Skip to content

verbs: Add new unordered MR access flag - #1780

Open
yishaih wants to merge 4 commits into
linux-rdma:masterfrom
yishaih:verbs_misc
Open

verbs: Add new unordered MR access flag#1780
yishaih wants to merge 4 commits into
linux-rdma:masterfrom
yishaih:verbs_misc

Conversation

@yishaih

@yishaih yishaih commented Jul 29, 2026

Copy link
Copy Markdown
Member

This series introduces a new optional MR access flag named IBV_ACCESS_UNORDERED.

This access flag allows additional relaxing of the order of memory accesses to the memory region.

Using this access flag allows for concurrent reads and writes of this MR to be reordered.

Extra details exist as part of the man page and along the commit messages in the series.

The matching kernel part was sent already to rdma-next.

To commit: 3fbf618b8897 ("RDMA/uverbs: Add new Unordered MR access flag").

Signed-off-by: Michael Gur <michaelgur@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Fix typos and grammar.

Signed-off-by: Michael Gur <michaelgur@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Add a new Unordered access flag for MR registrations.
This access flag allows additional relaxing of the order of memory
accesses to the memory region.  Using this access flag allows for
concurrent reads and writes of this MR to be reordered.

This access flag is part of the optional access flags range, drivers
shouldn't fail registration for any value in this range and are expected
to treat them as best-effort.

The new access flag is defined as a mask of the new unordered bit and
the existing relaxed bit. This guarantees backward compatibility:
applications opting into unordered access will still benefit from
relaxed ordering on older kernels that don't handle the new bit.

Update the man page to document both ordering access flags under a new
ORDERING SEMANTICS section.

Signed-off-by: Michael Gur <michaelgur@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Add unordered access option to all five pingpong example applications.

Relaxing access to data is recommended for performance reasons for all
RDMA applications that don't poll data and rely on completions for their
memory access logic.
Additionally, without this flag the examples will fail to register
memory on HW that restricts strong-ordered memory registration.

Signed-off-by: Michael Gur <michaelgur@nvidia.com>
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants