXCEEDS_WQE] = "res_exceeds_wqe",
+ [BNXT_RE_RES_OPCODE_ERR] = "res_opcode_err",
+ [BNXT_RE_RES_RX_INVALID_RKEY] = "res_rx_invalid_rkey",
+ [BNXT_RE_RES_RX_DOMAIN_ERR] = "res_rx_domain_err",
+ [BNXT_RE_RES_RX_NO_PERM] = "res_rx_no_perm",
+ [BNXT_RE_RES_RX_RANGE_ERR] = "res_rx_range_err",
+ [BNXT_RE_RES_TX_INVALID_RKEY] = "res_tx_invalid_rkey",
+ [BNXT_RE_RES_TX_DOMAIN_ERR] = "res_tx_domain_err",
+ [BNXT_RE_RES_TX_NO_PERM] = "res_tx_no_perm",
+ [BNXT_RE_RES_TX_RANGE_ERR] = "res_tx_range_err",
+ [BNXT_RE_RES_IRRQ_OFLOW] = "res_irrq_oflow",
+ [BNXT_RE_RES_UNSUP_OPCODE] = "res_unsup_opcode",
+ [BNXT_RE_RES_UNALIGNED_ATOMIC] = "res_unaligned_atomic",
+ [BNXT_RE_RES_REM_INV_ERR] = "res_rem_inv_err",
+ [BNXT_RE_RES_MEM_ERROR] = "res_mem_err",
+ [BNXT_RE_RES_SRQ_ERR] = "res_srq_err",
+ [BNXT_RE_RES_CMP_ERR] = "res_cmp_err",
+ [BNXT_RE_RES_INVALID_DUP_RKEY] = "res_invalid_dup_rkey",
+ [BNXT_RE_RES_WQE_FORMAT_ERR] = "res_wqe_format_err",
+ [BNXT_RE_RES_CQ_LOAD_ERR] = "res_cq_load_err",
+ [BNXT_RE_RES_SRQ_LOAD_ERR] = "res_srq_load_err",
+ [BNXT_RE_RES_TX_PCI_ERR] = "res_tx_pci_err",
+ [BNXT_RE_RES_RX_PCI_ERR] = "res_rx_pci_err"
};
int bnxt_re_ib_get_hw_stats(struct ib_device *ibdev,
@@ -76,6 +115,7 @@ int bnxt_re_ib_get_hw_stats(struct ib_device *ibdev,
{
struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
struct ctx_hw_stats *bnxt_re_stats = rdev->qplib_ctx.stats.dma;
+ int rc = 0;
if (!port || !stats)
return -EINVAL;
@@ -97,6 +137,91 @@ int bnxt_re_ib_get_hw_stats(struct ib_device *ibdev,
stats->value[BNXT_RE_TX_BYTES] =
le64_to_cpu(bnxt_re_stats->tx_ucast_bytes);
}
+ if (test_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS, &rdev->flags)) {
+ rc = bnxt_qplib_get_roce_stats(&rdev->rcfw, &rdev->stats);
+ if (rc)
+ clear_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS,
+ &rdev->flags);
+ stats->value[BNXT_RE_TO_RETRANSMITS] =
+ rdev->stats.to_retransmits;
+ stats->value[BNXT_RE_SEQ_ERR_NAKS_RCVD] =
+ rdev->stats.seq_err_naks_rcvd;
+ stats->value[BNXT_RE_MAX_RETRY_EXCEEDED] =
+ rdev->stats.max_retry_exceeded;
+ stats->value[BNXT_RE_RNR_NAKS_RCVD] =
+ rdev->stats.rnr_naks_rcvd;
+ stats->value[BNXT_RE_MISSING_RESP] =
+ rdev->stats.missing_resp;
+ stats->value[BNXT_RE_UNRECOVERABLE_ERR] =
+ rdev->stats.unrecoverable_err;
+ stats->value[BNXT_RE_BAD_RESP_ERR] =
+ rdev->stats.bad_resp_err;
+ stats->value[BNXT_RE_LOCAL_QP_OP_ERR] =
+ rdev->stats.local_qp_op_err;
+ stats->value[BNXT_RE_LOCAL_PROTECTION_ERR] =
+ rdev->stats.local_protection_err;
+ stats->value[BNXT_RE_MEM_MGMT_OP_ERR] =
+ rdev->stats.mem_mgmt_op_err;
+ stats->value[BNXT_RE_REMOTE_INVALID_REQ_ERR] =
+ rdev->stats.remote_invalid_req_err;
+ stats->value[BNXT_RE_REMOTE_ACCESS_ERR] =
+ rdev->stats.remote_access_err;
+ stats->value[BNXT_RE_REMOTE_OP_ERR] =
+ rdev->stats.remote_op_err;
+ stats->value[BNXT_RE_DUP_REQ] =
+ rdev->stats.dup_req;
+ stats->value[BNXT_RE_RES_EXCEED_MAX] =
+ rdev->stats.res_exceed_max;
+ stats->value[BNXT_RE_RES_LENGTH_MISMATCH] =
+ rdev->stats.res_length_mismatch;
+ stats->value[BNXT_RE_RES_EXCEEDS_WQE] =
+ rdev->stats.res_exceeds_wqe;
+ stats->value[BNXT_RE_RES_OPCODE_ERR] =
+ rdev->stats.res_opcode_err;
+ stats->value[BNXT_RE_RES_RX_INVALID_RKEY] =
+ rdev->stats.res_rx_invalid_rkey;
+ stats->value[BNXT_RE_RES_RX_DOMAIN_ERR] =
+ rdev->stats.res_rx_domain_err;
+ stats->value[BNXT_RE_RES_RX_NO_PERM] =
+ rdev->stats.res_rx_no_perm;
+ stats->value[BNXT_RE_RES_RX_RANGE_ERR] =
+ rdev->stats.res_rx_range_err;
+ stats->value[BNXT_RE_RES_TX_INVALID_RKEY] =
+ rdev->stats.res_tx_invalid_rkey;
+ stats->value[BNXT_RE_RES_TX_DOMAIN_ERR] =
+ rdev->stats.res_tx_domain_err;
+ stats->value[BNXT_RE_RES_TX_NO_PERM] =
+ rdev->stats.res_tx_no_perm;
+ stats->value[BNXT_RE_RES_TX_RANGE_ERR] =
+ rdev->stats.res_tx_range_err;
+ stats->value[BNXT_RE_RES_IRRQ_OFLOW] =
+ rdev->stats.res_irrq_oflow;
+ stats->value[BNXT_RE_RES_UNSUP_OPCODE] =
+ rdev->stats.res_unsup_opcode;
+ stats->value[BNXT_RE_RES_UNALIGNED_ATOMIC] =
+ rdev->stats.res_unaligned_atomic;
+ stats->value[BNXT_RE_RES_REM_INV_ERR] =
+ rdev->stats.res_rem_inv_err;
+ stats->value[BNXT_RE_RES_MEM_ERROR] =
+ rdev->stats.res_mem_error;
+ stats->value[BNXT_RE_RES_SRQ_ERR] =
+ rdev->stats.res_srq_err;
+ stats->value[BNXT_RE_RES_CMP_ERR] =
+ rdev->stats.res_cmp_err;
+ stats->value[BNXT_RE_RES_INVALID_DUP_RKEY] =
+ rdev->stats.res_invalid_dup_rkey;
+ stats->value[BNXT_RE_RES_WQE_FORMAT_ERR] =
+ rdev->stats.res_wqe_format_err;
+ stats->value[BNXT_RE_RES_CQ_LOAD_ERR] =
+ rdev->stats.res_cq_load_err;
+ stats->value[BNXT_RE_RES_SRQ_LOAD_ERR] =
+ rdev->stats.res_srq_load_err;
+ stats->value[BNXT_RE_RES_TX_PCI_ERR] =
+ rdev->stats.res_tx_pci_err;
+ stats->value[BNXT_RE_RES_RX_PCI_ERR] =
+ rdev->stats.res_rx_pci_err;
+ }
+
return ARRAY_SIZE(bnxt_re_stat_name);
}
diff --git a/drivers/infiniband/hw/bnxt_re/hw_counters.h b/drivers/infiniband/hw/bnxt_re/hw_counters.h
index be0dc00..a01a922 100644
--- a/drivers/infiniband/hw/bnxt_re/hw_counters.h
+++ b/drivers/infiniband/hw/bnxt_re/hw_counters.h
@@ -51,6 +51,45 @@ enum bnxt_re_hw_stats {
BNXT_RE_TX_PKTS,
BNXT_RE_TX_BYTES,
BNXT_RE_RECOVERABLE_ERRORS,
+ BNXT_RE_TO_RETRANSMITS,
+ BNXT_RE_SEQ_ERR_NAKS_RCVD,
+ BNXT_RE_MAX_RETRY_EXCEEDED,
+ BNXT_RE_RNR_NAKS_RCVD,
+ BNXT_RE_MISSING_RESP,
+ BNXT_RE_UNRECOVERABLE_ERR,
+ BNXT_RE_BAD_RESP_ERR,
+ BNXT_RE_LOCAL_QP_OP_ERR,
+ BNXT_RE_LOCAL_PROTECTION_ERR,
+ BNXT_RE_MEM_MGMT_OP_ERR,
+ BNXT_RE_REMOTE_INVALID_REQ_ERR,
+ BNXT_RE_REMOTE_ACCESS_ERR,
+ BNXT_RE_REMOTE_OP_ERR,
+ BNXT_RE_DUP_REQ,
+ BNXT_RE_RES_EXCEED_MAX,
+ BNXT_RE_RES_LENGTH_MISMATCH,
+ BNXT_RE_RES_EXCEEDS_WQE,
+ BNXT_RE_RES_OPCODE_ERR,
+ BNXT_RE_RES_RX_INVALID_RKEY,
+ BNXT_RE_RES_RX_DOMAIN_ERR,
+ BNXT_RE_RES_RX_NO_PERM,
+ BNXT_RE_RES_RX_RANGE_ERR,
+ BNXT_RE_RES_TX_INVALID_RKEY,
+ BNXT_RE_RES_TX_DOMAIN_ERR,
+ BNXT_RE_RES_TX_NO_PERM,
+ BNXT_RE_RES_TX_RANGE_ERR,
