[PATCH V3 1/3] nvmet: Add get_mdts op for controllers

[PATCH V3 1/3] nvmet: Add get_mdts op for controllers

[Max Gurtovoy]

Some transports, such as RDMA, would like to set the Maximum Data
Transfer Size (MDTS) according to device/port/ctrl characteristics.
This will enable the transport to set the optimal MDTS according to
controller needs and device capabilities. Add a new nvmet transport
op that is called during ctrl identification. This will not effect
transports that don't implement this option. The return value of the new
op is according to the NVMe spec definition for MDTS.

Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
---

changes from V2:
 - added Reviewed-by signature (Sagi)

changes from V1:
 - renamed set_mdts to get_mdts (Bart)
 - added const for ctrl argument (Bart)
 - updated commit message to explain return value of the new op (Bart)

---
 drivers/nvme/target/admin-cmd.c | 8 ++++++--
 drivers/nvme/target/nvmet.h     | 1 +
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c
index c0aa9c3..b9ec489 100644
--- a/drivers/nvme/target/admin-cmd.c
+++ b/drivers/nvme/target/admin-cmd.c
@@ -369,8 +369,12 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req)
 	/* we support multiple ports, multiples hosts and ANA: */
 	id->cmic = (1 << 0) | (1 << 1) | (1 << 3);
 
-	/* no limit on data transfer sizes for now */
-	id->mdts = 0;
+	/* Limit MDTS according to transport capability */
+	if (ctrl->ops->get_mdts)
+		id->mdts = ctrl->ops->get_mdts(ctrl);
+	else
+		id->mdts = 0;
+
 	id->cntlid = cpu_to_le16(ctrl->cntlid);
 	id->ver = cpu_to_le32(ctrl->subsys->ver);
 
diff --git a/drivers/nvme/target/nvmet.h b/drivers/nvme/target/nvmet.h
index 42ba2dd..421dff3 100644
--- a/drivers/nvme/target/nvmet.h
+++ b/drivers/nvme/target/nvmet.h
@@ -289,6 +289,7 @@ struct nvmet_fabrics_ops {
 			struct nvmet_port *port, char *traddr);
 	u16 (*install_queue)(struct nvmet_sq *nvme_sq);
 	void (*discovery_chg)(struct nvmet_port *port);
+	u8 (*get_mdts)(const struct nvmet_ctrl *ctrl);
 };
 
 #define NVMET_MAX_INLINE_BIOVEC	8
-- 
1.8.3.1


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

[PATCH V3 2/3] nvmet-rdma: Implement get_mdts controller op

[Max Gurtovoy]

Set the maximal data transfer size to be 1MB (currently mdts is
unlimited). This will allow calculating the amount of MR's that
one ctrl should allocate to fulfill it's capabilities.

Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
---

changes from V2:
 - move mdts explenation comment to define entry (Sagi)

changes from V1:
 - renamed nvmet_rdma_set_mdts to nvmet_rdma_get_mdts
 - align to get_mdts callback changes

---
 drivers/nvme/target/rdma.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 37d262a..f47a79b 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -31,6 +31,9 @@
 #define NVMET_RDMA_MAX_INLINE_SGE		4
 #define NVMET_RDMA_MAX_INLINE_DATA_SIZE		max_t(int, SZ_16K, PAGE_SIZE)
 
+/* Assume mpsmin == device_page_size == 4KB */
+#define NVMET_RDMA_MAX_MDTS			8
+
 struct nvmet_rdma_cmd {
 	struct ib_sge		sge[NVMET_RDMA_MAX_INLINE_SGE + 1];
 	struct ib_cqe		cqe;
@@ -1602,6 +1605,11 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
 	}
 }
 
+static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl)
+{
+	return NVMET_RDMA_MAX_MDTS;
+}
+
 static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
 	.owner			= THIS_MODULE,
 	.type			= NVMF_TRTYPE_RDMA,
@@ -1612,6 +1620,7 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
 	.queue_response		= nvmet_rdma_queue_response,
 	.delete_ctrl		= nvmet_rdma_delete_ctrl,
 	.disc_traddr		= nvmet_rdma_disc_port_addr,
+	.get_mdts		= nvmet_rdma_get_mdts,
 };
 
 static void nvmet_rdma_remove_one(struct ib_device *ib_device, void *client_data)
-- 
1.8.3.1


