Explanation of the bug: when processing its first request, rarpd
opens a routing socket to send requests to the arp table. It keeps
that socket open afterwards, while waiting for new RARP requests.
Meanwhile, the data received on the routing socket fill up until
they are about 8Kbytes in size. Any additional data is lost.
When rarpd receives its next RARP request, it tries to access the
ARP table via a routing socket call, then waits for the answer to
its own request. This answer is lost because the received data is
already filled: when looking for the reply, rarpd receives only
8kbytes worth of data, then loops waiting forever.
Someone please test it on -STABLE and commit it. We can close the PR
when testing on STABLE is done.
PR: bin/5669
Submitted by: Pierre Beyssac <pb@fasterix.freenix.org>
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.
Boy, I'm glad we're not using sup anymore. This update would have been
insane otherwise.
Fenner was kind enough to point out the error of my ways. This incorporates
diffs from him which:
- Keep everything in network order.
- Log the booted ether & ip address, instead of my address on that net
- change several exit()'s to return()'s, so that rarpd continues running
even if it thinks it's in a weird state.
One small tweak by me: in rarp_bootable(), we have to make sure to
construct 'ipname' in host byte order (if we don't, we have to
specify /tftpboot/<remote IP in hex> with <remote IP in hex> in
network byte order, which is confusing).
Also restored use of <dirent.h> rather than <sys/dir.h> as pointed
out by bde.
Also updated the man page so that the -v flag is documented.
With any luck, I won't have to touch this thing again.
This includes the following changes:
- Support for poking ARP entries into the local table is now built
in, so the arptab.c module I hacked together is no longer needed.
- rarp_process() and rarp_reply() now accept a len argument which is
passed down from rarp_loop() which tells rarp_reply() exactly how
long the original RARP frame was. (Usually, it's 60 bytes, which is
the minimum.) Previously, the length was calculated using the sum
of sizeof(struct ether_header) + sizeof(struct ether_arp) (plus the
ethernet frame header, I think). The result was a total packet
length of 42 bytes. Now, rarp_reply() sends out packets that are
the same size as those it recieves (60 bytes). This agrees with the
behavior of rarpd on SunOS (as observed with tcpdump). The unused
extra bytes are zeroed.
frame type in network byte order. The htons() that wasn't needed in
2.1 is now required in 2.2.
Ultimately, this rarpd should be replaced with the more recent one
supplied with the new BPF distribution.
declaration for ether_ntohost(). (Does anyone know what header file
is supposed to contain the declarations for the ether_addr functions?
I can't them in the SunOS includes anywhere.)
- Implement ether_hostton()
- Implement ether_aton()
- Modify ether_aton() and ether_ntoa() to match the semantics of the
SunOS versions of these functions.
- Neaten up ether_hostton() and ether_ntohost() a little.
- Get rid of ether_print() since it isn't needed for rarpd and it isn't
documented as a standard ethers(5) function.
rarpd.8:
- Make it clear that the 'ipaddr' that rarpd looks for in /tftpboot
is actually in hexadecimal (as in /tftpboot/803B4032) since those who
are not versed in the black art of system administration are not likely
to know this.
table; arptab.c is really a hacked up version of arp.c that only
supports adding temporary entries. (This stuff is nasty -- I wish I
knew what was so wrong with SIOCSARP/SIOCGARP/etc... that made the
BSD developers decide to take it out.) The idea here is that the
client issuing the rarp is expected to be in the middle of booting
and would therefore be unable to answer arp queries from other machines
on the wire. Having rarpd stuff a temporary entry for the booting
host into the local arp table helps keep arp requests from going unanswered.
Also added ether_print() and ether_ntoa() to the ether_addr.c module.
Eventually I'll get ether_aton() and ether_hostton() written and
then this file can be dropped straight into libc. (Assuming no one
objects, of course. :)
"Yes Virginia, there is a rarpd."
(Before anyone asks, this *not* the rarpd from NetBSD. It did come from
the same place as theirs, however.)
This is a port of the rarpd program included with the tcpdump-2.0
source code (which I finally unearthed after scrounging around
some of the darker corners of the Internet). It's as close to the
original as I could keep it except for the following changes:
- The original program was based on an older version of the Berkeley
Packet Filter which used different filter programming instructions.
Fortunately, an updated RARP packet filter is available right in the
BPF man page so this was easy to fix.
- The old code didn't know how to deal with variable length addresses
in ifreq buffers. This has been fixed.
- Some byte order weirdness had to be fixed. The sanity checks in
rarp_check() needed some htons()es, and the rarp_reply() function
needed to properly set the ether_type field in the ethernet header
to ETHERTYPE_REVARP before transmitting the packet, otherwise
the bytes in ether_type would wind up reversed. It is important to note
that using htons(ETHERTYPE_REVARP) will not work. This is odd, because
the NetBSD rarpd uses htons(ETHERTYPE_REVARP). (Praise be to tcpdump:
I would never have been able to track this silliness down without it.)
- The update_arptab() function has been castrated. It depends on
SIOCSARP which has been deprecated in 4.4BSD. The NetBSD people
don't seem to be using this function either. It wouldn't be too
hard to replace this with equivalent code from arp.c, but it
might not be necessary.
- I put together an ether_ntohost() support function that allows
both local (/etc/ethers) and NIS lookups. This stuff should go
in libc at some point, but nothing else seems to need it for now,
so it can wait a while.
As you may have guessed, you need to have the Berkeley Packet Filter in
your kernel in order to use this program. The good news is that together
with the recently added bootparamd, you can use finally use a FreeBSD
box to boot Sun boxes over the network. (This was my whole motivation
for getting this stuff to work: I have this one subnet that has a whole
bunch of Sun3 X-terminals on it with only two Sun4 workstations, both
of which are locked in peoples' offices. If those two machines crash
(and they do every so often) then none of the X-terms will boot. Now I
can use a spare PC that I have as a boot server. :)
