Skip to contentSkip to conversation

DEVX asynchronous events

PATCH Superseded Leon Romanovsky 4 participants 35 messages Jun 18, 2019 – Jun 30, 2019

Subsystems (7)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover DEVX asynchronous events

From: Leon Romanovsky <[email protected]>

Changelog:
v0 -> v1:
* Fix the unbind / hot unplug flows to work properly.
* Fix Ref count handling on the eventfd mode in some flow.
* Rebased to latest rdma-next

Thanks

------------------------------------------------------------------------------------

Quoted 1 line

From Yishai:

This series enables RDMA applications that use the DEVX interface to subscribe and read device asynchronous events.

The solution is designed to allow extension of events in the future without need to perform any changes in the driver code.

To enable that few changes had been done in mlx5_core, it includes:
* Reading device event capabilities that are user related (affiliated and un-affiliated) and set the matching mask upon creating the matching EQ.
* Enable DEVX/mlx5_ib to register for ANY event instead of the option to get some hard-coded ones.
* Enable DEVX/mlx5_ib to get the device raw data for CQ completion events.
* Enhance mlx5_core_create/destroy CQ to enable DEVX using them so that CQ events will be reported as well.

In mlx5_ib layer the below changes were done:
* A new DEVX API was introduced to allocate an event channel by using the uverbs FD object type.
* Implement the FD channel operations to enable read/poo/close over it.
* A new DEVX API was introduced to subscribe for specific events over an event channel.
* Manage an internal data structure over XA(s) to subscribe/dispatch events over the different event channels.
* Use from DEVX the mlx5_core APIs to create/destroy a CQ to be able to get its relevant events.

Yishai

Yishai Hadas (12):
net/mlx5: Fix mlx5_core_destroy_cq() error flow net/mlx5: Use event mask based on device capabilities net/mlx5: Expose the API to register for ANY event net/mlx5: mlx5_core_create_cq() enhancements net/mlx5: Report a CQ error event only when a handler was set net/mlx5: Report EQE data upon CQ completion net/mlx5: Expose device definitions for object events IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD IB/mlx5: Register DEVX with mlx5_core to get async events IB/mlx5: Enable subscription for device events over DEVX IB/mlx5: Implement DEVX dispatching event
IB/mlx5: Add DEVX support for CQ events

 drivers/infiniband/hw/mlx5/cq.c               |    5 +-
 drivers/infiniband/hw/mlx5/devx.c             | 1082 ++++++++++++++++-
 drivers/infiniband/hw/mlx5/main.c             |   10 +-
 drivers/infiniband/hw/mlx5/mlx5_ib.h          |   12 +
 drivers/infiniband/hw/mlx5/odp.c              |    3 +-
 drivers/infiniband/hw/mlx5/qp.c               |    2 +-
 drivers/net/ethernet/mellanox/mlx5/core/cq.c  |   21 +-
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |    2 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |    3 +-
 .../net/ethernet/mellanox/mlx5/core/en_txrx.c |    2 +-
 drivers/net/ethernet/mellanox/mlx5/core/eq.c  |   68 +-
 .../ethernet/mellanox/mlx5/core/fpga/conn.c   |    6 +-
 drivers/net/ethernet/mellanox/mlx5/core/fw.c  |    6 +
 .../net/ethernet/mellanox/mlx5/core/lib/eq.h  |    5 +-
 include/linux/mlx5/cq.h                       |    6 +-
 include/linux/mlx5/device.h                   |    6 +-
 include/linux/mlx5/driver.h                   |    2 +
 include/linux/mlx5/eq.h                       |    4 +-
 include/linux/mlx5/mlx5_ifc.h                 |   34 +-
 include/uapi/rdma/mlx5_user_ioctl_cmds.h      |   19 +
 include/uapi/rdma/mlx5_user_ioctl_verbs.h     |    9 +
 21 files changed, 1237 insertions(+), 70 deletions(-)
#1 lore raw

PATCH 1/12 net/mlx5: Fix mlx5_core_destroy_cq() error flow

From: Yishai Hadas <[email protected]>

The firmware command to destroy a CQ might fail when the object is referenced by other object and the ref count is managed by the firmware.

To enable a second successful destruction post the first failure need to change mlx5_eq_del_cq() to be a void function.

As an error in mlx5_eq_del_cq() is quite fatal from the option to recover, a debug message inside it should be good enougth and it was changed to be void.

---
 drivers/net/ethernet/mellanox/mlx5/core/cq.c     |  9 ++-------
 drivers/net/ethernet/mellanox/mlx5/core/eq.c     | 16 +++++++---------
 drivers/net/ethernet/mellanox/mlx5/core/lib/eq.h |  2 +-
 3 files changed, 10 insertions(+), 17 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/cq.c +2 -7
index 713a17ee3751..703d88332bc6 100644
@@ -158,13 +158,8 @@ int mlx5_core_destroy_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq)
 	u32 in[MLX5_ST_SZ_DW(destroy_cq_in)] = {0};
 	int err;
 
-	err = mlx5_eq_del_cq(mlx5_get_async_eq(dev), cq);
-	if (err)
-		return err;
-
-	err = mlx5_eq_del_cq(&cq->eq->core, cq);
-	if (err)
-		return err;
+	mlx5_eq_del_cq(mlx5_get_async_eq(dev), cq);
+	mlx5_eq_del_cq(&cq->eq->core, cq);
 
 	MLX5_SET(destroy_cq_in, in, opcode, MLX5_CMD_OP_DESTROY_CQ);
 	MLX5_SET(destroy_cq_in, in, cqn, cq->cqn);
drivers/net/ethernet/mellanox/mlx5/core/eq.c +7 -9
index 58fff2f39b38..8000d2a4a7e2 100644
@@ -389,7 +389,7 @@ int mlx5_eq_add_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq)
 	return err;
 }
 
-int mlx5_eq_del_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq)
+void mlx5_eq_del_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq)
 {
 	struct mlx5_cq_table *table = &eq->cq_table;
 	struct mlx5_core_cq *tmp;
@@ -399,16 +399,14 @@ int mlx5_eq_del_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq)
 	spin_unlock(&table->lock);
 
 	if (!tmp) {
-		mlx5_core_warn(eq->dev, "cq 0x%x not found in eq 0x%x tree\n", eq->eqn, cq->cqn);
-		return -ENOENT;
+		mlx5_core_dbg(eq->dev, "cq 0x%x not found in eq 0x%x tree\n",
+			      eq->eqn, cq->cqn);
+		return;
 	}
 
-	if (tmp != cq) {
-		mlx5_core_warn(eq->dev, "corruption on cqn 0x%x in eq 0x%x\n", eq->eqn, cq->cqn);
-		return -EINVAL;
-	}
-
-	return 0;
+	if (tmp != cq)
+		mlx5_core_dbg(eq->dev, "corruption on cqn 0x%x in eq 0x%x\n",
+			      eq->eqn, cq->cqn);
 }
 
 int mlx5_eq_table_init(struct mlx5_core_dev *dev)
drivers/net/ethernet/mellanox/mlx5/core/lib/eq.h +1 -1
index 24bd991a727e..d826e63d5a17 100644
@@ -75,7 +75,7 @@ int mlx5_eq_table_create(struct mlx5_core_dev *dev);
 void mlx5_eq_table_destroy(struct mlx5_core_dev *dev);
 
 int mlx5_eq_add_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq);
-int mlx5_eq_del_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq);
+void mlx5_eq_del_cq(struct mlx5_eq *eq, struct mlx5_core_cq *cq);
 struct mlx5_eq_comp *mlx5_eqn2comp_eq(struct mlx5_core_dev *dev, int eqn);
 struct mlx5_eq *mlx5_get_async_eq(struct mlx5_core_dev *dev);
 void mlx5_cq_tasklet_cb(unsigned long data);
Signature · 1 line
2.20.1
#2 lore raw

PATCH 2/12 net/mlx5: Use event mask based on device capabilities

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Use the reported device capabilities for the supported user events (i.e. affiliated and un-affiliated) to set the EQ mask.

As the event mask can be up to 256 defined by 4 entries of u64 change the applicable code to work accordingly.

---
 drivers/infiniband/hw/mlx5/odp.c             |  3 +-
 drivers/net/ethernet/mellanox/mlx5/core/eq.c | 45 ++++++++++++++++----
 drivers/net/ethernet/mellanox/mlx5/core/fw.c |  6 +++
 include/linux/mlx5/device.h                  |  6 ++-
 include/linux/mlx5/eq.h                      |  4 +-
 include/linux/mlx5/mlx5_ifc.h                | 13 ++++--
 6 files changed, 63 insertions(+), 14 deletions(-)
drivers/infiniband/hw/mlx5/odp.c +2 -1
index 600fe23e2eae..a6740ec308ed 100644
@@ -1559,10 +1559,11 @@ mlx5_ib_create_pf_eq(struct mlx5_ib_dev *dev, struct mlx5_ib_pf_eq *eq)
 	eq->irq_nb.notifier_call = mlx5_ib_eq_pf_int;
 	param = (struct mlx5_eq_param) {
 		.irq_index = 0,
-		.mask = 1 << MLX5_EVENT_TYPE_PAGE_FAULT,
 		.nent = MLX5_IB_NUM_PF_EQE,
 	};
 	eq->core = mlx5_eq_create_generic(dev->mdev, &param);
+
+	param.mask[0] = 1ull << MLX5_EVENT_TYPE_PAGE_FAULT;
 	if (IS_ERR(eq->core)) {
 		err = PTR_ERR(eq->core);
 		goto err_wq;
drivers/net/ethernet/mellanox/mlx5/core/eq.c +37 -8

104 lines not on this page · raw message

drivers/net/ethernet/mellanox/mlx5/core/fw.c +6 -0
index 1ab6f7e3bec6..05367f15c3a7 100644
@@ -202,6 +202,12 @@ int mlx5_query_hca_caps(struct mlx5_core_dev *dev)
 			return err;
 	}
 
+	if (MLX5_CAP_GEN(dev, event_cap)) {
+		err = mlx5_core_get_caps(dev, MLX5_CAP_DEV_EVENT);
+		if (err)
+			return err;
+	}
+
 	return 0;
 }
 
include/linux/mlx5/device.h +5 -1
index 5e760067ac41..0d1abe097627 100644
@@ -351,7 +351,7 @@ enum mlx5_event {
 
 	MLX5_EVENT_TYPE_DEVICE_TRACER      = 0x26,
 
-	MLX5_EVENT_TYPE_MAX                = MLX5_EVENT_TYPE_DEVICE_TRACER + 1,
+	MLX5_EVENT_TYPE_MAX                = 0x100,
 };
 
 enum {
@@ -1077,6 +1077,7 @@ enum mlx5_cap_type {
 	MLX5_CAP_DEBUG,
 	MLX5_CAP_RESERVED_14,
 	MLX5_CAP_DEV_MEM,
+	MLX5_CAP_DEV_EVENT = 0x14,
 	/* NUM OF CAP Types */
 	MLX5_CAP_NUM
 };
@@ -1255,6 +1256,9 @@ enum mlx5_qcam_feature_groups {
 #define MLX5_CAP64_DEV_MEM(mdev, cap)\
 	MLX5_GET64(device_mem_cap, mdev->caps.hca_cur[MLX5_CAP_DEV_MEM], cap)
 
+#define MLX5_CAP_DEV_EVENT(mdev, cap)\
+	MLX5_ADDR_OF(device_event_cap, (mdev)->caps.hca_cur[MLX5_CAP_DEV_EVENT], cap)
+
 enum {
 	MLX5_CMD_STAT_OK			= 0x0,
 	MLX5_CMD_STAT_INT_ERR			= 0x1,
include/linux/mlx5/eq.h +3 -1
index 70e16dcfb4c4..202df2e5fe8c 100644
@@ -15,7 +15,9 @@ struct mlx5_core_dev;
 struct mlx5_eq_param {
 	u8             irq_index;
 	int            nent;
-	u64            mask;
+	u64            mask[4];
+	void          *context;
+	irq_handler_t  handler;
 };
 
 struct mlx5_eq *
include/linux/mlx5/mlx5_ifc.h +10 -3
index 16348528fef6..3ef716c054c2 100644
@@ -823,6 +823,12 @@ struct mlx5_ifc_device_mem_cap_bits {
 	u8         reserved_at_180[0x680];
 };
 
+struct mlx5_ifc_device_event_cap_bits {
+	u8         user_affiliated_events[4][0x40];
+
+	u8         user_unaffiliated_events[4][0x40];
+};
+
 enum {
 	MLX5_ATOMIC_CAPS_ATOMIC_SIZE_QP_1_BYTE     = 0x0,
 	MLX5_ATOMIC_CAPS_ATOMIC_SIZE_QP_2_BYTES    = 0x2,
@@ -980,7 +986,8 @@ struct mlx5_ifc_cmd_hca_cap_bits {
 
 	u8         log_max_srq_sz[0x8];
 	u8         log_max_qp_sz[0x8];
-	u8         reserved_at_90[0x8];
+	u8         event_cap[0x1];
+	u8         reserved_at_91[0x7];
 	u8         prio_tag_required[0x1];
 	u8         reserved_at_99[0x2];
 	u8         log_max_qp[0x5];
@@ -7364,9 +7371,9 @@ struct mlx5_ifc_create_eq_in_bits {
 
 	u8         reserved_at_280[0x40];
 
-	u8         event_bitmask[0x40];
+	u8         event_bitmask[4][0x40];
 
-	u8         reserved_at_300[0x580];
+	u8         reserved_at_3c0[0x4c0];
 
 	u8         pas[0][0x40];
 };
Signature · 1 line
2.20.1
#4 lore raw

PATCH 3/12 net/mlx5: Expose the API to register for ANY event

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Expose the API to register for ANY event, mlx5_ib will be able to use this functionality for its needs.

---
 drivers/net/ethernet/mellanox/mlx5/core/eq.c     | 2 ++
 drivers/net/ethernet/mellanox/mlx5/core/lib/eq.h | 3 ---
 include/linux/mlx5/driver.h                      | 2 ++
 3 files changed, 4 insertions(+), 3 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eq.c +2 -0
index 9d07add38940..a7a8bf73e465 100644
@@ -954,6 +954,7 @@ int mlx5_eq_notifier_register(struct mlx5_core_dev *dev, struct mlx5_nb *nb)
 
 	return atomic_notifier_chain_register(&eqt->nh[nb->event_type], &nb->nb);
 }
+EXPORT_SYMBOL(mlx5_eq_notifier_register);
 
 int mlx5_eq_notifier_unregister(struct mlx5_core_dev *dev, struct mlx5_nb *nb)
 {
@@ -964,3 +965,4 @@ int mlx5_eq_notifier_unregister(struct mlx5_core_dev *dev, struct mlx5_nb *nb)
 
 	return atomic_notifier_chain_unregister(&eqt->nh[nb->event_type], &nb->nb);
 }
+EXPORT_SYMBOL(mlx5_eq_notifier_unregister);
drivers/net/ethernet/mellanox/mlx5/core/lib/eq.h +0 -3
index d826e63d5a17..3dfab91ae5f2 100644
@@ -97,7 +97,4 @@ void mlx5_core_eq_free_irqs(struct mlx5_core_dev *dev);
 struct cpu_rmap *mlx5_eq_table_get_rmap(struct mlx5_core_dev *dev);
 #endif
 
-int mlx5_eq_notifier_register(struct mlx5_core_dev *dev, struct mlx5_nb *nb);
-int mlx5_eq_notifier_unregister(struct mlx5_core_dev *dev, struct mlx5_nb *nb);
-
 #endif
include/linux/mlx5/driver.h +2 -0
index d8ab633406c2..3f8b9d8e2070 100644
@@ -1047,6 +1047,8 @@ int mlx5_register_interface(struct mlx5_interface *intf);
 void mlx5_unregister_interface(struct mlx5_interface *intf);
 int mlx5_notifier_register(struct mlx5_core_dev *dev, struct notifier_block *nb);
 int mlx5_notifier_unregister(struct mlx5_core_dev *dev, struct notifier_block *nb);
+int mlx5_eq_notifier_register(struct mlx5_core_dev *dev, struct mlx5_nb *nb);
+int mlx5_eq_notifier_unregister(struct mlx5_core_dev *dev, struct mlx5_nb *nb);
 
 int mlx5_core_query_vendor_id(struct mlx5_core_dev *mdev, u32 *vendor_id);
 
Signature · 1 line
2.20.1
#5 lore raw

PATCH 4/12 net/mlx5: mlx5_core_create_cq() enhancements

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Enhance mlx5_core_create_cq() to get the command out buffer from the callers to let them use the output.

---
 drivers/infiniband/hw/mlx5/cq.c                     | 3 ++-
 drivers/net/ethernet/mellanox/mlx5/core/cq.c        | 7 +++----
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c   | 3 ++-
 drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c | 3 ++-
 include/linux/mlx5/cq.h                             | 2 +-
 5 files changed, 10 insertions(+), 8 deletions(-)
drivers/infiniband/hw/mlx5/cq.c +2 -1
index 07b73df0e1a3..9f39e7b9dd1b 100644
@@ -892,6 +892,7 @@ int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 	int vector = attr->comp_vector;
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	struct mlx5_ib_cq *cq = to_mcq(ibcq);
+	u32 out[MLX5_ST_SZ_DW(create_cq_out)];
 	int uninitialized_var(index);
 	int uninitialized_var(inlen);
 	u32 *cqb = NULL;
@@ -954,7 +955,7 @@ int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 	if (cq->create_flags & IB_UVERBS_CQ_FLAGS_IGNORE_OVERRUN)
 		MLX5_SET(cqc, cqc, oi, 1);
 
-	err = mlx5_core_create_cq(dev->mdev, &cq->mcq, cqb, inlen);
+	err = mlx5_core_create_cq(dev->mdev, &cq->mcq, cqb, inlen, out, sizeof(out));
 	if (err)
 		goto err_cqb;
 
drivers/net/ethernet/mellanox/mlx5/core/cq.c +3 -4
index 703d88332bc6..1bd4336392a2 100644
@@ -87,11 +87,10 @@ static void mlx5_add_cq_to_tasklet(struct mlx5_core_cq *cq)
 }
 
 int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
-			u32 *in, int inlen)
+			u32 *in, int inlen, u32 *out, int outlen)
 {
 	int eqn = MLX5_GET(cqc, MLX5_ADDR_OF(create_cq_in, in, cq_context), c_eqn);
 	u32 dout[MLX5_ST_SZ_DW(destroy_cq_out)];
-	u32 out[MLX5_ST_SZ_DW(create_cq_out)];
 	u32 din[MLX5_ST_SZ_DW(destroy_cq_in)];
 	struct mlx5_eq_comp *eq;
 	int err;
@@ -100,9 +99,9 @@ int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
 	if (IS_ERR(eq))
 		return PTR_ERR(eq);
 
-	memset(out, 0, sizeof(out));
+	memset(out, 0, outlen);
 	MLX5_SET(create_cq_in, in, opcode, MLX5_CMD_OP_CREATE_CQ);
-	err = mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
+	err = mlx5_cmd_exec(dev, in, inlen, out, outlen);
 	if (err)
 		return err;
 
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +2 -1
index 457cc39423f2..7186282785df 100644
@@ -1517,6 +1517,7 @@ static void mlx5e_free_cq(struct mlx5e_cq *cq)
 
 static int mlx5e_create_cq(struct mlx5e_cq *cq, struct mlx5e_cq_param *param)
 {
+	u32 out[MLX5_ST_SZ_DW(create_cq_out)];
 	struct mlx5_core_dev *mdev = cq->mdev;
 	struct mlx5_core_cq *mcq = &cq->mcq;
 
@@ -1551,7 +1552,7 @@ static int mlx5e_create_cq(struct mlx5e_cq *cq, struct mlx5e_cq_param *param)
 					    MLX5_ADAPTER_PAGE_SHIFT);
 	MLX5_SET64(cqc, cqc, dbr_addr,      cq->wq_ctrl.db.dma);
 
-	err = mlx5_core_create_cq(mdev, mcq, in, inlen);
+	err = mlx5_core_create_cq(mdev, mcq, in, inlen, out, sizeof(out));
 
 	kvfree(in);
 
drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c +2 -1
index ca2296a2f9ee..dc7b9d9f274d 100644
@@ -429,6 +429,7 @@ static int mlx5_fpga_conn_create_cq(struct mlx5_fpga_conn *conn, int cq_size)
 	struct mlx5_fpga_device *fdev = conn->fdev;
 	struct mlx5_core_dev *mdev = fdev->mdev;
 	u32 temp_cqc[MLX5_ST_SZ_DW(cqc)] = {0};
+	u32 out[MLX5_ST_SZ_DW(create_cq_out)];
 	struct mlx5_wq_param wqp;
 	struct mlx5_cqe64 *cqe;
 	int inlen, err, eqn;
@@ -476,7 +477,7 @@ static int mlx5_fpga_conn_create_cq(struct mlx5_fpga_conn *conn, int cq_size)
 	pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, in, pas);
 	mlx5_fill_page_frag_array(&conn->cq.wq_ctrl.buf, pas);
 
-	err = mlx5_core_create_cq(mdev, &conn->cq.mcq, in, inlen);
+	err = mlx5_core_create_cq(mdev, &conn->cq.mcq, in, inlen, out, sizeof(out));
 	kvfree(in);
 
 	if (err)
include/linux/mlx5/cq.h +1 -1
index 769326ea1d9b..e44157a2b7db 100644
@@ -185,7 +185,7 @@ static inline void mlx5_cq_put(struct mlx5_core_cq *cq)
 }
 
 int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
-			u32 *in, int inlen);
+			u32 *in, int inlen, u32 *out, int outlen);
 int mlx5_core_destroy_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq);
 int mlx5_core_query_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
 		       u32 *out, int outlen);
Signature · 1 line
2.20.1
#6 lore raw

PATCH 5/12 net/mlx5: Report a CQ error event only when a handler was set

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Report a CQ error event only when a handler was set.

This enables mlx5_ib to not set a handler upon CQ creation and use some other mechanism to get this event as of other events by the mlx5_eq_notifier_register API.

---
 drivers/net/ethernet/mellanox/mlx5/core/eq.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/eq.c +2 -1
index a7a8bf73e465..316fbed81470 100644
@@ -503,7 +503,8 @@ static int cq_err_event_notifier(struct notifier_block *nb,
 		return NOTIFY_OK;
 	}
 
-	cq->event(cq, type);
+	if (cq->event)
+		cq->event(cq, type);
 
 	mlx5_cq_put(cq);
 
Signature · 1 line
2.20.1
#7 lore raw

PATCH 6/12 net/mlx5: Report EQE data upon CQ completion

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Report EQE data upon CQ completion to let upper layers use this data.