+ BNXT_RE_RES_IRRQ_OFLOW,
+ BNXT_RE_RES_UNSUP_OPCODE,
+ BNXT_RE_RES_UNALIGNED_ATOMIC,
+ BNXT_RE_RES_REM_INV_ERR,
+ BNXT_RE_RES_MEM_ERROR,
+ BNXT_RE_RES_SRQ_ERR,
+ BNXT_RE_RES_CMP_ERR,
+ BNXT_RE_RES_INVALID_DUP_RKEY,
+ BNXT_RE_RES_WQE_FORMAT_ERR,
+ BNXT_RE_RES_CQ_LOAD_ERR,
+ BNXT_RE_RES_SRQ_LOAD_ERR,
+ BNXT_RE_RES_TX_PCI_ERR,
+ BNXT_RE_RES_RX_PCI_ERR,
BNXT_RE_NUM_COUNTERS
};
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index 38a5d4b..f637ebe 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -1245,6 +1245,7 @@ static int bnxt_re_ib_reg(struct bnxt_re_dev *rdev)
set_bit(BNXT_RE_FLAG_IBDEV_REGISTERED, &rdev->flags);
ib_get_eth_speed(&rdev->ibdev, 1, &rdev->active_speed,
&rdev->active_width);
+ set_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS, &rdev->flags);
bnxt_re_dispatch_event(&rdev->ibdev, NULL, 1, IB_EVENT_PORT_ACTIVE);
bnxt_re_dispatch_event(&rdev->ibdev, NULL, 1, IB_EVENT_GID_CHANGE);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
index b805495..b1b7ded4 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
@@ -793,3 +793,73 @@ int bnxt_qplib_map_tc2cos(struct bnxt_qplib_res *res, u16 *cids)
0);
return 0;
}
+
+int bnxt_qplib_get_roce_stats(struct bnxt_qplib_rcfw *rcfw,
+ struct bnxt_qplib_roce_stats *stats)
+{
+ struct cmdq_query_roce_stats req;
+ struct creq_query_roce_stats_resp resp;
+ struct bnxt_qplib_rcfw_sbuf *sbuf;
+ struct creq_query_roce_stats_resp_sb *sb;
+ u16 cmd_flags = 0;
+ int rc = 0;
+
+ RCFW_CMD_PREP(req, QUERY_ROCE_STATS, cmd_flags);
+
+ sbuf = bnxt_qplib_rcfw_alloc_sbuf(rcfw, sizeof(*sb));
+ if (!sbuf) {
+ dev_err(&rcfw->pdev->dev,
+ "QPLIB: SP: QUERY_ROCE_STATS alloc side buffer failed");
+ return -ENOMEM;
+ }
+
+ sb = sbuf->sb;
+ req.resp_size = sizeof(*sb) / BNXT_QPLIB_CMDQE_UNITS;
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req, (void *)&resp,
+ (void *)sbuf, 0);
+ if (rc)
+ goto bail;
+ /* Extract the context from the side buffer */
+ stats->to_retransmits = le64_to_cpu(sb->to_retransmits);
+ stats->seq_err_naks_rcvd = le64_to_cpu(sb->seq_err_naks_rcvd);
+ stats->max_retry_exceeded = le64_to_cpu(sb->max_retry_exceeded);
+ stats->rnr_naks_rcvd = le64_to_cpu(sb->rnr_naks_rcvd);
+ stats->missing_resp = le64_to_cpu(sb->missing_resp);
+ stats->unrecoverable_err = le64_to_cpu(sb->unrecoverable_err);
+ stats->bad_resp_err = le64_to_cpu(sb->bad_resp_err);
+ stats->local_qp_op_err = le64_to_cpu(sb->local_qp_op_err);
+ stats->local_protection_err = le64_to_cpu(sb->local_protection_err);
+ stats->mem_mgmt_op_err = le64_to_cpu(sb->mem_mgmt_op_err);
+ stats->remote_invalid_req_err = le64_to_cpu(sb->remote_invalid_req_err);
+ stats->remote_access_err = le64_to_cpu(sb->remote_access_err);
+ stats->remote_op_err = le64_to_cpu(sb->remote_op_err);
+ stats->dup_req = le64_to_cpu(sb->dup_req);
+ stats->res_exceed_max = le64_to_cpu(sb->res_exceed_max);
+ stats->res_length_mismatch = le64_to_cpu(sb->res_length_mismatch);
+ stats->res_exceeds_wqe = le64_to_cpu(sb->res_exceeds_wqe);
+ stats->res_opcode_err = le64_to_cpu(sb->res_opcode_err);
+ stats->res_rx_invalid_rkey = le64_to_cpu(sb->res_rx_invalid_rkey);
+ stats->res_rx_domain_err = le64_to_cpu(sb->res_rx_domain_err);
+ stats->res_rx_no_perm = le64_to_cpu(sb->res_rx_no_perm);
+ stats->res_rx_range_err = le64_to_cpu(sb->res_rx_range_err);
+ stats->res_tx_invalid_rkey = le64_to_cpu(sb->res_tx_invalid_rkey);
+ stats->res_tx_domain_err = le64_to_cpu(sb->res_tx_domain_err);
+ stats->res_tx_no_perm = le64_to_cpu(sb->res_tx_no_perm);
+ stats->res_tx_range_err = le64_to_cpu(sb->res_tx_range_err);
+ stats->res_irrq_oflow = le64_to_cpu(sb->res_irrq_oflow);
+ stats->res_unsup_opcode = le64_to_cpu(sb->res_unsup_opcode);
+ stats->res_unaligned_atomic = le64_to_cpu(sb->res_unaligned_atomic);
+ stats->res_rem_inv_err = le64_to_cpu(sb->res_rem_inv_err);
+ stats->res_mem_error = le64_to_cpu(sb->res_mem_error);
+ stats->res_srq_err = le64_to_cpu(sb->res_srq_err);
+ stats->res_cmp_err = le64_to_cpu(sb->res_cmp_err);
+ stats->res_invalid_dup_rkey = le64_to_cpu(sb->res_invalid_dup_rkey);
+ stats->res_wqe_format_err = le64_to_cpu(sb->res_wqe_format_err);
+ stats->res_cq_load_err = le64_to_cpu(sb->res_cq_load_err);
+ stats->res_srq_load_err = le64_to_cpu(sb->res_srq_load_err);
+ stats->res_tx_pci_err = le64_to_cpu(sb->res_tx_pci_err);
+ stats->res_rx_pci_err = le64_to_cpu(sb->res_rx_pci_err);
+bail:
+ bnxt_qplib_rcfw_free_sbuf(rcfw, sbuf);
+ return rc;
+}
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.h b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
index 074e5e3..9d3e8b9 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.h
@@ -128,6 +128,85 @@ struct bnxt_qplib_frpl {
#define BNXT_QPLIB_ACCESS_ZERO_BASED BIT(5)
#define BNXT_QPLIB_ACCESS_ON_DEMAND BIT(6)
+struct bnxt_qplib_roce_stats {
+ u64 to_retransmits;
+ u64 seq_err_naks_rcvd;
+ /* seq_err_naks_rcvd is 64 b */
+ u64 max_retry_exceeded;
+ /* max_retry_exceeded is 64 b */
+ u64 rnr_naks_rcvd;
+ /* rnr_naks_rcvd is 64 b */
+ u64 missing_resp;
+ u64 unrecoverable_err;
+ /* unrecoverable_err is 64 b */
+ u64 bad_resp_err;
+ /* bad_resp_err is 64 b */
+ u64 local_qp_op_err;
+ /* local_qp_op_err is 64 b */
+ u64 local_protection_err;
+ /* local_protection_err is 64 b */
+ u64 mem_mgmt_op_err;
+ /* mem_mgmt_op_err is 64 b */
+ u64 remote_invalid_req_err;
+ /* remote_invalid_req_err is 64 b */
+ u64 remote_access_err;
+ /* remote_access_err is 64 b */
+ u64 remote_op_err;
+ /* remote_op_err is 64 b */
+ u64 dup_req;
+ /* dup_req is 64 b */
+ u64 res_exceed_max;
+ /* res_exceed_max is 64 b */
+ u64 res_length_mismatch;
+ /* res_length_mismatch is 64 b */
+ u64 res_exceeds_wqe;
+ /* res_exceeds_wqe is 64 b */
+ u64 res_opcode_err;
+ /* res_opcode_err is 64 b */
+ u64 res_rx_invalid_rkey;
+ /* res_rx_invalid_rkey is 64 b */
+ u64 res_rx_domain_err;
+ /* res_rx_domain_err is 64 b */
