fix sign.

Add a sanity check for the ip header length (from FreeBSD)

fix various small typos, mainly in comments.

fix various typos in comments.

ping, ping6: fix comment about ID field

Since ping.c 1.76 and ping6.c 1.58 from 2004-04-22, the ID field
contains random bits instead of a process ID.

un-ifdef sgi. NFC

Fix cksum calculation for clearing the cached route.

In ping.c:r1.104, the size of echoreply packet was changed to ICMP_MINLEN,
Therefore also calculation size must be ICMP_MINLEN.

branches: 1.115.2;
Use an explicitly aligned buffer for ip.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Reject packet intervals < 1 ns as they lead to infinite loops adding
zero timespecs.

Fix the behind-schedule behavior so it doesn't spend all its time in
that loop adding very small timespecs. Try ping -c 500 -i 0.000000001
to see this in action with the old ping.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Check for values between INT_MAX and LONG_MAX (if they're different)
when using strtol to get an int. This applies to the -c and -l options;
the other uses were already checked.

Also limit the inter-packet interval given with -i to values that
don't cause integer overflow calling poll() with milliseconds.

Really large intervals (the number is read as floating point) can
produce positive poll() values but negative integers when converted to
struct timespec; this produces behavior akin to using -l at first and
could be construed as a local DoS vulnerability.

Make optspace an array of uint8_t, since some of the bytes written into
it are outside the signed char range.

Manipulate the signal masks so signals only arrive while we're waiting,
so they can't illegally reenter libc.

Fixes the problem Greg A. Woods describes in PR 51267, which might or
might not be the original submitter's problem.

Don't call variables that are used "dummy".

branches: 1.109.2;
Adjust default packet size to 56 data bytes (64 total).
Make error messages consistent.

PR/49423: Martin Husemann: ping for small packets does not work in -7 or
-current
XXX: pullup 7?

branches: 1.107.4;
avoid bogus uninitialized gcc warning

more precision for F_TIMING64

deduct phdrlen after computing max size.

1. Allocate the max packet size before accounting for phdrlen, harmless.
2. In the clear-route-cache sendto, don't send 0 bytes (if -s was specified
with < 8, phdrlen would be 0).
3. Always send ICMP_MINLEN packets; this is what everyone else does. Makes
ping -s n where n < 8 work.
4. The condition for checking the data bytes was completely wrong. only check
the data bytes if we got all of them.
5. The condition for printing a newline was wrong; before it would not print
a newline before printing the data bytes, and it would append to the previous
error message.

Fix a bug that misunderstand F_TIMING64, F_POLICY and F_AUTHHDR.

branches: 1.102.2; 1.102.6; 1.102.8;
include <netipsec/ipsec.h> rather than <netinet6/ipsec.h> from userland
where possible, for consistency and compatibility to FreeBSD
(exception: KAME specific statistics gathering in netstat(1) and systat(1))

branches: 1.101.2;
print summary stats.

Sort options and descriptions in standard order.
Remove unecessary Bk/Ek in SYNOPSIS.

PR/45257: Instead of passing a 32bit sec and 32 bit usec timestamp in little
endian format by default, pass a struct timespec in native host format. Add
-C flag to produce a compatible timestamp like before.

static + __dead

handle the first loop where time == 0.

we need difftime to return a signed result, so clamp the timeval's to
uint32_t but then, return a signed result of their difference.

PR/45257: Ryo Shimizu: ping(8) prints bogus round-trip times after Year 2038
Clamp all time computations to 32 bits; idea from dsl@

Don't use array access to compute addresses that are definitely beyond
the static array boundaries.

Make get/setuid prog ops (so that i can do rump.ping -f regardless
of host privs).

RUMP_ACTION -> RUMPPRG

add compile-conditional rumpclient support

remove ancient ifdefs (Aleksej Saushev)

fix sign-compare issues

PR/41111: Ed Ravin: ping -R gives misleading error when remote side doesn't
support record route
Deal with source route and record route specially giving a meaningful error
message.
Pullup to 5.0?

branches: 1.87.10; 1.87.12; 1.87.16;
Shutdown sloop socket for read as it's never read on to
prevent mbufs from queueing in its recv buf.
PR:21459

branches: 1.86.4; 1.86.6;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

Change error message to say "setuid", forgotten in previous commit.

Pointed out by Ray Lai, thanks!

On 2nd thought, setuid() is better here.

PR/28741: Michael Santos: ping does [not] drop root privileges

Moved socket calls way up, and called seteuid(getuid()).

prtsig: remove a mysterious abort().

Use SA_NOKERNINFO instead of playing with the tty.

Factor out the tty code and explain we prefer to stop in the ^Z bg case
when we exit, rather than kill the tty setting.

PR/33623: Chuck Cranor: Ping stops when ran in the background
because it tries to set the tty not to print kerninfo. Change it
to only only play with the tty when ping is running in the foreground
(and will not stop when calling tcsetattr()). In my opinion, it is
preferable to print the kerninfo line with the ping info message
rather than to mess with the tty, but that's just me.

change (mostly) int to socklen_t. GCC 4 doesn't like that int and
socklen_t are different signness.

branches: 1.77.2; 1.77.4; 1.77.6;
There's no point in bringing PRNG output to network byteorder prior to
transmission.

do not disclose endian. henning@openbsd

branches: 1.75.2; 1.75.4; 1.75.6;
do not disclose endianness/sizeof(long) on probe packet. idea from openbsd

Homogenize usage messages: make the 'usage' word all lowercase, as this seems
to be the most common practice in our tree.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22308, verified by myself.

permit -s 0. from Maxim Konovalov

set sa_len right

die if strdup fails

use strlcpy. initialize sa_len correctly.

