Skip to content

Fix JACCL GID index for multi-node Thunderbolt 5 setups #3389

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

Closed
qubitcontracting wants to merge 1 commit into from
+24 −3
Closed

Fix JACCL GID index for multi-node Thunderbolt 5 setups #3389

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 24 additions & 3 deletions mlx/distributed/jaccl/utils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,8 +181,20 @@ const Destination& Connection::info() {

ibv_port_attr port_attr;
ibv().query_port(ctx, 1, &port_attr);
ibv_gid gid;
ibv().query_gid(ctx, 1, 1, &gid);
// Scan GID table for a valid entry. On some Thunderbolt interfaces the
// IPv4-mapped GID is not at index 1 (e.g. index 2 on certain TB5 configs).
ibv_gid gid = {};
for (int idx = 1; idx <= 3; idx++) {
ibv_gid tmp;
if (ibv().query_gid(ctx, 1, idx, &tmp) == 0 &&
tmp.global.interface_id != 0) {
gid = tmp;
break;
}
}
if (gid.global.interface_id == 0) {
ibv().query_gid(ctx, 1, 0, &gid);
}

src.local_id = port_attr.lid;
src.queue_pair_number = queue_pair->qp_num;
Expand Down Expand Up @@ -227,7 +239,16 @@ void Connection::queue_pair_rtr(const Destination& dst) {
attr.ah_attr.is_global = 1;
attr.ah_attr.grh.hop_limit = 1;
attr.ah_attr.grh.dgid = dst.global_identifier;
attr.ah_attr.grh.sgid_index = 1;
// Find valid local sgid_index dynamically (matches info() scan)
attr.ah_attr.grh.sgid_index = 0;
for (int idx = 1; idx <= 3; idx++) {
ibv_gid tmp;
if (ibv().query_gid(ctx, 1, idx, &tmp) == 0 &&
tmp.global.interface_id != 0) {
attr.ah_attr.grh.sgid_index = idx;
break;
}
}
}

int mask = IBV_QP_STATE | IBV_QP_AV | IBV_QP_PATH_MTU | IBV_QP_DEST_QPN |
Expand Down