gcc memcpy "knows" about types that are supposed to be actually already
aligned and triggers alignment errors doing the memcpy itself.
"Fix" this by changing it to a bcopy(). In this case, we had:
struct timeval *tp;
struct timeval tv1;
memcpy(&tv1,tp,sizeof(tv1));
.. and since gcc *knows* that a pointer to a timeval is longword aligned
and that tv1 is longword aligned, then it can use an inline that assumes
alignment. The following works too:
cp = (char *)tp;
memcpy(&tv1,cp,sizeof(tv1));
Simply casting (char *)tp for the memcpy doesn't work. :-(
This affected different 64 bit platforms in different ways and depends
a lot on gcc as well. I've seen this on alpha and ia64 at least, although
alpha isn't doing it right now.
o Fix zero payloading, unbreak ping -s 0.
o Increase socket recieve buffer, ping -s 65467 is working now.
Submitted by: anti-magic sweep based on kris's patch
Reviewed by: bde, silence on -audit
MFC after: 2 months
It does not help modern compilers, and some may take some hit from it.
(I also found several functions that listed *every* of its 10 local vars with
"register" -- just how many free registers do people think machines have?)
right; after a single packet was dropped it beeped after every
transmission.
Change its implementation to only output a bell when there is an
increase in the maximum value of the number of packets that were
sent but not yet received. This has the benefit that even for very
long round-trip times, ping -A will do roughly the right thing
after a few inital false-positives.
Reviewed by: ru
COPTS towards the end of final CFLAGS so that it can be used to
override Makefile and other defaults. Using it in Makefiles risks
having options set using it clobbered when somebody uses it on the
command line.
Approved by: bde
Avoid using parenthesis enclosure macros (.Pq and .Po/.Pc) with plain text.
Not only this slows down the mdoc(7) processing significantly, but it also
has an undesired (in this case) effect of disabling hyphenation within the
entire enclosed block.
The original code was certainly broken; it knows that whereto is
to be used for a sockaddr_in, so it should be declared as such.
To support multiple protocols, there is also a sockaddr_storage
struct that can be used; I don't think struct sockaddr is supposed
to be used anywhere other than for casts and pointers.
Submitted by: Ian Dowse <iedowse@maths.tcd.ie>
MFC after: 3 weeks
This one is strange and goes against my rusty compiler knowledge.
The global declaration
struct sockaddr whereto;
produces for both i386 && alpha:
.comm whereto,16,1
which means common storage, byte aligned. Ahem. I though structs
were supposed to be ALDOUBLE always? I mean, w/o pragma packed?
Later on, this address is coerced to:
to = (struct sockaddr_in *)&whereto;
Up until now, we've been fine on alpha because the address
just ended up aligned to a 4 byte boundary. Lately, though,
it end up as:
0000000120027b0f B whereto
And, tra la, you get unaligned access faults. The solution I picked, in
lieu of understanding what the compiler was doing, is to put whereto
as a union of a sockaddr and sockaddr_in. That's more formally correct
if somewhat awkward looking.
user runs with privilege, allowing the sending of icmp packets with
larger size (up to 48k, the default receive buffer size in ping),
which is useful for network driver development testing, as well
as experimentation with fragmentation.
Reviewed by: wpaul
add an upper limit to -t
match the types of return values and the variables they are stuffed in
make the man page and usage() a little more consistantly ugly
less obfuscation.
Submitted by: adrian, billf
o main returns int not void
o use braces to avoid potentially ambiguous else
Note: The fix to natd is potentially functional in nature since I used
the indentation as the right thing rather than the struct semantics.
Someone more familiar with the code should double check me on this one.
Reviewed by: obrien and chuckr
In the words of the submitter:
"The patch below allows to ping from any address on the multihomed host.
The man page is also updated, the text was cutted from traceroute(8)."
Submitted by: Ruslan Ermilov
PR: 6832
extremely useful for networking testing. Other options secured from
user-level D.O.S. attacks. -f, -s now root-only. -i wait times < 1.0
root-only. -c count limited to 100 and defaults to 16 when ping run
by non-root user.