Touch up error/warning messages.

select() -> poll()

Add support for printing out messages about missing packet gaps. This is
done with the -v switch. If -q is also given with -v, then only the
packet gap messages will be printed along with summary after each gap
message.

u_short cleansing (use u_int16_t where appropriate)

clarify comment when nreceived > ntransmitted. from deraadt

Sync getopt() / man page with actual getopt options.

fix a couple of -Wshadow warnings

- use IP_MAXPACKET instead of 65535.
- change max packet size from 65468 to 65467(= IP_MAXPACKET-60-8).

- fix overrun bug.
- bump bufsize to fix "packet loss" with large packet.

convert to use getprogname()

correct use of howmany() for fd_set. commented by deraadt

Format string cleanups by Bill Sommerfeld.

avoid fd_set overflow. see openbsd select(2).

branches: 1.55.4;
sync with latest libipsec and kernel.

Fix data comparison to not overrun array bounds.

do not pass NULL to errx().

From: Chris Joness <cjones@rupert.honors.montana.edu>

do not bark even if ipsec policy setting for "route flushing 127.0.0.1"
ping socket is not successful. it usually means that the kernel is
not ipsec ready, and in that case this should be just okay.

0 -> STDIN_FILENO

IPsec support (specify policy on ping - quite useful to test
if IPsec is working or not) is added.

Support for ping -a: beep when packet received.
Slightly tweaked from version submitted by andrew@untraceable.net in PR7815

Correct nokerninfo tty flag being left set in most error cases.
Closes PR/7388 by Dave Huang <khym@bga.com>.

branches: 1.47.2;
Don't bother doing stddev calculation when n==1, since that results in
a computation of 0.0/0.0.
On IEEE systems you get a NaN from this; non-IEEE systems probably
give you a SIGFPE.

Once more around the hickory shrub. Faster and simpler stddev calculation from
Bill Studenmund.

Fix calculation of standard deviation. From Brian C. Grayson <bgrayson@orac.ece.utexas.edu>

Add code from Daniel Hagerty <hag@ai.mit.edu> to print standard deviation
when printing out final statistics line.

Closes PR bin/6198

char -> unsigned char

Fix usage string;
- option to specify maxwait is -w, not -i.
- add missing -h host and -Q option.

Update to the 980911 version from ftp.rhyolite.com:/src/ping.tar.Z

Add getuid() check if -f is set, fixes pr 6222 by Uwe Arndt.

Restrict -l to the superuser as suggested in PR 3973 from <soren@t.dk>.
Also distinguish "-f" from "-i 0.x" in the superuser restriction messages.

Nuke the termios stuff from the 1997/03/11 import, which prevents ping
from being run in the background of most shells without redirecting stdin
from /dev/null. (AFAICT, the termios stuff is useless and does not
interfere with SIGINFO handling.)

__AUDIT__ cleanup.

Take into consideration that icmp_nextmtu is stored in network byte order;
from Leonard Samuelson in PR bin/5230.

PR/4615: David Jones: Ping usage does not show -o option.

Make the "failed to clear cached route" a warning only if verbose is set.
This always happens when lo0 is not configured and confuses new users.

lint

branches: 1.32.2;
* cleanup .Nm
* deprecate register
* cleanup for WARNS=1
* bzero(), bcopy() -> memset(), memmove()
* prefix hex numbers with '0x'

convert to host byte order when printing stats of received ICMP
packets other than echo-replies; from Dave Huang in PR bin/3876

Update from Vernon Schryver, fixing the kluge to flush the route cache.

Make ping obey the -n flag in places.

Don't allow < 1 sec ping interval to regular users.

Try adding `o' option again; connection died.

Take care of alignment problems on the alpha: From John Birell, via cgd.

PR/3329: From David Jones: Add F_ONCE

Fix PR/3327: rcvd_tbl index was assumed to be in host order, creating
incorrect dup reports. From enami tsugutomo

- remove -g from Makefile
- fix setsockopt and recvfrom casts.

- Use err, warn, errx
- Don't use floor, because we would need libm
- Add maxwait option from old ping
- Add a centralized gethost() to get hostnames
- Fix PR/1014 and PR/2017; (adds -D, -T options, and -I and -t for unicast)
- Document exit status

make ping(8) print out statistics when given a SIGINFO.

fix a couple of bugs pointed out by John Birrell, though fix some
differently he did. (1) don't assume that the timeval in
the packet is correctly aligned, (2) don't compare beyond the end
of the packet, (3) minor cast for printf happiness.

Add and document options -w (to specify a maximum wait timeout, useful
for testing network interfaces) and -S (to specify a source address
on a machine with multiple interfaces).

sys/signal.h -> signal.h

Fill in sin_len.

Use inet_aton(), not inet_addr().

if no packets were received, exit with status 1. this should be
documented somewhere. As suggested by Rens Troost.

Update to use timer{add,sub}().

convert to new RCS Id conventions; reduce my headache

-Wall cleanups from Jim Jegers

will need <sys/queue.h>

Remove some more uses of obsolete functions.

Eliminate uses of some obsolete functions.

Merge with 4.4-Lite version.

Another multicast patch.
>From LBL and Steve Mccanne

Use doubles to rather than longs to avoid tsum overflowing so quickly.