+ u64 res_rx_no_perm;
+ /* res_rx_no_perm is 64 b */
+ u64 res_rx_range_err;
+ /* res_rx_range_err is 64 b */
+ u64 res_tx_invalid_rkey;
+ /* res_tx_invalid_rkey is 64 b */
+ u64 res_tx_domain_err;
+ /* res_tx_domain_err is 64 b */
+ u64 res_tx_no_perm;
+ /* res_tx_no_perm is 64 b */
+ u64 res_tx_range_err;
+ /* res_tx_range_err is 64 b */
+ u64 res_irrq_oflow;
+ /* res_irrq_oflow is 64 b */
+ u64 res_unsup_opcode;
+ /* res_unsup_opcode is 64 b */
+ u64 res_unaligned_atomic;
+ /* res_unaligned_atomic is 64 b */
+ u64 res_rem_inv_err;
+ /* res_rem_inv_err is 64 b */
+ u64 res_mem_error;
+ /* res_mem_error is 64 b */
+ u64 res_srq_err;
+ /* res_srq_err is 64 b */
+ u64 res_cmp_err;
+ /* res_cmp_err is 64 b */
+ u64 res_invalid_dup_rkey;
+ /* res_invalid_dup_rkey is 64 b */
+ u64 res_wqe_format_err;
+ /* res_wqe_format_err is 64 b */
+ u64 res_cq_load_err;
+ /* res_cq_load_err is 64 b */
+ u64 res_srq_load_err;
+ /* res_srq_load_err is 64 b */
+ u64 res_tx_pci_err;
+ /* res_tx_pci_err is 64 b */
+ u64 res_rx_pci_err;
+ /* res_rx_pci_err is 64 b */
+};
+
int bnxt_qplib_get_sgid(struct bnxt_qplib_res *res,
struct bnxt_qplib_sgid_tbl *sgid_tbl, int index,
struct bnxt_qplib_gid *gid);
@@ -168,4 +247,6 @@ int bnxt_qplib_alloc_fast_reg_page_list(struct bnxt_qplib_res *res,
int bnxt_qplib_free_fast_reg_page_list(struct bnxt_qplib_res *res,
struct bnxt_qplib_frpl *frpl);
int bnxt_qplib_map_tc2cos(struct bnxt_qplib_res *res, u16 *cids);
+int bnxt_qplib_get_roce_stats(struct bnxt_qplib_rcfw *rcfw,
+ struct bnxt_qplib_roce_stats *stats);
#endif /* __BNXT_QPLIB_SP_H__*/
diff --git a/drivers/infiniband/hw/bnxt_re/roce_hsi.h b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
index 5cd31de..2d7ea09 100644
--- a/drivers/infiniband/hw/bnxt_re/roce_hsi.h
+++ b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
@@ -954,6 +954,7 @@ struct cmdq_base {
#define CMDQ_BASE_OPCODE_QUERY_VERSION 0x8bUL
#define CMDQ_BASE_OPCODE_MODIFY_CC 0x8cUL
#define CMDQ_BASE_OPCODE_QUERY_CC 0x8dUL
+ #define CMDQ_BASE_OPCODE_QUERY_ROCE_STATS 0x8eUL
u8 cmd_size;
__le16 flags;
__le16 cookie;
@@ -2049,6 +2050,20 @@ struct creq_modify_qp_resp {
__le16 reserved48[3];
};
+/* cmdq_query_roce_stats (size:128b/16B) */
+struct cmdq_query_roce_stats {
+ u8 opcode;
+ #define CMDQ_QUERY_ROCE_STATS_OPCODE_QUERY_ROCE_STATS 0x8eUL
+ #define CMDQ_QUERY_ROCE_STATS_OPCODE_LAST \
+ CMDQ_QUERY_ROCE_STATS_OPCODE_QUERY_ROCE_STATS
+ u8 cmd_size;
+ __le16 flags;
+ __le16 cookie;
+ u8 resp_size;
+ u8 reserved8;
+ __le64 resp_addr;
+};
+
/* Query QP command response (16 bytes) */
struct creq_query_qp_resp {
u8 type;
@@ -2819,6 +2834,80 @@ struct creq_query_cc_resp_sb {
__le64 reserved64_1;
};
+/* creq_query_roce_stats_resp (size:128b/16B) */
+struct creq_query_roce_stats_resp {
+ u8 type;
+ #define CREQ_QUERY_ROCE_STATS_RESP_TYPE_MASK 0x3fUL
+ #define CREQ_QUERY_ROCE_STATS_RESP_TYPE_SFT 0
+ #define CREQ_QUERY_ROCE_STATS_RESP_TYPE_QP_EVENT 0x38UL
+ #define CREQ_QUERY_ROCE_STATS_RESP_TYPE_LAST \
+ CREQ_QUERY_ROCE_STATS_RESP_TYPE_QP_EVENT
+ u8 status;
+ __le16 cookie;
+ __le32 size;
+ u8 v;
+ #define CREQ_QUERY_ROCE_STATS_RESP_V 0x1UL
+ u8 event;
+ #define CREQ_QUERY_ROCE_STATS_RESP_EVENT_QUERY_ROCE_STATS 0x8eUL
+ #define CREQ_QUERY_ROCE_STATS_RESP_EVENT_LAST \
+ CREQ_QUERY_ROCE_STATS_RESP_EVENT_QUERY_ROCE_STATS
+ u8 reserved48[6];
+};
+
+/* creq_query_roce_stats_resp_sb (size:2624b/328B) */
+struct creq_query_roce_stats_resp_sb {
+ u8 opcode;
+ #define CREQ_QUERY_ROCE_STATS_RESP_SB_OPCODE_QUERY_ROCE_STATS 0x8eUL
+ #define CREQ_QUERY_ROCE_STATS_RESP_SB_OPCODE_LAST \
+ CREQ_QUERY_ROCE_STATS_RESP_SB_OPCODE_QUERY_ROCE_STATS
+ u8 status;
+ __le16 cookie;
+ __le16 flags;
+ u8 resp_size;
+ u8 rsvd;
+ __le32 num_counters;
+ __le32 rsvd1;
+ __le64 to_retransmits;
+ __le64 seq_err_naks_rcvd;
+ __le64 max_retry_exceeded;
+ __le64 rnr_naks_rcvd;
+ __le64 missing_resp;
+ __le64 unrecoverable_err;
+ __le64 bad_resp_err;
+ __le64 local_qp_op_err;
+ __le64 local_protection_err;
+ __le64 mem_mgmt_op_err;
+ __le64 remote_invalid_req_err;
+ __le64 remote_access_err;
+ __le64 remote_op_err;
+ __le64 dup_req;
+ __le64 res_exceed_max;
+ __le64 res_length_mismatch;
+ __le64 res_exceeds_wqe;
+ __le64 res_opcode_err;
+ __le64 res_rx_invalid_rkey;
+ __le64 res_rx_domain_err;
+ __le64 res_rx_no_perm;
+ __le64 res_rx_range_err;
+ __le64 res_tx_invalid_rkey;
+ __le64 res_tx_domain_err;
+ __le64 res_tx_no_perm;
+ __le64 res_tx_range_err;
+ __le64 res_irrq_oflow;
+ __le64 res_unsup_opcode;
+ __le64 res_unaligned_atomic;
+ __le64 res_rem_inv_err;
+ __le64 res_mem_error;
+ __le64 res_srq_err;
+ __le64 res_cmp_err;
+ __le64 res_invalid_dup_rkey;
+ __le64 res_wqe_format_err;
+ __le64 res_cq_load_err;
+ __le64 res_srq_load_err;
+ __le64 res_tx_pci_err;
+ __le64 res_rx_pci_err;
+};
+
/* QP error notification event (16 bytes) */
struct creq_qp_error_notification {
u8 type;
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 14+ messages in thread* [for-next 5/5] RDMA/bnxt_re: Add SRQ support for Broadcom adapters
[not found] ` <1515152434-13105-1-git-send-email-devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
` (3 preceding siblings ...)
2018-01-05 11:40 ` [for-next 4/5] RDMA/bnxt_re: expose detailed stats retrieved from HW Devesh Sharma
@ 2018-01-05 11:40 ` Devesh Sharma
4 siblings, 0 replies; 14+ messages in thread
From: Devesh Sharma @ 2018-01-05 11:40 UTC (permalink / raw)
To: dledford-H+wXaHxf7aLQT0dZR+AlfA, jgg-VPRAkNaXOzVWk0Htik3J/w
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA, Devesh Sharma
Shared receive queue (SRQ) is defined as a pool of
receive buffers shared among multiple QPs which belong
to same protection domain in a given process context.
Use of SRQ reduces the memory foot print of IB applications.
Broadcom adapters support SRQ, adding code-changes to enable
shared receive queue.