---
 drivers/infiniband/hw/mlx5/cq.c                     | 2 +-
 drivers/infiniband/hw/mlx5/main.c                   | 2 +-
 drivers/infiniband/hw/mlx5/qp.c                     | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/cq.c        | 5 +++--
 drivers/net/ethernet/mellanox/mlx5/core/en.h        | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c   | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/eq.c        | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c | 3 ++-
 include/linux/mlx5/cq.h                             | 4 ++--
 9 files changed, 13 insertions(+), 11 deletions(-)
drivers/infiniband/hw/mlx5/cq.c +1 -1
index 9f39e7b9dd1b..fa6d49d583fe 100644
@@ -37,7 +37,7 @@
 #include "mlx5_ib.h"
 #include "srq.h"
 
-static void mlx5_ib_cq_comp(struct mlx5_core_cq *cq)
+static void mlx5_ib_cq_comp(struct mlx5_core_cq *cq, struct mlx5_eqe *eqe)
 {
 	struct ib_cq *ibcq = &to_mibcq(cq)->ibcq;
 
drivers/infiniband/hw/mlx5/main.c +1 -1
index 7fb1f775e3d4..c9c6710afc6b 100644
@@ -4463,7 +4463,7 @@ static void mlx5_ib_handle_internal_error(struct mlx5_ib_dev *ibdev)
 	 * lock/unlock above locks Now need to arm all involved CQs.
 	 */
 	list_for_each_entry(mcq, &cq_armed_list, reset_notify) {
-		mcq->comp(mcq);
+		mcq->comp(mcq, NULL);
 	}
 	spin_unlock_irqrestore(&ibdev->reset_flow_resource_lock, flags);
 }
drivers/infiniband/hw/mlx5/qp.c +1 -1
index 63d8f61e50e0..f240e022682b 100644
@@ -6409,7 +6409,7 @@ static void handle_drain_completion(struct ib_cq *cq,
 		/* Run the CQ handler - this makes sure that the drain WR will
 		 * be processed if wasn't processed yet.
 		 */
-		mcq->mcq.comp(&mcq->mcq);
+		mcq->mcq.comp(&mcq->mcq, NULL);
 	}
 
 	wait_for_completion(&sdrain->done);
drivers/net/ethernet/mellanox/mlx5/core/cq.c +3 -2
index 1bd4336392a2..818edc63e428 100644
@@ -58,7 +58,7 @@ void mlx5_cq_tasklet_cb(unsigned long data)
 	list_for_each_entry_safe(mcq, temp, &ctx->process_list,
 				 tasklet_ctx.list) {
 		list_del_init(&mcq->tasklet_ctx.list);
-		mcq->tasklet_ctx.comp(mcq);
+		mcq->tasklet_ctx.comp(mcq, NULL);
 		mlx5_cq_put(mcq);
 		if (time_after(jiffies, end))
 			break;
@@ -68,7 +68,8 @@ void mlx5_cq_tasklet_cb(unsigned long data)
 		tasklet_schedule(&ctx->task);
 }
 
-static void mlx5_add_cq_to_tasklet(struct mlx5_core_cq *cq)
+static void mlx5_add_cq_to_tasklet(struct mlx5_core_cq *cq,
+				   struct mlx5_eqe *eqe)
 {
 	unsigned long flags;
 	struct mlx5_eq_tasklet *tasklet_ctx = cq->tasklet_ctx.priv;
drivers/net/ethernet/mellanox/mlx5/core/en.h +1 -1
index 3a183d690e23..16753f263079 100644
@@ -780,7 +780,7 @@ netdev_tx_t mlx5e_sq_xmit(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 			  struct mlx5e_tx_wqe *wqe, u16 pi, bool xmit_more);
 
 void mlx5e_trigger_irq(struct mlx5e_icosq *sq);
-void mlx5e_completion_event(struct mlx5_core_cq *mcq);
+void mlx5e_completion_event(struct mlx5_core_cq *mcq, struct mlx5_eqe *eqe);
 void mlx5e_cq_error_event(struct mlx5_core_cq *mcq, enum mlx5_event event);
 int mlx5e_napi_poll(struct napi_struct *napi, int budget);
 bool mlx5e_poll_tx_cq(struct mlx5e_cq *cq, int napi_budget);
drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c +1 -1
index f9862bf75491..c665ae0f22bd 100644
@@ -136,7 +136,7 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
 	return work_done;
 }
 
-void mlx5e_completion_event(struct mlx5_core_cq *mcq)
+void mlx5e_completion_event(struct mlx5_core_cq *mcq, struct mlx5_eqe *eqe)
 {
 	struct mlx5e_cq *cq = container_of(mcq, struct mlx5e_cq, mcq);
 
drivers/net/ethernet/mellanox/mlx5/core/eq.c +1 -1
index 316fbed81470..185f99e30bd5 100644
@@ -153,7 +153,7 @@ static int mlx5_eq_comp_int(struct notifier_block *nb,
 		cq = mlx5_eq_cq_get(eq, cqn);
 		if (likely(cq)) {
 			++cq->arm_sn;
-			cq->comp(cq);
+			cq->comp(cq, eqe);
 			mlx5_cq_put(cq);
 		} else {
 			mlx5_core_warn(eq->dev, "Completion event for bogus CQ 0x%x\n", cqn);
drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c +2 -1
index dc7b9d9f274d..028891823f32 100644
@@ -414,7 +414,8 @@ static void mlx5_fpga_conn_cq_tasklet(unsigned long data)
 	mlx5_fpga_conn_cqes(conn, MLX5_FPGA_CQ_BUDGET);
 }
 
-static void mlx5_fpga_conn_cq_complete(struct mlx5_core_cq *mcq)
+static void mlx5_fpga_conn_cq_complete(struct mlx5_core_cq *mcq,
+				       struct mlx5_eqe *eqe)
 {
 	struct mlx5_fpga_conn *conn;
 
include/linux/mlx5/cq.h +2 -2
index e44157a2b7db..40748fc1b11b 100644
@@ -47,7 +47,7 @@ struct mlx5_core_cq {
 	struct completion	free;
 	unsigned		vector;
 	unsigned int		irqn;
-	void (*comp)		(struct mlx5_core_cq *);
+	void (*comp)(struct mlx5_core_cq *cq, struct mlx5_eqe *eqe);
 	void (*event)		(struct mlx5_core_cq *, enum mlx5_event);
 	u32			cons_index;
 	unsigned		arm_sn;
@@ -55,7 +55,7 @@ struct mlx5_core_cq {
 	int			pid;
 	struct {
 		struct list_head list;
-		void (*comp)(struct mlx5_core_cq *);
+		void (*comp)(struct mlx5_core_cq *cq, struct mlx5_eqe *eqe);
 		void		*priv;
 	} tasklet_ctx;
 	int			reset_notify_added;
Signature · 1 line
2.20.1
#8 lore raw

PATCH 7/12 net/mlx5: Expose device definitions for object events

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Expose an extra device definitions for objects events.

It includes: object_type values for legacy objects and generic data header for any other object.

---
 include/linux/mlx5/mlx5_ifc.h | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
include/linux/mlx5/mlx5_ifc.h +21 -0
index 3ef716c054c2..80a453a1e949 100644
@@ -91,6 +91,20 @@ enum {
 
 enum {
 	MLX5_OBJ_TYPE_GENEVE_TLV_OPT = 0x000b,
+	MLX5_OBJ_TYPE_MKEY = 0xff01,
+	MLX5_OBJ_TYPE_QP = 0xff02,
+	MLX5_OBJ_TYPE_PSV = 0xff03,
+	MLX5_OBJ_TYPE_RMP = 0xff04,
+	MLX5_OBJ_TYPE_XRC_SRQ = 0xff05,
+	MLX5_OBJ_TYPE_RQ = 0xff06,
+	MLX5_OBJ_TYPE_SQ = 0xff07,
+	MLX5_OBJ_TYPE_TIR = 0xff08,
+	MLX5_OBJ_TYPE_TIS = 0xff09,
+	MLX5_OBJ_TYPE_DCT = 0xff0a,
+	MLX5_OBJ_TYPE_XRQ = 0xff0b,
+	MLX5_OBJ_TYPE_RQT = 0xff0e,
+	MLX5_OBJ_TYPE_FLOW_COUNTER = 0xff0f,
+	MLX5_OBJ_TYPE_CQ = 0xff10,
 };
 
 enum {
@@ -9755,4 +9769,11 @@ struct mlx5_ifc_query_esw_functions_out_bits {
 	u8         reserved_at_280[0x180];
 };
 
+struct mlx5_ifc_affiliated_event_header_bits {
+	u8         reserved_at_0[0x10];
+	u8         obj_type[0x10];
+
+	u8         obj_id[0x10];
+};
+
 #endif /* MLX5_IFC_H */
Signature · 1 line
2.20.1
#9 lore raw

PATCH 8/12 IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD and its initial implementation.

This object is from type class FD and will be used to read DEVX async events.

---
 drivers/infiniband/hw/mlx5/devx.c         | 112 ++++++++++++++++++++--
 include/uapi/rdma/mlx5_user_ioctl_cmds.h  |  10 ++
 include/uapi/rdma/mlx5_user_ioctl_verbs.h |   4 +
 3 files changed, 116 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +102 -10
index 80b42d069328..1815ce0f8daf 100644
@@ -33,6 +33,24 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+struct devx_async_event_queue {
+	spinlock_t		lock;
+	wait_queue_head_t	poll_wait;
+	struct list_head	event_list;
+	atomic_t		bytes_in_use;
+	u8			is_destroyed:1;
+	u32			flags;
+};
+
+struct devx_async_event_file {
+	struct ib_uobject		uobj;
+	struct list_head subscribed_events_list; /* Head of events that are
+						  * subscribed to this FD
+						  */
+	struct devx_async_event_queue	ev_queue;
+	struct mlx5_ib_dev *dev;
+};
+
 #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
 struct devx_obj {
 	struct mlx5_core_dev	*mdev;
@@ -1336,27 +1354,21 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)(
 			      cmd_out, cmd_out_len);
 }
 
-struct devx_async_event_queue {
-	spinlock_t		lock;
-	wait_queue_head_t	poll_wait;
-	struct list_head	event_list;
-	atomic_t		bytes_in_use;
-	u8			is_destroyed:1;
-};
-
 struct devx_async_cmd_event_file {
 	struct ib_uobject		uobj;
 	struct devx_async_event_queue	ev_queue;
 	struct mlx5_async_ctx		async_ctx;
 };
 
-static void devx_init_event_queue(struct devx_async_event_queue *ev_queue)
+static void devx_init_event_queue(struct devx_async_event_queue *ev_queue,
+				  u32 flags)
 {
 	spin_lock_init(&ev_queue->lock);
 	INIT_LIST_HEAD(&ev_queue->event_list);
 	init_waitqueue_head(&ev_queue->poll_wait);
 	atomic_set(&ev_queue->bytes_in_use, 0);
 	ev_queue->is_destroyed = 0;
+	ev_queue->flags = flags;
 }
 
 static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC)(
@@ -1370,11 +1382,38 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC)(
 
 	ev_file = container_of(uobj, struct devx_async_cmd_event_file,
 			       uobj);
-	devx_init_event_queue(&ev_file->ev_queue);
+	devx_init_event_queue(&ev_file->ev_queue, 0);
 	mlx5_cmd_init_async_ctx(mdev->mdev, &ev_file->async_ctx);
 	return 0;
 }
 
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_ASYNC_EVENT_FD_ALLOC)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+		attrs, MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_HANDLE);
+	struct devx_async_event_file *ev_file;
+	struct mlx5_ib_ucontext *c = rdma_udata_to_drv_context(
+		&attrs->driver_udata, struct mlx5_ib_ucontext, ibucontext);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	u32 flags;
+	int err;
+
+	err = uverbs_get_flags32(&flags, attrs,
+		MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_FLAGS,
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA);
+
+	if (err)
+		return err;
+
+	ev_file = container_of(uobj, struct devx_async_event_file,
+			       uobj);
+	devx_init_event_queue(&ev_file->ev_queue, flags);
+	INIT_LIST_HEAD(&ev_file->subscribed_events_list);
+	ev_file->dev = dev;
+	return 0;
+}
+
 static void devx_query_callback(int status, struct mlx5_async_work *context)
 {
 	struct devx_async_data *async_data =
@@ -1729,6 +1768,32 @@ static const struct file_operations devx_async_cmd_event_fops = {
 	.llseek	 = no_llseek,
 };
 
+static ssize_t devx_async_event_read(struct file *filp, char __user *buf,
+				     size_t count, loff_t *pos)
+{
+	return -EINVAL;
+}
+
+static __poll_t devx_async_event_poll(struct file *filp,
+				      struct poll_table_struct *wait)
+{
+	return 0;
+}
+
+static int devx_async_event_close(struct inode *inode, struct file *filp)
+{
+	uverbs_close_fd(filp);
+	return 0;
+}
+
+static const struct file_operations devx_async_event_fops = {
+	.owner	 = THIS_MODULE,
+	.read	 = devx_async_event_read,
+	.poll    = devx_async_event_poll,
+	.release = devx_async_event_close,
+	.llseek	 = no_llseek,
+};
+
 static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
 						   enum rdma_remove_reason why)
 {
@@ -1748,6 +1813,12 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
 	return 0;
 };
 
+static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
+					    enum rdma_remove_reason why)
+{
+	return 0;
+};
+
 DECLARE_UVERBS_NAMED_METHOD(
 	MLX5_IB_METHOD_DEVX_UMEM_REG,
 	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_UMEM_REG_HANDLE,
@@ -1913,6 +1984,24 @@ DECLARE_UVERBS_NAMED_OBJECT(
 			     O_RDONLY),
 	&UVERBS_METHOD(MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC));
 
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_ASYNC_EVENT_FD_ALLOC,
+	UVERBS_ATTR_FD(MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_HANDLE,
+			MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY),
+	UVERBS_ATTR_FLAGS_IN(MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_FLAGS,
+			enum mlx5_ib_uapi_devx_create_event_channel_flags,
+			UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_OBJECT(
+	MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD,
+	UVERBS_TYPE_ALLOC_FD(sizeof(struct devx_async_event_file),
+			     devx_hot_unplug_async_event_file,
+			     &devx_async_event_fops, "[devx_async_event]",
+			     O_RDONLY),
+	&UVERBS_METHOD(MLX5_IB_METHOD_DEVX_ASYNC_EVENT_FD_ALLOC));
+
 static bool devx_is_supported(struct ib_device *device)
 {
 	struct mlx5_ib_dev *dev = to_mdev(device);
@@ -1933,5 +2022,8 @@ const struct uapi_definition mlx5_ib_devx_defs[] = {
 	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(
 		MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
 		UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)),
+	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(
+		MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD,
+		UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)),
 	{},
 };
include/uapi/rdma/mlx5_user_ioctl_cmds.h +10 -0
index d404c951954c..6ad8f4f11ddd 100644
@@ -127,16 +127,26 @@ enum mlx5_ib_devx_async_cmd_fd_alloc_attrs {
 	MLX5_IB_ATTR_DEVX_ASYNC_CMD_FD_ALLOC_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
 };
 
+enum mlx5_ib_devx_async_event_fd_alloc_attrs {
+	MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	MLX5_IB_ATTR_DEVX_ASYNC_EVENT_FD_ALLOC_FLAGS,
+};
+
 enum mlx5_ib_devx_async_cmd_fd_methods {
 	MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC = (1U << UVERBS_ID_NS_SHIFT),
 };
 
+enum mlx5_ib_devx_async_event_fd_methods {
+	MLX5_IB_METHOD_DEVX_ASYNC_EVENT_FD_ALLOC = (1U << UVERBS_ID_NS_SHIFT),
+};
+
 enum mlx5_ib_objects {
 	MLX5_IB_OBJECT_DEVX = (1U << UVERBS_ID_NS_SHIFT),
 	MLX5_IB_OBJECT_DEVX_OBJ,
 	MLX5_IB_OBJECT_DEVX_UMEM,
 	MLX5_IB_OBJECT_FLOW_MATCHER,
 	MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
+	MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD,
 };
 
 enum mlx5_ib_flow_matcher_create_attrs {
include/uapi/rdma/mlx5_user_ioctl_verbs.h +4 -0
index a8f34c237458..57beea4589e4 100644
@@ -63,5 +63,9 @@ enum mlx5_ib_uapi_dm_type {
 	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM,
 };
 
+enum mlx5_ib_uapi_devx_create_event_channel_flags {
+	MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA = 1 << 0,
+};
+
 #endif
 
Signature · 1 line
2.20.1
#13 lore raw

PATCH 9/12 IB/mlx5: Register DEVX with mlx5_core to get async events

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Register DEVX with with mlx5_core to get async events. This will enable to dispatch the applicable events to its consumers in down stream patches.

---
 drivers/infiniband/hw/mlx5/devx.c    | 30 ++++++++++++++++++++++++++++
 drivers/infiniband/hw/mlx5/main.c    |  8 ++++++--
 drivers/infiniband/hw/mlx5/mlx5_ib.h | 12 +++++++++++
 3 files changed, 48 insertions(+), 2 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +30 -0
index 1815ce0f8daf..e9b9ba5a3e9a 100644
@@ -1670,6 +1670,36 @@ static int devx_umem_cleanup(struct ib_uobject *uobject,
 	return 0;
 }
 
+static int devx_event_notifier(struct notifier_block *nb,
+			       unsigned long event_type, void *data)
+{
+	return NOTIFY_DONE;
+}
+
+void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_devx_event_table *table = &dev->devx_event_table;
+
+	xa_init(&table->event_xa);
+	mutex_init(&table->event_xa_lock);
+	MLX5_NB_INIT(&table->devx_nb, devx_event_notifier, NOTIFY_ANY);
+	mlx5_eq_notifier_register(dev->mdev, &table->devx_nb);
+}
+
+void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev)
+{
+	struct mlx5_devx_event_table *table = &dev->devx_event_table;
+	void *entry;
+	unsigned long id;
+
+	mlx5_eq_notifier_unregister(dev->mdev, &table->devx_nb);
+
+	xa_for_each(&table->event_xa, id, entry)
+		kfree(entry);
+
+	xa_destroy(&table->event_xa);
+}
+
 static ssize_t devx_async_cmd_event_read(struct file *filp, char __user *buf,
 					 size_t count, loff_t *pos)
 {
drivers/infiniband/hw/mlx5/main.c +6 -2
index c9c6710afc6b..67b9e7ac569a 100644
@@ -6653,15 +6653,19 @@ static int mlx5_ib_stage_devx_init(struct mlx5_ib_dev *dev)
 	int uid;
 
 	uid = mlx5_ib_devx_create(dev, false);
-	if (uid > 0)
+	if (uid > 0) {
 		dev->devx_whitelist_uid = uid;
+		mlx5_ib_devx_init_event_table(dev);
+	}
 
 	return 0;
 }
 static void mlx5_ib_stage_devx_cleanup(struct mlx5_ib_dev *dev)
 {
-	if (dev->devx_whitelist_uid)
+	if (dev->devx_whitelist_uid) {
+		mlx5_ib_devx_cleanup_event_table(dev);
 		mlx5_ib_devx_destroy(dev, dev->devx_whitelist_uid);
+	}
 }
 
 void __mlx5_ib_remove(struct mlx5_ib_dev *dev,
drivers/infiniband/hw/mlx5/mlx5_ib.h +12 -0
index 9cf23ae6324e..556af34b788b 100644
@@ -944,6 +944,13 @@ struct mlx5_ib_pf_eq {
 	mempool_t *pool;
 };
 
+struct mlx5_devx_event_table {
+	struct mlx5_nb devx_nb;
+	/* serialize updating the event_xa */
+	struct mutex event_xa_lock;
+	struct xarray event_xa;
+};
+
 struct mlx5_ib_dev {
 	struct ib_device		ib_dev;
 	struct mlx5_core_dev		*mdev;
@@ -994,6 +1001,7 @@ struct mlx5_ib_dev {
 	struct mlx5_srq_table   srq_table;
 	struct mlx5_async_ctx   async_ctx;
 	int			free_port;
+	struct mlx5_devx_event_table devx_event_table;
 };
 
 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
@@ -1333,6 +1341,8 @@ void mlx5_ib_put_native_port_mdev(struct mlx5_ib_dev *dev,
 #if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)
 int mlx5_ib_devx_create(struct mlx5_ib_dev *dev, bool is_user);
 void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid);
+void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev);
+void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev);
 const struct uverbs_object_tree_def *mlx5_ib_get_devx_tree(void);
 extern const struct uapi_definition mlx5_ib_devx_defs[];
 extern const struct uapi_definition mlx5_ib_flow_defs[];
@@ -1349,6 +1359,8 @@ static inline int
 mlx5_ib_devx_create(struct mlx5_ib_dev *dev,
 			   bool is_user) { return -EOPNOTSUPP; }
 static inline void mlx5_ib_devx_destroy(struct mlx5_ib_dev *dev, u16 uid) {}
+static inline void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev) {}
+static inline void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev) {}
 static inline bool mlx5_ib_devx_is_flow_dest(void *obj, int *dest_id,
 					     int *dest_type)
 {
Signature · 1 line
2.20.1
#17 lore raw

PATCH 10/12 IB/mlx5: Enable subscription for device events over DEVX

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Enable subscription for device events over DEVX.

Each subscription is added to the two level XA data structure according to its event number and the DEVX object information in case was given with the given target fd.

Those events will be reported over the given fd once will occur. Downstream patches will mange the dispatching to any subscription.