Add RCS identifiers.

have ping do the right thing, wrt microseconds.
actually, now it'll overflow its counters sooner than before, and everything
should *really* be converted to doubles and/or "unsigned long long"s

changed "Id" to "Header" for rcsids

added rcs ids to all files

branches: 1.1.1;
Initial revision

fix various small typos, mainly in comments.

fix various typos in comments.

ping, ping6: fix comment about ID field

Since ping.c 1.76 and ping6.c 1.58 from 2004-04-22, the ID field
contains random bits instead of a process ID.

un-ifdef sgi. NFC

Fix cksum calculation for clearing the cached route.

In ping.c:r1.104, the size of echoreply packet was changed to ICMP_MINLEN,
Therefore also calculation size must be ICMP_MINLEN.

branches: 1.115.2;
Use an explicitly aligned buffer for ip.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Reject packet intervals < 1 ns as they lead to infinite loops adding
zero timespecs.

Fix the behind-schedule behavior so it doesn't spend all its time in
that loop adding very small timespecs. Try ping -c 500 -i 0.000000001
to see this in action with the old ping.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Check for values between INT_MAX and LONG_MAX (if they're different)
when using strtol to get an int. This applies to the -c and -l options;
the other uses were already checked.

Also limit the inter-packet interval given with -i to values that
don't cause integer overflow calling poll() with milliseconds.

Really large intervals (the number is read as floating point) can
produce positive poll() values but negative integers when converted to
struct timespec; this produces behavior akin to using -l at first and
could be construed as a local DoS vulnerability.

Make optspace an array of uint8_t, since some of the bytes written into
it are outside the signed char range.

Manipulate the signal masks so signals only arrive while we're waiting,
so they can't illegally reenter libc.

Fixes the problem Greg A. Woods describes in PR 51267, which might or
might not be the original submitter's problem.

Don't call variables that are used "dummy".

branches: 1.109.2;
Adjust default packet size to 56 data bytes (64 total).
Make error messages consistent.

PR/49423: Martin Husemann: ping for small packets does not work in -7 or
-current
XXX: pullup 7?

branches: 1.107.4;
avoid bogus uninitialized gcc warning

more precision for F_TIMING64

deduct phdrlen after computing max size.

1. Allocate the max packet size before accounting for phdrlen, harmless.
2. In the clear-route-cache sendto, don't send 0 bytes (if -s was specified
with < 8, phdrlen would be 0).
3. Always send ICMP_MINLEN packets; this is what everyone else does. Makes
ping -s n where n < 8 work.
4. The condition for checking the data bytes was completely wrong. only check
the data bytes if we got all of them.
5. The condition for printing a newline was wrong; before it would not print
a newline before printing the data bytes, and it would append to the previous
error message.

Fix a bug that misunderstand F_TIMING64, F_POLICY and F_AUTHHDR.

branches: 1.102.2; 1.102.6; 1.102.8;
include <netipsec/ipsec.h> rather than <netinet6/ipsec.h> from userland
where possible, for consistency and compatibility to FreeBSD
(exception: KAME specific statistics gathering in netstat(1) and systat(1))

branches: 1.101.2;
print summary stats.

Sort options and descriptions in standard order.
Remove unecessary Bk/Ek in SYNOPSIS.

PR/45257: Instead of passing a 32bit sec and 32 bit usec timestamp in little
endian format by default, pass a struct timespec in native host format. Add
-C flag to produce a compatible timestamp like before.

static + __dead

handle the first loop where time == 0.

we need difftime to return a signed result, so clamp the timeval's to
uint32_t but then, return a signed result of their difference.

PR/45257: Ryo Shimizu: ping(8) prints bogus round-trip times after Year 2038
Clamp all time computations to 32 bits; idea from dsl@

Don't use array access to compute addresses that are definitely beyond
the static array boundaries.

Make get/setuid prog ops (so that i can do rump.ping -f regardless
of host privs).

RUMP_ACTION -> RUMPPRG

add compile-conditional rumpclient support

remove ancient ifdefs (Aleksej Saushev)

fix sign-compare issues

PR/41111: Ed Ravin: ping -R gives misleading error when remote side doesn't
support record route
Deal with source route and record route specially giving a meaningful error
message.
Pullup to 5.0?

branches: 1.87.10; 1.87.12; 1.87.16;
Shutdown sloop socket for read as it's never read on to
prevent mbufs from queueing in its recv buf.
PR:21459

branches: 1.86.4; 1.86.6;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

Change error message to say "setuid", forgotten in previous commit.

Pointed out by Ray Lai, thanks!

On 2nd thought, setuid() is better here.

PR/28741: Michael Santos: ping does [not] drop root privileges

Moved socket calls way up, and called seteuid(getuid()).

prtsig: remove a mysterious abort().

Use SA_NOKERNINFO instead of playing with the tty.

Factor out the tty code and explain we prefer to stop in the ^Z bg case
when we exit, rather than kill the tty setting.

PR/33623: Chuck Cranor: Ping stops when ran in the background
because it tries to set the tty not to print kerninfo. Change it
to only only play with the tty when ping is running in the foreground
(and will not stop when calling tcsetattr()). In my opinion, it is
preferable to print the kerninfo line with the ping info message
rather than to mess with the tty, but that's just me.

change (mostly) int to socklen_t. GCC 4 doesn't like that int and
socklen_t are different signness.

branches: 1.77.2; 1.77.4; 1.77.6;
There's no point in bringing PRNG output to network byteorder prior to
transmission.

do not disclose endian. henning@openbsd

branches: 1.75.2; 1.75.4; 1.75.6;
do not disclose endianness/sizeof(long) on probe packet. idea from openbsd

Homogenize usage messages: make the 'usage' word all lowercase, as this seems
to be the most common practice in our tree.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22308, verified by myself.

permit -s 0. from Maxim Konovalov

set sa_len right

die if strdup fails

use strlcpy. initialize sa_len correctly.

