IPv6 MDNS #35

twinshadow · 2019-11-08T07:37:53Z

This PR is split into 3 commits, but they may not be entirely distinct. It basically goes like this:

Add support for interfaces to store multiple IP addresses.
Add support for using multiple addresses in service and control functions.
Add support for sending and receiving IPv6 packets.

I've kept my FreeBSD patch in here as well, since I tested this with FreeBSD-current as well as OpenBSD-current.

Samples

Lookup of IPv6 addresses from mdnsctl

$ mdnsctl lookup Android.local
Address: 192.168.86.164
$ mdnsctl lookup -6 Android.local
Address: fe80::aa6b:adff:fe16:13e7

mdnsd advertisiment appearing on mdns-sd from another system. tutti.local is running mdnsd with these patches.

dns-sd -G v4v6 tutti.local
DATE: ---Thu 07 Nov 2019---
20:57:57.205  ...STARTING...
Timestamp     A/R Flags if Hostname                               Address                                      TTL
20:57:57.207  Add     3  1 tutti.local.                           FDCC:F80E:8097:0000:0000:0000:0000:0001%<0>  120
20:57:57.208  Add     2  1 tutti.local.                           192.168.32.5                                 120

Issues

The daemon only responds with the first address on an interface. I might be missing something obvious, but it looks like there would have to be substantial changes needed to handle packet fragmenting in pkt_handle_qst to respond with multiple A and/or AAAA records.

Fixes #4

Cheers!

- Multiple addresses per interface (no effect) - Get address changes from route socket - remove all kif_* functions - use getifaddrs to get all addresses for an interface - Define udp4/udp6 structures for MDNS FD and events

nbriggs · 2019-12-01T00:14:45Z

mdnsd/kiface.c

+#ifdef __FreeBSD__
+#define rtm_hdrlen rtm_msglen
+#endif


I suspect this is wrong -- on OpenBSD rtm_hdrlen is documented as being

u_short rtm_hdrlen; /* sizeof(rt_msghdr) to skip over the header */

while rtm_msglen on both FreeBSD and OpenBSD is documented as being

u_short rtm_msglen; /* to skip over non-understood messages */

so at kiface.c:150

sa = (struct sockaddr *)(buf + rtm->rtm_hdrlen); get_rtaddrs(rtm->rtm_addrs, sa, rti_info);

you'll be skipping the whole message rather than just the header of the message.

twinshadow · 2019-12-01T01:09:46Z

Yup, my patch does the wrong thing, and should be doing sizeof(&rtm) instead to accomplish the same effect.

…

On November 30, 2019 4:14:46 PM PST, Nick Briggs ***@***.***> wrote: nbriggs commented on this pull request. > +#ifdef __FreeBSD__ +#define rtm_hdrlen rtm_msglen +#endif I suspect this is wrong -- on OpenBSD rtm_hdrlen is documented as being ``` u_short rtm_hdrlen; /* sizeof(rt_msghdr) to skip over the header */ ``` while rtm_msglen on both FreeBSD and OpenBSD is documented as being ``` u_short rtm_msglen; /* to skip over non-understood messages */ ``` so at kiface.c:150 ``` sa = (struct sockaddr *)(buf + rtm->rtm_hdrlen); get_rtaddrs(rtm->rtm_addrs, sa, rti_info); ``` you'll be skipping the whole message rather than just the header of the message. -- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: #35 (review)

-- Sent from my Android device with K-9 Mail. Please excuse my brevity.

nbriggs · 2019-12-01T01:50:34Z

sizeof(*rtm) rather than sizeof(&rtm) if you're not going to use sizeof(struct rt_msghdr)

But even with that correction, it still takes a segmentation fault (on FreeBSD) if you add or delete an address on the interface that mdnsd is monitoring.

nbriggs · 2019-12-01T02:38:11Z

The documentation for FreeBSD route(4) says

The RTM_IFINFO message uses a if_msghdr header, the RTM_NEWADDR and
RTM_DELADDR messages use a ifa_msghdr header, the RTM_NEWMADDR and
RTM_DELMADDR messages use a ifma_msghdr header, the RTM_IFANNOUNCE
message uses a if_announcemsghdr header, and all other messages use the
rt_msghdr header.

So while you can walk the messages by the rtm_msglen you have to get the type of message to know what the header length is -- a problem you don't have in OpenBSD because you've got the variable header length in the common part of the header(s).

nbriggs · 2019-12-01T04:31:47Z

mdnsd/kiface.c

@@ -275,6 +141,9 @@ kev_dispatch_msg(int fd, short event, void *bula)
 		if (iface == NULL) /* this interface isn't configured */
 			continue;

+		sa = (struct sockaddr *)(buf + rtm->rtm_hdrlen);


If there are actually multiple messages in buf, presumably this should be next + rtm->rtm_hdrlen or you're going to pick up the wrong socket addresses from the 2nd and subsequent messages.

twinshadow · 2019-12-01T10:19:29Z

In OpenBSD, struct rt_msghdr->rtm_addrs and struct ifa_msghdr->ifam_addrs point to the same offset in the buffer. This is not true in FreeBSD, so the bitmask wasn't being passed to get_rtaddrs with the right value, even though the initial offset was correct.

nbriggs · 2019-12-01T22:08:19Z

No SIGSEGVs now. It does get a bit confused if you delete the only IPv4 address off the interface it's listening on -- a situation that wouldn't have made sense before IPv6 support.

nbriggs · 2019-12-02T23:17:01Z

mdnsd/kiface.c

 void
 kev_init(void)
 {
 	int		opt = 0, rcvbuf, default_rcvbuf;
+	int		rtfilter;


If you're going to #ifdef __OpenBSD__ the rtfilter = ... perhaps you could #ifdef the declaration for rtfilter.

twinshadow · 2019-12-03T20:12:26Z

No SIGSEGVs now. It does get a bit confused if you delete the only IPv4 address off the interface it's listening on -- a situation that wouldn't have made sense before IPv6 support.