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

[PATCH 3/3] nvmet-rdma: allocate RW ctxs according to mdts

[Max Gurtovoy]

Current nvmet-rdma code allocates MR pool budget based on queue size,
assuming both host and target use the same "max_pages_per_mr" count.
After limiting the mdts value for RDMA controllers, we know the factor
of maximum MR's per IO operation. Thus, make sure MR pool will be
sufficient for the required IO depth and IO size.

That is, say host's SQ size is 100, then the MR pool budget allocated
currently at target will also be 100 MRs. But 100 IO WRITE Requests
with 256 sg_count(IO size above 1MB) require 200 MRs when target's
"max_pages_per_mr" is 128.

Reported-by: Krishnamraju Eraparaju <krishna2@chelsio.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
---

changes from V2:
 - added "Reviewed-by" signature (Sagi)

changes from V1:
 - added "Reviewed-by" signature (Christoph)

---
 drivers/nvme/target/rdma.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index f47a79b..9e1b8c6 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -978,7 +978,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
 {
 	struct ib_qp_init_attr qp_attr;
 	struct nvmet_rdma_device *ndev = queue->dev;
-	int comp_vector, nr_cqe, ret, i;
+	int comp_vector, nr_cqe, ret, i, factor;
 
 	/*
 	 * Spread the io queues across completion vectors,
@@ -1011,7 +1011,9 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
 	qp_attr.qp_type = IB_QPT_RC;
 	/* +1 for drain */
 	qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
-	qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
+	factor = rdma_rw_mr_factor(ndev->device, queue->cm_id->port_num,
+				   1 << NVMET_RDMA_MAX_MDTS);
+	qp_attr.cap.max_rdma_ctxs = queue->send_queue_size * factor;
 	qp_attr.cap.max_send_sge = max(ndev->device->attrs.max_sge_rd,
 					ndev->device->attrs.max_send_sge);
 
-- 
1.8.3.1


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: [PATCH V3 1/3] nvmet: Add get_mdts op for controllers

[Christoph Hellwig]

On Sun, Mar 08, 2020 at 12:55:03PM +0200, Max Gurtovoy wrote:
> Some transports, such as RDMA, would like to set the Maximum Data
> Transfer Size (MDTS) according to device/port/ctrl characteristics.
> This will enable the transport to set the optimal MDTS according to
> controller needs and device capabilities. Add a new nvmet transport
> op that is called during ctrl identification. This will not effect
> transports that don't implement this option. The return value of the new
> op is according to the NVMe spec definition for MDTS.
> 
> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
> Signed-off-by: Israel Rukshin <israelr@mellanox.com>
> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: [PATCH V3 2/3] nvmet-rdma: Implement get_mdts controller op

[Christoph Hellwig]

On Sun, Mar 08, 2020 at 12:55:04PM +0200, Max Gurtovoy wrote:
> Set the maximal data transfer size to be 1MB (currently mdts is
> unlimited). This will allow calculating the amount of MR's that
> one ctrl should allocate to fulfill it's capabilities.
> 
> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: [PATCH V3 1/3] nvmet: Add get_mdts op for controllers

[Keith Busch]

On Sun, Mar 08, 2020 at 12:55:03PM +0200, Max Gurtovoy wrote:
> Some transports, such as RDMA, would like to set the Maximum Data
> Transfer Size (MDTS) according to device/port/ctrl characteristics.
> This will enable the transport to set the optimal MDTS according to
> controller needs and device capabilities. Add a new nvmet transport
> op that is called during ctrl identification. This will not effect
> transports that don't implement this option. The return value of the new
> op is according to the NVMe spec definition for MDTS.
> 
> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
> Signed-off-by: Israel Rukshin <israelr@mellanox.com>
> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

Series queued up for 5.7.

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme