[Bug 280074] em(4) temporarily hangs and re-enables TX CSUM if a bridge it's a member of is modified

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=280074

--- Comment #23 from Zhenlei Huang <[email protected]> ---
(In reply to Joshua Kinard from comment #12)
> (In reply to Zhenlei Huang from comment #9)
> I am actually running 15.0-RELEASE on this particular machine (NUC8i5BEK), so I 
> can't downgrade back to 14.x.  Has anyone looked at why the net/intel-em-kmod driver
> doesn't cause the issue I am describing, but the in-tree one does?  That's why I am
> leaning towards it being a problem in em(4) itself, not bridge(4) or epair(4).  If
> it were one of those two, I would expect a similar issue, if not the same one, to 
> manifest under both drivers.

Ah I thought you're still running 14.x .

I do not have em(4) device at my hand, but have igc(4) which also use iflib to
test.
Currently those drivers use iflib will do full restart on changing the
capabilities.

From my test with igc(4) from ssh session, after `ifconfig igc0 -txcsum` type
ENTER every seconds,

```
# ifconfig igc0 -txcsum
#
#
#
#
#
#
```
The ssh session stalls for about 6 seconds.

From the log `/var/log/message`
```
Feb 20 01:08:33  kernel: igc0: link state changed to DOWN
Feb 20 01:08:36  kernel: igc0: link state changed to UP
```
It appears that the igc0 requires about 3 seconds to have a full HW reset.

-- 
You are receiving this mail because:
You are the assignee for the bug.