Signed-off-by: Devesh Sharma <devesh.sharma-dY08KVG/lbpWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/bnxt_re/ib_verbs.c | 244 ++++++++++++++-
drivers/infiniband/hw/bnxt_re/ib_verbs.h | 19 ++
drivers/infiniband/hw/bnxt_re/main.c | 102 ++++++-
drivers/infiniband/hw/bnxt_re/qplib_fp.c | 462 +++++++++++++++++++++++++----
drivers/infiniband/hw/bnxt_re/qplib_fp.h | 78 +++--
drivers/infiniband/hw/bnxt_re/qplib_rcfw.c | 2 +-
drivers/infiniband/hw/bnxt_re/qplib_rcfw.h | 7 +-
include/uapi/rdma/bnxt_re-abi.h | 9 +
8 files changed, 825 insertions(+), 98 deletions(-)
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index c3135f6..7220322 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -1027,6 +1027,7 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd,
struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
struct bnxt_re_qp *qp;
struct bnxt_re_cq *cq;
+ struct bnxt_re_srq *srq;
int rc, entries;
if ((qp_init_attr->cap.max_send_wr > dev_attr->max_qp_wqes) ||
@@ -1082,9 +1083,15 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd,
}
if (qp_init_attr->srq) {
- dev_err(rdev_to_dev(rdev), "SRQ not supported");
- rc = -ENOTSUPP;
- goto fail;
+ srq = container_of(qp_init_attr->srq, struct bnxt_re_srq,
+ ib_srq);
+ if (!srq) {
+ dev_err(rdev_to_dev(rdev), "SRQ not found");
+ rc = -EINVAL;
+ goto fail;
+ }
+ qp->qplib_qp.srq = &srq->qplib_srq;
+ qp->qplib_qp.rq.max_wqe = 0;
} else {
/* Allocate 1 more than what's provided so posting max doesn't
* mean empty
@@ -1289,6 +1296,237 @@ static enum ib_mtu __to_ib_mtu(u32 mtu)
}
}
+/* Shared Receive Queues */
+int bnxt_re_destroy_srq(struct ib_srq *ib_srq)
+{
+ struct bnxt_re_srq *srq = container_of(ib_srq, struct bnxt_re_srq,
+ ib_srq);
+ struct bnxt_re_dev *rdev = srq->rdev;
+ struct bnxt_qplib_srq *qplib_srq = &srq->qplib_srq;
+ struct bnxt_qplib_nq *nq = NULL;
+ int rc;
+
+ if (qplib_srq->cq)
+ nq = qplib_srq->cq->nq;
+ rc = bnxt_qplib_destroy_srq(&rdev->qplib_res, qplib_srq);
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "Destroy HW SRQ failed!");
+ return rc;
+ }
+
+ if (srq->umem && !IS_ERR(srq->umem))
+ ib_umem_release(srq->umem);
+ kfree(srq);
+ atomic_dec(&rdev->srq_count);
+ if (nq)
+ nq->budget--;
+ return 0;
+}
+
+static int bnxt_re_init_user_srq(struct bnxt_re_dev *rdev,
+ struct bnxt_re_pd *pd,
+ struct bnxt_re_srq *srq,
+ struct ib_udata *udata)
+{
+ struct bnxt_re_srq_req ureq;
+ struct bnxt_qplib_srq *qplib_srq = &srq->qplib_srq;
+ struct ib_umem *umem;
+ int bytes = 0;
+ struct ib_ucontext *context = pd->ib_pd.uobject->context;
+ struct bnxt_re_ucontext *cntx = container_of(context,
+ struct bnxt_re_ucontext,
+ ib_uctx);
+ if (ib_copy_from_udata(&ureq, udata, sizeof(ureq)))
+ return -EFAULT;
+
+ bytes = (qplib_srq->max_wqe * BNXT_QPLIB_MAX_RQE_ENTRY_SIZE);
+ bytes = PAGE_ALIGN(bytes);
+ umem = ib_umem_get(context, ureq.srqva, bytes,
+ IB_ACCESS_LOCAL_WRITE, 1);
+ if (IS_ERR(umem))
+ return PTR_ERR(umem);
+
+ srq->umem = umem;
+ qplib_srq->nmap = umem->nmap;
+ qplib_srq->sglist = umem->sg_head.sgl;
+ qplib_srq->srq_handle = ureq.srq_handle;
+ qplib_srq->dpi = &cntx->dpi;
+
+ return 0;
+}
+
+struct ib_srq *bnxt_re_create_srq(struct ib_pd *ib_pd,
+ struct ib_srq_init_attr *srq_init_attr,
+ struct ib_udata *udata)
+{
+ struct bnxt_re_pd *pd = container_of(ib_pd, struct bnxt_re_pd, ib_pd);
+ struct bnxt_re_dev *rdev = pd->rdev;
+ struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
+ struct bnxt_re_srq *srq;
+ struct bnxt_qplib_nq *nq = NULL;
+ int rc, entries;
+
+ if (srq_init_attr->attr.max_wr >= dev_attr->max_srq_wqes) {
+ dev_err(rdev_to_dev(rdev), "Create CQ failed - max exceeded");
+ rc = -EINVAL;
+ goto exit;
+ }
+
+ if (srq_init_attr->srq_type != IB_SRQT_BASIC) {
+ rc = -ENOTSUPP;
+ goto exit;
+ }
+
+ srq = kzalloc(sizeof(*srq), GFP_KERNEL);
+ if (!srq) {
+ rc = -ENOMEM;
+ goto exit;
+ }
+ srq->rdev = rdev;
+ srq->qplib_srq.pd = &pd->qplib_pd;
+ srq->qplib_srq.dpi = &rdev->dpi_privileged;
+ /* Allocate 1 more than what's provided so posting max doesn't
+ * mean empty
+ */
+ entries = roundup_pow_of_two(srq_init_attr->attr.max_wr + 1);
+ if (entries > dev_attr->max_srq_wqes + 1)
+ entries = dev_attr->max_srq_wqes + 1;
+
+ srq->qplib_srq.max_wqe = entries;
+ srq->qplib_srq.max_sge = srq_init_attr->attr.max_sge;
+ srq->qplib_srq.threshold = srq_init_attr->attr.srq_limit;
+ srq->srq_limit = srq_init_attr->attr.srq_limit;
+ srq->qplib_srq.eventq_hw_ring_id = rdev->nq[0].ring_id;
+ nq = &rdev->nq[0];
+
+ if (udata) {
+ rc = bnxt_re_init_user_srq(rdev, pd, srq, udata);
+ if (rc)
+ goto fail;
+ }
+
+ rc = bnxt_qplib_create_srq(&rdev->qplib_res, &srq->qplib_srq);
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "Create HW SRQ failed!");
+ goto fail;
+ }
+
+ if (udata) {
+ struct bnxt_re_srq_resp resp;
+
+ resp.srqid = srq->qplib_srq.id;
+ rc = ib_copy_to_udata(udata, &resp, sizeof(resp));
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "SRQ copy to udata failed!");
+ bnxt_qplib_destroy_srq(&rdev->qplib_res,
+ &srq->qplib_srq);
+ goto exit;
+ }
+ }
+ if (nq)
+ nq->budget++;
+ atomic_inc(&rdev->srq_count);
+
+ return &srq->ib_srq;
+
+fail:
+ if (udata && srq->umem && !IS_ERR(srq->umem)) {
+ ib_umem_release(srq->umem);
+ srq->umem = NULL;
+ }
+
+ kfree(srq);
+exit:
+ return ERR_PTR(rc);
+}
+
+int bnxt_re_modify_srq(struct ib_srq *ib_srq, struct ib_srq_attr *srq_attr,
+ enum ib_srq_attr_mask srq_attr_mask,
+ struct ib_udata *udata)
+{
+ struct bnxt_re_srq *srq = container_of(ib_srq, struct bnxt_re_srq,
+ ib_srq);
+ struct bnxt_re_dev *rdev = srq->rdev;
+ int rc;
+
+ switch (srq_attr_mask) {
+ case IB_SRQ_MAX_WR:
+ /* SRQ resize is not supported */
+ break;
+ case IB_SRQ_LIMIT:
+ /* Change the SRQ threshold */
+ if (srq_attr->srq_limit > srq->qplib_srq.max_wqe)
+ return -EINVAL;
+
+ srq->qplib_srq.threshold = srq_attr->srq_limit;
+ rc = bnxt_qplib_modify_srq(&rdev->qplib_res, &srq->qplib_srq);
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "Modify HW SRQ failed!");
+ return rc;
+ }
+ /* On success, update the shadow */
+ srq->srq_limit = srq_attr->srq_limit;
+ /* No need to Build and send response back to udata */
+ break;
+ default:
+ dev_err(rdev_to_dev(rdev),
+ "Unsupported srq_attr_mask 0x%x", srq_attr_mask);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+int bnxt_re_query_srq(struct ib_srq *ib_srq, struct ib_srq_attr *srq_attr)
+{
+ struct bnxt_re_srq *srq = container_of(ib_srq, struct bnxt_re_srq,
+ ib_srq);
+ struct bnxt_re_srq tsrq;
+ struct bnxt_re_dev *rdev = srq->rdev;
+ int rc;
+
+ /* Get live SRQ attr */
+ tsrq.qplib_srq.id = srq->qplib_srq.id;
+ rc = bnxt_qplib_query_srq(&rdev->qplib_res, &tsrq.qplib_srq);
+ if (rc) {
+ dev_err(rdev_to_dev(rdev), "Query HW SRQ failed!");
+ return rc;
+ }
+ srq_attr->max_wr = srq->qplib_srq.max_wqe;
+ srq_attr->max_sge = srq->qplib_srq.max_sge;
+ srq_attr->srq_limit = tsrq.qplib_srq.threshold;
+
+ return 0;
+}
+
+int bnxt_re_post_srq_recv(struct ib_srq *ib_srq, struct ib_recv_wr *wr,
+ struct ib_recv_wr **bad_wr)
+{
+ struct bnxt_re_srq *srq = container_of(ib_srq, struct bnxt_re_srq,
