Skip to contentSkip to conversation

Optimize SGL registration

PATCH In review Leon Romanovsky 3 participants 9 messages Oct 6, 2019 – Oct 7, 2019

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover Optimize SGL registration

From: Leon Romanovsky <[email protected]>

Hi,

This series from Yamin implements long standing "TODO" existed in rw.c.

Thanks

Yamin Friedman (3):
net/mlx5: Expose optimal performance scatter entries capability RDMA/mlx5: Add capability for max sge to get optimized performance RDMA/rw: Support threshold for registration vs scattering to local
pages

 drivers/infiniband/core/rw.c      | 12 ++++++------
 drivers/infiniband/hw/mlx5/main.c |  2 ++
 include/linux/mlx5/mlx5_ifc.h     |  2 +-
 include/rdma/ib_verbs.h           |  2 ++
 4 files changed, 11 insertions(+), 7 deletions(-)

--
2.20.1

#1 lore raw

PATCH 1/3 net/mlx5: Expose optimal performance scatter entries capability

From: Yamin Friedman <[email protected]>

Expose maximum scatter entries per RDMA READ for optimal performance.

---
 include/linux/mlx5/mlx5_ifc.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/mlx5/mlx5_ifc.h +1 -1
index 138c50d5a353..c0bfb1d90dd2 100644
@@ -1153,7 +1153,7 @@ struct mlx5_ifc_cmd_hca_cap_bits {
 	u8         log_max_srq[0x5];
 	u8         reserved_at_b0[0x10];
 
-	u8         reserved_at_c0[0x8];
+	u8         max_sgl_for_optimized_performance[0x8];
 	u8         log_max_cq_sz[0x8];
 	u8         reserved_at_d0[0xb];
 	u8         log_max_cq[0x5];
Signature · 1 line
2.20.1
#2 lore raw

PATCH 2/3 RDMA/mlx5: Add capability for max sge to get optimized performance

In reply to Leon Romanovsky

From: Yamin Friedman <[email protected]>

Allows the IB device to provide a value of maximum scatter gather entries per RDMA READ.

In certain cases it may be preferable for a device to perform UMR memory registration rather than have many scatter entries in a single RDMA READ. This provides a significant performance increase in devices capable of using different memory registration schemes based on the number of scatter gather entries. This general capability allows each device vendor to fine tune when it is better to use memory registration.

---
 drivers/infiniband/hw/mlx5/main.c | 2 ++
 include/rdma/ib_verbs.h           | 2 ++
 2 files changed, 4 insertions(+)
drivers/infiniband/hw/mlx5/main.c +2 -0
index fa23c8e7043b..39d54e285ae9 100644
@@ -1012,6 +1012,8 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		1 << MLX5_CAP_GEN(mdev, log_max_klm_list_size);
 	props->max_pi_fast_reg_page_list_len =
 		props->max_fast_reg_page_list_len / 2;
+	props->max_sgl_rd =
+		MLX5_CAP_GEN(mdev, max_sgl_for_optimized_performance);
 	get_atomic_caps_qp(dev, props);
 	props->masked_atomic_cap   = IB_ATOMIC_NONE;
 	props->max_mcast_grp	   = 1 << MLX5_CAP_GEN(mdev, log_max_mcg);
include/rdma/ib_verbs.h +2 -0
index 4f671378dbfc..60fd98a9b7e8 100644
@@ -445,6 +445,8 @@ struct ib_device_attr {
 	struct ib_tm_caps	tm_caps;
 	struct ib_cq_caps       cq_caps;
 	u64			max_dm_size;
+	/* Max entries for sgl for optimized performance per READ */
+	u32			max_sgl_rd;
 };
 
 enum ib_mtu {
Signature · 1 line
2.20.1
#3 lore raw

Re: [PATCH rdma-next 2/3] RDMA/mlx5: Add capability for max sge to get optimized performance

include/rdma/ib_verbs.h struct ib_device_attr { · line 449

Quoted 10 lines · Leon Romanovsky
include/rdma/ib_verbs.h +2 -0
index 4f671378dbfc..60fd98a9b7e8 100644
@@ -445,6 +445,8 @@ struct ib_device_attr {
 	struct ib_tm_caps	tm_caps;
 	struct ib_cq_caps       cq_caps;
 	u64			max_dm_size;
+	/* Max entries for sgl for optimized performance per READ */
+	u32			max_sgl_rd;

This either needs to go into what current is patch 3 or an entirely separate one.

#4 lore raw

Re: [PATCH rdma-next 2/3] RDMA/mlx5: Add capability for max sge to get optimized performance

Quoted 14 lines · Christoph Hellwig

include/rdma/ib_verbs.h struct ib_device_attr { · line 449

Quoted 10 lines · Leon Romanovsky
include/rdma/ib_verbs.h +2 -0
index 4f671378dbfc..60fd98a9b7e8 100644
@@ -445,6 +445,8 @@ struct ib_device_attr {
 	struct ib_tm_caps	tm_caps;
 	struct ib_cq_caps       cq_caps;
 	u64			max_dm_size;
+	/* Max entries for sgl for optimized performance per READ */
+	u32			max_sgl_rd;

This either needs to go into what current is patch 3 or an entirely separate one.

I'll reshuffle.

Thanks

#5 lore raw

PATCH 3/3 RDMA/rw: Support threshold for registration vs scattering to local pages

In reply to Leon Romanovsky

From: Yamin Friedman <[email protected]>

If there are more scatter entries than the recommended limit provided by the ib device, UMR registration is used. This will provide optimal performance when performing large RDMA READs over devices that advertise the threshold capability.

