Date: Fri, 15 Nov 1996 14:30:01 +0900 (JST) From: seki@sysrap.cs.fujitsu.co.jp To: FreeBSD-gnats-submit@freebsd.org Cc: seki@sysrap.cs.fujitsu.co.jp Subject: i386/2011: Wrong stats (SNMP Dot3 MIBs) in ed driver Message-ID: <199611150530.OAA00358@fuzz.sysrap.cs.fujitsu.co.jp> Resent-Message-ID: <199611150540.VAA16301@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 2011
>Category: i386
>Synopsis: Ed driver puts wrong value in some of its SNMP Dot3 MIBs.
>Confidential: no
>Severity: non-critical
>Priority: low
>Responsible: freebsd-bugs
>State: open
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Thu Nov 14 21:40:01 PST 1996
>Last-Modified:
>Originator: Masahiro Sekiguchi
>Organization:
Fujitsu Limited
>Release: FreeBSD 2.2-ALPHA i386
>Environment:
2.2 ALPHA with any of ed compatible Ethernet cards.
>Description:
Ed driver counts some collision-related events in SNMP Dot3 MIBs
wrong way.
RFC 1650 specifies that dot3StatsDeferredTRansmissions,
dot3StatsSingleCollisionFrames, dot3StatsMultipleCollisionFrames,
and dot3StatsExcessiveCollisions are exclusive; i.e., a frame
may cause one of the above MIBs be incremented, but cannot
cause two or more of the above.
Ed driver in 2.2 ALPHA, however, increments two or more for a
frame. It conflicts with RFC 1650.
For example, the definition of dot3StatsDeferredTransmissions
in RFC 1650 says: "The count represented by an instance of this
object does not include frames involved in collisions." However,
Ed in 2.2 ALPHA increments both DeferredTransmissions and
SimgleCollisionsFrames for one single-collided frame.
One more related note.
RFC 1650 also says that error conditions are also exclusive.
Currently, ed driver tests for some bits in a hardware register,
and update error-related MIBs independently. *If* two or more
error bits in the register can be set for a frame, it causes
another wrong stat. I'm not sure it can be happen, or never
happen due to the hardware specification.
>How-To-Repeat:
Transmit an Ethernet frame through an ed compatible card, and
make it collide.
>Fix:
The following is a suggested patch to sys/i386/isa/if_ed.c.
This patch fixes bugs on collision-related stats. Handling of
error related MIBs are not touched.
--- cut here ---
--- if_ed.c.orig Thu Oct 17 22:42:13 1996
+++ if_ed.c Fri Nov 15 13:58:29 1996
@@ -2410,7 +2410,6 @@
* TSR_ABT is set.
*/
collisions = 16;
- sc->mibdata.dot3StatsMultipleCollisionFrames++;
sc->mibdata.dot3StatsExcessiveCollisions++;
sc->mibdata.dot3StatsCollFrequencies[15]++;
}
@@ -2458,12 +2457,10 @@
break;
case 1:
sc->mibdata.dot3StatsSingleCollisionFrames++;
- sc->mibdata.dot3StatsDeferredTransmissions++;
sc->mibdata.dot3StatsCollFrequencies[0]++;
break;
default:
sc->mibdata.dot3StatsMultipleCollisionFrames++;
- sc->mibdata.dot3StatsDeferredTransmissions++;
sc->mibdata.
dot3StatsCollFrequencies[collisions-1]
++;
MIBs
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611150530.OAA00358>