Touch up error/warning messages.

select() -> poll()

Add support for printing out messages about missing packet gaps. This is
done with the -v switch. If -q is also given with -v, then only the
packet gap messages will be printed along with summary after each gap
message.

u_short cleansing (use u_int16_t where appropriate)

clarify comment when nreceived > ntransmitted. from deraadt

Sync getopt() / man page with actual getopt options.

fix a couple of -Wshadow warnings

- use IP_MAXPACKET instead of 65535.
- change max packet size from 65468 to 65467(= IP_MAXPACKET-60-8).

- fix overrun bug.
- bump bufsize to fix "packet loss" with large packet.

convert to use getprogname()

correct use of howmany() for fd_set. commented by deraadt

Format string cleanups by Bill Sommerfeld.

avoid fd_set overflow. see openbsd select(2).

branches: 1.55.4;
sync with latest libipsec and kernel.

Fix data comparison to not overrun array bounds.

do not pass NULL to errx().

From: Chris Joness <cjones@rupert.honors.montana.edu>

do not bark even if ipsec policy setting for "route flushing 127.0.0.1"
ping socket is not successful. it usually means that the kernel is
not ipsec ready, and in that case this should be just okay.

0 -> STDIN_FILENO

IPsec support (specify policy on ping - quite useful to test
if IPsec is working or not) is added.

Support for ping -a: beep when packet received.
Slightly tweaked from version submitted by andrew@untraceable.net in PR7815

Correct nokerninfo tty flag being left set in most error cases.
Closes PR/7388 by Dave Huang <khym@bga.com>.

branches: 1.47.2;
Don't bother doing stddev calculation when n==1, since that results in
a computation of 0.0/0.0.
On IEEE systems you get a NaN from this; non-IEEE systems probably
give you a SIGFPE.

Once more around the hickory shrub. Faster and simpler stddev calculation from
Bill Studenmund.

Fix calculation of standard deviation. From Brian C. Grayson <bgrayson@orac.ece.utexas.edu>

Add code from Daniel Hagerty <hag@ai.mit.edu> to print standard deviation
when printing out final statistics line.

Closes PR bin/6198

char -> unsigned char

Fix usage string;
- option to specify maxwait is -w, not -i.
- add missing -h host and -Q option.

Update to the 980911 version from ftp.rhyolite.com:/src/ping.tar.Z

Add getuid() check if -f is set, fixes pr 6222 by Uwe Arndt.

Restrict -l to the superuser as suggested in PR 3973 from <soren@t.dk>.
Also distinguish "-f" from "-i 0.x" in the superuser restriction messages.

Nuke the termios stuff from the 1997/03/11 import, which prevents ping
from being run in the background of most shells without redirecting stdin
from /dev/null. (AFAICT, the termios stuff is useless and does not
interfere with SIGINFO handling.)

__AUDIT__ cleanup.

Take into consideration that icmp_nextmtu is stored in network byte order;
from Leonard Samuelson in PR bin/5230.

PR/4615: David Jones: Ping usage does not show -o option.

Make the "failed to clear cached route" a warning only if verbose is set.
This always happens when lo0 is not configured and confuses new users.

lint

branches: 1.32.2;
* cleanup .Nm
* deprecate register
* cleanup for WARNS=1
* bzero(), bcopy() -> memset(), memmove()
* prefix hex numbers with '0x'

convert to host byte order when printing stats of received ICMP
packets other than echo-replies; from Dave Huang in PR bin/3876

Update from Vernon Schryver, fixing the kluge to flush the route cache.

Make ping obey the -n flag in places.

Don't allow < 1 sec ping interval to regular users.

Try adding `o' option again; connection died.

Take care of alignment problems on the alpha: From John Birell, via cgd.

PR/3329: From David Jones: Add F_ONCE

Fix PR/3327: rcvd_tbl index was assumed to be in host order, creating
incorrect dup reports. From enami tsugutomo

- remove -g from Makefile
- fix setsockopt and recvfrom casts.

- Use err, warn, errx
- Don't use floor, because we would need libm
- Add maxwait option from old ping
- Add a centralized gethost() to get hostnames
- Fix PR/1014 and PR/2017; (adds -D, -T options, and -I and -t for unicast)
- Document exit status

make ping(8) print out statistics when given a SIGINFO.

fix a couple of bugs pointed out by John Birrell, though fix some
differently he did. (1) don't assume that the timeval in
the packet is correctly aligned, (2) don't compare beyond the end
of the packet, (3) minor cast for printf happiness.

Add and document options -w (to specify a maximum wait timeout, useful
for testing network interfaces) and -S (to specify a source address
on a machine with multiple interfaces).

sys/signal.h -> signal.h

Fill in sin_len.

Use inet_aton(), not inet_addr().

if no packets were received, exit with status 1. this should be
documented somewhere. As suggested by Rens Troost.

Update to use timer{add,sub}().

convert to new RCS Id conventions; reduce my headache

-Wall cleanups from Jim Jegers

will need <sys/queue.h>

Remove some more uses of obsolete functions.

