ofed: don't stop removing stale RoCE GIDs at the first hole
Needs ReviewPublic
Actions

Authored by aehrenberg_nvidia.com on Tue, Jul 28, 8:57 AM.

Details

Reviewers

kib
slavash
jhb

Summary

When cleaning up stale GIDs the scan stopped as soon as
rdma_get_gid_attr() failed. But that can also happen for empty entries
in the middle of the table, so a single gap left everything after it
behind and the GID entries could eventually run out.

Now the whole table is scanned and the empty slots are simply skipped.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Event Timeline

aehrenberg_nvidia.com created this revision.Tue, Jul 28, 8:57 AM

Herald added a subscriber: imp. · View Herald TranscriptTue, Jul 28, 8:57 AM

aehrenberg_nvidia.com requested review of this revision.Tue, Jul 28, 8:57 AM

aehrenberg_nvidia.com edited the summary of this revision. (Show Details)Tue, Jul 28, 9:10 AM

aehrenberg_nvidia.com edited the test plan for this revision. (Show Details)

kib added inline comments.Tue, Jul 28, 9:43 PM

sys/ofed/drivers/infiniband/core/ib_roce_gid_mgmt.c
216	Why is the value initialized so early? IMO it should be done right before the loop that uses it. BTW, what guarantees the stability of the gid table?

Revision Contents
Changeset List

Path

Size

sys/

ofed/

drivers/

infiniband/

core/

ib_roce_gid_mgmt.c

5 lines

Diff 182921

View Options

sys/ofed/drivers/infiniband/core/ib_roce_gid_mgmt.c

ofed: don't stop removing stale RoCE GIDs at the first holeNeeds ReviewPublicActions