On 12/21/12 19:01, Mike Marciniszyn wrote:
> 0b088e00 ("RDS: Use page_remainder_alloc() for recv bufs")
> added uses of sg_dma_len() and sg_dma_address(). This makes
> RDS DOA with the qib driver.
>
> IB ulps should use ib_sg_dma_len() and ib_sg_dma_address
> respectively since some HCAs overload ib_sg_dma* operations.
>
> Signed-off-by: Mike Marciniszyn <[email protected]>
> ---
> net/rds/ib_recv.c | 9 ++++++---
> 1 files changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/net/rds/ib_recv.c b/net/rds/ib_recv.c
> index 8c5bc85..8eb9501 100644
> --- a/net/rds/ib_recv.c
> +++ b/net/rds/ib_recv.c
> @@ -339,8 +339,8 @@ static int rds_ib_recv_refill_one(struct rds_connection
> *conn,
> sge->length = sizeof(struct rds_header);
>
> sge = &recv->r_sge[1];
> - sge->addr = sg_dma_address(&recv->r_frag->f_sg);
> - sge->length = sg_dma_len(&recv->r_frag->f_sg);
> + sge->addr = ib_sg_dma_address(ic->i_cm_id->device, &recv->r_frag->f_sg);
> + sge->length = ib_sg_dma_len(ic->i_cm_id->device, &recv->r_frag->f_sg);
>
> ret = 0;
> out:
> @@ -381,7 +381,10 @@ void rds_ib_recv_refill(struct rds_connection *conn, int
> prefill)
> ret = ib_post_recv(ic->i_cm_id->qp, &recv->r_wr, &failed_wr);
> rdsdebug("recv %p ibinc %p page %p addr %lu ret %d\n", recv,
> recv->r_ibinc, sg_page(&recv->r_frag->f_sg),
> - (long) sg_dma_address(&recv->r_frag->f_sg), ret);
> + (long) ib_sg_dma_address(
> + ic->i_cm_id->device,
> + &recv->r_frag->f_sg),
> + ret);
> if (ret) {
> rds_ib_conn_error(conn, "recv post on "
> "%pI4 returned %d, disconnecting and "
The above patch will slow down RDS for anyone who is not using QLogic
HCA's, isn' it ? How about replacing the above patch with the
(untested) patch below ?
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index 8784486..ec9adf8 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -2588,16 +2588,6 @@ static void ehca_dma_unmap_sg(struct ib_device *dev,
struct scatterlist *sg,
/* This is only a stub; nothing to be done here */
}
-static u64 ehca_dma_address(struct ib_device *dev, struct scatterlist *sg)
-{
- return sg->dma_address;
-}
-
-static unsigned int ehca_dma_len(struct ib_device *dev, struct scatterlist *sg)
-{
- return sg->length;
-}
-
static void ehca_dma_sync_single_for_cpu(struct ib_device *dev, u64 addr,
size_t size,
enum dma_data_direction dir)
@@ -2650,8 +2640,6 @@ struct ib_dma_mapping_ops ehca_dma_mapping_ops = {
.unmap_page = ehca_dma_unmap_page,
.map_sg = ehca_dma_map_sg,
.unmap_sg = ehca_dma_unmap_sg,
- .dma_address = ehca_dma_address,
- .dma_len = ehca_dma_len,
.sync_single_for_cpu = ehca_dma_sync_single_for_cpu,
.sync_single_for_device = ehca_dma_sync_single_for_device,
.alloc_coherent = ehca_dma_alloc_coherent,
diff --git a/drivers/infiniband/hw/qib/qib_dma.c
b/drivers/infiniband/hw/qib/qib_dma.c
index 2920bb3..3620c6c 100644
--- a/drivers/infiniband/hw/qib/qib_dma.c
+++ b/drivers/infiniband/hw/qib/qib_dma.c
@@ -104,7 +104,12 @@ static int qib_map_sg(struct ib_device *dev, struct
scatterlist *sgl,
for_each_sg(sgl, sg, nents, i) {
addr = (u64) page_address(sg_page(sg));
/* TODO: handle highmem pages */
- if (!addr) {
+ if (addr) {
+ sg->dma_address = addr + sg->offset;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+ sg->dma_length = sg->length;
+#endif
+ } else {
ret = 0;
break;
}
@@ -119,21 +124,6 @@ static void qib_unmap_sg(struct ib_device *dev,
BUG_ON(!valid_dma_direction(direction));
}
-static u64 qib_sg_dma_address(struct ib_device *dev, struct scatterlist *sg)
-{
- u64 addr = (u64) page_address(sg_page(sg));
-
- if (addr)
- addr += sg->offset;
- return addr;
-}
-
-static unsigned int qib_sg_dma_len(struct ib_device *dev,
- struct scatterlist *sg)
-{
- return sg->length;
-}
-
static void qib_sync_single_for_cpu(struct ib_device *dev, u64 addr,
size_t size, enum dma_data_direction dir)
{
@@ -173,8 +163,6 @@ struct ib_dma_mapping_ops qib_dma_mapping_ops = {
.unmap_page = qib_dma_unmap_page,
.map_sg = qib_map_sg,
.unmap_sg = qib_unmap_sg,
- .dma_address = qib_sg_dma_address,
- .dma_len = qib_sg_dma_len,
.sync_single_for_cpu = qib_sync_single_for_cpu,
.sync_single_for_device = qib_sync_single_for_device,
.alloc_coherent = qib_dma_alloc_coherent,
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 46bc045..54a0689 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -1049,10 +1049,6 @@ struct ib_dma_mapping_ops {
void (*unmap_sg)(struct ib_device *dev,
struct scatterlist *sg, int nents,
enum dma_data_direction direction);
- u64 (*dma_address)(struct ib_device *dev,
- struct scatterlist *sg);
- unsigned int (*dma_len)(struct ib_device *dev,
- struct scatterlist *sg);
void (*sync_single_for_cpu)(struct ib_device *dev,
u64 dma_handle,
size_t size,
@@ -1863,12 +1859,13 @@ static inline void ib_dma_unmap_sg_attrs(struct
ib_device *dev,
* ib_sg_dma_address - Return the DMA address from a scatter/gather entry
* @dev: The device for which the DMA addresses were created
* @sg: The scatter/gather entry
+ *
+ * Note: this function is obsolete. To do: change all occurrences of
+ * ib_sg_dma_address() into sg_dma_address().
*/
static inline u64 ib_sg_dma_address(struct ib_device *dev,
struct scatterlist *sg)
{
- if (dev->dma_ops)
- return dev->dma_ops->dma_address(dev, sg);
return sg_dma_address(sg);
}
@@ -1876,12 +1873,13 @@ static inline u64 ib_sg_dma_address(struct ib_device
*dev,
* ib_sg_dma_len - Return the DMA length from a scatter/gather entry
* @dev: The device for which the DMA addresses were created
* @sg: The scatter/gather entry
+ *
+ * Note: this function is obsolete. To do: change all occurrences of
+ * ib_sg_dma_len() into sg_dma_len().
*/
static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
struct scatterlist *sg)
{
- if (dev->dma_ops)
- return dev->dma_ops->dma_len(dev, sg);
return sg_dma_len(sg);
}
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