Eliminate uses of some obsolete functions.

Merge with 4.4-Lite version.

Another multicast patch.
>From LBL and Steve Mccanne

Use doubles to rather than longs to avoid tsum overflowing so quickly.

Add RCS identifiers.

have ping do the right thing, wrt microseconds.
actually, now it'll overflow its counters sooner than before, and everything
should *really* be converted to doubles and/or "unsigned long long"s

changed "Id" to "Header" for rcsids

added rcs ids to all files

branches: 1.1.1;
Initial revision

fix various typos in comments.

ping, ping6: fix comment about ID field

Since ping.c 1.76 and ping6.c 1.58 from 2004-04-22, the ID field
contains random bits instead of a process ID.

un-ifdef sgi. NFC

Fix cksum calculation for clearing the cached route.

In ping.c:r1.104, the size of echoreply packet was changed to ICMP_MINLEN,
Therefore also calculation size must be ICMP_MINLEN.

branches: 1.115.2;
Use an explicitly aligned buffer for ip.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Reject packet intervals < 1 ns as they lead to infinite loops adding
zero timespecs.

Fix the behind-schedule behavior so it doesn't spend all its time in
that loop adding very small timespecs. Try ping -c 500 -i 0.000000001
to see this in action with the old ping.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Check for values between INT_MAX and LONG_MAX (if they're different)
when using strtol to get an int. This applies to the -c and -l options;
the other uses were already checked.

Also limit the inter-packet interval given with -i to values that
don't cause integer overflow calling poll() with milliseconds.

Really large intervals (the number is read as floating point) can
produce positive poll() values but negative integers when converted to
struct timespec; this produces behavior akin to using -l at first and
could be construed as a local DoS vulnerability.

Make optspace an array of uint8_t, since some of the bytes written into
it are outside the signed char range.

Manipulate the signal masks so signals only arrive while we're waiting,
so they can't illegally reenter libc.

Fixes the problem Greg A. Woods describes in PR 51267, which might or
might not be the original submitter's problem.

Don't call variables that are used "dummy".

branches: 1.109.2;
Adjust default packet size to 56 data bytes (64 total).
Make error messages consistent.

PR/49423: Martin Husemann: ping for small packets does not work in -7 or
-current
XXX: pullup 7?

branches: 1.107.4;
avoid bogus uninitialized gcc warning

more precision for F_TIMING64

deduct phdrlen after computing max size.

1. Allocate the max packet size before accounting for phdrlen, harmless.
2. In the clear-route-cache sendto, don't send 0 bytes (if -s was specified
with < 8, phdrlen would be 0).
3. Always send ICMP_MINLEN packets; this is what everyone else does. Makes
ping -s n where n < 8 work.
4. The condition for checking the data bytes was completely wrong. only check
the data bytes if we got all of them.
5. The condition for printing a newline was wrong; before it would not print
a newline before printing the data bytes, and it would append to the previous
error message.

Fix a bug that misunderstand F_TIMING64, F_POLICY and F_AUTHHDR.

branches: 1.102.2; 1.102.6; 1.102.8;
include <netipsec/ipsec.h> rather than <netinet6/ipsec.h> from userland
where possible, for consistency and compatibility to FreeBSD
(exception: KAME specific statistics gathering in netstat(1) and systat(1))

branches: 1.101.2;
print summary stats.

Sort options and descriptions in standard order.
Remove unecessary Bk/Ek in SYNOPSIS.

PR/45257: Instead of passing a 32bit sec and 32 bit usec timestamp in little
endian format by default, pass a struct timespec in native host format. Add
-C flag to produce a compatible timestamp like before.

static + __dead

handle the first loop where time == 0.

we need difftime to return a signed result, so clamp the timeval's to
uint32_t but then, return a signed result of their difference.

PR/45257: Ryo Shimizu: ping(8) prints bogus round-trip times after Year 2038
Clamp all time computations to 32 bits; idea from dsl@

Don't use array access to compute addresses that are definitely beyond
the static array boundaries.

Make get/setuid prog ops (so that i can do rump.ping -f regardless
of host privs).

RUMP_ACTION -> RUMPPRG

add compile-conditional rumpclient support

remove ancient ifdefs (Aleksej Saushev)

fix sign-compare issues

PR/41111: Ed Ravin: ping -R gives misleading error when remote side doesn't
support record route
Deal with source route and record route specially giving a meaningful error
message.
Pullup to 5.0?

branches: 1.87.10; 1.87.12; 1.87.16;
Shutdown sloop socket for read as it's never read on to
prevent mbufs from queueing in its recv buf.
PR:21459

branches: 1.86.4; 1.86.6;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

Change error message to say "setuid", forgotten in previous commit.

Pointed out by Ray Lai, thanks!

On 2nd thought, setuid() is better here.

PR/28741: Michael Santos: ping does [not] drop root privileges

Moved socket calls way up, and called seteuid(getuid()).

prtsig: remove a mysterious abort().

Use SA_NOKERNINFO instead of playing with the tty.

Factor out the tty code and explain we prefer to stop in the ^Z bg case
when we exit, rather than kill the tty setting.