---
 drivers/infiniband/hw/mlx5/devx.c        | 564 ++++++++++++++++++++++-
 include/uapi/rdma/mlx5_user_ioctl_cmds.h |   9 +
 2 files changed, 566 insertions(+), 7 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +557 -7
index e9b9ba5a3e9a..304b13e7a265 100644
@@ -14,6 +14,7 @@
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/fs.h>
 #include "mlx5_ib.h"
+#include <linux/xarray.h>
 
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
@@ -33,6 +34,37 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+/* first level XA value data structure */
+struct devx_event {
+	struct xarray object_ids; /* second XA level, Key = object id */
+	struct list_head unaffiliated_list;
+};
+
+/* second level XA value data structure */
+struct devx_obj_event {
+	struct rcu_head rcu;
+	struct list_head obj_sub_list;
+};
+
+struct devx_event_subscription {
+	struct list_head file_list; /* headed in private_data->
+				     * subscribed_events_list
+				     */
+	struct list_head xa_list; /* headed in devx_event->unaffiliated_list or
+				   * devx_obj_event->obj_sub_list
+				   */
+	struct list_head obj_list; /* headed in devx_object */
+
+	u32 xa_key_level1;
+	u32 xa_key_level2;
+	struct rcu_head	rcu;
+	u64 cookie;
+	bool is_obj_related;
+	struct ib_uobject *fd_uobj;
+	void *object;	/* May need direct access upon hot unplug */
+	struct eventfd_ctx *eventfd;
+};
+
 struct devx_async_event_queue {
 	spinlock_t		lock;
 	wait_queue_head_t	poll_wait;
@@ -62,6 +94,7 @@ struct devx_obj {
 		struct mlx5_ib_devx_mr	devx_mr;
 		struct mlx5_core_dct	core_dct;
 	};
+	struct list_head event_sub; /* holds devx_event_subscription entries */
 };
 
 struct devx_umem {
@@ -167,6 +200,105 @@ bool mlx5_ib_devx_is_flow_counter(void *obj, u32 *counter_id)
 	return false;
 }
 
+static bool is_legacy_unaffiliated_event_num(u16 event_num)
+{
+	switch (event_num) {
+	case MLX5_EVENT_TYPE_PORT_CHANGE:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static bool is_legacy_obj_event_num(u16 event_num)
+{
+	switch (event_num) {
+	case MLX5_EVENT_TYPE_PATH_MIG:
+	case MLX5_EVENT_TYPE_COMM_EST:
+	case MLX5_EVENT_TYPE_SQ_DRAINED:
+	case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+	case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+	case MLX5_EVENT_TYPE_CQ_ERROR:
+	case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+	case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+	case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+	case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+	case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+	case MLX5_EVENT_TYPE_DCT_DRAINED:
+	case MLX5_EVENT_TYPE_COMP:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static u16 get_legacy_obj_type(u16 opcode)
+{
+	switch (opcode) {
+	case MLX5_CMD_OP_CREATE_RQ:
+	case MLX5_CMD_OP_CREATE_XRC_SRQ:
+		return MLX5_EVENT_QUEUE_TYPE_RQ;
+	case MLX5_CMD_OP_CREATE_QP:
+		return MLX5_EVENT_QUEUE_TYPE_QP;
+	case MLX5_CMD_OP_CREATE_SQ:
+		return MLX5_EVENT_QUEUE_TYPE_SQ;
+	case MLX5_CMD_OP_CREATE_DCT:
+		return MLX5_EVENT_QUEUE_TYPE_DCT;
+	default:
+		return 0;
+	}
+}
+
+static u16 get_dec_obj_type(struct devx_obj *obj, u16 event_num)
+{
+	u16 opcode;
+
+	opcode = (obj->obj_id >> 32) & 0xffff;
+
+	if (is_legacy_obj_event_num(event_num))
+		return get_legacy_obj_type(opcode);
+
+	switch (opcode) {
+	case MLX5_CMD_OP_CREATE_GENERAL_OBJECT:
+		return (obj->obj_id >> 48);
+	case MLX5_CMD_OP_CREATE_RQ:
+		return MLX5_OBJ_TYPE_RQ;
+	case MLX5_CMD_OP_CREATE_QP:
+		return MLX5_OBJ_TYPE_QP;
+	case MLX5_CMD_OP_CREATE_SQ:
+		return MLX5_OBJ_TYPE_SQ;
+	case MLX5_CMD_OP_CREATE_DCT:
+		return MLX5_OBJ_TYPE_DCT;
+	case MLX5_CMD_OP_CREATE_TIR:
+		return MLX5_OBJ_TYPE_TIR;
+	case MLX5_CMD_OP_CREATE_TIS:
+		return MLX5_OBJ_TYPE_TIS;
+	case MLX5_CMD_OP_CREATE_PSV:
+		return MLX5_OBJ_TYPE_PSV;
+	case MLX5_OBJ_TYPE_MKEY:
+		return MLX5_OBJ_TYPE_MKEY;
+	case MLX5_CMD_OP_CREATE_RMP:
+		return MLX5_OBJ_TYPE_RMP;
+	case MLX5_CMD_OP_CREATE_XRC_SRQ:
+		return MLX5_OBJ_TYPE_XRC_SRQ;
+	case MLX5_CMD_OP_CREATE_XRQ:
+		return MLX5_OBJ_TYPE_XRQ;
+	case MLX5_CMD_OP_CREATE_RQT:
+		return MLX5_OBJ_TYPE_RQT;
+	case MLX5_CMD_OP_ALLOC_FLOW_COUNTER:
+		return MLX5_OBJ_TYPE_FLOW_COUNTER;
+	case MLX5_CMD_OP_CREATE_CQ:
+		return MLX5_OBJ_TYPE_CQ;
+	default:
+		return 0;
+	}
+}
+
+static u32 get_dec_obj_id(u64 obj_id)
+{
+	return (obj_id & 0xffffffff);
+}
+
 /*
  * As the obj_id in the firmware is not globally unique the object type
  * must be considered upon checking for a valid object id.
@@ -1143,14 +1275,53 @@ static void devx_cleanup_mkey(struct devx_obj *obj)
 	write_unlock_irqrestore(&table->lock, flags);
 }
 
+static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
+				      struct devx_event_subscription *sub)
+{
+	list_del_rcu(&sub->file_list);
+	list_del_rcu(&sub->xa_list);
+
+	if (sub->is_obj_related) {
+		struct devx_event *event;
+		struct devx_obj_event *xa_val_level2;
+
+		list_del_rcu(&sub->obj_list);
+
+		/* check whether the key level 1 for this obj_sub_list
+		 * is empty
+		 */
+		event = xa_load(&dev->devx_event_table.event_xa,
+				sub->xa_key_level1);
+		WARN_ON(!event);
+
+		xa_val_level2 = xa_load(&event->object_ids,
+				sub->xa_key_level2);
+		if (list_empty(&xa_val_level2->obj_sub_list)) {
+			xa_erase(&event->object_ids,
+				 sub->xa_key_level2);
+			kfree_rcu(xa_val_level2, rcu);
+		}
+	}
+
+	if (sub->eventfd)
+		eventfd_ctx_put(sub->eventfd);
+
+	kfree_rcu(sub, rcu);
+}
+
 static int devx_obj_cleanup(struct ib_uobject *uobject,
 			    enum rdma_remove_reason why,
 			    struct uverbs_attr_bundle *attrs)
 {
 	u32 out[MLX5_ST_SZ_DW(general_obj_out_cmd_hdr)];
+	struct mlx5_devx_event_table *devx_event_table;
 	struct devx_obj *obj = uobject->object;
+	struct devx_event_subscription *sub_entry, *tmp;
+	struct mlx5_ib_dev *dev;
+	u32 obj_id;
 	int ret;
 
+	dev = mlx5_udata_to_mdev(&attrs->driver_udata);
 	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY)
 		devx_cleanup_mkey(obj);
 
@@ -1162,10 +1333,15 @@ static int devx_obj_cleanup(struct ib_uobject *uobject,
 	if (ib_is_destroy_retryable(ret, why, uobject))
 		return ret;
 
-	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) {
-		struct mlx5_ib_dev *dev =
-			mlx5_udata_to_mdev(&attrs->driver_udata);
+	devx_event_table = &dev->devx_event_table;
+	obj_id = get_dec_obj_id(obj->obj_id);
 
+	mutex_lock(&devx_event_table->event_xa_lock);
+	list_for_each_entry_safe(sub_entry, tmp, &obj->event_sub, obj_list)
+		devx_cleanup_subscription(dev, sub_entry);
+	mutex_unlock(&devx_event_table->event_xa_lock);
+
+	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) {
 		call_srcu(&dev->mr_srcu, &obj->devx_mr.rcu,
 			  devx_free_indirect_mkey);
 		return ret;
@@ -1237,6 +1413,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
 
 	uobj->object = obj;
 	obj->mdev = dev->mdev;
+	INIT_LIST_HEAD(&obj->event_sub);
 	devx_obj_build_destroy_cmd(cmd_in, cmd_out, obj->dinbox, &obj->dinlen,
 				   &obj_id);
 	WARN_ON(obj->dinlen > MLX5_MAX_DESTROY_INBOX_SIZE_DW * sizeof(u32));
@@ -1523,6 +1700,350 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)(
 	return err;
 }
 
+static void
+subscribe_event_xa_dealloc(struct mlx5_devx_event_table *devx_event_table,
+			   u32 key_level1,
+			   u32 key_level2,
+			   struct devx_obj_event *alloc_obj_event)
+{
+	struct devx_event *event;
+
+	/* Level 1 is valid for future use - no need to free */
+	if (!alloc_obj_event)
+		return;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	WARN_ON(!event);
+
+	xa_erase(&event->object_ids, key_level2);
+	kfree(alloc_obj_event);
+}
+
+static int
+subscribe_event_xa_alloc(struct mlx5_devx_event_table *devx_event_table,
+			 u32 key_level1,
+			 bool is_level2,
+			 u32 key_level2,
+			 struct devx_obj_event **alloc_obj_event)
+{
+	struct devx_obj_event *obj_event;
+	struct devx_event *event;
+	bool new_entry_level1 = false;
+	int err;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	if (!event) {
+		event = kzalloc(sizeof(*event), GFP_KERNEL);
+		if (!event)
+			return -ENOMEM;
+
+		new_entry_level1 = true;
+		INIT_LIST_HEAD(&event->unaffiliated_list);
+		xa_init(&event->object_ids);
+
+		err = xa_insert(&devx_event_table->event_xa,
+				key_level1,
+				event,
+				GFP_KERNEL);
+		if (err)
+			goto end;
+	}
+
+	if (!is_level2)
+		return 0;
+
+	obj_event = xa_load(&event->object_ids, key_level2);
+	if (!obj_event) {
+		err = xa_reserve(&event->object_ids, key_level2, GFP_KERNEL);
+		if (err)
+			goto err_level1;
+
+		obj_event = kzalloc(sizeof(*obj_event), GFP_KERNEL);
+		if (!obj_event) {
+			err = -ENOMEM;
+			goto err_level2;
+		}
+
+		INIT_LIST_HEAD(&obj_event->obj_sub_list);
+		*alloc_obj_event = obj_event;
+	}
+
+	return 0;
+
+err_level2:
+	xa_erase(&event->object_ids, key_level2);
+
+err_level1:
+	if (new_entry_level1)
+		xa_erase(&devx_event_table->event_xa, key_level1);
+end:
+	if (new_entry_level1)
+		kfree(event);
+	return err;
+}
+
+static bool is_valid_events_legacy(int num_events, u16 *event_type_num_list,
+				   struct devx_obj *obj)
+{
+	int i;
+
+	for (i = 0; i < num_events; i++) {
+		if (obj) {
+			if (!is_legacy_obj_event_num(event_type_num_list[i]))
+				return false;
+		} else if (!is_legacy_unaffiliated_event_num(
+				event_type_num_list[i])) {
+			return false;
+		}
+	}
+
+	return true;
+}
+
+#define MAX_SUPP_EVENT_NUM 255
+static bool is_valid_events(struct mlx5_core_dev *dev,
+			    int num_events, u16 *event_type_num_list,
+			    struct devx_obj *obj)
+{
+	__be64 *aff_events;
+	__be64 *unaff_events;
+	int mask_entry;
+	int mask_bit;
+	int i;
+
+	if (MLX5_CAP_GEN(dev, event_cap)) {
+		aff_events = MLX5_CAP_DEV_EVENT(dev,
+						user_affiliated_events);
+		unaff_events = MLX5_CAP_DEV_EVENT(dev,
+						  user_unaffiliated_events);
+	} else {
+		return is_valid_events_legacy(num_events, event_type_num_list,
+					      obj);
+	}
+
+	for (i = 0; i < num_events; i++) {
+		if (event_type_num_list[i] > MAX_SUPP_EVENT_NUM)
+			return false;
+
+		mask_entry = event_type_num_list[i] / 64;
+		mask_bit = event_type_num_list[i] % 64;
+
+		if (obj) {
+			/* CQ completion */
+			if (event_type_num_list[i] == 0)
+				continue;
+
+			if (!(be64_to_cpu(aff_events[mask_entry]) &
+					(1ull << mask_bit)))
+				return false;
+
+			continue;
+		}
+
+		if (!(be64_to_cpu(unaff_events[mask_entry]) &
+				(1ull << mask_bit)))
+			return false;
+	}
+
+	return true;
+}
+
+#define MAX_NUM_EVENTS 16
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *devx_uobj = uverbs_attr_get_uobject(
+				attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_OBJ_HANDLE);
+	struct mlx5_ib_ucontext *c = rdma_udata_to_drv_context(
+		&attrs->driver_udata, struct mlx5_ib_ucontext, ibucontext);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	struct ib_uobject *fd_uobj;
+	struct devx_obj *obj = NULL;
+	struct devx_async_event_file *ev_file;
+	struct mlx5_devx_event_table *devx_event_table = &dev->devx_event_table;
+	u16 *event_type_num_list;
+	struct devx_event_subscription **event_sub_arr;
+	struct devx_obj_event  **event_obj_array_alloc;
+	int redirect_fd;
+	bool use_eventfd = false;
+	int num_events;
+	int num_alloc_xa_entries = 0;
+	u16 obj_type = 0;
+	u64 cookie = 0;
+	u32 obj_id = 0;
+	int err;
+	int i;
+
+	if (!c->devx_uid)
+		return -EINVAL;
+
+	if (!IS_ERR(devx_uobj)) {
+		obj = (struct devx_obj *)devx_uobj->object;
+		if (obj)
+			obj_id = get_dec_obj_id(obj->obj_id);
+	}
+
+	fd_uobj = uverbs_attr_get_uobject(attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_HANDLE);
+	if (IS_ERR(fd_uobj))
+		return PTR_ERR(fd_uobj);
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM)) {
+		err = uverbs_copy_from(&redirect_fd, attrs,
+			       MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM);
+		if (err)
+			return err;
+
+		use_eventfd = true;
+	}
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE)) {
+		if (use_eventfd)
+			return -EINVAL;
+
+		err = uverbs_copy_from(&cookie, attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE);
+		if (err)
+			return err;
+	}
+
+	num_events = uverbs_attr_ptr_get_array_size(
+		attrs, MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST,
+		sizeof(u16));
+
+	if (num_events < 0)
+		return num_events;
+
+	if (num_events > MAX_NUM_EVENTS)
+		return -EINVAL;
+
+	event_type_num_list = uverbs_attr_get_alloced_ptr(attrs,
+			MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST);
+
+	if (!is_valid_events(dev->mdev, num_events, event_type_num_list, obj))
+		return -EINVAL;
+
+	event_sub_arr = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_event_subscription *));
+	event_obj_array_alloc = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_obj_event *));
+
+	if (!event_sub_arr || !event_obj_array_alloc)
+		return -ENOMEM;
+
+	/* Protect from concurrent subscriptions to same XA entries to allow
+	 * both to succeed
+	 */
+	mutex_lock(&devx_event_table->event_xa_lock);
+	for (i = 0; i < num_events; i++) {
+		u32 key_level1;
+
+		if (obj)
+			obj_type = get_dec_obj_type(obj,
+						    event_type_num_list[i]);
+		key_level1 = event_type_num_list[i] | obj_type << 16;
+
+		err = subscribe_event_xa_alloc(devx_event_table,
+					       key_level1,
+					       obj ? true : false,
+					       obj_id,
+					       &event_obj_array_alloc[i]);
+		if (err)
+			goto err;
+
+		num_alloc_xa_entries++;
+		event_sub_arr[i] = kzalloc(sizeof(*event_sub_arr[i]),
+					   GFP_KERNEL);
+		if (!event_sub_arr[i])
+			goto err;
+
+		if (use_eventfd) {
+			event_sub_arr[i]->eventfd =
+				eventfd_ctx_fdget(redirect_fd);
+
+			if (IS_ERR(event_sub_arr[i]->eventfd)) {
+				err = PTR_ERR(event_sub_arr[i]->eventfd);
+				event_sub_arr[i]->eventfd = NULL;
+				goto err;
+			}
+		}
+
+		event_sub_arr[i]->cookie = cookie;
+		event_sub_arr[i]->fd_uobj = fd_uobj;
+		event_sub_arr[i]->object = fd_uobj->object;
+		/* May be needed upon cleanup the devx object/subscription */
+		event_sub_arr[i]->xa_key_level1 = key_level1;
+		event_sub_arr[i]->xa_key_level2 = obj_id;
+		event_sub_arr[i]->is_obj_related = obj ? true : false;
+	}
+
+	/* Once all the allocations and the reservations for level 2
+	 * have been done we can go ahead and insert the data without any
+	 * concern of a failure.
+	 */
+	for (i = 0; i < num_events; i++) {
+		struct devx_event *event;
+		struct devx_obj_event  *obj_event;
+
+		spin_lock_irq(&ev_file->ev_queue.lock);
+		list_add_tail_rcu(&event_sub_arr[i]->file_list,
+				  &ev_file->subscribed_events_list);
+		spin_unlock_irq(&ev_file->ev_queue.lock);
+
+		event = xa_load(&devx_event_table->event_xa,
+				event_sub_arr[i]->xa_key_level1);
+		WARN_ON(!event);
+
+		if (!obj) {
+			list_add_tail_rcu(&event_sub_arr[i]->xa_list,
+					  &event->unaffiliated_list);
+			continue;
+		}
+
+		/* If was no entry prior calling this API need to insert
+		 * this XA key level 2 entry.
+		 */
+		if (event_obj_array_alloc[i]) {
+			err = xa_err(xa_store(&event->object_ids, obj_id,
+					      event_obj_array_alloc[i],
+					      GFP_KERNEL));
+			WARN_ON(err);
+		}
+
+		obj_event = xa_load(&event->object_ids, obj_id);
+		WARN_ON(!obj_event);
+		list_add_tail_rcu(&event_sub_arr[i]->xa_list,
+				  &obj_event->obj_sub_list);
+		list_add_tail_rcu(&event_sub_arr[i]->obj_list,
+				  &obj->event_sub);
+	}
+
+	mutex_unlock(&devx_event_table->event_xa_lock);
+	return 0;
+
+err:
+	for (i = 0; i < num_alloc_xa_entries; i++) {
+		subscribe_event_xa_dealloc(devx_event_table,
+					   event_sub_arr[i]->xa_key_level1,
+					   obj_id,
+					   event_obj_array_alloc[i]);
+
+		if (event_sub_arr[i]->eventfd)
+			eventfd_ctx_put(event_sub_arr[i]->eventfd);
+
+		kvfree(event_sub_arr[i]);
+	}
+
+	mutex_unlock(&devx_event_table->event_xa_lock);
+	return err;
+}
+
 static int devx_umem_get(struct mlx5_ib_dev *dev, struct ib_ucontext *ucontext,
 			 struct uverbs_attr_bundle *attrs,
 			 struct devx_umem *obj)
@@ -1689,14 +2210,21 @@ void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev)
 void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev)
 {
 	struct mlx5_devx_event_table *table = &dev->devx_event_table;
+	struct devx_event_subscription *sub, *tmp;
+	struct devx_event *event;
 	void *entry;
 	unsigned long id;
 
 	mlx5_eq_notifier_unregister(dev->mdev, &table->devx_nb);
-
-	xa_for_each(&table->event_xa, id, entry)
+	mutex_lock(&dev->devx_event_table.event_xa_lock);
+	xa_for_each(&table->event_xa, id, entry) {
+		event = entry;
+		list_for_each_entry_safe(sub, tmp, &event->unaffiliated_list,
+					 xa_list)
+			devx_cleanup_subscription(dev, sub);
 		kfree(entry);
-
+	}
+	mutex_unlock(&dev->devx_event_table.event_xa_lock);
 	xa_destroy(&table->event_xa);
 }
 
@@ -1980,10 +2508,32 @@ DECLARE_UVERBS_NAMED_METHOD(
 		UVERBS_ATTR_TYPE(u64),
 		UA_MANDATORY));
 
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT,
+	UVERBS_ATTR_FD(MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_HANDLE,
+		MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD,
+		UVERBS_ACCESS_READ,
+		UA_MANDATORY),
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_OBJ_HANDLE,
+		MLX5_IB_OBJECT_DEVX_OBJ,
+		UVERBS_ACCESS_READ,
+		UA_OPTIONAL),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST,
+		UVERBS_ATTR_MIN_SIZE(sizeof(u16)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE,
+		UVERBS_ATTR_TYPE(u64),
+		UA_OPTIONAL),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM,
+		UVERBS_ATTR_TYPE(u32),
+		UA_OPTIONAL));
+
 DECLARE_UVERBS_GLOBAL_METHODS(MLX5_IB_OBJECT_DEVX,
 			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OTHER),
 			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_UAR),
-			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_EQN));
+			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_EQN),
+			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT));
 
 DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_OBJ,
 			    UVERBS_TYPE_ALLOC_IDR(devx_obj_cleanup),
include/uapi/rdma/mlx5_user_ioctl_cmds.h +9 -0
index 6ad8f4f11ddd..d0da070cf0ab 100644
@@ -51,6 +51,7 @@ enum mlx5_ib_devx_methods {
 	MLX5_IB_METHOD_DEVX_OTHER  = (1U << UVERBS_ID_NS_SHIFT),
 	MLX5_IB_METHOD_DEVX_QUERY_UAR,
 	MLX5_IB_METHOD_DEVX_QUERY_EQN,
+	MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT,
 };
 
 enum  mlx5_ib_devx_other_attrs {
@@ -93,6 +94,14 @@ enum mlx5_ib_devx_obj_query_async_attrs {
 	MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_OUT_LEN,
 };
 
+enum mlx5_ib_devx_subscribe_event_attrs {
+	MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_OBJ_HANDLE,
+	MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST,
+	MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM,
+	MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE,
+};
+
 enum  mlx5_ib_devx_query_eqn_attrs {
 	MLX5_IB_ATTR_DEVX_QUERY_EQN_USER_VEC = (1U << UVERBS_ID_NS_SHIFT),
 	MLX5_IB_ATTR_DEVX_QUERY_EQN_DEV_EQN,
Signature · 1 line
2.20.1
#21 lore raw

PATCH 11/12 IB/mlx5: Implement DEVX dispatching event

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Implement DEVX dispatching event by looking up for the applicable subscriptions for the reported event and using their target fd to signal/set the event.

---
 drivers/infiniband/hw/mlx5/devx.c         | 362 +++++++++++++++++++++-
 include/uapi/rdma/mlx5_user_ioctl_verbs.h |   5 +
 2 files changed, 357 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +352 -10
index 304b13e7a265..49fdce95d6d9 100644
@@ -34,6 +34,11 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+struct devx_async_event_data {
+	struct list_head list; /* headed in ev_queue->event_list */
+	struct mlx5_ib_uapi_devx_async_event_hdr hdr;
+};
+
 /* first level XA value data structure */
 struct devx_event {
 	struct xarray object_ids; /* second XA level, Key = object id */
@@ -54,7 +59,9 @@ struct devx_event_subscription {
 				   * devx_obj_event->obj_sub_list
 				   */
 	struct list_head obj_list; /* headed in devx_object */
+	struct list_head event_list; /* headed in ev_queue->event_list */
 
+	u8  is_cleaned:1;
 	u32 xa_key_level1;
 	u32 xa_key_level2;
 	struct rcu_head	rcu;
@@ -71,6 +78,7 @@ struct devx_async_event_queue {
 	struct list_head	event_list;
 	atomic_t		bytes_in_use;
 	u8			is_destroyed:1;
+	u8			is_overflow_err:1;
 	u32			flags;
 };
 
@@ -294,6 +302,46 @@ static u16 get_dec_obj_type(struct devx_obj *obj, u16 event_num)
 	}
 }
 
+/* Any future affiliated event should have a fixed header to get the obj
+ * type and id including events on legacy objects.
+ */
+static u32 get_affiliated_event_obj_id(struct mlx5_eqe *eqe)
+{
+	u32 obj_id = MLX5_GET(affiliated_event_header, eqe, obj_id);
+
+	return obj_id;
+}
+
+static u16 get_affiliated_event_obj_type(struct mlx5_eqe *eqe)
+{
+	u16 obj_type = MLX5_GET(affiliated_event_header, eqe, obj_type);
+
+	return obj_type;
+}
+
+static u16 get_event_obj_type(unsigned long event_type, struct mlx5_eqe *eqe)
+{
+	switch (event_type) {
+	case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+	case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+	case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+	case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+	case MLX5_EVENT_TYPE_PATH_MIG:
+	case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+	case MLX5_EVENT_TYPE_COMM_EST:
+	case MLX5_EVENT_TYPE_SQ_DRAINED:
+	case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+	case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+		return eqe->data.qp_srq.type;
+	case MLX5_EVENT_TYPE_CQ_ERROR:
+		return 0;
+	case MLX5_EVENT_TYPE_DCT_DRAINED:
+		return MLX5_EVENT_QUEUE_TYPE_DCT;
+	default:
+		return get_affiliated_event_obj_type(eqe);
+	}
+}
+
 static u32 get_dec_obj_id(u64 obj_id)
 {
 	return (obj_id & 0xffffffff);
@@ -1276,9 +1324,13 @@ static void devx_cleanup_mkey(struct devx_obj *obj)
 }
 
 static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
-				      struct devx_event_subscription *sub)
+				      struct devx_event_subscription *sub,
+				      bool file_close)
 {
-	list_del_rcu(&sub->file_list);
+	if (sub->is_cleaned)
+		goto end;
+
+	sub->is_cleaned = 1;
 	list_del_rcu(&sub->xa_list);
 
 	if (sub->is_obj_related) {
@@ -1303,10 +1355,15 @@ static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
 		}
 	}
 
-	if (sub->eventfd)
-		eventfd_ctx_put(sub->eventfd);
+end:
+	if (file_close) {
+		if (sub->eventfd)
+			eventfd_ctx_put(sub->eventfd);
 
-	kfree_rcu(sub, rcu);
+		list_del_rcu(&sub->file_list);
+		/* subscription may not be used by the read API any more */
+		kfree_rcu(sub, rcu);
+	}
 }
 
 static int devx_obj_cleanup(struct ib_uobject *uobject,
@@ -1338,7 +1395,7 @@ static int devx_obj_cleanup(struct ib_uobject *uobject,
 
 	mutex_lock(&devx_event_table->event_xa_lock);
 	list_for_each_entry_safe(sub_entry, tmp, &obj->event_sub, obj_list)
-		devx_cleanup_subscription(dev, sub_entry);
+		devx_cleanup_subscription(dev, sub_entry, false);
 	mutex_unlock(&devx_event_table->event_xa_lock);
 
 	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) {
@@ -1588,6 +1645,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_ASYNC_EVENT_FD_ALLOC)(
 	devx_init_event_queue(&ev_file->ev_queue, flags);
 	INIT_LIST_HEAD(&ev_file->subscribed_events_list);
 	ev_file->dev = dev;
+	get_device(&dev->ib_dev.dev);
 	return 0;
 }
 
@@ -1981,6 +2039,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT)(
 		event_sub_arr[i]->xa_key_level1 = key_level1;
 		event_sub_arr[i]->xa_key_level2 = obj_id;
 		event_sub_arr[i]->is_obj_related = obj ? true : false;
+		INIT_LIST_HEAD(&event_sub_arr[i]->event_list);
 	}
 
 	/* Once all the allocations and the reservations for level 2
@@ -2191,10 +2250,174 @@ static int devx_umem_cleanup(struct ib_uobject *uobject,
 	return 0;
 }
 
+static bool is_unaffiliated_event(struct mlx5_core_dev *dev,
+				  unsigned long event_type)
+{
+	__be64 *unaff_events;
+	int mask_entry;
+	int mask_bit;
+
+	if (!MLX5_CAP_GEN(dev, event_cap))
+		return is_legacy_unaffiliated_event_num(event_type);
+
+	unaff_events = MLX5_CAP_DEV_EVENT(dev,
+					  user_unaffiliated_events);
+	WARN_ON(event_type > MAX_SUPP_EVENT_NUM);
+
+	mask_entry = event_type / 64;
+	mask_bit = event_type % 64;
+
+	if (!(be64_to_cpu(unaff_events[mask_entry]) & (1ull << mask_bit)))
+		return false;
+
+	return true;
+}
+
+static u32 devx_get_obj_id_from_event(unsigned long event_type, void *data)
+{
+	struct mlx5_eqe *eqe = data;
+	u32 obj_id = 0;
+
+	switch (event_type) {
+	case MLX5_EVENT_TYPE_SRQ_CATAS_ERROR:
+	case MLX5_EVENT_TYPE_SRQ_RQ_LIMIT:
+	case MLX5_EVENT_TYPE_PATH_MIG:
+	case MLX5_EVENT_TYPE_COMM_EST:
+	case MLX5_EVENT_TYPE_SQ_DRAINED:
+	case MLX5_EVENT_TYPE_SRQ_LAST_WQE:
+	case MLX5_EVENT_TYPE_WQ_CATAS_ERROR:
+	case MLX5_EVENT_TYPE_PATH_MIG_FAILED:
+	case MLX5_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
+	case MLX5_EVENT_TYPE_WQ_ACCESS_ERROR:
+		obj_id = be32_to_cpu(eqe->data.qp_srq.qp_srq_n) & 0xffffff;
+		break;
+	case MLX5_EVENT_TYPE_DCT_DRAINED:
+		obj_id = be32_to_cpu(eqe->data.dct.dctn) & 0xffffff;
+		break;
+	case MLX5_EVENT_TYPE_CQ_ERROR:
+		obj_id = be32_to_cpu(eqe->data.cq_err.cqn) & 0xffffff;
+		break;
+	default:
+		obj_id = get_affiliated_event_obj_id(data);
+		break;
+	}
+
+	return obj_id;
+}
+
+static int deliver_event(struct devx_event_subscription *event_sub,
+			 const void *data)
+{
+	struct ib_uobject *fd_uobj = event_sub->fd_uobj;
+	struct devx_async_event_file *ev_file;
+	struct devx_async_event_queue *ev_queue;
+	struct devx_async_event_data *event_data;
+	unsigned long flags;
+	bool omit_data;
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+	ev_queue = &ev_file->ev_queue;
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	if (omit_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		if (!list_empty(&event_sub->event_list)) {
+			spin_unlock_irqrestore(&ev_queue->lock, flags);
+			return 0;
+		}
+
+		list_add_tail(&event_sub->event_list, &ev_queue->event_list);
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		wake_up_interruptible(&ev_queue->poll_wait);
+		return 0;
+	}
+
+	event_data = kzalloc(sizeof(*event_data) +
+			     (omit_data ? 0 : sizeof(struct mlx5_eqe)),
+			     GFP_ATOMIC);
+	if (!event_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		ev_queue->is_overflow_err = 1;
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		return -ENOMEM;
+	}
+
+	event_data->hdr.cookie = event_sub->cookie;
+	memcpy(event_data->hdr.out_data, data, sizeof(struct mlx5_eqe));
+
+	spin_lock_irqsave(&ev_queue->lock, flags);
+	list_add_tail(&event_data->list, &ev_queue->event_list);
+	spin_unlock_irqrestore(&ev_queue->lock, flags);
+	wake_up_interruptible(&ev_queue->poll_wait);
+
+	return 0;
+}
+
+static void dispatch_event_fd(struct list_head *fd_list,
+			      const void *data)
+{
+	struct devx_event_subscription *item;
+
+	list_for_each_entry_rcu(item, fd_list, xa_list) {
+		if (!get_file_rcu((struct file *)item->object))
+			continue;
+
+		if (item->eventfd) {
+			eventfd_signal(item->eventfd, 1);
+			fput(item->object);
+			continue;
+		}
+
+		deliver_event(item, data);
+		fput(item->object);
+	}
+}
+
 static int devx_event_notifier(struct notifier_block *nb,
 			       unsigned long event_type, void *data)
 {
-	return NOTIFY_DONE;
+	struct mlx5_devx_event_table *table;
+	struct mlx5_ib_dev *dev;
+	struct devx_event *event;
+	struct devx_obj_event *obj_event;
+	u16 obj_type = 0;
+	bool is_unaffiliated;
+	u32 obj_id;
+
+	/* Explicit filtering to kernel events which may occur frequently */
+	if (event_type == MLX5_EVENT_TYPE_CMD ||
+	    event_type == MLX5_EVENT_TYPE_PAGE_REQUEST)
+		return NOTIFY_OK;
+
+	table = container_of(nb, struct mlx5_devx_event_table, devx_nb.nb);
+	dev = container_of(table, struct mlx5_ib_dev, devx_event_table);
+	is_unaffiliated = is_unaffiliated_event(dev->mdev, event_type);
+
+	if (!is_unaffiliated)
+		obj_type = get_event_obj_type(event_type, data);
+	event = xa_load(&table->event_xa, event_type | (obj_type << 16));
+	if (!event)
+		return NOTIFY_DONE;
+
+	if (is_unaffiliated) {
+		dispatch_event_fd(&event->unaffiliated_list, data);
+		return NOTIFY_OK;
+	}
+
+	obj_id = devx_get_obj_id_from_event(event_type, data);
+	rcu_read_lock();
+	obj_event = xa_load(&event->object_ids, obj_id);
+	if (!obj_event) {
+		rcu_read_unlock();
+		return NOTIFY_DONE;
+	}
+
+	dispatch_event_fd(&obj_event->obj_sub_list, data);
+
+	rcu_read_unlock();
+	return NOTIFY_OK;
 }
 
 void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev)
@@ -2221,7 +2444,7 @@ void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev)
 		event = entry;
 		list_for_each_entry_safe(sub, tmp, &event->unaffiliated_list,
 					 xa_list)
-			devx_cleanup_subscription(dev, sub);
+			devx_cleanup_subscription(dev, sub, false);
 		kfree(entry);
 	}
 	mutex_unlock(&dev->devx_event_table.event_xa_lock);
@@ -2329,18 +2552,126 @@ static const struct file_operations devx_async_cmd_event_fops = {
 static ssize_t devx_async_event_read(struct file *filp, char __user *buf,
 				     size_t count, loff_t *pos)
 {
-	return -EINVAL;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	struct devx_event_subscription *event_sub;
+	struct devx_async_event_data *uninitialized_var(event);
+	int ret = 0;
+	size_t eventsz;
+	bool omit_data;
+	void *event_data;
+
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	spin_lock_irq(&ev_queue->lock);
+
+	if (ev_queue->is_overflow_err) {
+		ev_queue->is_overflow_err = 0;
+		spin_unlock_irq(&ev_queue->lock);
+		return -EOVERFLOW;
+	}
+
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		if (wait_event_interruptible(ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;
+
+		spin_lock_irq(&ev_queue->lock);
+	}
+
+	if (omit_data) {
+		event_sub = list_first_entry(&ev_queue->event_list,
+					struct devx_event_subscription,
+					event_list);
+		eventsz = sizeof(event_sub->cookie);
+		event_data = &event_sub->cookie;
+	} else {
+		event = list_first_entry(&ev_queue->event_list,
+				      struct devx_async_event_data, list);
+		eventsz = sizeof(struct mlx5_eqe) +
+			sizeof(struct mlx5_ib_uapi_devx_async_event_hdr);
+		event_data = &event->hdr;
+	}
+
+	if (eventsz > count) {
+		spin_unlock_irq(&ev_queue->lock);
+		return -ENOSPC;
+	}
+
+	if (omit_data)
+		list_del_init(&event_sub->event_list);
+	else
+		list_del(&event->list);
+
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (copy_to_user(buf, event_data, eventsz))
+		ret = -EFAULT;
+	else
+		ret = eventsz;
+
+	if (!omit_data)
+		kfree(event);
+	return ret;
 }
 
 static __poll_t devx_async_event_poll(struct file *filp,
 				      struct poll_table_struct *wait)
 {
-	return 0;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	__poll_t pollflags = 0;
+
+	poll_wait(filp, &ev_queue->poll_wait, wait);
+
+	spin_lock_irq(&ev_queue->lock);
+	if (ev_queue->is_destroyed)
+		pollflags = EPOLLIN | EPOLLRDNORM | EPOLLRDHUP;
+	else if (!list_empty(&ev_queue->event_list))
+		pollflags = EPOLLIN | EPOLLRDNORM;
+	spin_unlock_irq(&ev_queue->lock);
+
+	return pollflags;
 }
 
 static int devx_async_event_close(struct inode *inode, struct file *filp)
 {
+	struct ib_uobject *uobj = filp->private_data;
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file, uobj);
+	struct devx_event_subscription *event_sub, *event_sub_tmp;
+	struct devx_async_event_data *entry, *tmp;
+
+	mutex_lock(&ev_file->dev->devx_event_table.event_xa_lock);
+	/* delete the subscriptions which are related to this FD */
+	list_for_each_entry_safe(event_sub, event_sub_tmp,
+				 &ev_file->subscribed_events_list, file_list)
+		devx_cleanup_subscription(ev_file->dev, event_sub, true);
+	mutex_unlock(&ev_file->dev->devx_event_table.event_xa_lock);
+
+	/* free the pending events allocation */
+	if (!(ev_file->ev_queue.flags &
+	    MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA)) {
+		spin_lock_irq(&ev_file->ev_queue.lock);
+		list_for_each_entry_safe(entry, tmp,
+					 &ev_file->ev_queue.event_list, list)
+			kfree(entry); /* read can't come any nore */
+		spin_unlock_irq(&ev_file->ev_queue.lock);
+	}
 	uverbs_close_fd(filp);
+	put_device(&ev_file->dev->ib_dev.dev);
 	return 0;
 }
 
@@ -2374,6 +2705,17 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
 static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
 					    enum rdma_remove_reason why)
 {
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);
 	return 0;
 };
 
include/uapi/rdma/mlx5_user_ioctl_verbs.h +5 -0
index 57beea4589e4..9500ff7363ef 100644
@@ -67,5 +67,10 @@ enum mlx5_ib_uapi_devx_create_event_channel_flags {
 	MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA = 1 << 0,
 };
 
+struct mlx5_ib_uapi_devx_async_event_hdr {
+	__aligned_u64	cookie;
+	__u8		out_data[];
+};
+
 #endif
 
Signature · 1 line
2.20.1
#27 lore raw

PATCH 12/12 IB/mlx5: Add DEVX support for CQ events

In reply to Leon Romanovsky

From: Yishai Hadas <[email protected]>

Add DEVX support for CQ events by creating and destroying the CQ via mlx5_core and set an handler to manage its completions.

---
 drivers/infiniband/hw/mlx5/devx.c | 40 +++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
drivers/infiniband/hw/mlx5/devx.c +40 -0
index 49fdce95d6d9..91ccd58ebc05 100644
@@ -19,9 +19,12 @@
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
 
+static void dispatch_event_fd(struct list_head *fd_list, const void *data);
+
 enum devx_obj_flags {
 	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,
 	DEVX_OBJ_FLAGS_DCT = 1 << 1,
+	DEVX_OBJ_FLAGS_CQ = 1 << 2,
 };
 
 struct devx_async_data {
@@ -94,6 +97,7 @@ struct devx_async_event_file {
 #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
 struct devx_obj {
 	struct mlx5_core_dev	*mdev;
+	struct mlx5_ib_dev	*ib_dev;
 	u64			obj_id;
 	u32			dinlen; /* destroy inbox length */
 	u32			dinbox[MLX5_MAX_DESTROY_INBOX_SIZE_DW];
@@ -101,6 +105,7 @@ struct devx_obj {
 	union {
 		struct mlx5_ib_devx_mr	devx_mr;
 		struct mlx5_core_dct	core_dct;
+		struct mlx5_core_cq	core_cq;
 	};
 	struct list_head event_sub; /* holds devx_event_subscription entries */
 };
@@ -1384,6 +1389,8 @@ static int devx_obj_cleanup(struct ib_uobject *uobject,
 
 	if (obj->flags & DEVX_OBJ_FLAGS_DCT)
 		ret = mlx5_core_destroy_dct(obj->mdev, &obj->core_dct);
+	else if (obj->flags & DEVX_OBJ_FLAGS_CQ)
+		ret = mlx5_core_destroy_cq(obj->mdev, &obj->core_cq);
 	else
 		ret = mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out,
 				    sizeof(out));
@@ -1408,6 +1415,30 @@ static int devx_obj_cleanup(struct ib_uobject *uobject,
 	return ret;
 }
 
+static void devx_cq_comp(struct mlx5_core_cq *mcq, struct mlx5_eqe *eqe)
+{
+	struct devx_obj *obj = container_of(mcq, struct devx_obj, core_cq);
+	struct mlx5_devx_event_table *table;
+	struct devx_event *event;
+	struct devx_obj_event *obj_event;
+	u32 obj_id = mcq->cqn;
+
+	table = &obj->ib_dev->devx_event_table;
+	event = xa_load(&table->event_xa, MLX5_EVENT_TYPE_COMP);
+	if (!event)
+		return;
+
+	rcu_read_lock();
+	obj_event = xa_load(&event->object_ids, obj_id);
+	if (!obj_event) {
+		rcu_read_unlock();
+		return;
+	}
+
+	dispatch_event_fd(&obj_event->obj_sub_list, eqe);
+	rcu_read_unlock();
+}
+
 static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
 	struct uverbs_attr_bundle *attrs)
 {
@@ -1459,6 +1490,12 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
 		err = mlx5_core_create_dct(dev->mdev, &obj->core_dct,
 					   cmd_in, cmd_in_len,
 					   cmd_out, cmd_out_len);
+	} else if (opcode == MLX5_CMD_OP_CREATE_CQ) {
+		obj->flags |= DEVX_OBJ_FLAGS_CQ;
+		obj->core_cq.comp = devx_cq_comp;
+		err = mlx5_core_create_cq(dev->mdev, &obj->core_cq,
+					  cmd_in, cmd_in_len, cmd_out,
+					  cmd_out_len);
 	} else {
 		err = mlx5_cmd_exec(dev->mdev, cmd_in,
 				    cmd_in_len,
@@ -1471,6 +1508,7 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
 	uobj->object = obj;
 	obj->mdev = dev->mdev;
 	INIT_LIST_HEAD(&obj->event_sub);
+	obj->ib_dev = dev;
 	devx_obj_build_destroy_cmd(cmd_in, cmd_out, obj->dinbox, &obj->dinlen,
 				   &obj_id);
 	WARN_ON(obj->dinlen > MLX5_MAX_DESTROY_INBOX_SIZE_DW * sizeof(u32));
@@ -1498,6 +1536,8 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)(
 obj_destroy:
 	if (obj->flags & DEVX_OBJ_FLAGS_DCT)
 		mlx5_core_destroy_dct(obj->mdev, &obj->core_dct);
+	else if (obj->flags & DEVX_OBJ_FLAGS_CQ)
+		mlx5_core_destroy_cq(obj->mdev, &obj->core_cq);
 	else
 		mlx5_cmd_exec(obj->mdev, obj->dinbox, obj->dinlen, out,
 			      sizeof(out));
Signature · 1 line
2.20.1
Acked-by on the v1 series
#31 lore raw
In reply to Leon Romanovsky
Quoted 4 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

Changelog:
v0 -> v1:

Normally 1st submission is V1 and 2nd is V2.
so this should have been v1->v2.

For mlx5-next patches:

Quoted 88 lines

* Fix the unbind / hot unplug flows to work properly.
* Fix Ref count handling on the eventfd mode in some flow.
* Rebased to latest rdma-next

Thanks

-------------------------------------------------------------------
-----------------
From Yishai:

This series enables RDMA applications that use the DEVX interface to subscribe and read device asynchronous events.

The solution is designed to allow extension of events in the future without need to perform any changes in the driver code.

To enable that few changes had been done in mlx5_core, it includes:
* Reading device event capabilities that are user related (affiliated and un-affiliated) and set the matching mask upon creating the matching EQ.
* Enable DEVX/mlx5_ib to register for ANY event instead of the
option to
get some hard-coded ones.
* Enable DEVX/mlx5_ib to get the device raw data for CQ completion
events.
* Enhance mlx5_core_create/destroy CQ to enable DEVX using them so
that CQ
events will be reported as well.

In mlx5_ib layer the below changes were done:
* A new DEVX API was introduced to allocate an event channel by
using
the uverbs FD object type.
* Implement the FD channel operations to enable read/poo/close over
it.
* A new DEVX API was introduced to subscribe for specific events
over an
event channel.
* Manage an internal data structure over XA(s) to
subscribe/dispatch events
over the different event channels.
* Use from DEVX the mlx5_core APIs to create/destroy a CQ to be able
to
get its relevant events.

Yishai

Yishai Hadas (12):
net/mlx5: Fix mlx5_core_destroy_cq() error flow net/mlx5: Use event mask based on device capabilities net/mlx5: Expose the API to register for ANY event net/mlx5: mlx5_core_create_cq() enhancements net/mlx5: Report a CQ error event only when a handler was set net/mlx5: Report EQE data upon CQ completion net/mlx5: Expose device definitions for object events IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD IB/mlx5: Register DEVX with mlx5_core to get async events IB/mlx5: Enable subscription for device events over DEVX IB/mlx5: Implement DEVX dispatching event
IB/mlx5: Add DEVX support for CQ events

 drivers/infiniband/hw/mlx5/cq.c               |    5 +-
 drivers/infiniband/hw/mlx5/devx.c             | 1082
++++++++++++++++-
 drivers/infiniband/hw/mlx5/main.c             |   10 +-
 drivers/infiniband/hw/mlx5/mlx5_ib.h          |   12 +
 drivers/infiniband/hw/mlx5/odp.c              |    3 +-
 drivers/infiniband/hw/mlx5/qp.c               |    2 +-
 drivers/net/ethernet/mellanox/mlx5/core/cq.c  |   21 +-
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |    2 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |    3 +-
 .../net/ethernet/mellanox/mlx5/core/en_txrx.c |    2 +-
 drivers/net/ethernet/mellanox/mlx5/core/eq.c  |   68 +-
 .../ethernet/mellanox/mlx5/core/fpga/conn.c   |    6 +-
 drivers/net/ethernet/mellanox/mlx5/core/fw.c  |    6 +
 .../net/ethernet/mellanox/mlx5/core/lib/eq.h  |    5 +-
 include/linux/mlx5/cq.h                       |    6 +-
 include/linux/mlx5/device.h                   |    6 +-
 include/linux/mlx5/driver.h                   |    2 +
 include/linux/mlx5/eq.h                       |    4 +-
 include/linux/mlx5/mlx5_ifc.h                 |   34 +-
 include/uapi/rdma/mlx5_user_ioctl_cmds.h      |   19 +
 include/uapi/rdma/mlx5_user_ioctl_verbs.h     |    9 +
 21 files changed, 1237 insertions(+), 70 deletions(-)

--
2.20.1

#32 lore raw
Quoted 8 lines · Saeed Mahameed
Quoted 4 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

Changelog:
v0 -> v1:

Normally 1st submission is V1 and 2nd is V2.
so this should have been v1->v2.

"Normally" depends on the language you are using. In C, everything starts from 0, including version of patches :).

Quoted 4 lines

For mlx5-next patches:

Thanks

#10 lore raw

Re: [PATCH rdma-next v1 08/12] IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD

In reply to Leon Romanovsky
Quoted 19 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD and its initial implementation.

This object is from type class FD and will be used to read DEVX async events.

 drivers/infiniband/hw/mlx5/devx.c         | 112 ++++++++++++++++++++--
 include/uapi/rdma/mlx5_user_ioctl_cmds.h  |  10 ++
 include/uapi/rdma/mlx5_user_ioctl_verbs.h |   4 +
 3 files changed, 116 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +1 -0
index 80b42d069328..1815ce0f8daf 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 36

Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +1 -0
@@ -33,6 +33,24 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+struct devx_async_event_queue {

It seems to be a mistake to try and re-use the async_event_queue for both cmd and event, as they use it very differently and don't even store the same things in the event_list. I think it is bettter to just inline this into devx_async_event_file (and inline the old struct in the cmd file

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 43

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
+	spinlock_t		lock;
+	wait_queue_head_t	poll_wait;
+	struct list_head	event_list;
+	atomic_t		bytes_in_use;
+	u8			is_destroyed:1;
+	u32			flags;
+};

All the flags testing is ugly, why not just add another bitfield?