Fixes bin/6649 and removes the last abusive signal handler.
Use SO_TIMESTAMP to get the kernel to timestamp packets on reception.
Fixes bin/5658 and provides slightly better accuracy.
Explicitly zero and terminate the IP options when using -R.
PR: bin/5658
PR: bin/6649
are unaligned for access by the alpha, so copy the value to a variable
that is aligned.
When checking the returned data, be careful to avoid confusing the
size of the icmp header with the size of a timeval. On i386 these
are both 8, but on alpha, a timeval is 16 bytes. This means that
a packet sent from an alpha contains 48 bytes of data, not 56 like
on i386.
This isn't necessarily the best statistic, but it is by far the easiest to
calculate. Update the man page to be more explicit about precisely which
statistics are printed out. Revert some of jmg's bogus man page changes from
rev 1.11.
unreachable hosts. Note that most of this consists of telling SIGINT
and SIGALRM to interrupt the system call, instead of restarting them.
Also try to get rid of some potential races Bruce didn't like; hopefully
they aren't a problem (potential or otherwise) now.
Reviewed by: julian
this is a NO-NO
re-arange to just set a "please die immediatly" flag in the signal handler
and handle this in the normal thread.
also handle ping -f better on slow links by backing off a bit when
we get a ENOBUFFS from the sendto().
2578 from Julian A. Likely not strictly needed, but it doesn't hurt
and protects ping against possible buffer overflows if the resolver
were to return large IP addresses.
used spaces to align the second line under the program name.
2) Cache uid after call to setuid(getuid()) so we don't waste a system call
for each packet with a call to getuid for the -v case.
3) Update manual to reflect new restriction on -l from last delta.
Suggested by: bde, Bill Fenner
2) Must be root to run preload (OpenBSD ping.c 1.8)
3) Don't print all replies unless verbose and root (from idea in
OpenBSD ping.c 1.10 and 1.11) to avoid leaking information available
only to root.
4) Remove unused h: from option string to getopt.
5) Make the compiler happy with exit(0) (Lite-2?)
Reviewed by: Dan Cross <tenser@spitfire.ecsel.psu.edu>
Good candidate for 2.2 and 2.1 (as are many of the 1.17 changes).
Do a better job of argument parsing.
Don't permit ping -f to a multicast address (very antisocial).
Don't permit -L, -I, -T options with unicast addresses.
Ensure that we ask for only AF_INET addresses (should close PR#2584).
Return <sysexits.h> error codes for failures. Document this.
Fix man page to identify the author and put sections in correct order.
Submitted by: Bruce Murphy <packrat@iinet.net.au>
Add '-a' audible flag, so terminal will beep upon receipt of a reply
packet. Useful for debugging ethernet runs, among other things.
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.
Simplified average calculations. This also fixes potential overflow errors
and minor rounding inconsistencies (always round to nearest now).
Don't lose more SIGINFOs than necessary.
Print \r at the start of the status message. This clears the dots that
are sometimes left by ping -f.
Reviewed by: sef
Turn this behavior off using '-Q'. This makes '-v' useless other than as
an ICMP-sniffer, which tcpdump is better at anyway.
Print out another couple of ICMP messages, and fix the printing of the
original packet (mostly byte order problems).
Note, this is not really a security risk, because the buffer in question
is a static variable in the data segment and not on the stack, and hence
cannot subert the flow of execution in any way. About the worst case was
that if you pinged a long hostname, ping could coredump.
Pointed out on: bugtraq (listserv@netspace.org)
determine whether a connection to a given machine is up or not.
In FreeBSD 2.0 (and therefore, I assume, BSD 4.4) the exit code of ping
is always zero, even if no packets were received.
I would like to propose the following change to /usr/src/sbin/ping/ping.c
to restore this useful behaviour:
Submitted by: Denis Fortin
Print out summary information on receipt of SIGINFO; also, stop the
kernel printing of information and restore it on exit. Now, it needs
an option to be quiet. ;)
