Skip to content

UCT/IB/ROCE: Use ipv6 non link-local GID in priority #10939

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 3 commits into
base: master
Choose a base branch
from
Open

UCT/IB/ROCE: Use ipv6 non link-local GID in priority #10939

wants to merge 3 commits into from

Conversation

@tvegas1
Copy link
Contributor

@tvegas1 tvegas1 commented Oct 7, 2025

What?

When selecting RoCE GID, search for IPv6 non link-local before trying to use IPv6 link-local GID.

Why?

Some clusters only have gid_index=3 usable. Manually setting UCX_IB_GID_INDEX=3 is cumbersome and error-prone.

How?

Relevant cli: show_gids. Tested on two clusters:

UCX_LOG_LEVEL=debug ./rfs/bin/ucx_perftest -l -t ucp_put_bw | grep gid_index

# For all interfaces (mlx5_x/roce_railsx):
UCX_NET_DEVICES=mlx5_0:1 UCX_PROTO_INFO=y ./rfs/bin/ucx_perftest -l -t ucp_put_bw
UCX_NET_DEVICES=mlx5_0:1 UCX_PROTO_INFO=y ./rfs/bin/ucx_perftest -t ucp_put_bw remote-node

brminich
brminich reviewed Oct 17, 2025
View reviewed changes
@brminich brminich requested a review from amastbaum October 17, 2025 14:17
yosefe
yosefe reviewed Oct 23, 2025
View reviewed changes
src/uct/ib/base/ib_device.c Outdated
{UCT_IB_DEVICE_ROCE_V1, AF_INET6}
struct {
uct_ib_roce_version_info_t info;
int skip_ll; /* link-local not allowed when true */
Copy link
Contributor

@yosefe yosefe Oct 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. maybe inverse to simplify - "allow_ll" or "is_ll"
  2. i'd explicitly initialize this value to 0 when applicable (line 1086, 1088 etc)

Copy link
Contributor

@brminich brminich Oct 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

src/uct/ib/base/ib_device.c
}

static int uct_ib_gid_is_ipv6_ll(const union ibv_gid *gid)
{
Copy link
Contributor

@yosefe yosefe Oct 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can reuse IN6_IS_ADDR_LINKLOCAL?

Copy link
Contributor

@brminich brminich Oct 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

amastbaum
amastbaum previously approved these changes Oct 23, 2025
View reviewed changes
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@yosefe yosefe yosefe left review comments

@brminich brminich brminich left review comments

@amastbaum amastbaum amastbaum left review comments

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@tvegas1 @yosefe @brminich @amastbaum