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 53

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+struct devx_async_event_file {
+	struct ib_uobject		uobj;
+	struct list_head subscribed_events_list; /* Head of events that are
+						  * subscribed to this FD
+						  */
+	struct devx_async_event_queue	ev_queue;
+	struct mlx5_ib_dev *dev;
+};
+

Crazy indenting

include/uapi/rdma/mlx5_user_ioctl_verbs.h enum mlx5_ib_uapi_dm_type { · line 68

Quoted 10 lines
include/uapi/rdma/mlx5_user_ioctl_verbs.h +3 -0
index a8f34c237458..57beea4589e4 100644
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -63,5 +63,9 @@ enum mlx5_ib_uapi_dm_type {
 	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM,
 };
 
+enum mlx5_ib_uapi_devx_create_event_channel_flags {
+	MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA = 1
 << 0,

Maybe this name is too long

Jason

#14 lore raw

Re: [PATCH rdma-next v1 09/12] IB/mlx5: Register DEVX with mlx5_core to get async events

In reply to Leon Romanovsky
Quoted 19 lines · Leon Romanovsky

void __mlx5_ib_remove(struct mlx5_ib_dev *dev,

drivers/infiniband/hw/mlx5/mlx5_ib.h +8 -0
index 9cf23ae6324e..556af34b788b 100644
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -944,6 +944,13 @@ struct mlx5_ib_pf_eq {
 	mempool_t *pool;
 };
 
+struct mlx5_devx_event_table {
+	struct mlx5_nb devx_nb;
+	/* serialize updating the event_xa */
+	struct mutex event_xa_lock;
+	struct xarray event_xa;
+};
+
 struct mlx5_ib_dev {
 	struct ib_device		ib_dev;
 	struct mlx5_core_dev		*mdev;

drivers/infiniband/hw/mlx5/mlx5_ib.h struct mlx5_ib_dev { · line 1004

Quoted 4 lines
drivers/infiniband/hw/mlx5/mlx5_ib.h +1 -0
@@ -994,6 +1001,7 @@ struct mlx5_ib_dev {
 	struct mlx5_srq_table   srq_table;
 	struct mlx5_async_ctx   async_ctx;
 	int			free_port;
+	struct mlx5_devx_event_table devx_event_table;

I really question if adding all these structs really does anything for readability..

Jason

#18 lore raw

Re: [PATCH rdma-next v1 10/12] IB/mlx5: Enable subscription for device events over DEVX

In reply to Leon Romanovsky
Quoted 50 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Enable subscription for device events over DEVX.

Each subscription is added to the two level XA data structure according to its event number and the DEVX object information in case was given with the given target fd.

Those events will be reported over the given fd once will occur. Downstream patches will mange the dispatching to any subscription.

 drivers/infiniband/hw/mlx5/devx.c        | 564 ++++++++++++++++++++++-
 include/uapi/rdma/mlx5_user_ioctl_cmds.h |   9 +
 2 files changed, 566 insertions(+), 7 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +20 -0
index e9b9ba5a3e9a..304b13e7a265 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -14,6 +14,7 @@
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/fs.h>
 #include "mlx5_ib.h"
+#include <linux/xarray.h>
 
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
@@ -33,6 +34,37 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+/* first level XA value data structure */
+struct devx_event {
+	struct xarray object_ids; /* second XA level, Key = object id */
+	struct list_head unaffiliated_list;
+};
+
+/* second level XA value data structure */
+struct devx_obj_event {
+	struct rcu_head rcu;
+	struct list_head obj_sub_list;
+};
+
+struct devx_event_subscription {
+	struct list_head file_list; /* headed in private_data->
+				     * subscribed_events_list
+				     */
+	struct list_head xa_list; /* headed in devx_event->unaffiliated_list or
+				   * devx_obj_event->obj_sub_list

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 64

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+				   */
+	struct list_head obj_list; /* headed in devx_object */
+
+	u32 xa_key_level1;
+	u32 xa_key_level2;
+	struct rcu_head	rcu;
+	u64 cookie;
+	bool is_obj_related;
+	struct ib_uobject *fd_uobj;
+	void *object;	/* May need direct access upon hot unplug */

This should be a 'struct file *' and have a better name.

And I'm unclear why we need to store both the ib_uobject and the struct file for the same thing? And why are we storing the uobj here instead of the struct devx_async_event_file *?

Since uobj->object == flip && filp->private_data == uobj, I have a hard time to understand why we need both things, it seems to me that if we get the fget on the filp then we can rely on the filp->private_data to get back to the devx_async_event_file.

Quoted 8 lines
drivers/infiniband/hw/mlx5/devx.c +3 -0
+	struct eventfd_ctx *eventfd;
+};
+
 
 /*
  * As the obj_id in the firmware is not globally unique the object type

* must be considered upon checking for a valid object id.

drivers/infiniband/hw/mlx5/devx.c static void devx_cleanup_mkey(struct devx_obj *obj) · line 1284

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
@@ -1143,14 +1275,53 @@ static void devx_cleanup_mkey(struct devx_obj *obj)
 	write_unlock_irqrestore(&table->lock, flags);
 }
 
+static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
+				      struct devx_event_subscription *sub)
+{
+	list_del_rcu(&sub->file_list);
+	list_del_rcu(&sub->xa_list);
+
+	if (sub->is_obj_related) {

is_obj_related looks like it is just list_empty(obj_list)??

Success oriented flow

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
@@ -1523,6 +1700,350 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)(
 	return err;
 }
 
+static void
+subscribe_event_xa_dealloc(struct mlx5_devx_event_table *devx_event_table,
+			   u32 key_level1,
+			   u32 key_level2,
+			   struct devx_obj_event *alloc_obj_event)
+{

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1718

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+	struct devx_event *event;
+
+	/* Level 1 is valid for future use - no need to free */
+	if (!alloc_obj_event)
+		return;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	WARN_ON(!event);
+
+	xa_erase(&event->object_ids, key_level2);

Shoulnd't this only erase if the value stored is NULL?

Quoted 40 lines
drivers/infiniband/hw/mlx5/devx.c +40 -0
+	kfree(alloc_obj_event);
+}
+
+static int
+subscribe_event_xa_alloc(struct mlx5_devx_event_table *devx_event_table,
+			 u32 key_level1,
+			 bool is_level2,
+			 u32 key_level2,
+			 struct devx_obj_event **alloc_obj_event)
+{
+	struct devx_obj_event *obj_event;
+	struct devx_event *event;
+	bool new_entry_level1 = false;
+	int err;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	if (!event) {
+		event = kzalloc(sizeof(*event), GFP_KERNEL);
+		if (!event)
+			return -ENOMEM;
+
+		new_entry_level1 = true;
+		INIT_LIST_HEAD(&event->unaffiliated_list);
+		xa_init(&event->object_ids);
+
+		err = xa_insert(&devx_event_table->event_xa,
+				key_level1,
+				event,
+				GFP_KERNEL);
+		if (err)
+			goto end;
+	}
+
+	if (!is_level2)
+		return 0;
+
+	obj_event = xa_load(&event->object_ids, key_level2);
+	if (!obj_event) {
+		err = xa_reserve(&event->object_ids, key_level2, GFP_KERNEL);
+		if (err)

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1768

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+			goto err_level1;
+
+		obj_event = kzalloc(sizeof(*obj_event), GFP_KERNEL);
+		if (!obj_event) {
+			err = -ENOMEM;
+			goto err_level2;
+		}
+
+		INIT_LIST_HEAD(&obj_event->obj_sub_list);
+		*alloc_obj_event = obj_event;

This is goofy, just store the empty obj_event in the xa instead of using xa_reserve, and when you go to do the error unwind just delete any level2' devx_obj_event' that has a list_empty(obj_sub_list), get rid of the wonky alloc_obj_event stuff.

The best configuration would be to use devx_cleanup_subscription to undo the partially ready subscription.

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1781

Quoted 13 lines
drivers/infiniband/hw/mlx5/devx.c +13 -0
+	}
+
+	return 0;
+
+err_level2:
+	xa_erase(&event->object_ids, key_level2);
+
+err_level1:
+	if (new_entry_level1)
+		xa_erase(&devx_event_table->event_xa, key_level1);
+end:
+	if (new_entry_level1)
+		kfree(event);

Can't do this, once the level1 is put in the tree it could be referenced by the irqs. At least it needs a kfree_rcu, most likely it is simpler to just leave it.

Quoted 75 lines
drivers/infiniband/hw/mlx5/devx.c +75 -0
+#define MAX_NUM_EVENTS 16
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *devx_uobj = uverbs_attr_get_uobject(
+				attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_OBJ_HANDLE);
+	struct mlx5_ib_ucontext *c = rdma_udata_to_drv_context(
+		&attrs->driver_udata, struct mlx5_ib_ucontext, ibucontext);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	struct ib_uobject *fd_uobj;
+	struct devx_obj *obj = NULL;
+	struct devx_async_event_file *ev_file;
+	struct mlx5_devx_event_table *devx_event_table = &dev->devx_event_table;
+	u16 *event_type_num_list;
+	struct devx_event_subscription **event_sub_arr;
+	struct devx_obj_event  **event_obj_array_alloc;
+	int redirect_fd;
+	bool use_eventfd = false;
+	int num_events;
+	int num_alloc_xa_entries = 0;
+	u16 obj_type = 0;
+	u64 cookie = 0;
+	u32 obj_id = 0;
+	int err;
+	int i;
+
+	if (!c->devx_uid)
+		return -EINVAL;
+
+	if (!IS_ERR(devx_uobj)) {
+		obj = (struct devx_obj *)devx_uobj->object;
+		if (obj)
+			obj_id = get_dec_obj_id(obj->obj_id);
+	}
+
+	fd_uobj = uverbs_attr_get_uobject(attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_HANDLE);
+	if (IS_ERR(fd_uobj))
+		return PTR_ERR(fd_uobj);
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM)) {
+		err = uverbs_copy_from(&redirect_fd, attrs,
+			       MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM);
+		if (err)
+			return err;
+
+		use_eventfd = true;
+	}
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE)) {
+		if (use_eventfd)
+			return -EINVAL;
+
+		err = uverbs_copy_from(&cookie, attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE);
+		if (err)
+			return err;
+	}
+
+	num_events = uverbs_attr_ptr_get_array_size(
+		attrs, MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST,
+		sizeof(u16));
+
+	if (num_events < 0)
+		return num_events;
+
+	if (num_events > MAX_NUM_EVENTS)
+		return -EINVAL;
+

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1935

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+	event_type_num_list = uverbs_attr_get_alloced_ptr(attrs,
+			MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST);
+
+	if (!is_valid_events(dev->mdev, num_events, event_type_num_list, obj))
+		return -EINVAL;
+
+	event_sub_arr = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_event_subscription *));
+	event_obj_array_alloc = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_obj_event *));

There are so many list_heads in the devx_event_subscription, why not use just one of them to store the allocated events instead of this temp array? ie event_list looks good for this purpose.

Quoted 11 lines
drivers/infiniband/hw/mlx5/devx.c +11 -0
+
+	if (!event_sub_arr || !event_obj_array_alloc)
+		return -ENOMEM;
+
+	/* Protect from concurrent subscriptions to same XA entries to allow
+	 * both to succeed
+	 */
+	mutex_lock(&devx_event_table->event_xa_lock);
+	for (i = 0; i < num_events; i++) {
+		u32 key_level1;
+

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1956

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+		if (obj)
+			obj_type = get_dec_obj_type(obj,
+						    event_type_num_list[i]);
+		key_level1 = event_type_num_list[i] | obj_type << 16;
+
+		err = subscribe_event_xa_alloc(devx_event_table,
+					       key_level1,
+					       obj ? true : false,
+					       obj_id,
+					       &event_obj_array_alloc[i]);

Usless ?:

Quoted 17 lines
drivers/infiniband/hw/mlx5/devx.c +17 -0
+		if (err)
+			goto err;
+
+		num_alloc_xa_entries++;
+		event_sub_arr[i] = kzalloc(sizeof(*event_sub_arr[i]),
+					   GFP_KERNEL);
+		if (!event_sub_arr[i])
+			goto err;
+
+		if (use_eventfd) {
+			event_sub_arr[i]->eventfd =
+				eventfd_ctx_fdget(redirect_fd);
+
+			if (IS_ERR(event_sub_arr[i]->eventfd)) {
+				err = PTR_ERR(event_sub_arr[i]->eventfd);
+				event_sub_arr[i]->eventfd = NULL;
+				goto err;

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1983

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+			}
+		}
+
+		event_sub_arr[i]->cookie = cookie;
+		event_sub_arr[i]->fd_uobj = fd_uobj;
+		event_sub_arr[i]->object = fd_uobj->object;
+		/* May be needed upon cleanup the devx object/subscription */
+		event_sub_arr[i]->xa_key_level1 = key_level1;
+		event_sub_arr[i]->xa_key_level2 = obj_id;
+		event_sub_arr[i]->is_obj_related = obj ? true : false;

Unneeded ?:

Jason

#22 lore raw

Re: [PATCH rdma-next v1 11/12] IB/mlx5: Implement DEVX dispatching event

In reply to Leon Romanovsky
Quoted 28 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Implement DEVX dispatching event by looking up for the applicable subscriptions for the reported event and using their target fd to signal/set the event.

 drivers/infiniband/hw/mlx5/devx.c         | 362 +++++++++++++++++++++-
 include/uapi/rdma/mlx5_user_ioctl_verbs.h |   5 +
 2 files changed, 357 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +6 -0
index 304b13e7a265..49fdce95d6d9 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -34,6 +34,11 @@ struct devx_async_data {
 	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
 };
 
+struct devx_async_event_data {
+	struct list_head list; /* headed in ev_queue->event_list */
+	struct mlx5_ib_uapi_devx_async_event_hdr hdr;
+};
+
 /* first level XA value data structure */
 struct devx_event {
 	struct xarray object_ids; /* second XA level, Key = object id */

drivers/infiniband/hw/mlx5/devx.c struct devx_event_subscription { · line 64

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +2 -0
@@ -54,7 +59,9 @@ struct devx_event_subscription {
 				   * devx_obj_event->obj_sub_list
 				   */
 	struct list_head obj_list; /* headed in devx_object */
+	struct list_head event_list; /* headed in ev_queue->event_list */
 
+	u8  is_cleaned:1;

There is a loose bool 'is_obj_related' that should be combined with this bool bitfield as well.

Quoted 19 lines
drivers/infiniband/hw/mlx5/devx.c +6 -4
 static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
-				      struct devx_event_subscription *sub)
+				      struct devx_event_subscription *sub,
+				      bool file_close)
 {
-	list_del_rcu(&sub->file_list);
+	if (sub->is_cleaned)
+		goto end;
+
+	sub->is_cleaned = 1;
 	list_del_rcu(&sub->xa_list);
 
 	if (sub->is_obj_related) {
@@ -1303,10 +1355,15 @@ static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
 		}
 	}
 
-	if (sub->eventfd)
-		eventfd_ctx_put(sub->eventfd);

drivers/infiniband/hw/mlx5/devx.c static void devx_cleanup_subscription(struct mlx5_ib_dev *dev, · line 1366

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +8 -1
+end:
+	if (file_close) {
+		if (sub->eventfd)
+			eventfd_ctx_put(sub->eventfd);
 
-	kfree_rcu(sub, rcu);
+		list_del_rcu(&sub->file_list);
+		/* subscription may not be used by the read API any more */
+		kfree_rcu(sub, rcu);
+	}

Dis like this confusing file_close stuff, just put this in the single place that calls this with the true bool

Quoted 22 lines
drivers/infiniband/hw/mlx5/devx.c +22 -0
+static int deliver_event(struct devx_event_subscription *event_sub,
+			 const void *data)
+{
+	struct ib_uobject *fd_uobj = event_sub->fd_uobj;
+	struct devx_async_event_file *ev_file;
+	struct devx_async_event_queue *ev_queue;
+	struct devx_async_event_data *event_data;
+	unsigned long flags;
+	bool omit_data;
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+	ev_queue = &ev_file->ev_queue;
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	if (omit_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		if (!list_empty(&event_sub->event_list)) {
+			spin_unlock_irqrestore(&ev_queue->lock, flags);
+			return 0;
+		}

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2339

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+		list_add_tail(&event_sub->event_list, &ev_queue->event_list);
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		wake_up_interruptible(&ev_queue->poll_wait);
+		return 0;
+	}
+
+	event_data = kzalloc(sizeof(*event_data) +
+			     (omit_data ? 0 : sizeof(struct mlx5_eqe)),
+			     GFP_ATOMIC);

omit_data is always false here

Quoted 54 lines
drivers/infiniband/hw/mlx5/devx.c +50 -1
+	if (!event_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		ev_queue->is_overflow_err = 1;
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		return -ENOMEM;
+	}
+
+	event_data->hdr.cookie = event_sub->cookie;
+	memcpy(event_data->hdr.out_data, data, sizeof(struct mlx5_eqe));
+
+	spin_lock_irqsave(&ev_queue->lock, flags);
+	list_add_tail(&event_data->list, &ev_queue->event_list);
+	spin_unlock_irqrestore(&ev_queue->lock, flags);
+	wake_up_interruptible(&ev_queue->poll_wait);
+
+	return 0;
+}
+
+static void dispatch_event_fd(struct list_head *fd_list,
+			      const void *data)
+{
+	struct devx_event_subscription *item;
+
+	list_for_each_entry_rcu(item, fd_list, xa_list) {
+		if (!get_file_rcu((struct file *)item->object))
+			continue;
+
+		if (item->eventfd) {
+			eventfd_signal(item->eventfd, 1);
+			fput(item->object);
+			continue;
+		}
+
+		deliver_event(item, data);
+		fput(item->object);
+	}
+}
+
 static int devx_event_notifier(struct notifier_block *nb,
 			       unsigned long event_type, void *data)
 {
-	return NOTIFY_DONE;
+	struct mlx5_devx_event_table *table;
+	struct mlx5_ib_dev *dev;
+	struct devx_event *event;
+	struct devx_obj_event *obj_event;
+	u16 obj_type = 0;
+	bool is_unaffiliated;
+	u32 obj_id;
+
+	/* Explicit filtering to kernel events which may occur frequently */
+	if (event_type == MLX5_EVENT_TYPE_CMD ||
+	    event_type == MLX5_EVENT_TYPE_PAGE_REQUEST)
+		return NOTIFY_OK;

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2402

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+	table = container_of(nb, struct mlx5_devx_event_table, devx_nb.nb);
+	dev = container_of(table, struct mlx5_ib_dev, devx_event_table);
+	is_unaffiliated = is_unaffiliated_event(dev->mdev, event_type);
+
+	if (!is_unaffiliated)
+		obj_type = get_event_obj_type(event_type, data);
+	event = xa_load(&table->event_xa, event_type | (obj_type << 16));
+	if (!event)
+		return NOTIFY_DONE;

event should be in the rcu as well

Quoted 59 lines
drivers/infiniband/hw/mlx5/devx.c +43 -2
+	if (is_unaffiliated) {
+		dispatch_event_fd(&event->unaffiliated_list, data);
+		return NOTIFY_OK;
+	}
+
+	obj_id = devx_get_obj_id_from_event(event_type, data);
+	rcu_read_lock();
+	obj_event = xa_load(&event->object_ids, obj_id);
+	if (!obj_event) {
+		rcu_read_unlock();
+		return NOTIFY_DONE;
+	}
+
+	dispatch_event_fd(&obj_event->obj_sub_list, data);
+
+	rcu_read_unlock();
+	return NOTIFY_OK;
 }
 
 void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev)
@@ -2221,7 +2444,7 @@ void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev)
 		event = entry;
 		list_for_each_entry_safe(sub, tmp, &event->unaffiliated_list,
 					 xa_list)
-			devx_cleanup_subscription(dev, sub);
+			devx_cleanup_subscription(dev, sub, false);
 		kfree(entry);
 	}
 	mutex_unlock(&dev->devx_event_table.event_xa_lock);
@@ -2329,18 +2552,126 @@ static const struct file_operations devx_async_cmd_event_fops = {
 static ssize_t devx_async_event_read(struct file *filp, char __user *buf,
 				     size_t count, loff_t *pos)
 {
-	return -EINVAL;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	struct devx_event_subscription *event_sub;
+	struct devx_async_event_data *uninitialized_var(event);
+	int ret = 0;
+	size_t eventsz;
+	bool omit_data;
+	void *event_data;
+
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	spin_lock_irq(&ev_queue->lock);
+
+	if (ev_queue->is_overflow_err) {
+		ev_queue->is_overflow_err = 0;
+		spin_unlock_irq(&ev_queue->lock);
+		return -EOVERFLOW;
+	}
+
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2589

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+		if (wait_event_interruptible(ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;

All these tests should be under the lock.

Why don't we return EIO as soon as is-destroyed happens? What is the point of flushing out the accumulated events?

Quoted 11 lines
drivers/infiniband/hw/mlx5/devx.c +11 -0
+
+		spin_lock_irq(&ev_queue->lock);
+	}
+
+	if (omit_data) {
+		event_sub = list_first_entry(&ev_queue->event_list,
+					struct devx_event_subscription,
+					event_list);
+		eventsz = sizeof(event_sub->cookie);
+		event_data = &event_sub->cookie;
+	} else {

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2610

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+		event = list_first_entry(&ev_queue->event_list,
+				      struct devx_async_event_data, list);
+		eventsz = sizeof(struct mlx5_eqe) +
+			sizeof(struct mlx5_ib_uapi_devx_async_event_hdr);
+		event_data = &event->hdr;
+	}
+
+	if (eventsz > count) {
+		spin_unlock_irq(&ev_queue->lock);
+		return -ENOSPC;

This is probably the wrong errno

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2623

Quoted 13 lines
drivers/infiniband/hw/mlx5/devx.c +13 -0
+	}
+
+	if (omit_data)
+		list_del_init(&event_sub->event_list);
+	else
+		list_del(&event->list);
+
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (copy_to_user(buf, event_data, eventsz))
+		ret = -EFAULT;
+	else
+		ret = eventsz;

This is really poorly ordered, EFAULT will cause the event to be lost. :(

Maybe the event should be re-added on error? Tricky.

Quoted 37 lines
drivers/infiniband/hw/mlx5/devx.c +27 -1
+	if (!omit_data)
+		kfree(event);
+	return ret;
 }
 
 static __poll_t devx_async_event_poll(struct file *filp,
 				      struct poll_table_struct *wait)
 {
-	return 0;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	__poll_t pollflags = 0;
+
+	poll_wait(filp, &ev_queue->poll_wait, wait);
+
+	spin_lock_irq(&ev_queue->lock);
+	if (ev_queue->is_destroyed)
+		pollflags = EPOLLIN | EPOLLRDNORM | EPOLLRDHUP;
+	else if (!list_empty(&ev_queue->event_list))
+		pollflags = EPOLLIN | EPOLLRDNORM;
+	spin_unlock_irq(&ev_queue->lock);
+
+	return pollflags;
 }
 
 static int devx_async_event_close(struct inode *inode, struct file *filp)
 {
+	struct ib_uobject *uobj = filp->private_data;
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file, uobj);
+	struct devx_event_subscription *event_sub, *event_sub_tmp;
+	struct devx_async_event_data *entry, *tmp;
+
+	mutex_lock(&ev_file->dev->devx_event_table.event_xa_lock);
+	/* delete the subscriptions which are related to this FD */
+	list_for_each_entry_safe(event_sub, event_sub_tmp,
+				 &ev_file->subscribed_events_list, file_list)

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2670

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+		devx_cleanup_subscription(ev_file->dev, event_sub, true);
+	mutex_unlock(&ev_file->dev->devx_event_table.event_xa_lock);
+
+	/* free the pending events allocation */
+	if (!(ev_file->ev_queue.flags &
+	    MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA)) {
+		spin_lock_irq(&ev_file->ev_queue.lock);
+		list_for_each_entry_safe(entry, tmp,
+					 &ev_file->ev_queue.event_list, list)
+			kfree(entry); /* read can't come any nore */

spelling

Quoted 12 lines
drivers/infiniband/hw/mlx5/devx.c +4 -0
+		spin_unlock_irq(&ev_file->ev_queue.lock);
+	}
 	uverbs_close_fd(filp);
+	put_device(&ev_file->dev->ib_dev.dev);
 	return 0;
 }
 
@@ -2374,6 +2705,17 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
 static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
 					    enum rdma_remove_reason why)
 {
+	struct devx_async_event_file *ev_file =

drivers/infiniband/hw/mlx5/devx.c static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj, · line 2718

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+		container_of(uobj, struct devx_async_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);

Why isn't this wakeup always done?

