[PATCH 1/5] rds: ib: drop unnecessary rdma_reject

Subsystems: networking [general], rds - reliable datagram sockets, the rest

STALE3432d

7 messages, 4 authors, 2017-03-09 · open the first message on its own page

[PATCH 1/5] rds: ib: drop unnecessary rdma_reject

From: Zhu Yanjun <hidden>
Date: 2017-03-09 07:26:41

When rdma_accept fails, rdma_reject is called in it. As such, it is
not necessary to execute rdma_reject again.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_cm.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c
index ce3775a..eca3d5f 100644
--- a/net/rds/ib_cm.c
+++ b/net/rds/ib_cm.c

@@ -677,8 +677,7 @@ int rds_ib_cm_handle_connect(struct rdma_cm_id *cm_id,
 		event->param.conn.initiator_depth);
 
 	/* rdma_accept() calls rdma_reject() internally if it fails */
-	err = rdma_accept(cm_id, &conn_param);
-	if (err)
+	if (rdma_accept(cm_id, &conn_param))
 		rds_ib_conn_error(conn, "rdma_accept failed (%d)\n", err);
 
 out:

-- 
2.7.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 3/5] rds: ib: remove redundant ib_dealloc_fmr

From: Zhu Yanjun <hidden>
Date: 2017-03-09 07:27:07

The function ib_dealloc_fmr will never be called. As such, it should
be removed.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_fmr.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c
index 4fe8f4f..b807df6 100644
--- a/net/rds/ib_fmr.c
+++ b/net/rds/ib_fmr.c

@@ -78,12 +78,10 @@ struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev, int npages)
 	return ibmr;
 
 out_no_cigar:
-	if (ibmr) {
-		if (fmr->fmr)
-			ib_dealloc_fmr(fmr->fmr);
+	if (ibmr)
 		kfree(ibmr);
-	}
 	atomic_dec(&pool->item_count);
+
 	return ERR_PTR(err);
 }

-- 
2.7.4

Re: [PATCH 3/5] rds: ib: remove redundant ib_dealloc_fmr

From: Johannes Thumshirn <hidden>
Date: 2017-03-09 08:17:44

On 03/09/2017 08:26 AM, Zhu Yanjun wrote:

quoted hunk

The function ib_dealloc_fmr will never be called. As such, it should
be removed.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_fmr.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c
index 4fe8f4f..b807df6 100644
--- a/net/rds/ib_fmr.c
+++ b/net/rds/ib_fmr.c

@@ -78,12 +78,10 @@ struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev, int npages)
 	return ibmr;
 
 out_no_cigar:
-	if (ibmr) {
-		if (fmr->fmr)
-			ib_dealloc_fmr(fmr->fmr);
+	if (ibmr)
 		kfree(ibmr);

kfree() already does a NULL check (see
http://lxr.free-electrons.com/source/mm/slab.c#L3811) so no need to do
it here.



-- 
Johannes Thumshirn                                          Storage
jthumshirn@suse.de                                +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

[PATCH 5/5] rds: ib: unmap the scatter/gather list when error

From: Zhu Yanjun <hidden>
Date: 2017-03-09 07:27:10

When some errors occur, the scatter/gather list mapped to DMA addresses
should be handled.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_fmr.c | 26 +++++++++++++++++++-------
 1 file changed, 19 insertions(+), 7 deletions(-)

diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c
index 1e00371..c60247f 100644
--- a/net/rds/ib_fmr.c
+++ b/net/rds/ib_fmr.c

@@ -113,29 +113,39 @@ static int rds_ib_map_fmr(struct rds_ib_device *rds_ibdev,
 		u64 dma_addr = ib_sg_dma_address(dev, &scat[i]);
 
 		if (dma_addr & ~PAGE_MASK) {
-			if (i > 0)
+			if (i > 0) {
+				ib_dma_unmap_sg(dev, sg, nents,
+						DMA_BIDIRECTIONAL);
 				return -EINVAL;
-			else
+			} else {
 				++page_cnt;
+			}
 		}
 		if ((dma_addr + dma_len) & ~PAGE_MASK) {
-			if (i < sg_dma_len - 1)
+			if (i < sg_dma_len - 1) {
+				ib_dma_unmap_sg(dev, sg, nents,
+						DMA_BIDIRECTIONAL);
 				return -EINVAL;
-			else
+			} else {
 				++page_cnt;
+			}
 		}
 
 		len += dma_len;
 	}
 
 	page_cnt += len >> PAGE_SHIFT;
-	if (page_cnt > ibmr->pool->fmr_attr.max_pages)
+	if (page_cnt > ibmr->pool->fmr_attr.max_pages) {
+		ib_dma_unmap_sg(dev, sg, nents, DMA_BIDIRECTIONAL);
 		return -EINVAL;
+	}
 