I've added some patches to tone-down the warnings from lack of IPv4 addresses. Still working on a patches to respond to remote DNS queries with all IP addresses and update the cache when changes occur to local interfaces.

If there are packets going out on multiple interfaces while operating on MDNS IPV4/6, interfaces should not be expected to be dual-stack. IPv6 only requires an index, and doesn't seem to have any problems if there are no IPv6 addresses on the interface. IPv4 requires an address on the interface to do any multicast operations.

nbriggs · 2019-12-10T03:20:01Z

mdnsctl/parser.c

@@ -448,6 +473,10 @@ parse_flags(const char *word, int *flags)
 			*flags |= F_A;
 			r++;
 			break;
+		case '6':


The usage info for the "lookup" subcommand wasn't updated to include the new "6" flag.

twinshadow · 2019-12-16T09:29:49Z

I just realized that bug #31 will more easily trigger with this PR, since more addresses are added to the fixed-size array. I'll have to add a fix for that here.

Adds a check when adding addresses in pge_initprimary to prevent writing to unallocated memory and causing the program to segfault. Fixes #31, but this code should be refactored in the future.

New feature or bug is that responses will contain both A and AAAA without regard for the request specifier for A or AAAA. I need to read the spec more.

Send the correct AF for the RR type and enumerate all interfaces for addresses.

twinshadow · 2019-12-18T17:26:49Z

DNS responses from the daemon now send all addresses from all interfaces and can respond to mdnsctl with the appropriate address from any interface (eg. em0 being IPv6 only and bridge0 being IPv4 only). I've tested out the MDNS packet changes with mdns-sd on NetBSD.

The only other things I can think to commit are style, grammar, and comment changes before rebasing everything for merging.

twinshadow added 4 commits November 11, 2019 12:16

Overhaul of interface functions

0ea863e

- Multiple addresses per interface (no effect) - Get address changes from route socket - remove all kif_* functions - use getifaddrs to get all addresses for an interface - Define udp4/udp6 structures for MDNS FD and events

Add IPv6 support to parts of libmdns and mdnsctl

cb6f86c

Process IP6 DNS packets

800ee86

FIXUP: FreeBSD tested fixes

4196b83

nbriggs reviewed Dec 1, 2019

View reviewed changes

FIXUP: FreeBSD segfault and other fixes

13d13ce

nbriggs reviewed Dec 2, 2019

View reviewed changes

twinshadow added 2 commits December 9, 2019 00:36

FIXUP: FREEBSD: ifdef rtfilter define

b7a49b2

nbriggs reviewed Dec 10, 2019

View reviewed changes

FIXUP: add -46 options to usage

e0ce0bb

twinshadow added 5 commits December 16, 2019 13:02

FIXUP: increase PGE_RR_MAX to 1024

a9e3e7c

Adds a check when adding addresses in pge_initprimary to prevent writing to unallocated memory and causing the program to segfault. Fixes #31, but this code should be refactored in the future.

FIXUP: update the man pages and mdnsctl lookup usage

5a669a4

FIXUP: newaddr/deladdr update pge cache

352ced1

FIXUP: fix rr overwriting from rr patches.

a871deb

New feature or bug is that responses will contain both A and AAAA without regard for the request specifier for A or AAAA. I need to read the spec more.

FIXUP: fix control_send_rr regression

fda1318

Send the correct AF for the RR type and enumerate all interfaces for addresses.

nbriggs mentioned this pull request Oct 2, 2020

Silence complaints generated by mdnsd when legacy AirDrop is used on macOS #33

Merged

twinshadow closed this by deleting the head repository Mar 4, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

IPv6 MDNS #35

IPv6 MDNS #35

twinshadow commented Nov 8, 2019 •

edited

Loading

nbriggs Dec 1, 2019

twinshadow commented Dec 1, 2019 via email

nbriggs commented Dec 1, 2019

nbriggs commented Dec 1, 2019 •

edited

Loading

nbriggs Dec 1, 2019

twinshadow commented Dec 1, 2019

nbriggs commented Dec 1, 2019

nbriggs Dec 2, 2019

twinshadow commented Dec 3, 2019 •

edited

Loading

nbriggs Dec 10, 2019

twinshadow commented Dec 16, 2019

twinshadow commented Dec 18, 2019

IPv6 MDNS #35

IPv6 MDNS #35

Conversation

twinshadow commented Nov 8, 2019 • edited Loading

Samples

Issues

nbriggs Dec 1, 2019

Choose a reason for hiding this comment

twinshadow commented Dec 1, 2019 via email

nbriggs commented Dec 1, 2019

nbriggs commented Dec 1, 2019 • edited Loading

nbriggs Dec 1, 2019

Choose a reason for hiding this comment

twinshadow commented Dec 1, 2019

nbriggs commented Dec 1, 2019

nbriggs Dec 2, 2019

Choose a reason for hiding this comment

twinshadow commented Dec 3, 2019 • edited Loading

nbriggs Dec 10, 2019

Choose a reason for hiding this comment

twinshadow commented Dec 16, 2019

twinshadow commented Dec 18, 2019

twinshadow commented Nov 8, 2019 •

edited

Loading

nbriggs commented Dec 1, 2019 •

edited

Loading

twinshadow commented Dec 3, 2019 •

edited

Loading