With ConnectX-5 running NVMeoF RDMA with FIO single QP 128KB writes:
Without use of cap: 70Gb/sec
With use of cap: 84Gb/sec

---
 drivers/infiniband/core/rw.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/infiniband/core/rw.c +6 -6
index 5337393d4dfe..ecff40efcb88 100644
@@ -20,9 +20,7 @@ module_param_named(force_mr, rdma_rw_force_mr, bool, 0);
 MODULE_PARM_DESC(force_mr, "Force usage of MRs for RDMA READ/WRITE operations");
 
 /*
- * Check if the device might use memory registration.  This is currently only
- * true for iWarp devices. In the future we can hopefully fine tune this based
- * on HCA driver input.
+ * Check if the device might use memory registration.
  */
 static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
 {
@@ -30,6 +28,8 @@ static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
 		return true;
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd)
+		return true;
 	return false;
 }
 
@@ -37,9 +37,6 @@ static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
  * Check if the device will use memory registration for this RW operation.
  * We currently always use memory registrations for iWarp RDMA READs, and
  * have a debug option to force usage of MRs.
- *
- * XXX: In the future we can hopefully fine tune this based on HCA driver
- * input.
  */
 static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num,
 		enum dma_data_direction dir, int dma_nents)
@@ -48,6 +45,9 @@ static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num,
 		return true;
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd && dir == DMA_FROM_DEVICE
+	    && dma_nents > dev->attrs.max_sgl_rd)
+		return true;
 	return false;
 }
 
Signature · 1 line
2.20.1
#6 lore raw

Re: [PATCH rdma-next 3/3] RDMA/rw: Support threshold for registration vs scattering to local pages

Quoted 6 lines

/*
- * Check if the device might use memory registration. This is currently only
- * true for iWarp devices. In the future we can hopefully fine tune this based
- * on HCA driver input.
+ * Check if the device might use memory registration.
*/

Please keep the important bits of this comments instead of just removing them.

drivers/infiniband/core/rw.c static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num) · line 32

Quoted 7 lines

{

drivers/infiniband/core/rw.c +2 -0
@@ -30,6 +28,8 @@ static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
 		return true;
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd)
+		return true;

Logically this should go before the rdma_rw_force_mr check.

drivers/infiniband/core/rw.c static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num, · line 49

Quoted 4 lines
drivers/infiniband/core/rw.c +2 -0
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd && dir == DMA_FROM_DEVICE
+	    && dma_nents > dev->attrs.max_sgl_rd)

Wrong indendation. The && belongs on the first line. And again, this logically belongs before the rdma_rw_force_mr check.

#7 lore raw

Re: [PATCH rdma-next 3/3] RDMA/rw: Support threshold for registration vs scattering to local pages

Quoted 20 lines · Christoph Hellwig
Quoted 6 lines

/*
- * Check if the device might use memory registration. This is currently only
- * true for iWarp devices. In the future we can hopefully fine tune this based
- * on HCA driver input.
+ * Check if the device might use memory registration.
*/

Please keep the important bits of this comments instead of just removing them.

drivers/infiniband/core/rw.c static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num) · line 32

Quoted 7 lines

{

drivers/infiniband/core/rw.c +2 -0
@@ -30,6 +28,8 @@ static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
 		return true;
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd)
+		return true;

Logically this should go before the rdma_rw_force_mr check.

Quoted 7 lines

drivers/infiniband/core/rw.c static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num, · line 49

Quoted 4 lines
drivers/infiniband/core/rw.c +2 -0
 	if (unlikely(rdma_rw_force_mr))
 		return true;
+	if (dev->attrs.max_sgl_rd && dir == DMA_FROM_DEVICE
+	    && dma_nents > dev->attrs.max_sgl_rd)

Wrong indendation. The && belongs on the first line. And again, this logically belongs before the rdma_rw_force_mr check.

I'll fix.

Thanks

#8 lore raw

Re: [PATCH rdma-next 3/3] RDMA/rw: Support threshold for registration vs scattering to local pages

Quoted 25 lines · Leon Romanovsky
Quoted 18 lines · Christoph Hellwig
Quoted 6 lines

/*
- * Check if the device might use memory registration. This is currently only
- * true for iWarp devices. In the future we can hopefully fine tune this based
- * on HCA driver input.
+ * Check if the device might use memory registration.
*/

Please keep the important bits of this comments instead of just removing them.

drivers/infiniband/core/rw.c static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num) · line 32

Quoted 7 lines
  {
drivers/infiniband/core/rw.c +2 -0
@@ -30,6 +28,8 @@ static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num)
  		return true;
  	if (unlikely(rdma_rw_force_mr))
  		return true;
+	if (dev->attrs.max_sgl_rd)
+		return true;

Logically this should go before the rdma_rw_force_mr check.

Quoted 6 lines

drivers/infiniband/core/rw.c static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num, · line 49

Quoted 4 lines
drivers/infiniband/core/rw.c +2 -0
  	if (unlikely(rdma_rw_force_mr))
  		return true;
+	if (dev->attrs.max_sgl_rd && dir == DMA_FROM_DEVICE
+	    && dma_nents > dev->attrs.max_sgl_rd)

Wrong indendation. The && belongs on the first line. And again, this logically belongs before the rdma_rw_force_mr check.

Quoted 3 lines

I'll fix.

Thanks

The above comments looks reasonable.

Nice optimization Yamin.