PR/33623: Chuck Cranor: Ping stops when ran in the background
because it tries to set the tty not to print kerninfo. Change it
to only only play with the tty when ping is running in the foreground
(and will not stop when calling tcsetattr()). In my opinion, it is
preferable to print the kerninfo line with the ping info message
rather than to mess with the tty, but that's just me.

change (mostly) int to socklen_t. GCC 4 doesn't like that int and
socklen_t are different signness.

branches: 1.77.2; 1.77.4; 1.77.6;
There's no point in bringing PRNG output to network byteorder prior to
transmission.

do not disclose endian. henning@openbsd

branches: 1.75.2; 1.75.4; 1.75.6;
do not disclose endianness/sizeof(long) on probe packet. idea from openbsd

Homogenize usage messages: make the 'usage' word all lowercase, as this seems
to be the most common practice in our tree.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22308, verified by myself.

permit -s 0. from Maxim Konovalov

set sa_len right

die if strdup fails

use strlcpy. initialize sa_len correctly.

Touch up error/warning messages.

select() -> poll()

Add support for printing out messages about missing packet gaps. This is
done with the -v switch. If -q is also given with -v, then only the
packet gap messages will be printed along with summary after each gap
message.

u_short cleansing (use u_int16_t where appropriate)

clarify comment when nreceived > ntransmitted. from deraadt

Sync getopt() / man page with actual getopt options.

fix a couple of -Wshadow warnings

- use IP_MAXPACKET instead of 65535.
- change max packet size from 65468 to 65467(= IP_MAXPACKET-60-8).

- fix overrun bug.
- bump bufsize to fix "packet loss" with large packet.

convert to use getprogname()

correct use of howmany() for fd_set. commented by deraadt

Format string cleanups by Bill Sommerfeld.

avoid fd_set overflow. see openbsd select(2).

branches: 1.55.4;
sync with latest libipsec and kernel.

Fix data comparison to not overrun array bounds.

do not pass NULL to errx().

From: Chris Joness <cjones@rupert.honors.montana.edu>

do not bark even if ipsec policy setting for "route flushing 127.0.0.1"
ping socket is not successful. it usually means that the kernel is
not ipsec ready, and in that case this should be just okay.

0 -> STDIN_FILENO

IPsec support (specify policy on ping - quite useful to test
if IPsec is working or not) is added.

Support for ping -a: beep when packet received.
Slightly tweaked from version submitted by andrew@untraceable.net in PR7815

Correct nokerninfo tty flag being left set in most error cases.
Closes PR/7388 by Dave Huang <khym@bga.com>.

branches: 1.47.2;
Don't bother doing stddev calculation when n==1, since that results in
a computation of 0.0/0.0.
On IEEE systems you get a NaN from this; non-IEEE systems probably
give you a SIGFPE.

Once more around the hickory shrub. Faster and simpler stddev calculation from
Bill Studenmund.

Fix calculation of standard deviation. From Brian C. Grayson <bgrayson@orac.ece.utexas.edu>

Add code from Daniel Hagerty <hag@ai.mit.edu> to print standard deviation
when printing out final statistics line.

Closes PR bin/6198

char -> unsigned char

Fix usage string;
- option to specify maxwait is -w, not -i.
- add missing -h host and -Q option.

Update to the 980911 version from ftp.rhyolite.com:/src/ping.tar.Z

Add getuid() check if -f is set, fixes pr 6222 by Uwe Arndt.

Restrict -l to the superuser as suggested in PR 3973 from <soren@t.dk>.
Also distinguish "-f" from "-i 0.x" in the superuser restriction messages.

Nuke the termios stuff from the 1997/03/11 import, which prevents ping
from being run in the background of most shells without redirecting stdin
from /dev/null. (AFAICT, the termios stuff is useless and does not
interfere with SIGINFO handling.)

__AUDIT__ cleanup.

Take into consideration that icmp_nextmtu is stored in network byte order;
from Leonard Samuelson in PR bin/5230.

PR/4615: David Jones: Ping usage does not show -o option.

Make the "failed to clear cached route" a warning only if verbose is set.
This always happens when lo0 is not configured and confuses new users.

lint

branches: 1.32.2;
* cleanup .Nm
* deprecate register
* cleanup for WARNS=1
* bzero(), bcopy() -> memset(), memmove()
* prefix hex numbers with '0x'

convert to host byte order when printing stats of received ICMP
packets other than echo-replies; from Dave Huang in PR bin/3876

Update from Vernon Schryver, fixing the kluge to flush the route cache.

Make ping obey the -n flag in places.

Don't allow < 1 sec ping interval to regular users.

Try adding `o' option again; connection died.

Take care of alignment problems on the alpha: From John Birell, via cgd.

PR/3329: From David Jones: Add F_ONCE

Fix PR/3327: rcvd_tbl index was assumed to be in host order, creating
incorrect dup reports. From enami tsugutomo

- remove -g from Makefile
- fix setsockopt and recvfrom casts.

- Use err, warn, errx
- Don't use floor, because we would need libm
- Add maxwait option from old ping
- Add a centralized gethost() to get hostnames
- Fix PR/1014 and PR/2017; (adds -D, -T options, and -I and -t for unicast)
- Document exit status

make ping(8) print out statistics when given a SIGINFO.

fix a couple of bugs pointed out by John Birrell, though fix some
differently he did. (1) don't assume that the timeval in
the packet is correctly aligned, (2) don't compare beyond the end
of the packet, (3) minor cast for printf happiness.

Add and document options -w (to specify a maximum wait timeout, useful
for testing network interfaces) and -S (to specify a source address
on a machine with multiple interfaces).

sys/signal.h -> signal.h

Fill in sin_len.

Use inet_aton(), not inet_addr().

if no packets were received, exit with status 1. this should be
documented somewhere. As suggested by Rens Troost.

