Both ways to create an artificial NDP record are broken: a call via rtsock sets wrong timeout, a netlink code does not check opts.expire_time value at all.
Patch addresses both issues.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Skipped - Unit
Tests Skipped
Event Timeline
Comment Actions
Are the tests that hard to add, or perhaps can it be proceeded with further without them?
Comment Actions
I will try to add them this or the next week, it takes a while to get familiar with test suite.
| usr.sbin/ndp/ndp.c | ||
|---|---|---|
| 960 | IMHO the comment isn't really needed. It just reiterates what we see in the code. If you don't like idea of using uptime for that purpose you'd better make a longer XXX comment that would elaborate on why using uptime is not the best idea. | |
| usr.sbin/ndp/ndp_netlink.c | ||
| 484–495 | ||
| 507 | Please declare int off; at the beginning of the block. P.S. IMHO, the entire API construct of adding a nested attribute should be a bit easier. Just comment into the air, not asking for anything here :) | |
Comment Actions
I had already done some quick testing of the previous version earlier, but today I rebased this patch onto the latest version and applied it to the 16.0-CURRENT sources to do some testing.
I’m not fully familiar with the NDP mechanics, but when an entry expires, shouldn’t it be removed and the NDP resolution performed again? That doesn’t seem to be happening here. The NDP entry is marked as expired, but the host continues to use it and sends packets as if it were a static entry.
I’m not sure whether this is caused by the patch or if the bug is elsewhere.