+ ib_srq);
+ struct bnxt_qplib_swqe wqe;
+ unsigned long flags;
+ int rc = 0, payload_sz = 0;
+
+ spin_lock_irqsave(&srq->lock, flags);
+ while (wr) {
+ /* Transcribe each ib_recv_wr to qplib_swqe */
+ wqe.num_sge = wr->num_sge;
+ payload_sz = bnxt_re_build_sgl(wr->sg_list, wqe.sg_list,
+ wr->num_sge);
+ wqe.wr_id = wr->wr_id;
+ wqe.type = BNXT_QPLIB_SWQE_TYPE_RECV;
+
+ rc = bnxt_qplib_post_srq_recv(&srq->qplib_srq, &wqe);
+ if (rc) {
+ *bad_wr = wr;
+ break;
+ }
+ wr = wr->next;
+ }
+ spin_unlock_irqrestore(&srq->lock, flags);
+
+ return rc;
+}
static int bnxt_re_modify_shadow_qp(struct bnxt_re_dev *rdev,
struct bnxt_re_qp *qp1_qp,
int qp_attr_mask)
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index 66dd8d2..423ebe0 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -68,6 +68,15 @@ struct bnxt_re_ah {
struct bnxt_qplib_ah qplib_ah;
};
+struct bnxt_re_srq {
+ struct bnxt_re_dev *rdev;
+ u32 srq_limit;
+ struct ib_srq ib_srq;
+ struct bnxt_qplib_srq qplib_srq;
+ struct ib_umem *umem;
+ spinlock_t lock; /* protect srq */
+};
+
struct bnxt_re_qp {
struct list_head list;
struct bnxt_re_dev *rdev;
@@ -165,6 +174,16 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *pd,
int bnxt_re_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
int bnxt_re_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
int bnxt_re_destroy_ah(struct ib_ah *ah);
+struct ib_srq *bnxt_re_create_srq(struct ib_pd *pd,
+ struct ib_srq_init_attr *srq_init_attr,
+ struct ib_udata *udata);
+int bnxt_re_modify_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr,
+ enum ib_srq_attr_mask srq_attr_mask,
+ struct ib_udata *udata);
+int bnxt_re_query_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr);
+int bnxt_re_destroy_srq(struct ib_srq *srq);
+int bnxt_re_post_srq_recv(struct ib_srq *srq, struct ib_recv_wr *recv_wr,
+ struct ib_recv_wr **bad_recv_wr);
struct ib_qp *bnxt_re_create_qp(struct ib_pd *pd,
struct ib_qp_init_attr *qp_init_attr,
struct ib_udata *udata);
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index f637ebe..ed84dd1 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -588,6 +588,12 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
ibdev->query_ah = bnxt_re_query_ah;
ibdev->destroy_ah = bnxt_re_destroy_ah;
+ ibdev->create_srq = bnxt_re_create_srq;
+ ibdev->modify_srq = bnxt_re_modify_srq;
+ ibdev->query_srq = bnxt_re_query_srq;
+ ibdev->destroy_srq = bnxt_re_destroy_srq;
+ ibdev->post_srq_recv = bnxt_re_post_srq_recv;
+
ibdev->create_qp = bnxt_re_create_qp;
ibdev->modify_qp = bnxt_re_modify_qp;
ibdev->query_qp = bnxt_re_query_qp;
@@ -689,10 +695,10 @@ static struct bnxt_re_dev *bnxt_re_dev_add(struct net_device *netdev,
return rdev;
}
-static int bnxt_re_aeq_handler(struct bnxt_qplib_rcfw *rcfw,
- struct creq_func_event *aeqe)
+static int bnxt_re_handle_unaffi_async_event(struct creq_func_event
+ *unaffi_async)
{
- switch (aeqe->event) {
+ switch (unaffi_async->event) {
case CREQ_FUNC_EVENT_EVENT_TX_WQE_ERROR:
break;
case CREQ_FUNC_EVENT_EVENT_TX_DATA_ERROR:
@@ -721,6 +727,93 @@ static int bnxt_re_aeq_handler(struct bnxt_qplib_rcfw *rcfw,
return 0;
}
+static int bnxt_re_handle_qp_async_event(struct creq_qp_event *qp_event,
+ struct bnxt_re_qp *qp)
+{
+ struct ib_event event;
+
+ memset(&event, 0, sizeof(event));
+ if (qp->qplib_qp.srq) {
+ event.device = &qp->rdev->ibdev;
+ event.element.qp = &qp->ib_qp;
+ event.event = IB_EVENT_QP_LAST_WQE_REACHED;
+ }
+
+ if (event.device && qp->ib_qp.event_handler)
+ qp->ib_qp.event_handler(&event, qp->ib_qp.qp_context);
+
+ return 0;
+}
+
+static int bnxt_re_handle_affi_async_event(struct creq_qp_event *affi_async,
+ void *obj)
+{
+ int rc = 0;
+ u8 event;
+
+ if (!obj)
+ return rc; /* QP was already dead, still return success */
+
+ event = affi_async->event;
+ if (event == CREQ_QP_EVENT_EVENT_QP_ERROR_NOTIFICATION) {
+ struct bnxt_qplib_qp *lib_qp = obj;
+ struct bnxt_re_qp *qp = container_of(lib_qp, struct bnxt_re_qp,
+ qplib_qp);
+ rc = bnxt_re_handle_qp_async_event(affi_async, qp);
+ }
+ return rc;
+}
+
+static int bnxt_re_aeq_handler(struct bnxt_qplib_rcfw *rcfw,
+ void *aeqe, void *obj)
+{
+ struct creq_qp_event *affi_async;
+ struct creq_func_event *unaffi_async;
+ u8 type;
+ int rc;
+
+ type = ((struct creq_base *)aeqe)->type;
+ if (type == CREQ_BASE_TYPE_FUNC_EVENT) {
+ unaffi_async = aeqe;
+ rc = bnxt_re_handle_unaffi_async_event(unaffi_async);
+ } else {
+ affi_async = aeqe;
+ rc = bnxt_re_handle_affi_async_event(affi_async, obj);
+ }
+
+ return rc;
+}
+
+static int bnxt_re_srqn_handler(struct bnxt_qplib_nq *nq,
+ struct bnxt_qplib_srq *handle, u8 event)
+{
+ struct bnxt_re_srq *srq = container_of(handle, struct bnxt_re_srq,
+ qplib_srq);
+ struct ib_event ib_event;
+ int rc = 0;
+
+ if (!srq) {
+ dev_err(NULL, "%s: SRQ is NULL, SRQN not handled",
+ ROCE_DRV_MODULE_NAME);
+ rc = -EINVAL;
+ goto done;
+ }
+ ib_event.device = &srq->rdev->ibdev;
+ ib_event.element.srq = &srq->ib_srq;
+ if (event == NQ_SRQ_EVENT_EVENT_SRQ_THRESHOLD_EVENT)
+ ib_event.event = IB_EVENT_SRQ_LIMIT_REACHED;
+ else
+ ib_event.event = IB_EVENT_SRQ_ERR;
+
+ if (srq->ib_srq.event_handler) {
+ /* Lock event_handler? */
+ (*srq->ib_srq.event_handler)(&ib_event,
+ srq->ib_srq.srq_context);
+ }
+done:
+ return rc;
+}
+
static int bnxt_re_cqn_handler(struct bnxt_qplib_nq *nq,
struct bnxt_qplib_cq *handle)
{
@@ -763,7 +856,8 @@ static int bnxt_re_init_res(struct bnxt_re_dev *rdev)
rc = bnxt_qplib_enable_nq(rdev->en_dev->pdev, &rdev->nq[i - 1],
i - 1, rdev->msix_entries[i].vector,
rdev->msix_entries[i].db_offset,
- &bnxt_re_cqn_handler, NULL);
+ &bnxt_re_cqn_handler,
+ &bnxt_re_srqn_handler);
if (rc) {
dev_err(rdev_to_dev(rdev),
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.c b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
index eb7195c..8b5f11a 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
@@ -52,6 +52,7 @@
static void bnxt_qplib_arm_cq_enable(struct bnxt_qplib_cq *cq);
static void __clean_cq(struct bnxt_qplib_cq *cq, u64 qp);
+static void bnxt_qplib_arm_srq(struct bnxt_qplib_srq *srq, u32 arm_type);
static void bnxt_qplib_cancel_phantom_processing(struct bnxt_qplib_qp *qp)
{
@@ -278,6 +279,7 @@ static void bnxt_qplib_service_nq(unsigned long data)
struct nq_base *nqe, **nq_ptr;
struct bnxt_qplib_cq *cq;
int num_cqne_processed = 0;
+ int num_srqne_processed = 0;
u32 sw_cons, raw_cons;
u16 type;
int budget = nq->budget;
@@ -320,6 +322,26 @@ static void bnxt_qplib_service_nq(unsigned long data)
spin_unlock_bh(&cq->compl_lock);
break;
}
+ case NQ_BASE_TYPE_SRQ_EVENT:
+ {
+ struct nq_srq_event *nqsrqe =
+ (struct nq_srq_event *)nqe;
+
+ q_handle = le32_to_cpu(nqsrqe->srq_handle_low);
+ q_handle |= (u64)le32_to_cpu(nqsrqe->srq_handle_high)
+ << 32;
+ bnxt_qplib_arm_srq((struct bnxt_qplib_srq *)q_handle,
+ DBR_DBR_TYPE_SRQ_ARMENA);
+ if (!nq->srqn_handler(nq,
+ (struct bnxt_qplib_srq *)q_handle,
+ nqsrqe->event))
+ num_srqne_processed++;
+ else
+ dev_warn(&nq->pdev->dev,
+ "QPLIB: SRQ event 0x%x not handled",
+ nqsrqe->event);
+ break;
+ }
case NQ_BASE_TYPE_DBQ_EVENT:
break;
default:
