On Mon, Mar 22, 2010 at 4:54 PM, Dave Shield <[email protected]>wrote:
> On 21 March 2010 08:45, Dave Shield <[email protected]> wrote:
> > On 20 March 2010 15:13, Robert Story <[email protected]> wrote:
> >> Dave, have any other issues come up that might warrant a rc2?
> >
> > I've got a list of three or four possible patches sitting on my desk at
> work.
> > I'll send out a CFV tomorrow.
>
>
> Please find attached four proposed patches, which could potentially be
> applied to the 5.4.x (and 5.2.x) lines. We're currently in release-freeze
> for both of these branches, so could people please indicate whether or
> not they would like to see these included in the 5.4.3 and 5.2.6 releases.
>
> It would be helpful to have an explicit vote for each individual patch:
>
> 1) Latch disk statistics: no / 54x only / both 54x and 52x
> 2) Missing privKey crash: no / 54x only / both 54x and 52x
> 3) VACM best match: no / 54x only / both 54x and 52x
> 4) MIB dir path logging: no / 54x only / both 54x and 52x
>
> A word of explanation for each of these:
>
> 1) Latch disk statistics (disk-latch.patch)
> For large disks (>2Tb), the UCD-SNMP-MIB:dskTable currently
> reports the lower-32 bits of the total/usage/available statistics, which
> is effectively a random, meaningless (but apparently valid) value.
> This patch latches these three objects at the maximum value for an
> Integer32 object - thus indicating that the value is not actually useable.
>
> 2) Missing privKey crash (des-priv.patch)
> In certain circumstances (relating to sending SNMPv3 notifications),
> it's possible for sending a trap to follow a NULL pointer and crash the
> agent. This patch protects against this.
> (See the coders thread "snmpwalk causes agent to coldStart"
> from last Monday for details)
>
> 3) VACM best match (vacm-best-match.patch)
> The VACM MIB describes how to select which row from the
> vacmAccessTable to use when deciding whether to allow access.
> The Net-SNMP agent does not follow this algorithm properly.
> This patch corrects this.
> (See the users thread "vacm_getAccessEntry() doesn't look for best match"
> from December 2009 for details)
>
> 4) MIB dir path logging (mibdir-log.patch)
> A common query on the lists is asking why the agent isn't loading
> configuration or MIB files - particularly when switching from a
> vendor-supplied
> package to source-compiled version (or vice versa). This often comes
> down to changes in the search paths used, but determining what paths
> are being used requires turning on debugging tokens.
> This patch reports the relevant search path as part of the error message
> when a MIB module isn't found, or no access control is configured.
> (As well reporting these paths as part of the usage output message)
>
How well have these patches been tested, and on which platforms ? My remarks
for patches (1), (2) and (4) are as follows:
* (2) seems a good idea to me.
* (1) will only work when sizeof(long) > sizeof(int). This is not the case
for 32-bit platforms, and this is not the case for 64-bit Windows systems
(see also http://en.wikipedia.org/wiki/64-bit).
* (4) breaks the Windows build because netsnmp_get_mib_directory() is not
exported from the Net-SNMP DLL on the 5.4 branch (see also
http://net-snmp.svn.sourceforge.net/viewvc/net-snmp?view=rev&revision=18347
).
Bart.
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Net-snmp-coders mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