Jason

#28 lore raw

Re: [PATCH rdma-next v1 12/12] IB/mlx5: Add DEVX support for CQ events

In reply to Leon Romanovsky
Quoted 27 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Add DEVX support for CQ events by creating and destroying the CQ via mlx5_core and set an handler to manage its completions.

 drivers/infiniband/hw/mlx5/devx.c | 40 +++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
drivers/infiniband/hw/mlx5/devx.c +4 -0
index 49fdce95d6d9..91ccd58ebc05 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -19,9 +19,12 @@
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
 
+static void dispatch_event_fd(struct list_head *fd_list, const void *data);
+
 enum devx_obj_flags {
 	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,
 	DEVX_OBJ_FLAGS_DCT = 1 << 1,
+	DEVX_OBJ_FLAGS_CQ = 1 << 2,
 };
 
 struct devx_async_data {

drivers/infiniband/hw/mlx5/devx.c struct devx_async_event_file { · line 100

Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +1 -0
@@ -94,6 +97,7 @@ struct devx_async_event_file {
 #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
 struct devx_obj {
 	struct mlx5_core_dev	*mdev;
+	struct mlx5_ib_dev	*ib_dev;

This seems strange, why would we need to store the core_dev and the ib_dev in a struct when ibdev->mdev == core_dev?

Jason

#11 lore raw

Re: [PATCH rdma-next v1 08/12] IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD

In reply to Jason Gunthorpe
Quoted 23 lines · Jason Gunthorpe
Quoted 22 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD and its initial implementation.

This object is from type class FD and will be used to read DEVX async events.

  drivers/infiniband/hw/mlx5/devx.c         | 112 ++++++++++++++++++++--
  include/uapi/rdma/mlx5_user_ioctl_cmds.h  |  10 ++
  include/uapi/rdma/mlx5_user_ioctl_verbs.h |   4 +
  3 files changed, 116 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +1 -0
index 80b42d069328..1815ce0f8daf 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -33,6 +33,24 @@ struct devx_async_data {
  	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
  };
 
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 36

Quoted 1 line
drivers/infiniband/hw/mlx5/devx.c +1 -0
+struct devx_async_event_queue {

It seems to be a mistake to try and re-use the async_event_queue for both cmd and event, as they use it very differently and don't even store the same things in the event_list. I think it is bettter to just inline this into devx_async_event_file (and inline the old struct in the cmd file

How about having another struct with all the event's queue fields together ? this has the benefit of having all those related fields in one place and leave the cmd as is.

Alternatively,
We can inline the event stuff under devx_async_event_file and leave the cmd for now under a struct as it's not directly related to this series.

What do you think ?

Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 43

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
+	spinlock_t		lock;
+	wait_queue_head_t	poll_wait;
+	struct list_head	event_list;
+	atomic_t		bytes_in_use;
+	u8			is_destroyed:1;
+	u32			flags;
+};

All the flags testing is ugly, why not just add another bitfield?

The flags are coming from user space and have their different name space, I prefer to not mix with kernel ones. (i.e. is_destroyed). Makes sense ?

Quoted 14 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 53

Quoted 11 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+struct devx_async_event_file {
+	struct ib_uobject		uobj;
+	struct list_head subscribed_events_list; /* Head of events that are
+						  * subscribed to this FD
+						  */
+	struct devx_async_event_queue	ev_queue;
+	struct mlx5_ib_dev *dev;
+};
+

Crazy indenting

OK, will handle.

Quoted 7 lines
Quoted 7 lines
include/uapi/rdma/mlx5_user_ioctl_verbs.h +1 -0
index a8f34c237458..57beea4589e4 100644
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -63,5 +63,9 @@ enum mlx5_ib_uapi_dm_type {
  	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM,
  };
 
Quoted 5 lines

include/uapi/rdma/mlx5_user_ioctl_verbs.h enum mlx5_ib_uapi_dm_type { · line 68

Quoted 3 lines
include/uapi/rdma/mlx5_user_ioctl_verbs.h +2 -0
+enum mlx5_ib_uapi_devx_create_event_channel_flags {
+	MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA = 1
 << 0,

Maybe this name is too long

Quite long but follows the name scheme having the UAPI prefix. Any shorter suggestion ?

#15 lore raw

Re: [PATCH rdma-next v1 09/12] IB/mlx5: Register DEVX with mlx5_core to get async events

In reply to Jason Gunthorpe
Quoted 19 lines · Jason Gunthorpe
Quoted 18 lines · Leon Romanovsky

void __mlx5_ib_remove(struct mlx5_ib_dev *dev,

drivers/infiniband/hw/mlx5/mlx5_ib.h +8 -0
index 9cf23ae6324e..556af34b788b 100644
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -944,6 +944,13 @@ struct mlx5_ib_pf_eq {
  	mempool_t *pool;
  };
 
+struct mlx5_devx_event_table {
+	struct mlx5_nb devx_nb;
+	/* serialize updating the event_xa */
+	struct mutex event_xa_lock;
+	struct xarray event_xa;
+};
+
  struct mlx5_ib_dev {
  	struct ib_device		ib_dev;
  	struct mlx5_core_dev		*mdev;
Quoted 9 lines

drivers/infiniband/hw/mlx5/mlx5_ib.h struct mlx5_ib_dev { · line 1004

Quoted 5 lines
drivers/infiniband/hw/mlx5/mlx5_ib.h +1 -0
@@ -994,6 +1001,7 @@ struct mlx5_ib_dev {
  	struct mlx5_srq_table   srq_table;
  	struct mlx5_async_ctx   async_ctx;
  	int			free_port;
+	struct mlx5_devx_event_table devx_event_table;

I really question if adding all these structs really does anything for readability..

I would prefer this option to add only one structure (i.e. mlx5_devx_event_table) on ib_dev, it will hold internally the other related stuff.

#12 lore raw

Re: [PATCH rdma-next v1 08/12] IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD

In reply to Yishai Hadas
Quoted 32 lines · Yishai Hadas
Quoted 16 lines · Jason Gunthorpe
Quoted 15 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD and its initial implementation.

This object is from type class FD and will be used to read DEVX async events.

  drivers/infiniband/hw/mlx5/devx.c         | 112 ++++++++++++++++++++--
  include/uapi/rdma/mlx5_user_ioctl_cmds.h  |  10 ++
  include/uapi/rdma/mlx5_user_ioctl_verbs.h |   4 +
  3 files changed, 116 insertions(+), 10 deletions(-)
Quoted 14 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 35

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +2 -0
index 80b42d069328..1815ce0f8daf 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -33,6 +33,24 @@ struct devx_async_data {
  	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
  };
+struct devx_async_event_queue {

It seems to be a mistake to try and re-use the async_event_queue for both cmd and event, as they use it very differently and don't even store the same things in the event_list. I think it is bettter to just inline this into devx_async_event_file (and inline the old struct in the cmd file

Quoted 7 lines

How about having another struct with all the event's queue fields together ? this has the benefit of having all those related fields in one place and leave the cmd as is.

Alternatively,
We can inline the event stuff under devx_async_event_file and leave the cmd for now under a struct as it's not directly related to this series.

I would probbaly do this

Quoted 10 lines
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 43

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
+	spinlock_t		lock;
+	wait_queue_head_t	poll_wait;
+	struct list_head	event_list;
+	atomic_t		bytes_in_use;
+	u8			is_destroyed:1;
+	u32			flags;
+};

All the flags testing is ugly, why not just add another bitfield?

Quoted 3 lines

The flags are coming from user space and have their different name space, I prefer to not mix with kernel ones. (i.e. is_destroyed).
Makes sense ?

No, better to add a bitfield than store the raw flags and another bitfield.

Quoted 12 lines
Quoted 11 lines

include/uapi/rdma/mlx5_user_ioctl_verbs.h enum mlx5_ib_uapi_dm_type { · line 67

Quoted 9 lines
include/uapi/rdma/mlx5_user_ioctl_verbs.h +3 -0
index a8f34c237458..57beea4589e4 100644
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -63,5 +63,9 @@ enum mlx5_ib_uapi_dm_type {
  	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM,
  };
+enum mlx5_ib_uapi_devx_create_event_channel_flags {
+	MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA = 1
 << 0,

Maybe this name is too long

Quoted 3 lines

Quite long but follows the name scheme having the UAPI prefix. Any shorter suggestion ?

I think you should shorten it

Jason

#16 lore raw

Re: [PATCH rdma-next v1 09/12] IB/mlx5: Register DEVX with mlx5_core to get async events

In reply to Yishai Hadas
Quoted 29 lines · Yishai Hadas
Quoted 18 lines · Jason Gunthorpe
Quoted 17 lines · Leon Romanovsky

void __mlx5_ib_remove(struct mlx5_ib_dev *dev,

drivers/infiniband/hw/mlx5/mlx5_ib.h +8 -0
index 9cf23ae6324e..556af34b788b 100644
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -944,6 +944,13 @@ struct mlx5_ib_pf_eq {
  	mempool_t *pool;
  };
+struct mlx5_devx_event_table {
+	struct mlx5_nb devx_nb;
+	/* serialize updating the event_xa */
+	struct mutex event_xa_lock;
+	struct xarray event_xa;
+};
+
  struct mlx5_ib_dev {
  	struct ib_device		ib_dev;
  	struct mlx5_core_dev		*mdev;
Quoted 9 lines

drivers/infiniband/hw/mlx5/mlx5_ib.h struct mlx5_ib_dev { · line 1004

Quoted 5 lines
drivers/infiniband/hw/mlx5/mlx5_ib.h +1 -0
@@ -994,6 +1001,7 @@ struct mlx5_ib_dev {
  	struct mlx5_srq_table   srq_table;
  	struct mlx5_async_ctx   async_ctx;
  	int			free_port;
+	struct mlx5_devx_event_table devx_event_table;

I really question if adding all these structs really does anything for readability..

Quoted 3 lines

I would prefer this option to add only one structure (i.e. mlx5_devx_event_table) on ib_dev, it will hold internally the other related stuff.

It seems confounding but generally is the style in this struct :\

Jason

#19 lore raw

Re: [PATCH rdma-next v1 10/12] IB/mlx5: Enable subscription for device events over DEVX

In reply to Jason Gunthorpe
Quoted 55 lines · Jason Gunthorpe
Quoted 54 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Enable subscription for device events over DEVX.

Each subscription is added to the two level XA data structure according to its event number and the DEVX object information in case was given with the given target fd.

Those events will be reported over the given fd once will occur. Downstream patches will mange the dispatching to any subscription.

  drivers/infiniband/hw/mlx5/devx.c        | 564 ++++++++++++++++++++++-
  include/uapi/rdma/mlx5_user_ioctl_cmds.h |   9 +
  2 files changed, 566 insertions(+), 7 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +24 -0
index e9b9ba5a3e9a..304b13e7a265 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -14,6 +14,7 @@
  #include <linux/mlx5/driver.h>
  #include <linux/mlx5/fs.h>
  #include "mlx5_ib.h"
+#include <linux/xarray.h>
 
  #define UVERBS_MODULE_NAME mlx5_ib
  #include <rdma/uverbs_named_ioctl.h>
@@ -33,6 +34,37 @@ struct devx_async_data {
  	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
  };
 
+/* first level XA value data structure */
+struct devx_event {
+	struct xarray object_ids; /* second XA level, Key = object id */
+	struct list_head unaffiliated_list;
+};
+
+/* second level XA value data structure */
+struct devx_obj_event {
+	struct rcu_head rcu;
+	struct list_head obj_sub_list;
+};
+
+struct devx_event_subscription {
+	struct list_head file_list; /* headed in private_data->
+				     * subscribed_events_list
+				     */
+	struct list_head xa_list; /* headed in devx_event->unaffiliated_list or
+				   * devx_obj_event->obj_sub_list
+				   */
+	struct list_head obj_list; /* headed in devx_object */
+
+	u32 xa_key_level1;
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 64

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+	u32 xa_key_level2;
+	struct rcu_head	rcu;
+	u64 cookie;
+	bool is_obj_related;
+	struct ib_uobject *fd_uobj;
+	void *object;	/* May need direct access upon hot unplug */

This should be a 'struct file *' and have a better name.

OK, will change.

Quoted 2 lines

And I'm unclear why we need to store both the ib_uobject and the struct file for the same thing?

Post hot unplug/unbind the uobj can't be accessed any more to reach the object as it will be set to NULL by ib_core layer [1]. As the filp is still open we need a direct access to it down the road and for that we have it separately.

This was the comment that I have just put above in the code, I may improve it with more details as pointed here.

[1]
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/rdma_core.c#L149

And why are we storing the uobj here

Quoted 2 lines

instead of the struct devx_async_event_file *?

Basically storing the uobj is the same as of storing the devx_async_event_file, as we just use container_of to get it from. There is no direct access from uobj to any of its fields so that we should be fine.

Quoted 5 lines

Since uobj->object == flip && filp->private_data == uobj, I have a hard time to understand why we need both things, it seems to me that if we get the fget on the filp then we can rely on the filp->private_data to get back to the devx_async_event_file.

The idea was to not take an extra ref count on the file (i.e. fget) per subscription, this will let the release option to be called once the file will be closed by the application. Otherwise we might need to consider having some unsubscribe method to put the ref count back with all its overhead and implications without a real justified reason.

Quoted 11 lines
Quoted 11 lines
drivers/infiniband/hw/mlx5/devx.c +3 -0
+	struct eventfd_ctx *eventfd;
+};
+
 
  /*
   * As the obj_id in the firmware is not globally unique the object type

* must be considered upon checking for a valid object id.

drivers/infiniband/hw/mlx5/devx.c +0 -0
@@ -1143,14 +1275,53 @@ static void devx_cleanup_mkey(struct devx_obj *obj)
  	write_unlock_irqrestore(&table->lock, flags);
  }
 
Quoted 10 lines

drivers/infiniband/hw/mlx5/devx.c static void devx_cleanup_mkey(struct devx_obj *obj) · line 1284

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
+static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
+				      struct devx_event_subscription *sub)
+{
+	list_del_rcu(&sub->file_list);
+	list_del_rcu(&sub->xa_list);
+
+	if (sub->is_obj_related) {

is_obj_related looks like it is just list_empty(obj_list)??

Yes, in that approach we may need to call INIT_LIST_HEAD(sub->obj_list) in case it wasn't an object upon subscription, will do that.

Quoted 16 lines

Success oriented flow

Quoted 14 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
@@ -1523,6 +1700,350 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)(
  	return err;
  }
 
+static void
+subscribe_event_xa_dealloc(struct mlx5_devx_event_table *devx_event_table,
+			   u32 key_level1,
+			   u32 key_level2,
+			   struct devx_obj_event *alloc_obj_event)
+{
+	struct devx_event *event;
+
+	/* Level 1 is valid for future use - no need to free */
+	if (!alloc_obj_event)
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1718

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+		return;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	WARN_ON(!event);
+
+	xa_erase(&event->object_ids, key_level2);

Shoulnd't this only erase if the value stored is NULL?

If this key_level2 wasn't allocated by the subscribe flow and exists before we may not reach here and return from the above lines in this function [1], otherwise we need to erase as done here.

[1]
if (!alloc_obj_event)

	return;
Quoted 47 lines
Quoted 47 lines
drivers/infiniband/hw/mlx5/devx.c +47 -0
+	kfree(alloc_obj_event);
+}
+
+static int
+subscribe_event_xa_alloc(struct mlx5_devx_event_table *devx_event_table,
+			 u32 key_level1,
+			 bool is_level2,
+			 u32 key_level2,
+			 struct devx_obj_event **alloc_obj_event)
+{
+	struct devx_obj_event *obj_event;
+	struct devx_event *event;
+	bool new_entry_level1 = false;
+	int err;
+
+	event = xa_load(&devx_event_table->event_xa, key_level1);
+	if (!event) {
+		event = kzalloc(sizeof(*event), GFP_KERNEL);
+		if (!event)
+			return -ENOMEM;
+
+		new_entry_level1 = true;
+		INIT_LIST_HEAD(&event->unaffiliated_list);
+		xa_init(&event->object_ids);
+
+		err = xa_insert(&devx_event_table->event_xa,
+				key_level1,
+				event,
+				GFP_KERNEL);
+		if (err)
+			goto end;
+	}
+
+	if (!is_level2)
+		return 0;
+
+	obj_event = xa_load(&event->object_ids, key_level2);
+	if (!obj_event) {
+		err = xa_reserve(&event->object_ids, key_level2, GFP_KERNEL);
+		if (err)
+			goto err_level1;
+
+		obj_event = kzalloc(sizeof(*obj_event), GFP_KERNEL);
+		if (!obj_event) {
+			err = -ENOMEM;
+			goto err_level2;
+		}
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1768

Quoted 3 lines
drivers/infiniband/hw/mlx5/devx.c +3 -0
+
+		INIT_LIST_HEAD(&obj_event->obj_sub_list);
+		*alloc_obj_event = obj_event;

This is goofy, just store the empty obj_event in the xa instead of using xa_reserve, and when you go to do the error unwind just delete any level2' devx_obj_event' that has a list_empty(obj_sub_list), get rid of the wonky alloc_obj_event stuff.

Please see my answer above about how level2 is managed by this alloc_obj_event, is that really worth a change ? I found current logic to be clear. I may put some note here if we can stay with that.

Quoted 3 lines

The best configuration would be to use devx_cleanup_subscription to undo the partially ready subscription.

This partially ready subscription might not match the devx_cleanup_subscription(), e.g. it wasn't added to xa_list and can't be deleted without any specific flag to ignore ..

Quoted 9 lines
Quoted 9 lines
drivers/infiniband/hw/mlx5/devx.c +9 -0
+	}
+
+	return 0;
+
+err_level2:
+	xa_erase(&event->object_ids, key_level2);
+
+err_level1:
+	if (new_entry_level1)
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1781

Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +4 -0
+		xa_erase(&devx_event_table->event_xa, key_level1);
+end:
+	if (new_entry_level1)
+		kfree(event);

Can't do this, once the level1 is put in the tree it could be referenced by the irqs. At least it needs a kfree_rcu, most likely it is simpler to just leave it.

Agree, it looks simpler just to leave it, will handle.

Quoted 81 lines
Quoted 81 lines
drivers/infiniband/hw/mlx5/devx.c +81 -0
+#define MAX_NUM_EVENTS 16
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_SUBSCRIBE_EVENT)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *devx_uobj = uverbs_attr_get_uobject(
+				attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_OBJ_HANDLE);
+	struct mlx5_ib_ucontext *c = rdma_udata_to_drv_context(
+		&attrs->driver_udata, struct mlx5_ib_ucontext, ibucontext);
+	struct mlx5_ib_dev *dev = to_mdev(c->ibucontext.device);
+	struct ib_uobject *fd_uobj;
+	struct devx_obj *obj = NULL;
+	struct devx_async_event_file *ev_file;
+	struct mlx5_devx_event_table *devx_event_table = &dev->devx_event_table;
+	u16 *event_type_num_list;
+	struct devx_event_subscription **event_sub_arr;
+	struct devx_obj_event  **event_obj_array_alloc;
+	int redirect_fd;
+	bool use_eventfd = false;
+	int num_events;
+	int num_alloc_xa_entries = 0;
+	u16 obj_type = 0;
+	u64 cookie = 0;
+	u32 obj_id = 0;
+	int err;
+	int i;
+
+	if (!c->devx_uid)
+		return -EINVAL;
+
+	if (!IS_ERR(devx_uobj)) {
+		obj = (struct devx_obj *)devx_uobj->object;
+		if (obj)
+			obj_id = get_dec_obj_id(obj->obj_id);
+	}
+
+	fd_uobj = uverbs_attr_get_uobject(attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_HANDLE);
+	if (IS_ERR(fd_uobj))
+		return PTR_ERR(fd_uobj);
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM)) {
+		err = uverbs_copy_from(&redirect_fd, attrs,
+			       MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_FD_NUM);
+		if (err)
+			return err;
+
+		use_eventfd = true;
+	}
+
+	if (uverbs_attr_is_valid(attrs,
+				 MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE)) {
+		if (use_eventfd)
+			return -EINVAL;
+
+		err = uverbs_copy_from(&cookie, attrs,
+				MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_COOKIE);
+		if (err)
+			return err;
+	}
+
+	num_events = uverbs_attr_ptr_get_array_size(
+		attrs, MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST,
+		sizeof(u16));
+
+	if (num_events < 0)
+		return num_events;
+
+	if (num_events > MAX_NUM_EVENTS)
+		return -EINVAL;
+
+	event_type_num_list = uverbs_attr_get_alloced_ptr(attrs,
+			MLX5_IB_ATTR_DEVX_SUBSCRIBE_EVENT_TYPE_NUM_LIST);
+
+	if (!is_valid_events(dev->mdev, num_events, event_type_num_list, obj))
+		return -EINVAL;
+
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1935

Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +4 -0
+	event_sub_arr = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_event_subscription *));
+	event_obj_array_alloc = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_obj_event *));

There are so many list_heads in the devx_event_subscription, why not use just one of them to store the allocated events instead of this temp array? ie event_list looks good for this purpose.

I'm using the array later on with direct access to the index that should be de-allocated. I would prefer staying with this array rather than using the 'event_list' which has other purpose down the road, it's used per subscription and doesn't look match to hold the devx_obj_event which has no list entry for this purpose..

Quoted 15 lines
Quoted 15 lines
drivers/infiniband/hw/mlx5/devx.c +15 -0
+
+	if (!event_sub_arr || !event_obj_array_alloc)
+		return -ENOMEM;
+
+	/* Protect from concurrent subscriptions to same XA entries to allow
+	 * both to succeed
+	 */
+	mutex_lock(&devx_event_table->event_xa_lock);
+	for (i = 0; i < num_events; i++) {
+		u32 key_level1;
+
+		if (obj)
+			obj_type = get_dec_obj_type(obj,
+						    event_type_num_list[i]);
+		key_level1 = event_type_num_list[i] | obj_type << 16;
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1956

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+
+		err = subscribe_event_xa_alloc(devx_event_table,
+					       key_level1,
+					       obj ? true : false,
+					       obj_id,
+					       &event_obj_array_alloc[i]);

Usless ?:

What do you suggest instead ?