Update to use timer{add,sub}().

convert to new RCS Id conventions; reduce my headache

-Wall cleanups from Jim Jegers

will need <sys/queue.h>

Remove some more uses of obsolete functions.

Eliminate uses of some obsolete functions.

Merge with 4.4-Lite version.

Another multicast patch.
>From LBL and Steve Mccanne

Use doubles to rather than longs to avoid tsum overflowing so quickly.

Add RCS identifiers.

have ping do the right thing, wrt microseconds.
actually, now it'll overflow its counters sooner than before, and everything
should *really* be converted to doubles and/or "unsigned long long"s

changed "Id" to "Header" for rcsids

added rcs ids to all files

branches: 1.1.1;
Initial revision

ping, ping6: fix comment about ID field

Since ping.c 1.76 and ping6.c 1.58 from 2004-04-22, the ID field
contains random bits instead of a process ID.

un-ifdef sgi. NFC

Fix cksum calculation for clearing the cached route.

In ping.c:r1.104, the size of echoreply packet was changed to ICMP_MINLEN,
Therefore also calculation size must be ICMP_MINLEN.

branches: 1.115.2;
Use an explicitly aligned buffer for ip.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Reject packet intervals < 1 ns as they lead to infinite loops adding
zero timespecs.

Fix the behind-schedule behavior so it doesn't spend all its time in
that loop adding very small timespecs. Try ping -c 500 -i 0.000000001
to see this in action with the old ping.

PR bin/36997 Zafer Aydogan: ping doesn't validate numeric inputs enough.

Check for values between INT_MAX and LONG_MAX (if they're different)
when using strtol to get an int. This applies to the -c and -l options;
the other uses were already checked.

Also limit the inter-packet interval given with -i to values that
don't cause integer overflow calling poll() with milliseconds.

Really large intervals (the number is read as floating point) can
produce positive poll() values but negative integers when converted to
struct timespec; this produces behavior akin to using -l at first and
could be construed as a local DoS vulnerability.

Make optspace an array of uint8_t, since some of the bytes written into
it are outside the signed char range.

Manipulate the signal masks so signals only arrive while we're waiting,
so they can't illegally reenter libc.

Fixes the problem Greg A. Woods describes in PR 51267, which might or
might not be the original submitter's problem.

Don't call variables that are used "dummy".

branches: 1.109.2;
Adjust default packet size to 56 data bytes (64 total).
Make error messages consistent.

PR/49423: Martin Husemann: ping for small packets does not work in -7 or
-current
XXX: pullup 7?

branches: 1.107.4;
avoid bogus uninitialized gcc warning

more precision for F_TIMING64

deduct phdrlen after computing max size.

1. Allocate the max packet size before accounting for phdrlen, harmless.
2. In the clear-route-cache sendto, don't send 0 bytes (if -s was specified
with < 8, phdrlen would be 0).
3. Always send ICMP_MINLEN packets; this is what everyone else does. Makes
ping -s n where n < 8 work.
4. The condition for checking the data bytes was completely wrong. only check
the data bytes if we got all of them.
5. The condition for printing a newline was wrong; before it would not print
a newline before printing the data bytes, and it would append to the previous
error message.

Fix a bug that misunderstand F_TIMING64, F_POLICY and F_AUTHHDR.

branches: 1.102.2; 1.102.6; 1.102.8;
include <netipsec/ipsec.h> rather than <netinet6/ipsec.h> from userland
where possible, for consistency and compatibility to FreeBSD
(exception: KAME specific statistics gathering in netstat(1) and systat(1))

branches: 1.101.2;
print summary stats.

Sort options and descriptions in standard order.
Remove unecessary Bk/Ek in SYNOPSIS.

PR/45257: Instead of passing a 32bit sec and 32 bit usec timestamp in little
endian format by default, pass a struct timespec in native host format. Add
-C flag to produce a compatible timestamp like before.

static + __dead

handle the first loop where time == 0.

we need difftime to return a signed result, so clamp the timeval's to
uint32_t but then, return a signed result of their difference.

PR/45257: Ryo Shimizu: ping(8) prints bogus round-trip times after Year 2038
Clamp all time computations to 32 bits; idea from dsl@

Don't use array access to compute addresses that are definitely beyond
the static array boundaries.

Make get/setuid prog ops (so that i can do rump.ping -f regardless
of host privs).

RUMP_ACTION -> RUMPPRG

add compile-conditional rumpclient support

remove ancient ifdefs (Aleksej Saushev)

fix sign-compare issues

PR/41111: Ed Ravin: ping -R gives misleading error when remote side doesn't
support record route
Deal with source route and record route specially giving a meaningful error
message.
Pullup to 5.0?

branches: 1.87.10; 1.87.12; 1.87.16;
Shutdown sloop socket for read as it's never read on to
prevent mbufs from queueing in its recv buf.
PR:21459

branches: 1.86.4; 1.86.6;
Remove more duplicate #includes,
from Slava Semushin <slava.semushin@gmail.com>

Change error message to say "setuid", forgotten in previous commit.

Pointed out by Ray Lai, thanks!

On 2nd thought, setuid() is better here.

PR/28741: Michael Santos: ping does [not] drop root privileges

Moved socket calls way up, and called seteuid(getuid()).

prtsig: remove a mysterious abort().

Use SA_NOKERNINFO instead of playing with the tty.

Factor out the tty code and explain we prefer to stop in the ^Z bg case
when we exit, rather than kill the tty setting.