 	dma_pages = kmalloc_node(sizeof(u64) * page_cnt, GFP_ATOMIC,
 				 rdsibdev_to_node(rds_ibdev));
-	if (!dma_pages)
+	if (!dma_pages) {
+		ib_dma_unmap_sg(dev, sg, nents, DMA_BIDIRECTIONAL);
 		return -ENOMEM;
+	}
 
 	page_cnt = 0;
 	for (i = 0; i < sg_dma_len; ++i) {

@@ -148,8 +158,10 @@ static int rds_ib_map_fmr(struct rds_ib_device *rds_ibdev,
 	}
 
 	ret = ib_map_phys_fmr(fmr->fmr, dma_pages, page_cnt, io_addr);
-	if (ret)
+	if (ret) {
+		ib_dma_unmap_sg(dev, sg, nents, DMA_BIDIRECTIONAL);
 		goto out;
+	}
 
 	/* Success - we successfully remapped the MR, so we can
 	 * safely tear down the old mapping.

-- 
2.7.4

[PATCH 4/5] rds: ib: add the static type to the function

From: Zhu Yanjun <hidden>
Date: 2017-03-09 07:27:11

The function rds_ib_map_fmr is used only in the ib_fmr.c
file. As such, the static type is added to limit it in this file.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_fmr.c | 5 +++--
 net/rds/ib_mr.h  | 2 --
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c
index b807df6..1e00371 100644
--- a/net/rds/ib_fmr.c
+++ b/net/rds/ib_fmr.c

@@ -85,8 +85,9 @@ struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev, int npages)
 	return ERR_PTR(err);
 }
 
-int rds_ib_map_fmr(struct rds_ib_device *rds_ibdev, struct rds_ib_mr *ibmr,
-		   struct scatterlist *sg, unsigned int nents)
+static int rds_ib_map_fmr(struct rds_ib_device *rds_ibdev,
+			  struct rds_ib_mr *ibmr, struct scatterlist *sg,
+			  unsigned int nents)
 {
 	struct ib_device *dev = rds_ibdev->dev;
 	struct rds_ib_fmr *fmr = &ibmr->u.fmr;

diff --git a/net/rds/ib_mr.h b/net/rds/ib_mr.h
index 5d6e98a..0ea4ab0 100644
--- a/net/rds/ib_mr.h
+++ b/net/rds/ib_mr.h

@@ -125,8 +125,6 @@ void rds_ib_mr_exit(void);
 void __rds_ib_teardown_mr(struct rds_ib_mr *);
 void rds_ib_teardown_mr(struct rds_ib_mr *);
 struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *, int);
-int rds_ib_map_fmr(struct rds_ib_device *, struct rds_ib_mr *,
-		   struct scatterlist *, unsigned int);
 struct rds_ib_mr *rds_ib_reuse_mr(struct rds_ib_mr_pool *);
 int rds_ib_flush_mr_pool(struct rds_ib_mr_pool *, int, struct rds_ib_mr **);
 struct rds_ib_mr *rds_ib_reg_fmr(struct rds_ib_device *, struct scatterlist *,

-- 
2.7.4

Re: [PATCH 4/5] rds: ib: add the static type to the function

From: Santosh Shilimkar <hidden>
Date: 2017-03-09 16:54:32

On 3/8/2017 11:26 PM, Zhu Yanjun wrote:

The function rds_ib_map_fmr is used only in the ib_fmr.c
file. As such, the static type is added to limit it in this file.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_fmr.c | 5 +++--
 net/rds/ib_mr.h  | 2 --
 2 files changed, 3 insertions(+), 4 deletions(-)

Acked-by: Santosh Shilimkar <redacted>

Re: [PATCH 1/5] rds: ib: drop unnecessary rdma_reject

From: Santosh Shilimkar <hidden>
Date: 2017-03-09 16:49:08

On 3/8/2017 11:26 PM, Zhu Yanjun wrote:

quoted hunk

When rdma_accept fails, rdma_reject is called in it. As such, it is
not necessary to execute rdma_reject again.

Cc: Joe Jin <redacted>
Cc: Junxiao Bi <redacted>
Signed-off-by: Zhu Yanjun <redacted>
---
 net/rds/ib_cm.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c
index ce3775a..eca3d5f 100644
--- a/net/rds/ib_cm.c
+++ b/net/rds/ib_cm.c

@@ -677,8 +677,7 @@ int rds_ib_cm_handle_connect(struct rdma_cm_id *cm_id,
 		event->param.conn.initiator_depth);

 	/* rdma_accept() calls rdma_reject() internally if it fails */
-	err = rdma_accept(cm_id, &conn_param);
-	if (err)
+	if (rdma_accept(cm_id, &conn_param))
 		rds_ib_conn_error(conn, "rdma_accept failed (%d)\n", err);

There are couple of consumer reject reasons needs to be conveyed.
Current code don't pass them, but downstream code has them and
hence I kept the code as is. Its good to avoid the reject on
current code so am fine with the change.

Acked-by: Santosh Shilimkar <redacted>

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help