Quoted 29 lines
Quoted 28 lines
drivers/infiniband/hw/mlx5/devx.c +27 -0
+		if (err)
+			goto err;
+
+		num_alloc_xa_entries++;
+		event_sub_arr[i] = kzalloc(sizeof(*event_sub_arr[i]),
+					   GFP_KERNEL);
+		if (!event_sub_arr[i])
+			goto err;
+
+		if (use_eventfd) {
+			event_sub_arr[i]->eventfd =
+				eventfd_ctx_fdget(redirect_fd);
+
+			if (IS_ERR(event_sub_arr[i]->eventfd)) {
+				err = PTR_ERR(event_sub_arr[i]->eventfd);
+				event_sub_arr[i]->eventfd = NULL;
+				goto err;
+			}
+		}
+
+		event_sub_arr[i]->cookie = cookie;
+		event_sub_arr[i]->fd_uobj = fd_uobj;
+		event_sub_arr[i]->object = fd_uobj->object;
+		/* May be needed upon cleanup the devx object/subscription */
+		event_sub_arr[i]->xa_key_level1 = key_level1;
+		event_sub_arr[i]->xa_key_level2 = obj_id;
+		event_sub_arr[i]->is_obj_related = obj ? true : false;
#23 lore raw

Re: [PATCH rdma-next v1 11/12] IB/mlx5: Implement DEVX dispatching event

In reply to Jason Gunthorpe
Quoted 34 lines · Jason Gunthorpe
Quoted 33 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Implement DEVX dispatching event by looking up for the applicable subscriptions for the reported event and using their target fd to signal/set the event.

  drivers/infiniband/hw/mlx5/devx.c         | 362 +++++++++++++++++++++-
  include/uapi/rdma/mlx5_user_ioctl_verbs.h |   5 +
  2 files changed, 357 insertions(+), 10 deletions(-)
drivers/infiniband/hw/mlx5/devx.c +7 -0
index 304b13e7a265..49fdce95d6d9 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -34,6 +34,11 @@ struct devx_async_data {
  	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
  };
 
+struct devx_async_event_data {
+	struct list_head list; /* headed in ev_queue->event_list */
+	struct mlx5_ib_uapi_devx_async_event_hdr hdr;
+};
+
  /* first level XA value data structure */
  struct devx_event {
  	struct xarray object_ids; /* second XA level, Key = object id */
@@ -54,7 +59,9 @@ struct devx_event_subscription {
  				   * devx_obj_event->obj_sub_list
  				   */
  	struct list_head obj_list; /* headed in devx_object */
+	struct list_head event_list; /* headed in ev_queue->event_list */
 
Quoted 5 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_event_subscription { · line 64

Quoted 1 line
drivers/infiniband/hw/mlx5/devx.c +1 -0
+	u8  is_cleaned:1;

There is a loose bool 'is_obj_related' that should be combined with this bool bitfield as well.

OK

Quoted 24 lines
Quoted 24 lines
drivers/infiniband/hw/mlx5/devx.c +10 -4
  static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
-				      struct devx_event_subscription *sub)
+				      struct devx_event_subscription *sub,
+				      bool file_close)
  {
-	list_del_rcu(&sub->file_list);
+	if (sub->is_cleaned)
+		goto end;
+
+	sub->is_cleaned = 1;
  	list_del_rcu(&sub->xa_list);
 
  	if (sub->is_obj_related) {
@@ -1303,10 +1355,15 @@ static void devx_cleanup_subscription(struct mlx5_ib_dev *dev,
  		}
  	}
 
-	if (sub->eventfd)
-		eventfd_ctx_put(sub->eventfd);
+end:
+	if (file_close) {
+		if (sub->eventfd)
+			eventfd_ctx_put(sub->eventfd);
 
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static void devx_cleanup_subscription(struct mlx5_ib_dev *dev, · line 1366

Quoted 5 lines
drivers/infiniband/hw/mlx5/devx.c +4 -1
-	kfree_rcu(sub, rcu);
+		list_del_rcu(&sub->file_list);
+		/* subscription may not be used by the read API any more */
+		kfree_rcu(sub, rcu);
+	}

Dis like this confusing file_close stuff, just put this in the single place that calls this with the true bool

OK, will do.

Quoted 26 lines
Quoted 26 lines
drivers/infiniband/hw/mlx5/devx.c +26 -0
+static int deliver_event(struct devx_event_subscription *event_sub,
+			 const void *data)
+{
+	struct ib_uobject *fd_uobj = event_sub->fd_uobj;
+	struct devx_async_event_file *ev_file;
+	struct devx_async_event_queue *ev_queue;
+	struct devx_async_event_data *event_data;
+	unsigned long flags;
+	bool omit_data;
+
+	ev_file = container_of(fd_uobj, struct devx_async_event_file,
+			       uobj);
+	ev_queue = &ev_file->ev_queue;
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	if (omit_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		if (!list_empty(&event_sub->event_list)) {
+			spin_unlock_irqrestore(&ev_queue->lock, flags);
+			return 0;
+		}
+
+		list_add_tail(&event_sub->event_list, &ev_queue->event_list);
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		wake_up_interruptible(&ev_queue->poll_wait);
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2339

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+		return 0;
+	}
+
+	event_data = kzalloc(sizeof(*event_data) +
+			     (omit_data ? 0 : sizeof(struct mlx5_eqe)),
+			     GFP_ATOMIC);

omit_data is always false here

Correct, will clean it up.

Quoted 58 lines
Quoted 58 lines
drivers/infiniband/hw/mlx5/devx.c +54 -1
+	if (!event_data) {
+		spin_lock_irqsave(&ev_queue->lock, flags);
+		ev_queue->is_overflow_err = 1;
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
+		return -ENOMEM;
+	}
+
+	event_data->hdr.cookie = event_sub->cookie;
+	memcpy(event_data->hdr.out_data, data, sizeof(struct mlx5_eqe));
+
+	spin_lock_irqsave(&ev_queue->lock, flags);
+	list_add_tail(&event_data->list, &ev_queue->event_list);
+	spin_unlock_irqrestore(&ev_queue->lock, flags);
+	wake_up_interruptible(&ev_queue->poll_wait);
+
+	return 0;
+}
+
+static void dispatch_event_fd(struct list_head *fd_list,
+			      const void *data)
+{
+	struct devx_event_subscription *item;
+
+	list_for_each_entry_rcu(item, fd_list, xa_list) {
+		if (!get_file_rcu((struct file *)item->object))
+			continue;
+
+		if (item->eventfd) {
+			eventfd_signal(item->eventfd, 1);
+			fput(item->object);
+			continue;
+		}
+
+		deliver_event(item, data);
+		fput(item->object);
+	}
+}
+
  static int devx_event_notifier(struct notifier_block *nb,
  			       unsigned long event_type, void *data)
  {
-	return NOTIFY_DONE;
+	struct mlx5_devx_event_table *table;
+	struct mlx5_ib_dev *dev;
+	struct devx_event *event;
+	struct devx_obj_event *obj_event;
+	u16 obj_type = 0;
+	bool is_unaffiliated;
+	u32 obj_id;
+
+	/* Explicit filtering to kernel events which may occur frequently */
+	if (event_type == MLX5_EVENT_TYPE_CMD ||
+	    event_type == MLX5_EVENT_TYPE_PAGE_REQUEST)
+		return NOTIFY_OK;
+
+	table = container_of(nb, struct mlx5_devx_event_table, devx_nb.nb);
+	dev = container_of(table, struct mlx5_ib_dev, devx_event_table);
+	is_unaffiliated = is_unaffiliated_event(dev->mdev, event_type);
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2402

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+
+	if (!is_unaffiliated)
+		obj_type = get_event_obj_type(event_type, data);
+	event = xa_load(&table->event_xa, event_type | (obj_type << 16));
+	if (!event)
+		return NOTIFY_DONE;

event should be in the rcu as well

Do we really need this ? I didn't see a flow that really requires that.

Quoted 64 lines
Quoted 64 lines
drivers/infiniband/hw/mlx5/devx.c +47 -2
+	if (is_unaffiliated) {
+		dispatch_event_fd(&event->unaffiliated_list, data);
+		return NOTIFY_OK;
+	}
+
+	obj_id = devx_get_obj_id_from_event(event_type, data);
+	rcu_read_lock();
+	obj_event = xa_load(&event->object_ids, obj_id);
+	if (!obj_event) {
+		rcu_read_unlock();
+		return NOTIFY_DONE;
+	}
+
+	dispatch_event_fd(&obj_event->obj_sub_list, data);
+
+	rcu_read_unlock();
+	return NOTIFY_OK;
  }
 
  void mlx5_ib_devx_init_event_table(struct mlx5_ib_dev *dev)
@@ -2221,7 +2444,7 @@ void mlx5_ib_devx_cleanup_event_table(struct mlx5_ib_dev *dev)
  		event = entry;
  		list_for_each_entry_safe(sub, tmp, &event->unaffiliated_list,
  					 xa_list)
-			devx_cleanup_subscription(dev, sub);
+			devx_cleanup_subscription(dev, sub, false);
  		kfree(entry);
  	}
  	mutex_unlock(&dev->devx_event_table.event_xa_lock);
@@ -2329,18 +2552,126 @@ static const struct file_operations devx_async_cmd_event_fops = {
  static ssize_t devx_async_event_read(struct file *filp, char __user *buf,
  				     size_t count, loff_t *pos)
  {
-	return -EINVAL;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	struct devx_event_subscription *event_sub;
+	struct devx_async_event_data *uninitialized_var(event);
+	int ret = 0;
+	size_t eventsz;
+	bool omit_data;
+	void *event_data;
+
+	omit_data = ev_queue->flags &
+		MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA;
+
+	spin_lock_irq(&ev_queue->lock);
+
+	if (ev_queue->is_overflow_err) {
+		ev_queue->is_overflow_err = 0;
+		spin_unlock_irq(&ev_queue->lock);
+		return -EOVERFLOW;
+	}
+
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		if (wait_event_interruptible(ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2589

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;

All these tests should be under the lock.

We can't call wait_event_interruptible() above which may sleep under the lock, correct ? are you referring to the list_empty() and is_destroyed ?

By the way looking in uverb code [1], similar code which is not done under the lock as of here..

[1] https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L244

Quoted 3 lines

Why don't we return EIO as soon as is-destroyed happens? What is the point of flushing out the accumulated events?

It follows the above uverb code/logic that returns existing events even in that case, also the async command events in this file follows that logic, I suggest to stay consistent.

Quoted 16 lines
Quoted 16 lines
drivers/infiniband/hw/mlx5/devx.c +15 -0
+
+		spin_lock_irq(&ev_queue->lock);
+	}
+
+	if (omit_data) {
+		event_sub = list_first_entry(&ev_queue->event_list,
+					struct devx_event_subscription,
+					event_list);
+		eventsz = sizeof(event_sub->cookie);
+		event_data = &event_sub->cookie;
+	} else {
+		event = list_first_entry(&ev_queue->event_list,
+				      struct devx_async_event_data, list);
+		eventsz = sizeof(struct mlx5_eqe) +
+			sizeof(struct mlx5_ib_uapi_devx_async_event_hdr);
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2610

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+		event_data = &event->hdr;
+	}
+
+	if (eventsz > count) {
+		spin_unlock_irq(&ev_queue->lock);
+		return -ENOSPC;

This is probably the wrong errno

OK, will change to -EINVAL as in uverbs
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L254

Quoted 16 lines
Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +3 -0
+	}
+
+	if (omit_data)

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2623

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+		list_del_init(&event_sub->event_list);
+	else
+		list_del(&event->list);
+
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (copy_to_user(buf, event_data, eventsz))
+		ret = -EFAULT;
+	else
+		ret = eventsz;

This is really poorly ordered, EFAULT will cause the event to be lost. :(

Agree but apparently rare case .. see also the below notes.

Quoted 2 lines

Maybe the event should be re-added on error? Tricky.

What will happen if another copy_to_user may then fail again (loop ?) ... not sure that we want to get into this tricky handling ...

As of above, It follows the logic from uverbs at that area.
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L267

Quoted 42 lines
Quoted 42 lines
drivers/infiniband/hw/mlx5/devx.c +31 -1
+	if (!omit_data)
+		kfree(event);
+	return ret;
  }
 
  static __poll_t devx_async_event_poll(struct file *filp,
  				      struct poll_table_struct *wait)
  {
-	return 0;
+	struct devx_async_event_file *ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+	__poll_t pollflags = 0;
+
+	poll_wait(filp, &ev_queue->poll_wait, wait);
+
+	spin_lock_irq(&ev_queue->lock);
+	if (ev_queue->is_destroyed)
+		pollflags = EPOLLIN | EPOLLRDNORM | EPOLLRDHUP;
+	else if (!list_empty(&ev_queue->event_list))
+		pollflags = EPOLLIN | EPOLLRDNORM;
+	spin_unlock_irq(&ev_queue->lock);
+
+	return pollflags;
  }
 
  static int devx_async_event_close(struct inode *inode, struct file *filp)
  {
+	struct ib_uobject *uobj = filp->private_data;
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file, uobj);
+	struct devx_event_subscription *event_sub, *event_sub_tmp;
+	struct devx_async_event_data *entry, *tmp;
+
+	mutex_lock(&ev_file->dev->devx_event_table.event_xa_lock);
+	/* delete the subscriptions which are related to this FD */
+	list_for_each_entry_safe(event_sub, event_sub_tmp,
+				 &ev_file->subscribed_events_list, file_list)
+		devx_cleanup_subscription(ev_file->dev, event_sub, true);
+	mutex_unlock(&ev_file->dev->devx_event_table.event_xa_lock);
+
+	/* free the pending events allocation */
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2670

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+	if (!(ev_file->ev_queue.flags &
+	    MLX5_IB_UAPI_DEVX_CREATE_EVENT_CHANNEL_FLAGS_OMIT_EV_DATA)) {
+		spin_lock_irq(&ev_file->ev_queue.lock);
+		list_for_each_entry_safe(entry, tmp,
+					 &ev_file->ev_queue.event_list, list)
+			kfree(entry); /* read can't come any nore */

spelling

OK

Quoted 17 lines
Quoted 17 lines
drivers/infiniband/hw/mlx5/devx.c +8 -0
+		spin_unlock_irq(&ev_file->ev_queue.lock);
+	}
  	uverbs_close_fd(filp);
+	put_device(&ev_file->dev->ib_dev.dev);
  	return 0;
  }
 
@@ -2374,6 +2705,17 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
  static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
  					    enum rdma_remove_reason why)
  {
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj, · line 2718

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);

Why isn't this wakeup always done?

Maybe you are right and this can be always done to wake up any readers as the 'is_destroyed' was set.

By the way, any idea why it was done as such in uverbs [1] for similar flow ? also the command events follows that.

[1]
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_std_types.c#L207

#29 lore raw

Re: [PATCH rdma-next v1 12/12] IB/mlx5: Add DEVX support for CQ events

In reply to Jason Gunthorpe
Quoted 26 lines · Jason Gunthorpe
Quoted 25 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Add DEVX support for CQ events by creating and destroying the CQ via mlx5_core and set an handler to manage its completions.

  drivers/infiniband/hw/mlx5/devx.c | 40 +++++++++++++++++++++++++++++++
  1 file changed, 40 insertions(+)
drivers/infiniband/hw/mlx5/devx.c +4 -0
index 49fdce95d6d9..91ccd58ebc05 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -19,9 +19,12 @@
  #define UVERBS_MODULE_NAME mlx5_ib
  #include <rdma/uverbs_named_ioctl.h>
 
+static void dispatch_event_fd(struct list_head *fd_list, const void *data);
+
  enum devx_obj_flags {
  	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,
  	DEVX_OBJ_FLAGS_DCT = 1 << 1,
+	DEVX_OBJ_FLAGS_CQ = 1 << 2,
  };
 
Quoted 10 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_event_file { · line 100

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +1 -0
  struct devx_async_data {
@@ -94,6 +97,7 @@ struct devx_async_event_file {
  #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
  struct devx_obj {
  	struct mlx5_core_dev	*mdev;
+	struct mlx5_ib_dev	*ib_dev;

This seems strange, why would we need to store the core_dev and the ib_dev in a struct when ibdev->mdev == core_dev?

We need to add the ib_dev as we can't access it from the core_dev.
Most of this patch we can probably go and drop the mdev and access it from ib_dev, I preferred to not handle that in this patch.

#20 lore raw

Re: [PATCH rdma-next v1 10/12] IB/mlx5: Enable subscription for device events over DEVX

In reply to Yishai Hadas
Quoted 11 lines · Yishai Hadas
Quoted 10 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_data { · line 64

Quoted 7 lines
drivers/infiniband/hw/mlx5/devx.c +7 -0
+	u32 xa_key_level1;
+	u32 xa_key_level2;
+	struct rcu_head	rcu;
+	u64 cookie;
+	bool is_obj_related;
+	struct ib_uobject *fd_uobj;
+	void *object;	/* May need direct access upon hot unplug */

This should be a 'struct file *' and have a better name.

Quoted 7 lines

OK, will change.

Quoted 2 lines

And I'm unclear why we need to store both the ib_uobject and the struct file for the same thing?

Post hot unplug/unbind the uobj can't be accessed any more to reach the object as it will be set to NULL by ib_core layer [1].

struct file users need to get the uobject from the file->private_data under a fget.

There is only place place that needed fd_uobj, and it was under the fget section, so it should simply use private_data.

This is why you should only store the struct file and not the uobject.

Quoted 5 lines

This was the comment that I have just put above in the code, I may improve it with more details as pointed here.

[1]
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/rdma_core.c#L149

I'm wondering if this is a bug to do this for fds?

Quoted 9 lines
Quoted 5 lines

Since uobj->object == flip && filp->private_data == uobj, I have a hard time to understand why we need both things, it seems to me that if we get the fget on the filp then we can rely on the filp->private_data to get back to the devx_async_event_file.

The idea was to not take an extra ref count on the file (i.e. fget) per subscription, this will let the release option to be called once the file will be closed by the application.

No extra ref is needed, the fget is already obtained in the only place that needs fd_uobj.

Quoted 21 lines
Quoted 11 lines
Quoted 11 lines
drivers/infiniband/hw/mlx5/devx.c +11 -0
+	obj_event = xa_load(&event->object_ids, key_level2);
+	if (!obj_event) {
+		err = xa_reserve(&event->object_ids, key_level2, GFP_KERNEL);
+		if (err)
+			goto err_level1;
+
+		obj_event = kzalloc(sizeof(*obj_event), GFP_KERNEL);
+		if (!obj_event) {
+			err = -ENOMEM;
+			goto err_level2;
+		}
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1768

Quoted 3 lines
drivers/infiniband/hw/mlx5/devx.c +3 -0
+
+		INIT_LIST_HEAD(&obj_event->obj_sub_list);
+		*alloc_obj_event = obj_event;

This is goofy, just store the empty obj_event in the xa instead of using xa_reserve, and when you go to do the error unwind just delete any level2' devx_obj_event' that has a list_empty(obj_sub_list), get rid of the wonky alloc_obj_event stuff.

Quoted 3 lines

Please see my answer above about how level2 is managed by this alloc_obj_event, is that really worth a change ? I found current logic to be clear. I may put some note here if we can stay with that.

I think it is alot cleaner/simpler than using this extra memory

Quoted 6 lines
Quoted 2 lines

The best configuration would be to use devx_cleanup_subscription to undo the partially ready subscription.

This partially ready subscription might not match the devx_cleanup_subscription(), e.g. it wasn't added to xa_list and can't be deleted without any specific flag to ignore ..

Maybe, but I suspect it can work out

Quoted 10 lines
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1935

Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +4 -0
+	event_sub_arr = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_event_subscription *));
+	event_obj_array_alloc = uverbs_zalloc(attrs,
+		MAX_NUM_EVENTS * sizeof(struct devx_obj_event *));

There are so many list_heads in the devx_event_subscription, why not use just one of them to store the allocated events instead of this temp array? ie event_list looks good for this purpose.

Quoted 5 lines

I'm using the array later on with direct access to the index that should be de-allocated. I would prefer staying with this array rather than using the 'event_list' which has other purpose down the road, it's used per subscription and doesn't look match to hold the devx_obj_event which has no list entry for this purpose..

Replace the event_obj_array_alloc by storing that data directly in the xarray

Replace the event_sub_arr by building them into a linked list - it always need to iterate over the whole list anyhow.

Quoted 24 lines
Quoted 15 lines
Quoted 15 lines
drivers/infiniband/hw/mlx5/devx.c +15 -0
+
+	if (!event_sub_arr || !event_obj_array_alloc)
+		return -ENOMEM;
+
+	/* Protect from concurrent subscriptions to same XA entries to allow
+	 * both to succeed
+	 */
+	mutex_lock(&devx_event_table->event_xa_lock);
+	for (i = 0; i < num_events; i++) {
+		u32 key_level1;
+
+		if (obj)
+			obj_type = get_dec_obj_type(obj,
+						    event_type_num_list[i]);
+		key_level1 = event_type_num_list[i] | obj_type << 16;
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)( · line 1956

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+
+		err = subscribe_event_xa_alloc(devx_event_table,
+					       key_level1,
+					       obj ? true : false,
+					       obj_id,
+					       &event_obj_array_alloc[i]);

Usless ?:

Quoted 1 line

What do you suggest instead ?

Nothing is needed, cast to implicit bool is always forced to true/false

Jason

#24 lore raw

Re: [PATCH rdma-next v1 11/12] IB/mlx5: Implement DEVX dispatching event

In reply to Yishai Hadas
Quoted 17 lines · Yishai Hadas
Quoted 8 lines
Quoted 8 lines
drivers/infiniband/hw/mlx5/devx.c +8 -0
+	/* Explicit filtering to kernel events which may occur frequently */
+	if (event_type == MLX5_EVENT_TYPE_CMD ||
+	    event_type == MLX5_EVENT_TYPE_PAGE_REQUEST)
+		return NOTIFY_OK;
+
+	table = container_of(nb, struct mlx5_devx_event_table, devx_nb.nb);
+	dev = container_of(table, struct mlx5_ib_dev, devx_event_table);
+	is_unaffiliated = is_unaffiliated_event(dev->mdev, event_type);
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2402

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+
+	if (!is_unaffiliated)
+		obj_type = get_event_obj_type(event_type, data);
+	event = xa_load(&table->event_xa, event_type | (obj_type << 16));
+	if (!event)
+		return NOTIFY_DONE;

event should be in the rcu as well

Quoted 2 lines

Do we really need this ? I didn't see a flow that really requires that.

I think there are no frees left? Even so it makes much more sense to include the event in the rcu as if we ever did need to kfree it would have to be via rcu

Quoted 18 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/hw/mlx5/devx.c +9 -0
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		if (wait_event_interruptible(ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2589

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;

All these tests should be under the lock.

Quoted 3 lines

We can't call wait_event_interruptible() above which may sleep under the lock, correct ? are you referring to the list_empty() and is_destroyed ?

yes

Quoted 4 lines

By the way looking in uverb code [1], similar code which is not done under the lock as of here..

[1] https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L244

Also not a good idea

Quoted 6 lines
Quoted 2 lines

Why don't we return EIO as soon as is-destroyed happens? What is the point of flushing out the accumulated events?

It follows the above uverb code/logic that returns existing events even in that case, also the async command events in this file follows that logic, I suggest to stay consistent.

Don't follow broken uverbs stuff...

Quoted 7 lines
Quoted 1 line

Maybe the event should be re-added on error? Tricky.

What will happen if another copy_to_user may then fail again (loop ?) ... not sure that we want to get into this tricky handling ...

