Re: [Openslp-devel] Patches for openslp

[Ian N. <in...@gm...>]

Hi John. I don't think any of us here do. The person you might be
remembering is Matt Pendlebury.
On 26 Mar 2014 15:04, "John Calcote" <joh...@gm...> wrote:

> Hi Richard,
>
> Thanks so much for all the work you've done. Don't take this wrong, but
> don't you have an employee there at Thales that has commit rights to our
> repo? Matthew Hird, I believe. Or has he left the company? Just curious.
>
> Thanks,
> John Calcote
>
> > -----Original Message-----
> > From: Richard Porter [mailto:Ric...@th...]
> > Sent: Tuesday, March 25, 2014 8:18 AM
> > To: ope...@li...
> > Cc: joh...@gm...
> > Subject: Patches for openslp
> >
> > Hi
> >
> > Here are some more patches we've developed for openslp v2.0.  Most are
> > bug fixes; but the largest implements network interface change monitoring
> > on Linux, using netlink.
> >
> > These build on top of the patches I sent to the list in December 2012
> (though
> > I'm not sure those ever made it through mailman, so I'm attaching the
> email
> > again here.)  They probably need to be applied in order too - or at least
> with
> > the netlink one last.
> >
> >
> > The files and associated change details are:
> >
> > 01-regfile-memory-corruption:
> >
> > If you're going to increment the buffer pointer, be sure to not then
> memcpy
> > off the end of the buffer...
> >
> >
> > 02-invalid-adverts-segfault:
> >
> > When registering a database entry, if the attribute parsing fails then
> > fail the registration.  This prevents a NULL pointer being assigned as
> > the attribute list, which will be dereferenced subsequently on database
> > lookups later.
> >
> >
> > 03-da-adverts-iface-down:
> >
> > Don't send DA advertisements for addresses of interfaces that aren't
> active.
> >
> > Only add active interfaces to the list.
> >
> > Only add active interfaces to the default list too
> >
> > The interface active check needs the address family to be set, so make
> > sure it's set before calling.
> >
> > When comparing IPv6 addresses, also compare the scope ID.
> >
> >
> > 05-try-all-das:
> >
> > When finding servers, make SLPFindSrvs() try again if there are more DAs
> > available to query.
> >
> > Add the same retry loop to findattrs as was added to findsrvs.
> >
> > KnownDABadDA() is expecting a sockaddr* not an in_addr*
> >
> >
> > 06-sin6_scope_id:
> >
> > Manage sin6_scope_id when dealing with link-local IPv6 addresses.  This
> > is especially important for listening on multicast addresses.
> >
> > Don't destroy the IPv6 multicast sockets until the last instance of a
> > particular service type has been removed from the database.
> >
> > Avoid a memory leak when reinitialising network interfaces.
> >
> > Add const qualifier to SLPIfaceGetIndex() to satisfy windows compiler
> >
> > Move the definition of SLPDDatabaseSrvtypeUsed to before it's now used.
> >   Shuts up the windows compiler.
> >
> >
> > 07-getaddrinfo-segfault:
> >
> > Check the return value of getaddrinfo() before dereferencing the pointer
> > it allocates when successful.
> >
> >
> > 08-netlink:
> >
> > Integrate libnl interface and address caches into the slpd IO loop.
> >
> > Manage the sin6_scope_id when binding IPv6 sockets
> >
> > When the netlink cache is updated, call HandleSigHup to get slpd to
> > rejig its interfaces
> >
> > Move the netlink cache management to the common/slp_iface.c file.
> >
> > Need to initialise the netlink cache before calling SLPDPropertyInit()
> >
> > Use the netlink cache for all network configuration lookup
> >
> > Add a set of callbacks between slpd_main and slp_iface to signal when
> > interfaces or addresses change.  Reset the interface list when signalled
> > by the callbacks.
> >
> > The database entries will time out as normal, so some 'stale'
> > registration entries might hang around for a few seconds.  Similarly,
> > entries for new addresses will be added at the next registration update.
> >
> > slptool needs to initialise the netlink caches.
> >
> > Require libnl-3.2.19
> >
> > Make sure G_SlpdProperty.interfaces is correctly set in
> > SLPDPropertyReinit() even if there are no known interfaces
> >
> > Fix a memory leak in SLPDIncomingReinit() when getting interface
> > information
> >
> > Add some extra logging when scanning interfaces
> >
> > If we're using netlink, there's a small window between receiving a new
> > IPv4 address notification and being able to actually bind to that
> > address.  bind(2) will fail with EADDRNOTAVAIL if the kernel's
> > Forwarding Information Base hasn't been updated to know that the address
> > is local.  IP_FREEBIND allows us to bind anyway.  This is safe here, as
> > we are careful to only use addresses that we know are local.
> >
> >
> > 09-attribute-value-div-10:
> >
> > The fancy method that figures out how many digits are needed to
> > represent an integer failed when the integer was divisible exactly by 10.
> >
> >
> > (04 was going to be a fix for slptool, but that has already been fixed
> > by change 1767.)
> >
> >
> > - Richard
>
>
>
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/13534_NeoTech
> _______________________________________________
> Openslp-devel mailing list
> Ope...@li...
> https://lists.sourceforge.net/lists/listinfo/openslp-devel
>