@@ -384,17 +406,19 @@ int bnxt_qplib_enable_nq(struct pci_dev *pdev, struct bnxt_qplib_nq *nq,
int (*cqn_handler)(struct bnxt_qplib_nq *nq,
struct bnxt_qplib_cq *),
int (*srqn_handler)(struct bnxt_qplib_nq *nq,
- void *, u8 event))
+ struct bnxt_qplib_srq *,
+ u8 event))
{
resource_size_t nq_base;
int rc = -1;
nq->pdev = pdev;
nq->vector = msix_vector;
+ if (cqn_handler)
+ nq->cqn_handler = cqn_handler;
- nq->cqn_handler = cqn_handler;
-
- nq->srqn_handler = srqn_handler;
+ if (srqn_handler)
+ nq->srqn_handler = srqn_handler;
tasklet_init(&nq->worker, bnxt_qplib_service_nq, (unsigned long)nq);
@@ -468,6 +492,238 @@ int bnxt_qplib_alloc_nq(struct pci_dev *pdev, struct bnxt_qplib_nq *nq)
return 0;
}
+/* SRQ */
+static void bnxt_qplib_arm_srq(struct bnxt_qplib_srq *srq, u32 arm_type)
+{
+ struct bnxt_qplib_hwq *srq_hwq = &srq->hwq;
+ struct dbr_dbr db_msg = { 0 };
+ void __iomem *db;
+ u32 sw_prod = 0;
+
+ /* Ring DB */
+ sw_prod = (arm_type == DBR_DBR_TYPE_SRQ_ARM) ? srq->threshold :
+ HWQ_CMP(srq_hwq->prod, srq_hwq);
+ db_msg.index = cpu_to_le32((sw_prod << DBR_DBR_INDEX_SFT) &
+ DBR_DBR_INDEX_MASK);
+ db_msg.type_xid = cpu_to_le32(((srq->id << DBR_DBR_XID_SFT) &
+ DBR_DBR_XID_MASK) | arm_type);
+ db = (arm_type == DBR_DBR_TYPE_SRQ_ARMENA) ?
+ srq->dbr_base : srq->dpi->dbr;
+ wmb(); /* barrier before db ring */
+ __iowrite64_copy(db, &db_msg, sizeof(db_msg) / sizeof(u64));
+}
+
+int bnxt_qplib_destroy_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq)
+{
+ struct bnxt_qplib_rcfw *rcfw = res->rcfw;
+ struct cmdq_destroy_srq req;
+ struct creq_destroy_srq_resp resp;
+ u16 cmd_flags = 0;
+ int rc;
+
+ RCFW_CMD_PREP(req, DESTROY_SRQ, cmd_flags);
+
+ /* Configure the request */
+ req.srq_cid = cpu_to_le32(srq->id);
+
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req,
+ (void *)&resp, NULL, 0);
+ if (rc)
+ return rc;
+
+ bnxt_qplib_free_hwq(res->pdev, &srq->hwq);
+ kfree(srq->swq);
+ return 0;
+}
+
+int bnxt_qplib_create_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq)
+{
+ struct bnxt_qplib_rcfw *rcfw = res->rcfw;
+ struct cmdq_create_srq req;
+ struct creq_create_srq_resp resp;
+ struct bnxt_qplib_pbl *pbl;
+ u16 cmd_flags = 0;
+ int rc, idx;
+
+ srq->hwq.max_elements = srq->max_wqe;
+ rc = bnxt_qplib_alloc_init_hwq(res->pdev, &srq->hwq, srq->sglist,
+ srq->nmap, &srq->hwq.max_elements,
+ BNXT_QPLIB_MAX_RQE_ENTRY_SIZE, 0,
+ PAGE_SIZE, HWQ_TYPE_QUEUE);
+ if (rc)
+ goto exit;
+
+ srq->swq = kcalloc(srq->hwq.max_elements, sizeof(*srq->swq),
+ GFP_KERNEL);
+ if (!srq->swq)
+ goto fail;
+
+ RCFW_CMD_PREP(req, CREATE_SRQ, cmd_flags);
+
+ /* Configure the request */
+ req.dpi = cpu_to_le32(srq->dpi->dpi);
+ req.srq_handle = cpu_to_le64(srq);
+
+ req.srq_size = cpu_to_le16((u16)srq->hwq.max_elements);
+ pbl = &srq->hwq.pbl[PBL_LVL_0];
+ req.pg_size_lvl = cpu_to_le16((((u16)srq->hwq.level &
+ CMDQ_CREATE_SRQ_LVL_MASK) <<
+ CMDQ_CREATE_SRQ_LVL_SFT) |
+ (pbl->pg_size == ROCE_PG_SIZE_4K ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_4K :
+ pbl->pg_size == ROCE_PG_SIZE_8K ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_8K :
+ pbl->pg_size == ROCE_PG_SIZE_64K ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_64K :
+ pbl->pg_size == ROCE_PG_SIZE_2M ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_2M :
+ pbl->pg_size == ROCE_PG_SIZE_8M ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_8M :
+ pbl->pg_size == ROCE_PG_SIZE_1G ?
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_1G :
+ CMDQ_CREATE_SRQ_PG_SIZE_PG_4K));
+ req.pbl = cpu_to_le64(pbl->pg_map_arr[0]);
+ req.pd_id = cpu_to_le32(srq->pd->id);
+ req.eventq_id = cpu_to_le16(srq->eventq_hw_ring_id);
+
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req,
+ (void *)&resp, NULL, 0);
+ if (rc)
+ goto fail;
+
+ spin_lock_init(&srq->lock);
+ srq->start_idx = 0;
+ srq->last_idx = srq->hwq.max_elements - 1;
+ for (idx = 0; idx < srq->hwq.max_elements; idx++)
+ srq->swq[idx].next_idx = idx + 1;
+ srq->swq[srq->last_idx].next_idx = -1;
+
+ srq->id = le32_to_cpu(resp.xid);
+ srq->dbr_base = res->dpi_tbl.dbr_bar_reg_iomem;
+ if (srq->threshold)
+ bnxt_qplib_arm_srq(srq, DBR_DBR_TYPE_SRQ_ARMENA);
+ srq->arm_req = false;
+
+ return 0;
+fail:
+ bnxt_qplib_free_hwq(res->pdev, &srq->hwq);
+ kfree(srq->swq);
+exit:
+ return rc;
+}
+
+int bnxt_qplib_modify_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq)
+{
+ struct bnxt_qplib_hwq *srq_hwq = &srq->hwq;
+ u32 sw_prod, sw_cons, count = 0;
+
+ sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
+ sw_cons = HWQ_CMP(srq_hwq->cons, srq_hwq);
+
+ count = sw_prod > sw_cons ? sw_prod - sw_cons :
+ srq_hwq->max_elements - sw_cons + sw_prod;
+ if (count > srq->threshold) {
+ srq->arm_req = false;
+ bnxt_qplib_arm_srq(srq, DBR_DBR_TYPE_SRQ_ARM);
+ } else {
+ /* Deferred arming */
+ srq->arm_req = true;
+ }
+
+ return 0;
+}
+
+int bnxt_qplib_query_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq)
+{
+ struct bnxt_qplib_rcfw *rcfw = res->rcfw;
+ struct cmdq_query_srq req;
+ struct creq_query_srq_resp resp;
+ struct bnxt_qplib_rcfw_sbuf *sbuf;
+ struct creq_query_srq_resp_sb *sb;
+ u16 cmd_flags = 0;
+ int rc = 0;
+
+ RCFW_CMD_PREP(req, QUERY_SRQ, cmd_flags);
+ req.srq_cid = cpu_to_le32(srq->id);
+
+ /* Configure the request */
+ sbuf = bnxt_qplib_rcfw_alloc_sbuf(rcfw, sizeof(*sb));
+ if (!sbuf)
+ return -ENOMEM;
+ sb = sbuf->sb;
+ rc = bnxt_qplib_rcfw_send_message(rcfw, (void *)&req, (void *)&resp,
+ (void *)sbuf, 0);
+ srq->threshold = le16_to_cpu(sb->srq_limit);
+ bnxt_qplib_rcfw_free_sbuf(rcfw, sbuf);
+
+ return rc;
+}
+
+int bnxt_qplib_post_srq_recv(struct bnxt_qplib_srq *srq,
+ struct bnxt_qplib_swqe *wqe)
+{
+ struct bnxt_qplib_hwq *srq_hwq = &srq->hwq;
+ struct rq_wqe *srqe, **srqe_ptr;
+ struct sq_sge *hw_sge;
+ u32 sw_prod, sw_cons, count = 0;
+ int i, rc = 0, next;
+
+ spin_lock(&srq_hwq->lock);
+ if (srq->start_idx == srq->last_idx) {
+ dev_err(&srq_hwq->pdev->dev, "QPLIB: FP: SRQ (0x%x) is full!",
+ srq->id);
+ rc = -EINVAL;
+ spin_unlock(&srq_hwq->lock);
+ goto done;
+ }
+ next = srq->start_idx;
+ srq->start_idx = srq->swq[next].next_idx;
+ spin_unlock(&srq_hwq->lock);
+
+ sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
+ srqe_ptr = (struct rq_wqe **)srq_hwq->pbl_ptr;
+ srqe = &srqe_ptr[RQE_PG(sw_prod)][RQE_IDX(sw_prod)];
+ memset(srqe, 0, BNXT_QPLIB_MAX_RQE_ENTRY_SIZE);
+ /* Calculate wqe_size16 and data_len */
+ for (i = 0, hw_sge = (struct sq_sge *)srqe->data;
+ i < wqe->num_sge; i++, hw_sge++) {
+ hw_sge->va_or_pa = cpu_to_le64(wqe->sg_list[i].addr);
+ hw_sge->l_key = cpu_to_le32(wqe->sg_list[i].lkey);
+ hw_sge->size = cpu_to_le32(wqe->sg_list[i].size);
+ }
+ srqe->wqe_type = wqe->type;
+ srqe->flags = wqe->flags;
+ srqe->wqe_size = wqe->num_sge +
+ ((offsetof(typeof(*srqe), data) + 15) >> 4);
+ srqe->wr_id[0] = cpu_to_le32((u32)next);
+ srq->swq[next].wr_id = wqe->wr_id;
+
+ srq_hwq->prod++;
+
+ spin_lock(&srq_hwq->lock);
+ sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
+ /* retaining srq_hwq->cons for this logic
+ * actually the lock is only required to
+ * read srq_hwq->cons.