As of above, It follows the logic from uverbs at that area.
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L267

again it is wrong...

There is no loop if you just stick the item back on the head of the list and exit, which is probably the right thing to do..

Quoted 18 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/hw/mlx5/devx.c +5 -0
@@ -2374,6 +2705,17 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
  static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
  					    enum rdma_remove_reason why)
  {
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj, · line 2718

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);

Why isn't this wakeup always done?

Quoted 5 lines

Maybe you are right and this can be always done to wake up any readers as the 'is_destroyed' was set.

By the way, any idea why it was done as such in uverbs [1] for similar flow ? also the command events follows that.

I don't know, it is probably pointless too.

If we don't need it here then we shouldn't have it.

These random pointless ifs bother me as we have to spend time trying to figure out that they are pointless down the road.

Jason

#30 lore raw

Re: [PATCH rdma-next v1 12/12] IB/mlx5: Add DEVX support for CQ events

In reply to Yishai Hadas
Quoted 36 lines · Yishai Hadas
Quoted 25 lines · Jason Gunthorpe
Quoted 24 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Add DEVX support for CQ events by creating and destroying the CQ via mlx5_core and set an handler to manage its completions.

  drivers/infiniband/hw/mlx5/devx.c | 40 +++++++++++++++++++++++++++++++
  1 file changed, 40 insertions(+)
drivers/infiniband/hw/mlx5/devx.c +4 -0
index 49fdce95d6d9..91ccd58ebc05 100644
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -19,9 +19,12 @@
  #define UVERBS_MODULE_NAME mlx5_ib
  #include <rdma/uverbs_named_ioctl.h>
+static void dispatch_event_fd(struct list_head *fd_list, const void *data);
+
  enum devx_obj_flags {
  	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,
  	DEVX_OBJ_FLAGS_DCT = 1 << 1,
+	DEVX_OBJ_FLAGS_CQ = 1 << 2,
  };
  struct devx_async_data {
Quoted 9 lines

drivers/infiniband/hw/mlx5/devx.c struct devx_async_event_file { · line 100

Quoted 5 lines
drivers/infiniband/hw/mlx5/devx.c +1 -0
@@ -94,6 +97,7 @@ struct devx_async_event_file {
  #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
  struct devx_obj {
  	struct mlx5_core_dev	*mdev;
+	struct mlx5_ib_dev	*ib_dev;

This seems strange, why would we need to store the core_dev and the ib_dev in a struct when ibdev->mdev == core_dev?

Quoted 3 lines

We need to add the ib_dev as we can't access it from the core_dev. Most of this patch we can probably go and drop the mdev and access it from ib_dev, I preferred to not handle that in this patch.

Should add a patch to revise it then

Jason

#33 lore raw
In reply to Leon Romanovsky
Quoted 4 lines · Leon Romanovsky
Quoted 3 lines · Saeed Mahameed
Quoted 2 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines

Changelog:
v0 -> v1:

Normally 1st submission is V1 and 2nd is V2.
so this should have been v1->v2.

"Normally" depends on the language you are using. In C, everything starts from 0, including version of patches :).

You are wrong:
quoting: https://kernelnewbies.org/PatchTipsAndTricks

"For example, if you're sending the second revision of a patch, you should use [PATCH v2]."

now don't tell me that second revision is actually 3rd revision or 1st is 2nd :)..

Quoted 5 lines
Quoted 3 lines

For mlx5-next patches:

Thanks

#25 lore raw

Re: [PATCH rdma-next v1 11/12] IB/mlx5: Implement DEVX dispatching event

In reply to Jason Gunthorpe
Quoted 19 lines · Jason Gunthorpe
Quoted 17 lines · Yishai Hadas
Quoted 16 lines
Quoted 4 lines
drivers/infiniband/hw/mlx5/devx.c +4 -0
+	/* Explicit filtering to kernel events which may occur frequently */
+	if (event_type == MLX5_EVENT_TYPE_CMD ||
+	    event_type == MLX5_EVENT_TYPE_PAGE_REQUEST)
+		return NOTIFY_OK;

drivers/infiniband/hw/mlx5/devx.c static int devx_umem_cleanup(struct ib_uobject *uobject, · line 2402

Quoted 10 lines
drivers/infiniband/hw/mlx5/devx.c +10 -0
+
+	table = container_of(nb, struct mlx5_devx_event_table, devx_nb.nb);
+	dev = container_of(table, struct mlx5_ib_dev, devx_event_table);
+	is_unaffiliated = is_unaffiliated_event(dev->mdev, event_type);
+
+	if (!is_unaffiliated)
+		obj_type = get_event_obj_type(event_type, data);
+	event = xa_load(&table->event_xa, event_type | (obj_type << 16));
+	if (!event)
+		return NOTIFY_DONE;

event should be in the rcu as well

Quoted 7 lines
Quoted 2 lines

Do we really need this ? I didn't see a flow that really requires that.

I think there are no frees left? Even so it makes much more sense to include the event in the rcu as if we ever did need to kfree it would have to be via rcu

OK

Quoted 31 lines
Quoted 18 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/hw/mlx5/devx.c +9 -0
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		if (wait_event_interruptible(ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static const struct file_operations devx_async_cmd_event_fops = { · line 2589

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;

All these tests should be under the lock.

Quoted 3 lines

We can't call wait_event_interruptible() above which may sleep under the lock, correct ? are you referring to the list_empty() and is_destroyed ?

yes

Quoted 4 lines

By the way looking in uverb code [1], similar code which is not done under the lock as of here..

[1] https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L244

Also not a good idea

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

Why don't we return EIO as soon as is-destroyed happens? What is the point of flushing out the accumulated events?

It follows the above uverb code/logic that returns existing events even in that case, also the async command events in this file follows that logic, I suggest to stay consistent.

Don't follow broken uverbs stuff...

May it be that there is some event that we still want to deliver post unbind/hot-unplug ? for example IB_EVENT_DEVICE_FATAL in uverbs and others from the driver code.

Not sure that we want to change this logic.
What do you think ?

Quoted 6 lines
Quoted 6 lines
Quoted 2 lines

Maybe the event should be re-added on error? Tricky.

What will happen if another copy_to_user may then fail again (loop ?) ... not sure that we want to get into this tricky handling ...

Quoted 8 lines
Quoted 2 lines

As of above, It follows the logic from uverbs at that area.
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L267

again it is wrong...

There is no loop if you just stick the item back on the head of the list and exit, which is probably the right thing to do..

What if copy_to_user() will fail again just later on ? we might end-up with loop of read(s) that always find an event as it was put back. I suggest to leave this flow as it's now, at least for this series submission.

Agree ?

Quoted 24 lines
Quoted 18 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/hw/mlx5/devx.c +5 -0
@@ -2374,6 +2705,17 @@ static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
   static int devx_hot_unplug_async_event_file(struct ib_uobject *uobj,
   					    enum rdma_remove_reason why)
   {
+	struct devx_async_event_file *ev_file =
+		container_of(uobj, struct devx_async_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &ev_file->ev_queue;
+
Quoted 8 lines

drivers/infiniband/hw/mlx5/devx.c static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj, · line 2718

Quoted 6 lines
drivers/infiniband/hw/mlx5/devx.c +6 -0
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);

Why isn't this wakeup always done?

Quoted 5 lines

Maybe you are right and this can be always done to wake up any readers as the 'is_destroyed' was set.

By the way, any idea why it was done as such in uverbs [1] for similar flow ? also the command events follows that.

Quoted 7 lines

I don't know, it is probably pointless too.

If we don't need it here then we shouldn't have it.

These random pointless ifs bother me as we have to spend time trying to figure out that they are pointless down the road.

OK, will drop this if.

#26 lore raw

Re: [PATCH rdma-next v1 11/12] IB/mlx5: Implement DEVX dispatching event

Quoted 7 lines · Yishai Hadas
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

Why don't we return EIO as soon as is-destroyed happens? What is the point of flushing out the accumulated events?

It follows the above uverb code/logic that returns existing events even in that case, also the async command events in this file follows that logic, I suggest to stay consistent.

Quoted 5 lines
Quoted 1 line

Don't follow broken uverbs stuff...

May it be that there is some event that we still want to deliver post unbind/hot-unplug ? for example IB_EVENT_DEVICE_FATAL in uverbs and others from the driver code.

EIO is DEVICE_FATAL.

Quoted 2 lines

Not sure that we want to change this logic.
What do you think ?

I think this code should exit immediately with EIO if the device is disassociated.

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 1 line

Maybe the event should be re-added on error? Tricky.

What will happen if another copy_to_user may then fail again (loop ?) ... not sure that we want to get into this tricky handling ...

As of above, It follows the logic from uverbs at that area.
https://elixir.bootlin.com/linux/latest/source/drivers/infiniband/core/uverbs_main.c#L267

Quoted 8 lines
Quoted 5 lines

again it is wrong...

There is no loop if you just stick the item back on the head of the list and exit, which is probably the right thing to do..

What if copy_to_user() will fail again just later on ? we might end-up with loop of read(s) that always find an event as it was put back.

That is clearly an application bug and is not the concern of the kernel..

Quoted 4 lines

I suggest to leave this flow as it's now, at least for this series submission.

Agree ?

I don't think you can actually fix this, so maybe we have to leave it. But add a comment explaining

Jason

#3 lore raw

Re: [PATCH mlx5-next v1 02/12] net/mlx5: Use event mask based on device capabilities

In reply to Leon Romanovsky
Quoted 27 lines · Leon Romanovsky

From: Yishai Hadas <[email protected]>

Use the reported device capabilities for the supported user events (i.e.
affiliated and un-affiliated) to set the EQ mask.

As the event mask can be up to 256 defined by 4 entries of u64 change the applicable code to work accordingly.

---
 drivers/infiniband/hw/mlx5/odp.c             |  3 +-
 drivers/net/ethernet/mellanox/mlx5/core/eq.c | 45 ++++++++++++++++
----
 drivers/net/ethernet/mellanox/mlx5/core/fw.c |  6 +++
 include/linux/mlx5/device.h                  |  6 ++-
 include/linux/mlx5/eq.h                      |  4 +-
 include/linux/mlx5/mlx5_ifc.h                | 13 ++++--
 6 files changed, 63 insertions(+), 14 deletions(-)
drivers/infiniband/hw/mlx5/odp.c +0 -0
index 600fe23e2eae..a6740ec308ed 100644

drivers/infiniband/hw/mlx5/odp.c mlx5_ib_create_pf_eq(struct mlx5_ib_dev *dev, · line 1567

Quoted 10 lines
drivers/infiniband/hw/mlx5/odp.c +2 -1
@@ -1559,10 +1559,11 @@ mlx5_ib_create_pf_eq(struct mlx5_ib_dev *dev,
 struct mlx5_ib_pf_eq *eq)
 	eq->irq_nb.notifier_call = mlx5_ib_eq_pf_int;
 	param = (struct mlx5_eq_param) {
 		.irq_index = 0,
-		.mask = 1 << MLX5_EVENT_TYPE_PAGE_FAULT,
 		.nent = MLX5_IB_NUM_PF_EQE,
 	};
 	eq->core = mlx5_eq_create_generic(dev->mdev, &param);
+
+	param.mask[0] = 1ull << MLX5_EVENT_TYPE_PAGE_FAULT;

As Yishai already pointer out, there is a regression here, the line above was merged in the wrong order, mask should be setup before calling mlx5_eq_create_generic.

I will expect V3.

Quoted 83 lines
drivers/infiniband/hw/mlx5/odp.c +0 -0
 	if (IS_ERR(eq->core)) {
 		err = PTR_ERR(eq->core);
drivers/infiniband/hw/mlx5/odp.c +0 -0
 		goto err_wq;
drivers/net/ethernet/mellanox/mlx5/core/eq.c +1 -0
index 8000d2a4a7e2..9d07add38940 100644
@@ -256,6 +256,7 @@ create_map_eq(struct mlx5_core_dev *dev, struct
 mlx5_eq *eq,
 	int inlen;
 	u32 *in;
 	int err;
+	int i;
 
 	/* Init CQ table */
drivers/net/ethernet/mellanox/mlx5/core/eq.c +4 -2
 	memset(cq_table, 0, sizeof(*cq_table));
@@ -283,10 +284,12 @@ create_map_eq(struct mlx5_core_dev *dev, struct
 mlx5_eq *eq,
 	mlx5_fill_page_array(&eq->buf, pas);
 
 	MLX5_SET(create_eq_in, in, opcode, MLX5_CMD_OP_CREATE_EQ);
-	if (!param->mask && MLX5_CAP_GEN(dev, log_max_uctx))
+	if (!param->mask[0] && MLX5_CAP_GEN(dev, log_max_uctx))
 		MLX5_SET(create_eq_in, in, uid,
 MLX5_SHARED_RESOURCE_UID);
 
-	MLX5_SET64(create_eq_in, in, event_bitmask, param->mask);
+	for (i = 0; i < 4; i++)
+		MLX5_ARRAY_SET64(create_eq_in, in, event_bitmask, i,
+				 param->mask[i]);
 
drivers/net/ethernet/mellanox/mlx5/core/eq.c +23 -1
 	eqc = MLX5_ADDR_OF(create_eq_in, in, eq_context_entry);
 	MLX5_SET(eqc, eqc, log_eq_size, ilog2(eq->nent));
@@ -507,10 +510,32 @@ static int cq_err_event_notifier(struct
 notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-static u64 gather_async_events_mask(struct mlx5_core_dev *dev)
+static void gather_async_events_from_cap(struct mlx5_core_dev *dev,
+					 u64 mask[4])
+{
+	__be64 *user_unaffiliated_events;
+	__be64 *user_affiliated_events;
+	int i;
+
+	user_affiliated_events =
+		MLX5_CAP_DEV_EVENT(dev, user_affiliated_events);
+	user_unaffiliated_events =
+		MLX5_CAP_DEV_EVENT(dev, user_unaffiliated_events);
+
+	for (i = 0; i < 4; i++)
+		mask[i] = be64_to_cpu(user_affiliated_events[i] |
+				      user_unaffiliated_events[i]);
+}
+
+static void gather_async_events_mask(struct mlx5_core_dev *dev, u64
 mask[4])
 {
 	u64 async_event_mask = MLX5_ASYNC_EVENT_MASK;
 
+	if (MLX5_CAP_GEN(dev, event_cap)) {
+		gather_async_events_from_cap(dev, mask);
+		return;
+	}
+
 	if (MLX5_VPORT_MANAGER(dev))

async_event_mask |= (1ull <<
MLX5_EVENT_TYPE_NIC_VPORT_CHANGE);

drivers/net/ethernet/mellanox/mlx5/core/eq.c +1 -1
@@ -544,7 +569,7 @@ static u64 gather_async_events_mask(struct
 mlx5_core_dev *dev)
 		async_event_mask |=
 			(1ull <<
 MLX5_EVENT_TYPE_ESW_FUNCTIONS_CHANGED);
 
-	return async_event_mask;
+	mask[0] = async_event_mask;
 }

static int create_async_eqs(struct mlx5_core_dev *dev)

drivers/net/ethernet/mellanox/mlx5/core/eq.c static int create_async_eqs(struct mlx5_core_dev · line 592

Quoted 10 lines
drivers/net/ethernet/mellanox/mlx5/core/eq.c +3 -1
@@ -559,9 +584,11 @@ static int create_async_eqs(struct mlx5_core_dev
 *dev)
 	table->cmd_eq.irq_nb.notifier_call = mlx5_eq_async_int;
 	param = (struct mlx5_eq_param) {
 		.irq_index = 0,
-		.mask = 1ull << MLX5_EVENT_TYPE_CMD,
+		.mask = {1ull << MLX5_EVENT_TYPE_CMD},
 		.nent = MLX5_NUM_CMD_EQE,
 	};
+
+	param.mask[0] = 1ull << MLX5_EVENT_TYPE_CMD;

No need to setup the mask twice,
here and everywhere in this patch, pick one approach to set it.

Quoted 151 lines
drivers/net/ethernet/mellanox/mlx5/core/eq.c +0 -0
 	err = create_async_eq(dev, &table->cmd_eq.core, &param);
 	if (err) {

mlx5_core_warn(dev, "failed to create cmd EQ %d\n",
err);

drivers/net/ethernet/mellanox/mlx5/core/eq.c +1 -1
@@ -577,9 +604,9 @@ static int create_async_eqs(struct mlx5_core_dev
 *dev)
 	table->async_eq.irq_nb.notifier_call = mlx5_eq_async_int;
 	param = (struct mlx5_eq_param) {
 		.irq_index = 0,
-		.mask = gather_async_events_mask(dev),
 		.nent = MLX5_NUM_ASYNC_EQE,
 	};
+	gather_async_events_mask(dev, param.mask);
 	err = create_async_eq(dev, &table->async_eq.core, &param);
 	if (err) {

mlx5_core_warn(dev, "failed to create async EQ %d\n",
err);

drivers/net/ethernet/mellanox/mlx5/core/eq.c +3 -1
@@ -595,9 +622,11 @@ static int create_async_eqs(struct mlx5_core_dev
 *dev)
 	table->pages_eq.irq_nb.notifier_call = mlx5_eq_async_int;
 	param = (struct mlx5_eq_param) {
 		.irq_index = 0,
-		.mask =  1 << MLX5_EVENT_TYPE_PAGE_REQUEST,
+		.mask = {1 << MLX5_EVENT_TYPE_PAGE_REQUEST},
 		.nent = /* TODO: sriov max_vf + */ 1,
 	};
+
+	param.mask[0] = 1ull << MLX5_EVENT_TYPE_PAGE_REQUEST;
 	err = create_async_eq(dev, &table->pages_eq.core, &param);
 	if (err) {

mlx5_core_warn(dev, "failed to create pages EQ %d\n",
err);

drivers/net/ethernet/mellanox/mlx5/core/eq.c +1 -1
@@ -789,7 +818,7 @@ static int create_comp_eqs(struct mlx5_core_dev
 *dev)
 		eq->irq_nb.notifier_call = mlx5_eq_comp_int;
 		param = (struct mlx5_eq_param) {
 			.irq_index = vecidx,
-			.mask = 0,
+			.mask = {0},
 			.nent = nent,
 		};
drivers/net/ethernet/mellanox/mlx5/core/eq.c +0 -0
 		err = create_map_eq(dev, &eq->core, &param);
drivers/net/ethernet/mellanox/mlx5/core/fw.c +6 -0
index 1ab6f7e3bec6..05367f15c3a7 100644
@@ -202,6 +202,12 @@ int mlx5_query_hca_caps(struct mlx5_core_dev
 *dev)
 			return err;
 	}
 
+	if (MLX5_CAP_GEN(dev, event_cap)) {
+		err = mlx5_core_get_caps(dev, MLX5_CAP_DEV_EVENT);
+		if (err)
+			return err;
+	}
+
 	return 0;
 }
include/linux/mlx5/device.h +1 -1
index 5e760067ac41..0d1abe097627 100644
@@ -351,7 +351,7 @@ enum mlx5_event {
 
 	MLX5_EVENT_TYPE_DEVICE_TRACER      = 0x26,
 
-	MLX5_EVENT_TYPE_MAX                =
 MLX5_EVENT_TYPE_DEVICE_TRACER + 1,
+	MLX5_EVENT_TYPE_MAX                = 0x100,
 };
 

enum {

include/linux/mlx5/device.h +4 -0
@@ -1077,6 +1077,7 @@ enum mlx5_cap_type {
 	MLX5_CAP_DEBUG,
 	MLX5_CAP_RESERVED_14,
 	MLX5_CAP_DEV_MEM,
+	MLX5_CAP_DEV_EVENT = 0x14,
 	/* NUM OF CAP Types */
 	MLX5_CAP_NUM
 };
@@ -1255,6 +1256,9 @@ enum mlx5_qcam_feature_groups {
 #define MLX5_CAP64_DEV_MEM(mdev, cap)\
 	MLX5_GET64(device_mem_cap, mdev-
 >caps.hca_cur[MLX5_CAP_DEV_MEM], cap)
 
+#define MLX5_CAP_DEV_EVENT(mdev, cap)\
+	MLX5_ADDR_OF(device_event_cap, (mdev)-
 >caps.hca_cur[MLX5_CAP_DEV_EVENT], cap)
+
 enum {
include/linux/mlx5/device.h +0 -0
 	MLX5_CMD_STAT_OK			= 0x0,
 	MLX5_CMD_STAT_INT_ERR			= 0x1,
include/linux/mlx5/eq.h +3 -1
index 70e16dcfb4c4..202df2e5fe8c 100644
@@ -15,7 +15,9 @@ struct mlx5_core_dev;
 struct mlx5_eq_param {
 	u8             irq_index;
 	int            nent;
-	u64            mask;
+	u64            mask[4];
+	void          *context;
+	irq_handler_t  handler;
 };
 
 struct mlx5_eq *
include/linux/mlx5/mlx5_ifc.h +10 -3
index 16348528fef6..3ef716c054c2 100644
@@ -823,6 +823,12 @@ struct mlx5_ifc_device_mem_cap_bits {
 	u8         reserved_at_180[0x680];
 };
 
+struct mlx5_ifc_device_event_cap_bits {
+	u8         user_affiliated_events[4][0x40];
+
+	u8         user_unaffiliated_events[4][0x40];
+};
+
 enum {
 	MLX5_ATOMIC_CAPS_ATOMIC_SIZE_QP_1_BYTE     = 0x0,
 	MLX5_ATOMIC_CAPS_ATOMIC_SIZE_QP_2_BYTES    = 0x2,
@@ -980,7 +986,8 @@ struct mlx5_ifc_cmd_hca_cap_bits {
 
 	u8         log_max_srq_sz[0x8];
 	u8         log_max_qp_sz[0x8];
-	u8         reserved_at_90[0x8];
+	u8         event_cap[0x1];
+	u8         reserved_at_91[0x7];
 	u8         prio_tag_required[0x1];
 	u8         reserved_at_99[0x2];
 	u8         log_max_qp[0x5];
@@ -7364,9 +7371,9 @@ struct mlx5_ifc_create_eq_in_bits {
 
 	u8         reserved_at_280[0x40];
 
-	u8         event_bitmask[0x40];
+	u8         event_bitmask[4][0x40];
 
-	u8         reserved_at_300[0x580];
+	u8         reserved_at_3c0[0x4c0];
 
 	u8         pas[0][0x40];
 };
#34 lore raw
In reply to Saeed Mahameed
Quoted 15 lines · Saeed Mahameed
Quoted 4 lines · Leon Romanovsky
Quoted 3 lines · Saeed Mahameed
Quoted 2 lines · Leon Romanovsky

From: Leon Romanovsky <[email protected]>

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines

Changelog:
v0 -> v1:

Normally 1st submission is V1 and 2nd is V2.
so this should have been v1->v2.

"Normally" depends on the language you are using. In C, everything starts from 0, including version of patches :).

Quoted 8 lines

You are wrong:
quoting: https://kernelnewbies.org/PatchTipsAndTricks

"For example, if you're sending the second revision of a patch, you should use [PATCH v2]."

now don't tell me that second revision is actually 3rd revision or 1st is 2nd :)..

:)

If you don't mind, I will stick to common sense (v0, v1, v2 ...) and official kernel documentation, which mentions existence of v1.

https://elixir.bootlin.com/linux/latest/source/Documentation/process/submitting-patches.rst#L682

Quoted 6 lines
Quoted 6 lines
Quoted 4 lines

For mlx5-next patches:

Thanks