Commit Graph

191 Commits

Author SHA1 Message Date
Todd C. Miller
c6f9f06c45 Check return value of dispatch_pending_signals() in case we received
SIGINT or SIGQUIT before executing the command.
2017-04-07 08:38:56 -06:00
Todd C. Miller
35e4bd3e9f Replace pipe_nonblock() with pipe2() 2017-03-13 12:11:52 -06:00
Todd C. Miller
6dba84dca9 Reorganize the command execution code to separate out the pty and
non-pty code paths into their own event loops.  The non-pty exec
code is now contained in exec_nopty.c and the pty exec code is split
between exec_pty.c (parent process) and exec_monitor.c (session leader).
This results in a small bit of duplicated code but improves readability.
Some of the duplicated code will fall out in future changes to the
event subsystem (the signal pipe).
2017-03-03 10:35:11 -07:00
Todd C. Miller
13fcdb4f5f Set the child pid to -1 after we've waited for it and take care to
avoid killing pid -1.  This makes it a bit more explicit and removes
the need for a separate variable to track the child's status.
Sudo already stops processing signals after it receives SIGCHLD so
it is not vulnerable to CVE-2017-2616.
2017-02-24 15:14:56 -07:00
Todd C. Miller
dde2b5eb2c Close execfd in parent processes where it is not needed. 2017-01-22 18:56:13 -08:00
Todd C. Miller
3f022419ae Be consistent with the naming of the variable used to store the
function return value.  Previously, some code used "rval", some
used "ret".  This standardizes on "ret" and uses "rc" for temporary
return codes.
2016-09-08 16:38:08 -06:00
Todd C. Miller
a9570e64ff fix brace style 2016-08-31 08:34:07 -06:00
Todd C. Miller
1ec4d9918d Move sudo_debug_execve() call into sudo_execve(). 2016-06-09 10:48:59 -06:00
Todd C. Miller
a2e541aef8 O_NOCTTY has no effect when opening /dev/tty as the open can only
succeed if there is already a controlling tty.
2016-05-16 11:17:20 -06:00
Todd C. Miller
00727a310c Ignore the result of send() on exec error, if it fails the other
end of the pipe is gone and we are headed for exit.
Coverity CID 104066.
2016-05-07 04:47:12 -06:00
Todd C. Miller
6473d55aa7 Cast the return value of fcntl() to void when setting FD_CLOEXEC.
Coverity CID 104063, 104064, 104069, 104070, 104071, 104072, 104073, 104074
2016-05-05 16:16:24 -06:00
Todd C. Miller
829917f008 pid_t is defined by POSIX as a signed integer type so we don't need
a cast when comparing to -1.
2016-05-04 14:14:38 -06:00
Todd C. Miller
c70da6dcdf In dispatch_signal() for stopped processes check for tcgetpgrp()
returning -1.  Also change checks from "saved_pgrp != -1" to
"fd != -1".  Coverity CID 104098.
2016-05-04 14:13:44 -06:00
Todd C. Miller
3e9702580e style fix; fork_cmnd should start on a new line 2016-04-25 09:58:32 -06:00
Todd C. Miller
b20977d445 Add support for using fexecve() if supported on commands that are
checksummed.
2016-01-04 10:35:18 -07:00
Todd C. Miller
3f6f1cd15c When a terminal device is closed, SIGHUP is sent to the controlling
process associated with that terminal.  It is not sent to the entire
process group so sudo needs to relay SIGHUP to the command when it
is not being run in a new pty.  Bug #719
2015-09-15 10:30:36 -06:00
Todd C. Miller
d08faa73a7 Fill in cstat if exec_setup() fails. Previously it was only filled
in for an execve() failure.  Fixes an unkillable sudo process when
exec_setup() fails and I/O logging is enabled.
2015-09-09 10:50:21 -06:00
Todd C. Miller
13869d349c Linux sets si_pid in struct siginfo to 0 when the process that sent
the signal is in a different container since the PID namespaces in
different conatiners are separate.  Avoid looking up the process
group by id when si_pid is 0 since getpgid(0) returns the process
group of the current process.  Since sudo ignores signals sent
by processes in its own process group, this had the effect of
ignoring signals sent from other containers.  From Maarten de Vries
2015-08-10 15:13:37 -06:00
Todd C. Miller
0b241088b3 There's no need to conditionalize the #include <unistd.h>, we require
a POSIX system.
2015-07-02 09:08:28 -06:00
Todd C. Miller
4a07b472f0 Only include stddef.h where it is needed. 2015-06-20 05:34:35 -06:00
Todd C. Miller
c36415417f Add function name to "unable to allocate memory" warnings. 2015-06-19 14:51:17 -06:00
Todd C. Miller
dc883f2454 We require ANSI C so stop using the obsolete STDC_HEADERS. 2015-06-19 14:29:27 -06:00
Todd C. Miller
cb63ca701c Avoid using exiting allocators in the front end. 2015-06-17 17:00:54 -06:00
Todd C. Miller
59ab26dbcc Go back to a 2 args debug_decl and just use the "default" instance,
now renamed "active".
