Skip to contentSkip to conversation

Patch "IB/core, RDMA RW API: Do not exceed QP SGE send limit" has been added to the 4.7-stable tree

PATCH [email protected] 1 participant 1 message Aug 18, 2016

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer Applied
#0 lore raw

This is a note to let you know that I've just added the patch titled

IB/core, RDMA RW API: Do not exceed QP SGE send limit

to the 4.7-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
ib-core-rdma-rw-api-do-not-exceed-qp-sge-send-limit.patch
and it can be found in the queue-4.7 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

Quoted 1 line

From 632bc3f65081dd1e2e5394a9161580a0f78e8839 Mon Sep 17 00:00:00 2001

From: Bart Van Assche <[email protected]>
Date: Thu, 21 Jul 2016 13:03:30 -0700
Subject: IB/core, RDMA RW API: Do not exceed QP SGE send limit

From: Bart Van Assche <[email protected]>

commit 632bc3f65081dd1e2e5394a9161580a0f78e8839 upstream.

Compute the SGE limit for RDMA READ and WRITE requests in ib_create_qp(). Use that limit in the RDMA RW API implementation.

Cc Christoph Hellwig <[email protected]>
Cc Sagi Grimberg <[email protected]>
Cc Steve Wise <[email protected]>
Cc Parav Pandit <[email protected]>
Cc Nicholas Bellinger <[email protected]>
Cc Laurence Oberman <[email protected]>
---
 drivers/infiniband/core/rw.c    |   10 ++--------
 drivers/infiniband/core/verbs.c |    9 +++++++++
 include/rdma/ib_verbs.h         |    6 ++++++
 3 files changed, 17 insertions(+), 8 deletions(-)
drivers/infiniband/core/rw.c +2 -8
@@ -58,13 +58,6 @@ static inline bool rdma_rw_io_needs_mr(s
 	return false;
 }
 
-static inline u32 rdma_rw_max_sge(struct ib_device *dev,
-		enum dma_data_direction dir)
-{
-	return dir == DMA_TO_DEVICE ?
-		dev->attrs.max_sge : dev->attrs.max_sge_rd;
-}
-
 static inline u32 rdma_rw_fr_page_list_len(struct ib_device *dev)
 {
 	/* arbitrary limit to avoid allocating gigantic resources */
@@ -186,7 +179,8 @@ static int rdma_rw_init_map_wrs(struct r
 		u64 remote_addr, u32 rkey, enum dma_data_direction dir)
 {
 	struct ib_device *dev = qp->pd->device;
-	u32 max_sge = rdma_rw_max_sge(dev, dir);
+	u32 max_sge = dir == DMA_TO_DEVICE ? qp->max_write_sge :
+		      qp->max_read_sge;
 	struct ib_sge *sge;
 	u32 total_len = 0, i, j;
 
drivers/infiniband/core/verbs.c +9 -0
@@ -814,6 +814,15 @@ struct ib_qp *ib_create_qp(struct ib_pd
 		}
 	}
 
+	/*
+	 * Note: all hw drivers guarantee that max_send_sge is lower than
+	 * the device RDMA WRITE SGE limit but not all hw drivers ensure that
+	 * max_send_sge <= max_sge_rd.
+	 */
+	qp->max_write_sge = qp_init_attr->cap.max_send_sge;
+	qp->max_read_sge = min_t(u32, qp_init_attr->cap.max_send_sge,
+				 device->attrs.max_sge_rd);
+
 	return qp;
 }
 EXPORT_SYMBOL(ib_create_qp);
include/rdma/ib_verbs.h +6 -0
@@ -1428,6 +1428,10 @@ struct ib_srq {
 	} ext;
 };
 
+/*
+ * @max_write_sge: Maximum SGE elements per RDMA WRITE request.
+ * @max_read_sge:  Maximum SGE elements per RDMA READ request.
+ */
 struct ib_qp {
 	struct ib_device       *device;
 	struct ib_pd	       *pd;
@@ -1449,6 +1453,8 @@ struct ib_qp {
 	void                  (*event_handler)(struct ib_event *, void *);
 	void		       *qp_context;
 	u32			qp_num;
+	u32			max_write_sge;
+	u32			max_read_sge;
 	enum ib_qp_type		qp_type;
 };
 

Patches currently in stable-queue which might be from [email protected] are

queue-4.7/dm-set-dmf_suspended-_before_-clearing-dmf_noflush_suspending.patch queue-4.7/ib-core-make-rdma_rw_ctx_init-initialize-all-used-fields.patch queue-4.7/ib-rdmavt-disable-by-default.patch queue-4.7/ib-srpt-limit-the-number-of-sg-elements-per-work-request.patch queue-4.7/ib-core-rdma-rw-api-do-not-exceed-qp-sge-send-limit.patch