Bug 118927

Summary:	[em] em(4) broken: link state changed to DOWN (/UP), link states coalesced errors [regression]
Product:	Base System	Reporter:	Pekka Savola <pekkas>
Component:	kern	Assignee:	jfv
Status:	Closed FIXED
Severity:	Affects Only Me	CC:	sbruno
Priority:	Normal	Keywords:	IntelNetworking
Version:	6.3-PRERELEASE
Hardware:	Any
OS:	Any

Description Pekka Savola 2007-12-21 11:00:00 UTC

After upgrading via sources from 6.2-STABLE (Aug 7) to 6.3-PRERELEASE
(snapshots from about November 29 and December 21 tested), em2 interface
started spewing out the following ad infinitum:

em2: 32 link states coalesced
em2: link state changed to UP
em2: 19 link states coalesced
em2: link state changed to DOWN
em2: 3 link states coalesced
em2: link state changed to UP
em2: 17 link states coalesced
em2: link state changed to DOWN
em2: link state changed to UP
em2: 20 link states coalesced

The link led is blinking.  Fibers are not connected.

Rebooting to 6.2-STABLE from Aug 7 works.

This happens on an add-on card with 1000FX (multimode) adapter only.
Intel cards on the same system which have UTP interface are OK.

Unfortunately I had to plug off the card, but I inserted into a Linux
and got the following information about it:

0b:01.0 Ethernet controller: Intel Corporation 82542 Gigabit Ethernet Controller (Fiber) (rev 02)

. but I guess that should be enough to figure out what's going on.

I suspect one of the em(4) driver updates broke this particular piece
of hardware.

Fix: 

Fix not known.  Using 6.2-STABLE from Aug 7 2007 does not yet introduce
the problem.
How-To-Repeat: Try to boot to a recent RELENG_6 kernel with Intel 82542 network card.

Comment 1 Pekka Savola 2007-12-21 11:32:33 UTC

Additional comment:  More information about the device:

PCI: 0b:01:0: chip 8086,1000 card 8086,1000 rev 02 class 02,00,00 hdr 00

Comment 2 Remko Lodder freebsd_committer

2007-12-21 18:38:52 UTC

Responsible Changed
From-To: freebsd-bugs->jfv

Reassign to the maintainer

Comment 3 Sean Bruno freebsd_committer

2015-06-29 18:02:19 UTC

My testing of stable/10 and -current seems to show across my test boxes that this is no longer happening.  If this is still testable and is still an issue, we would request that you test with the upcoming 10.2r beta and report back on a new ticket.