PR/33623: Chuck Cranor: Ping stops when ran in the background
because it tries to set the tty not to print kerninfo. Change it
to only only play with the tty when ping is running in the foreground
(and will not stop when calling tcsetattr()). In my opinion, it is
preferable to print the kerninfo line with the ping info message
rather than to mess with the tty, but that's just me.

change (mostly) int to socklen_t. GCC 4 doesn't like that int and
socklen_t are different signness.

branches: 1.77.2; 1.77.4; 1.77.6;
There's no point in bringing PRNG output to network byteorder prior to
transmission.

do not disclose endian. henning@openbsd

branches: 1.75.2; 1.75.4; 1.75.6;
do not disclose endianness/sizeof(long) on probe packet. idea from openbsd

Homogenize usage messages: make the 'usage' word all lowercase, as this seems
to be the most common practice in our tree.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22308, verified by myself.

permit -s 0. from Maxim Konovalov

set sa_len right

die if strdup fails

use strlcpy. initialize sa_len correctly.

Touch up error/warning messages.

select() -> poll()

Add support for printing out messages about missing packet gaps. This is
done with the -v switch. If -q is also given with -v, then only the
packet gap messages will be printed along with summary after each gap
message.

u_short cleansing (use u_int16_t where appropriate)

clarify comment when nreceived > ntransmitted. from deraadt

Sync getopt() / man page with actual getopt options.

fix a couple of -Wshadow warnings

- use IP_MAXPACKET instead of 65535.
- change max packet size from 65468 to 65467(= IP_MAXPACKET-60-8).

- fix overrun bug.
- bump bufsize to fix "packet loss" with large packet.

convert to use getprogname()

correct use of howmany() for fd_set. commented by deraadt

Format string cleanups by Bill Sommerfeld.

avoid fd_set overflow. see openbsd select(2).

branches: 1.55.4;
sync with latest libipsec and kernel.

Fix data comparison to not overrun array bounds.

do not pass NULL to errx().

From: Chris Joness <cjones@rupert.honors.montana.edu>

do not bark even if ipsec policy setting for "route flushing 127.0.0.1"
ping socket is not successful. it usually means that the kernel is
not ipsec ready, and in that case this should be just okay.

0 -> STDIN_FILENO

IPsec support (specify policy on ping - quite useful to test
if IPsec is working or not) is added.

Support for ping -a: beep when packet received.
Slightly tweaked from version submitted by andrew@untraceable.net in PR7815

Correct nokerninfo tty flag being left set in most error cases.
Closes PR/7388 by Dave Huang <khym@bga.com>.

branches: 1.47.2;
Don't bother doing stddev calculation when n==1, since that results in
a computation of 0.0/0.0.
On IEEE systems you get a NaN from this; non-IEEE systems probably
give you a SIGFPE.

Once more around the hickory shrub. Faster and simpler stddev calculation from
Bill Studenmund.

Fix calculation of standard deviation. From Brian C. Grayson <bgrayson@orac.ece.utexas.edu>

Add code from Daniel Hagerty <hag@ai.mit.edu> to print standard deviation
when printing out final statistics line.

Closes PR bin/6198

char -> unsigned char

Fix usage string;
- option to specify maxwait is -w, not -i.
- add missing -h host and -Q option.

Update to the 980911 version from ftp.rhyolite.com:/src/ping.tar.Z

Add getuid() check if -f is set, fixes pr 6222 by Uwe Arndt.

Restrict -l to the superuser as suggested in PR 3973 from <soren@t.dk>.
Also distinguish "-f" from "-i 0.x" in the superuser restriction messages.

Nuke the termios stuff from the 1997/03/11 import, which prevents ping
from being run in the background of most shells without redirecting stdin
from /dev/null. (AFAICT, the termios stuff is useless and does not
interfere with SIGINFO handling.)

__AUDIT__ cleanup.

Take into consideration that icmp_nextmtu is stored in network byte order;
from Leonard Samuelson in PR bin/5230.

PR/4615: David Jones: Ping usage does not show -o option.

Make the "failed to clear cached route" a warning only if verbose is set.
This always happens when lo0 is not configured and confuses new users.

lint

branches: 1.32.2;
* cleanup .Nm
* deprecate register
* cleanup for WARNS=1
* bzero(), bcopy() -> memset(), memmove()
* prefix hex numbers with '0x'

convert to host byte order when printing stats of received ICMP
packets other than echo-replies; from Dave Huang in PR bin/3876

Update from Vernon Schryver, fixing the kluge to flush the route cache.

Make ping obey the -n flag in places.

Don't allow < 1 sec ping interval to regular users.

Try adding `o' option again; connection died.

Take care of alignment problems on the alpha: From John Birell, via cgd.

PR/3329: From David Jones: Add F_ONCE

Fix PR/3327: rcvd_tbl index was assumed to be in host order, creating
incorrect dup reports. From enami tsugutomo

- remove -g from Makefile
- fix setsockopt and recvfrom casts.

- Use err, warn, errx
- Don't use floor, because we would need libm
- Add maxwait option from old ping
- Add a centralized gethost() to get hostnames
- Fix PR/1014 and PR/2017; (adds -D, -T options, and -I and -t for unicast)
- Document exit status

make ping(8) print out statistics when given a SIGINFO.

fix a couple of bugs pointed out by John Birrell, though fix some
differently he did. (1) don't assume that the timeval in
the packet is correctly aligned, (2) don't compare beyond the end
of the packet, (3) minor cast for printf happiness.