+ */
+ sw_cons = HWQ_CMP(srq_hwq->cons, srq_hwq);
+ count = sw_prod > sw_cons ? sw_prod - sw_cons :
+ srq_hwq->max_elements - sw_cons + sw_prod;
+ spin_unlock(&srq_hwq->lock);
+ /* Ring DB */
+ bnxt_qplib_arm_srq(srq, DBR_DBR_TYPE_SRQ);
+ if (srq->arm_req == true && count > srq->threshold) {
+ srq->arm_req = false;
+ bnxt_qplib_arm_srq(srq, DBR_DBR_TYPE_SRQ_ARM);
+ }
+done:
+ return rc;
+}
+
/* QP */
int bnxt_qplib_create_qp1(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp)
{
@@ -736,6 +992,12 @@ int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp)
pbl->pg_size == ROCE_PG_SIZE_1G ?
CMDQ_CREATE_QP_RQ_PG_SIZE_PG_1G :
CMDQ_CREATE_QP_RQ_PG_SIZE_PG_4K);
+ } else {
+ /* SRQ */
+ if (qp->srq) {
+ qp_flags |= CMDQ_CREATE_QP_QP_FLAGS_SRQ_USED;
+ req.srq_cid = cpu_to_le32(qp->srq->id);
+ }
}
if (qp->rcq)
@@ -2067,6 +2329,16 @@ static int bnxt_qplib_cq_process_req(struct bnxt_qplib_cq *cq,
return rc;
}
+static void bnxt_qplib_release_srqe(struct bnxt_qplib_srq *srq, u32 tag)
+{
+ spin_lock(&srq->hwq.lock);
+ srq->swq[srq->last_idx].next_idx = (int)tag;
+ srq->last_idx = (int)tag;
+ srq->swq[srq->last_idx].next_idx = -1;
+ srq->hwq.cons++; /* Support for SRQE counter */
+ spin_unlock(&srq->hwq.lock);
+}
+
static int bnxt_qplib_cq_process_res_rc(struct bnxt_qplib_cq *cq,
struct cq_res_rc *hwcqe,
struct bnxt_qplib_cqe **pcqe,
@@ -2074,6 +2346,7 @@ static int bnxt_qplib_cq_process_res_rc(struct bnxt_qplib_cq *cq,
{
struct bnxt_qplib_qp *qp;
struct bnxt_qplib_q *rq;
+ struct bnxt_qplib_srq *srq;
struct bnxt_qplib_cqe *cqe;
u32 wr_id_idx;
int rc = 0;
@@ -2101,27 +2374,46 @@ static int bnxt_qplib_cq_process_res_rc(struct bnxt_qplib_cq *cq,
wr_id_idx = le32_to_cpu(hwcqe->srq_or_rq_wr_id) &
CQ_RES_RC_SRQ_OR_RQ_WR_ID_MASK;
- rq = &qp->rq;
- if (wr_id_idx > rq->hwq.max_elements) {
- dev_err(&cq->hwq.pdev->dev, "QPLIB: FP: CQ Process RC ");
- dev_err(&cq->hwq.pdev->dev,
- "QPLIB: wr_id idx 0x%x exceeded RQ max 0x%x",
- wr_id_idx, rq->hwq.max_elements);
- return -EINVAL;
- }
-
- cqe->wr_id = rq->swq[wr_id_idx].wr_id;
- cqe++;
- (*budget)--;
- rq->hwq.cons++;
- *pcqe = cqe;
+ if (cqe->flags & CQ_RES_RC_FLAGS_SRQ_SRQ) {
+ srq = qp->srq;
+ if (!srq)
+ return -EINVAL;
+ if (wr_id_idx > srq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process RC ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: wr_id idx 0x%x exceeded SRQ max 0x%x",
+ wr_id_idx, srq->hwq.max_elements);
+ return -EINVAL;
+ }
+ cqe->wr_id = srq->swq[wr_id_idx].wr_id;
+ bnxt_qplib_release_srqe(srq, wr_id_idx);
+ cqe++;
+ (*budget)--;
+ *pcqe = cqe;
+ } else {
+ rq = &qp->rq;
+ if (wr_id_idx > rq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process RC ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: wr_id idx 0x%x exceeded RQ max 0x%x",
+ wr_id_idx, rq->hwq.max_elements);
+ return -EINVAL;
+ }
+ cqe->wr_id = rq->swq[wr_id_idx].wr_id;
+ cqe++;
+ (*budget)--;
+ rq->hwq.cons++;
+ *pcqe = cqe;
- if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
- qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
- /* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
+ qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
+ /* Add qp to flush list of the CQ */
+ bnxt_qplib_lock_buddy_cq(qp, cq);
+ __bnxt_qplib_add_flush_qp(qp);
+ bnxt_qplib_unlock_buddy_cq(qp, cq);
+ }
}
done:
@@ -2135,6 +2427,7 @@ static int bnxt_qplib_cq_process_res_ud(struct bnxt_qplib_cq *cq,
{
struct bnxt_qplib_qp *qp;
struct bnxt_qplib_q *rq;
+ struct bnxt_qplib_srq *srq;
struct bnxt_qplib_cqe *cqe;
u32 wr_id_idx;
int rc = 0;
@@ -2165,27 +2458,48 @@ static int bnxt_qplib_cq_process_res_ud(struct bnxt_qplib_cq *cq,
hwcqe->src_qp_high_srq_or_rq_wr_id) &
CQ_RES_UD_SRC_QP_HIGH_MASK) >> 8);
- rq = &qp->rq;
- if (wr_id_idx > rq->hwq.max_elements) {
- dev_err(&cq->hwq.pdev->dev, "QPLIB: FP: CQ Process UD ");
- dev_err(&cq->hwq.pdev->dev,
- "QPLIB: wr_id idx %#x exceeded RQ max %#x",
- wr_id_idx, rq->hwq.max_elements);
- return -EINVAL;
- }
+ if (cqe->flags & CQ_RES_RC_FLAGS_SRQ_SRQ) {
+ srq = qp->srq;
+ if (!srq)
+ return -EINVAL;
- cqe->wr_id = rq->swq[wr_id_idx].wr_id;
- cqe++;
- (*budget)--;
- rq->hwq.cons++;
- *pcqe = cqe;
+ if (wr_id_idx > srq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process UD ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: wr_id idx 0x%x exceeded SRQ max 0x%x",
+ wr_id_idx, srq->hwq.max_elements);
+ return -EINVAL;
+ }
+ cqe->wr_id = srq->swq[wr_id_idx].wr_id;
+ bnxt_qplib_release_srqe(srq, wr_id_idx);
+ cqe++;
+ (*budget)--;
+ *pcqe = cqe;
+ } else {
+ rq = &qp->rq;
+ if (wr_id_idx > rq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process UD ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: wr_id idx 0x%x exceeded RQ max 0x%x",
+ wr_id_idx, rq->hwq.max_elements);
+ return -EINVAL;
+ }
- if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
- qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
- /* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ cqe->wr_id = rq->swq[wr_id_idx].wr_id;
+ cqe++;
+ (*budget)--;
+ rq->hwq.cons++;
+ *pcqe = cqe;
+
+ if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
+ qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
+ /* Add qp to flush list of the CQ */
+ bnxt_qplib_lock_buddy_cq(qp, cq);
+ __bnxt_qplib_add_flush_qp(qp);
+ bnxt_qplib_unlock_buddy_cq(qp, cq);
+ }
}
done:
return rc;
@@ -2217,6 +2531,7 @@ static int bnxt_qplib_cq_process_res_raweth_qp1(struct bnxt_qplib_cq *cq,
{
struct bnxt_qplib_qp *qp;
struct bnxt_qplib_q *rq;
+ struct bnxt_qplib_srq *srq;
struct bnxt_qplib_cqe *cqe;
u32 wr_id_idx;
int rc = 0;
@@ -2255,26 +2570,49 @@ static int bnxt_qplib_cq_process_res_raweth_qp1(struct bnxt_qplib_cq *cq,
cqe->raweth_qp1_flags2 = le32_to_cpu(hwcqe->raweth_qp1_flags2);
cqe->raweth_qp1_metadata = le32_to_cpu(hwcqe->raweth_qp1_metadata);
- rq = &qp->rq;
- if (wr_id_idx > rq->hwq.max_elements) {
- dev_err(&cq->hwq.pdev->dev, "QPLIB: FP: CQ Process Raw/QP1 RQ wr_id ");