2015-02-01 08:24:49 -07:00
Todd C. Miller
29039859b7 Defer registration of the SIGCHLD handler until just before we exec
the command.  Fixes a problem where pam_gnome_keyring installs its
own SIGCHLD handler and may not restore the original one.  As a
result, we now have to explicitly wait for the askpass helper to
finish.  Bug #657
2014-11-18 14:05:51 -07:00
Todd C. Miller
21166d4da9 Make sure that SIGCHLD is not treated as a user-generated signal
in which case it could be ignored.  Bug #676
2014-11-04 09:24:41 -07:00
Todd C. Miller
17a2a27e46 Use generic bitmap macros instead of select-style fd_set. 2014-10-23 14:37:27 -06:00
Todd C. Miller
866cfc4fc3 Add support for multiple Debug lines per program. Callers may
register arbitrary debug facilities or use built-in defaults.  We
now use separate Debug statements for plugins and programs.
2014-10-22 13:23:05 -06:00
Todd C. Miller
328c563791 Check return value of sigaction(), even though it should never fail. 2014-09-04 10:13:26 -06:00
Todd C. Miller
0546f0b6a2 Convert a debug printf to a user-visible warning. 2014-08-31 20:32:18 -06:00
Todd C. Miller
cfa4cebf13 Remove signal_event from evbase before calling sudo_ev_loopexit()
when the command has exited or been killed.  It is possible that
we could receive another signal on the pipe if they are delivered
out of order.
2014-08-30 16:24:53 -06:00
Todd C. Miller
ae0014d6f4 Attempt to handle systems with SA_SIGINFO but that lack SI_USER. 2014-07-28 20:25:46 -06:00
Todd C. Miller
36a5767e3e efree -> sudo_efree for consistency 2014-07-10 15:35:04 -06:00
Todd C. Miller
2d61d38c23 Add sudo_ prefix to alloc.c functions and rename alloc.h -> sudo_alloc.h 2014-06-27 10:48:31 -06:00
Todd C. Miller
aecef4aa1d Rename warning/fatal -> sudo_warn/sudo_fatal to avoid namespace
pollution in libsudo_util.so.
2014-06-27 09:30:52 -06:00
Todd C. Miller
826e190f31 Ignore signals sent by the command's process group, not just the
command itself.  If we cannot determine the process group ID of the
sender (as it may no longer exist), just check the process ID.
2014-05-28 09:50:14 -06:00
Todd C. Miller
b121da2e21 In handler_user_only() only forward the signal if it was not generated
by the command.  This should fix a problem with programs that catch
SIGTSTP, perform cleanup, and then re-send the signal to their
process group (of which sudo is the leader).
2014-05-27 10:51:31 -06:00
Todd C. Miller
d628e17ead Handle EINTR from write(2) when writing to pipes and socket pairs. 2014-05-27 10:16:49 -06:00
Todd C. Miller
36b991348c Forward SIGINFO to running command if supported. If the command
is being run in the background (or exec_background is set in sudoers),
it is the sudo process, not the actual command, that receives the ^T.
2014-04-04 15:30:15 -06:00
Todd C. Miller
640a5ddb48 If the event loop exits due to an error and we are not logging I/O,
kill the command if still running.  Fixes a bug where sudo could
exit while the command was still running.
2014-01-14 20:34:20 -07:00
Todd C. Miller
d7277fc96e If not logging I/O we may get EOF when the command is executed and
the other end of the backchannel is closed.  Just remove the
backchannel event in this case or we will continue to receive
the event.  Bug #631
2014-01-14 10:54:56 -07:00
Todd C. Miller
27598bc0ab Update copyright year. 2014-01-07 10:43:31 -07:00
Todd C. Miller
d013b23ce3 If invoked as sudoedit we can't just exec the command directly since
the temporary files need to be updated before sudo exits.
2014-01-03 16:41:25 -07:00
Todd C. Miller
1adeda54ef Add support for preventing fds from getting clobbered by closefrom(). 2013-12-20 11:14:32 -07:00
Todd C. Miller
96eb2c4f8f Add warning_gettext() wrapper function that changes to the user locale,
then calls gettext().
Add U_ macro that calls warning_gettext() instead of gettext().
Rename warning2()/error2() back to warning_nodebug()/error_nodebug().
2013-11-18 08:59:57 -07:00
Todd C. Miller
99b7351de0 Fix some #if vs. #ifdef and remove an extraneous semicolon.
Bug #624; from Daniel Richard G.
2013-11-17 16:15:36 -07:00
Todd C. Miller
fd88ed42c8 Use MSG_WAITALL when receiving struct command_status over the Unix
domain socket since we no longer use datagrams.  This should avoid
the need to handle incomplete reads, though in theory it is still
possible.
2013-10-29 15:26:32 -06:00
Todd C. Miller
8861e01d16 Add support for libevent-style timed events. Adding a timed event
is currently O(n).  The only consumer of timed events is sudoreplay
which only used a singled one so O(n) == O(1) for now.  This also
allows us to remove the nanosleep compat function as we now use a
timeout event instead.
2013-10-28 10:00:09 -06:00
Todd C. Miller
d8b368b503 Now that sudo_ev_base_free() removes all events before freeing we
don't need to do this by hand.
2013-10-26 07:52:59 -06:00
Todd C. Miller
e8ce021e7d Quiet some llvm check false positives. The common idiom of using
TAILQ_FIRST, TAILQ_REMOVE and free in a loop to free each entry in
a TAILQ confuses llvm.  Use TAILQ_FOREACH_SAFE instead (which is
probably faster anyway).
2013-10-22 14:58:00 -06:00