The good folks at Red Hat QE have uncovered an issue in recent kernels related
to the network snmp procfs file. A new set of fields related to CRC checking
have been added, one of which is smack bang in the middle of existing columns
(Icmp). This causes offset assumptions to no longer hold, and the subsequent
failure in test 216.
[17%] 126 - output mismatch (see 126.out.bad)
3,6c3,6
< Icmp: InMsgs InErrors InDestUnreachs InTimeExcds InParmProbs InSrcQuenchs
InRedirects InEchos InEchoReps InTimestamps InTimestampReps InAddrMasks
InAddrMaskReps OutMsgs OutErrors OutDestUnreachs OutTimeExcds OutParmProbs
OutSrcQuenchs OutRedirects OutEchos OutEchoReps OutTimestamps OutTimestampReps
OutAddrMasks OutAddrMaskReps
< Tcp: RtoAlgorithm RtoMin RtoMax MaxConn ActiveOpens PassiveOpens AttemptFails
EstabResets CurrEstab InSegs OutSegs RetransSegs InErrs OutRsts
< Udp: InDatagrams NoPorts InErrors OutDatagrams RcvbufErrors SndbufErrors
< UdpLite: InDatagrams NoPorts InErrors OutDatagrams RcvbufErrors SndbufErrors
---
> Icmp: InMsgs InErrors InCsumErrors InDestUnreachs InTimeExcds InParmProbs InSrcQuenchs InRedirects InEchos InEchoReps InTimestamps InTimestampReps InAddrMasks InAddrMaskReps OutMsgs OutErrors OutDestUnreachs OutTimeExcds OutParmProbs OutSrcQuenchs OutRedirects OutEchos OutEchoReps OutTimestamps OutTimestampReps OutAddrMasks OutAddrMaskReps
> Tcp: RtoAlgorithm RtoMin RtoMax MaxConn ActiveOpens PassiveOpens AttemptFails EstabResets CurrEstab InSegs OutSegs RetransSegs InErrs OutRsts InCsumErrors
> Udp: InDatagrams NoPorts InErrors OutDatagrams RcvbufErrors SndbufErrors InCsumErrors
> UdpLite: InDatagrams NoPorts InErrors OutDatagrams RcvbufErrors SndbufErrors InCsumErrors
[32%] 216 - output mismatch (see 216.out.bad)
1a2,10
> network.icmp.inechos = 0 out of range 70..70
> network.icmp.inechoreps = 70 out of range 69..69
> network.icmp.intimestamps = 69 out of range 0..0
> network.icmp.outmsgs = 0 out of range 143..143
> network.icmp.outerrors = 143 out of range 0..0
> network.icmp.outdestunreachs = 0 out of range 5..5
> network.icmp.outtimeexcds = 5 out of range 0..0
> network.icmp.outechos = 0 out of range 69..69
> network.icmp.outtimestamps = 69 out of range 0..0
To address this, we will need to rework src/pmdas/linux/proc_net_snmp.c so that
it takes heed of the order in which the column headers are presented rather
than just assuming they will remain in the traditional order. i.e. fix this:
/*
* This is really bogus.
* We are consuming the heading line in the while (fgets(..)) and
* discarding it, then reading the next line and parsing the counters
* from there, with hard coded semantics as to the meaning of each
* counter found on this line.
* This does not work at all for IcmpMsg where the line we're
* ignoring provides the name of the counters on the line we're
* scanning (pairs appear to be like this, we're just lucky that
* the names in the ignored lines are not changing ... of course
* when they change we'll export complete nonsense.
* - Ken
*/
cheers.