On 2/12/2019 6:03 PM, Christoph Hellwig wrote:
On Thu, Feb 07, 2019 at 07:33:22PM +0200, Max Gurtovoy wrote:
This element will describe the needed characteristics for the signature
operation per signature enabled memory region (type IB_MR_TYPE_PI).
Signed-off-by: Max Gurtovoy <maxg@xxxxxxxxxxxx>
Signed-off-by: Israel Rukshin <israelr@xxxxxxxxxxxx>
---
drivers/infiniband/core/uverbs_cmd.c | 1 +
drivers/infiniband/core/verbs.c | 13 ++++++++++++-
include/rdma/ib_verbs.h | 2 +-
3 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 72c5a8daf558..699fbcf68d06 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -739,6 +739,7 @@ static int ib_uverbs_reg_mr(struct uverbs_attr_bundle *attrs)
mr->pd = pd;
mr->type = IB_MR_TYPE_MEM_REG;
mr->dm = NULL;
+ mr->sig_attrs = NULL;
Maybe this is an argument that the PI MR support should use a different
structure rathern than ib_mr? That would also remove the need for all
the error checking if the "right" kind of MR is passed in.
In general, this sounds reasonable. But if we do that, we should be
doing that for all the MR types.
I guess you meant to embed ib_mr into
ib_mr_integrity/ib_mr_aligned/ib_mr_gappy (or a much better naming...).
Well, we should consider that as an improvement to the rdma k-verbs API,
regardless this patch set.
Jason/Leon/Doug,
thoughts ?
if needed, we can work on new APIs after merging this series. the series
is already long enough...