- dev_err(&cq->hwq.pdev->dev, "QPLIB: ix 0x%x exceeded RQ max 0x%x",
- wr_id_idx, rq->hwq.max_elements);
- return -EINVAL;
- }
-
- cqe->wr_id = rq->swq[wr_id_idx].wr_id;
- cqe++;
- (*budget)--;
- rq->hwq.cons++;
- *pcqe = cqe;
+ if (cqe->flags & CQ_RES_RAWETH_QP1_FLAGS_SRQ_SRQ) {
+ srq = qp->srq;
+ if (!srq) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: SRQ used but not defined??");
+ return -EINVAL;
+ }
+ if (wr_id_idx > srq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process Raw/QP1 ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: wr_id idx 0x%x exceeded SRQ max 0x%x",
+ wr_id_idx, srq->hwq.max_elements);
+ return -EINVAL;
+ }
+ cqe->wr_id = srq->swq[wr_id_idx].wr_id;
+ bnxt_qplib_release_srqe(srq, wr_id_idx);
+ cqe++;
+ (*budget)--;
+ *pcqe = cqe;
+ } else {
+ rq = &qp->rq;
+ if (wr_id_idx > rq->hwq.max_elements) {
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: FP: CQ Process Raw/QP1 RQ wr_id ");
+ dev_err(&cq->hwq.pdev->dev,
+ "QPLIB: ix 0x%x exceeded RQ max 0x%x",
+ wr_id_idx, rq->hwq.max_elements);
+ return -EINVAL;
+ }
+ cqe->wr_id = rq->swq[wr_id_idx].wr_id;
+ cqe++;
+ (*budget)--;
+ rq->hwq.cons++;
+ *pcqe = cqe;
- if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
- qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
- /* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
+ qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
+ /* Add qp to flush list of the CQ */
+ bnxt_qplib_lock_buddy_cq(qp, cq);
+ __bnxt_qplib_add_flush_qp(qp);
+ bnxt_qplib_unlock_buddy_cq(qp, cq);
+ }
}
done:
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.h b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
index c582d4e..211b27a 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
@@ -39,6 +39,27 @@
#ifndef __BNXT_QPLIB_FP_H__
#define __BNXT_QPLIB_FP_H__
+struct bnxt_qplib_srq {
+ struct bnxt_qplib_pd *pd;
+ struct bnxt_qplib_dpi *dpi;
+ void __iomem *dbr_base;
+ u64 srq_handle;
+ u32 id;
+ u32 max_wqe;
+ u32 max_sge;
+ u32 threshold;
+ bool arm_req;
+ struct bnxt_qplib_cq *cq;
+ struct bnxt_qplib_hwq hwq;
+ struct bnxt_qplib_swq *swq;
+ struct scatterlist *sglist;
+ int start_idx;
+ int last_idx;
+ u32 nmap;
+ u16 eventq_hw_ring_id;
+ spinlock_t lock; /* protect SRQE link list */
+};
+
struct bnxt_qplib_sge {
u64 addr;
u32 lkey;
@@ -79,6 +100,7 @@ static inline u32 get_psne_idx(u32 val)
struct bnxt_qplib_swq {
u64 wr_id;
+ int next_idx;
u8 type;
u8 flags;
u32 start_psn;
@@ -404,29 +426,27 @@ struct bnxt_qplib_cq {
writel(NQ_DB_CP_FLAGS | ((raw_cons) & ((cp_bit) - 1)), db)
struct bnxt_qplib_nq {
- struct pci_dev *pdev;
-
- int vector;
- cpumask_t mask;
- int budget;
- bool requested;
- struct tasklet_struct worker;
- struct bnxt_qplib_hwq hwq;
-
- u16 bar_reg;
- u16 bar_reg_off;
- u16 ring_id;
- void __iomem *bar_reg_iomem;
-
- int (*cqn_handler)
- (struct bnxt_qplib_nq *nq,
- struct bnxt_qplib_cq *cq);
- int (*srqn_handler)
- (struct bnxt_qplib_nq *nq,
- void *srq,
- u8 event);
- struct workqueue_struct *cqn_wq;
- char name[32];
+ struct pci_dev *pdev;
+
+ int vector;
+ cpumask_t mask;
+ int budget;
+ bool requested;
+ struct tasklet_struct worker;
+ struct bnxt_qplib_hwq hwq;
+
+ u16 bar_reg;
+ u16 bar_reg_off;
+ u16 ring_id;
+ void __iomem *bar_reg_iomem;
+
+ int (*cqn_handler)(struct bnxt_qplib_nq *nq,
+ struct bnxt_qplib_cq *cq);
+ int (*srqn_handler)(struct bnxt_qplib_nq *nq,
+ struct bnxt_qplib_srq *srq,
+ u8 event);
+ struct workqueue_struct *cqn_wq;
+ char name[32];
};
struct bnxt_qplib_nq_work {
@@ -441,8 +461,18 @@ int bnxt_qplib_enable_nq(struct pci_dev *pdev, struct bnxt_qplib_nq *nq,
int (*cqn_handler)(struct bnxt_qplib_nq *nq,
struct bnxt_qplib_cq *cq),
int (*srqn_handler)(struct bnxt_qplib_nq *nq,
- void *srq,
+ struct bnxt_qplib_srq *srq,
u8 event));
+int bnxt_qplib_create_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq);
+int bnxt_qplib_modify_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq);
+int bnxt_qplib_query_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq);
+int bnxt_qplib_destroy_srq(struct bnxt_qplib_res *res,
+ struct bnxt_qplib_srq *srq);
+int bnxt_qplib_post_srq_recv(struct bnxt_qplib_srq *srq,
+ struct bnxt_qplib_swqe *wqe);
int bnxt_qplib_create_qp1(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp);
int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp);
int bnxt_qplib_modify_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
index 6a3633a..8329ec6 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
@@ -616,7 +616,7 @@ int bnxt_qplib_enable_rcfw_channel(struct pci_dev *pdev,
int msix_vector,
int cp_bar_reg_off, int virt_fn,
int (*aeq_handler)(struct bnxt_qplib_rcfw *,
- struct creq_func_event *))
+ void *, void *))
{
resource_size_t res_base;
struct cmdq_init init;
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
index 2946a7c..6bee6e3 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
@@ -167,7 +167,7 @@ struct bnxt_qplib_rcfw {
#define FIRMWARE_TIMED_OUT 3
wait_queue_head_t waitq;
int (*aeq_handler)(struct bnxt_qplib_rcfw *,
- struct creq_func_event *);
+ void *, void *);
u32 seq_num;
/* Bar region info */
@@ -199,9 +199,8 @@ int bnxt_qplib_enable_rcfw_channel(struct pci_dev *pdev,
struct bnxt_qplib_rcfw *rcfw,
int msix_vector,
int cp_bar_reg_off, int virt_fn,
- int (*aeq_handler)
- (struct bnxt_qplib_rcfw *,
- struct creq_func_event *));
+ int (*aeq_handler)(struct bnxt_qplib_rcfw *,
+ void *aeqe, void *obj));
struct bnxt_qplib_rcfw_sbuf *bnxt_qplib_rcfw_alloc_sbuf(
struct bnxt_qplib_rcfw *rcfw,
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 398a514..db54115 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -82,6 +82,15 @@ struct bnxt_re_qp_resp {
__u32 rsvd;
};
+struct bnxt_re_srq_req {
+ __u64 srqva;
+ __u64 srq_handle;
+};
+
+struct bnxt_re_srq_resp {
+ __u32 srqid;
+};
+
enum bnxt_re_shpg_offt {
BNXT_RE_BEG_RESV_OFFT = 0x00,
BNXT_RE_AVID_OFFT = 0x10,
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 14+ messages in thread