Perl Behavioral Differences Closing Child Process Spawned with open() vs. IPC::Open3 - perl

I'm trying to figure this out but haven't been able to wrap my head around it. I need to open a piped subprocess and read from its output. Originally I was using the standard open() call like this:
#!/usr/bin/perl;
use warnings;
use strict;
use Scalar::Util qw(openhandle);
use IPC::Open3;
my $fname = "/var/log/file.log.1.gz";
my $pid = open(my $fh, "-|:encoding(UTF-8)", "gunzip -c \"$fname\" | tac");
# Read one line from the file
while (my $row = <$fh>) {
print "Row: $row\n";
last; # Bail out early
}
# Check if the PID is valid and kill it if so
if (kill(0, $pid) == 1) {
kill(15, $pid);
waitpid($pid, 0);
$pid = 0;
}
# Close the filehandle if it is still open
if (openhandle($fh)) {
close $fh;
}
The above works, except that I get errors from tac in the logs saying:
tac: write error
From what I can tell from various testing and research that I've done, this is happening because killing the PID returned from open() just kills the first child process (but not the second) and so when I then close the filehandle, tac is still writing to it, thus the "write error" due to the broken pipe. The strange thing is, at times when I check ($? >> 8) if the close() call returns false, it will return 141, indicating it received a SIGPIPE (backing up my theory above). However, other times it returns 0 which is strange.
Furthermore, if I run the same command but without a double-pipe (only a single one), like this (everything else the same as above):
my $pid = open(my $fh, "-|:encoding(UTF-8)", "gunzip -c \"$fname\"");
...I'll get an error in the logs like this:
gzip: stdout: Broken pipe
...but in this case, gunzip/gzip was the only process (which I killed via the returned PID), so I'm not sure why it would still be writing to the pipe when I close the filehandle (since it was supposed to be killed already, AND waited for with waitpid()).
I'm trying to repro this in the Perl debugger but its difficult because I can't get the stderr of the child process with plain open() (the way I'm seeing the external process' stderr in prod is in the apache2 logs - this is a CGI script).
I understand from reading the docs that I can't get the PID of all child processes in a multi-piped open with open(), so I decided to try and resort to a different method so that I could close all processes cleanly. I tried open3(), and interestingly, without making any changes (literally running basically the same exact scenario as above but with open3() instead of open()):
my $pid = open3(my $in, my $fh, undef, "gunzip -c \"$fname\"");
...and then killing it just like I did above, I don't get any errors. This holds true for both the single piped process as shown above, as well as the double-piped process that involves piping to "tac".
Therefore, I'm wondering what I am missing here? I know there are differences in the way open() and open3() work, but are there differences in the way that child processes are spawned from them? In both cases I can see that the initial child (the PID returned) is itself a child of the Perl process. But its almost as if the process spawned by open(), is not getting properly killed and/or cleaned up (via waitpid()) while the same process spawned by open3() is, and that's the part I can't figure out.
And, more to the bigger picture and the issue at hand - what is the suggestion for the best way to cleanly close a multi-piped process in this sort of scenario? Am I spending more time than is warranted on this? The script itself works as it should aside from these errors, so if it turns out that the tac and gzip errors I'm seeing are inconsequential, should I just live with them and move on?
Any help is much appreciated!

If you just want to read the last line of a gzipped file, it's easy to do it in pure perl without calling an external program:
#!/usr/bin/env perl
use warnings;
use strict;
use feature qw/say/;
use IO::Uncompress::Gunzip qw/$GunzipError/;
my $fname = 'foo.txt.gz';
my $z = new IO::Uncompress::Gunzip $fname or die "Couldn't open file: $GunzipError\n";
my $row;
while (<$z>) {
$row = $_;
}
say "Row: $row";

This happens because either your perl script or its parent is ignoring the SIGPIPE signal, and the ignore signal dispositions are inherited by the children.
Here is a simpler testcase for your condition:
$ perl -e '$SIG{PIPE}="IGNORE"; open my $fh, "-|", "seq 100000 | tac; true"; print scalar <$fh>'
100000
tac: write error
$ (trap "" PIPE; perl -e 'open my $fh, "-|", "seq 100000 | tac"; print scalar <$fh>')
100000
tac: write error
$ (trap "" PIPE; perl -e 'my $pid = open my $fh, "-|", "seq 100000 | tac"; print scalar <$fh>; kill 15, $pid; waitpid $pid, 0')
100000
$ tac: write error
The latter version does the same kill as the version from the OP, which will not kill either the right or left side of the pipeline, but the shell running and waiting for both (some shells will exec through the left side of a pipeline; with such shells, a ; exit $? could be appended to the command in order to reproduce the example).
A case where SIGPIPE is ignored upon entering a perl script is when run via fastcgi -- which sets the SIGPIPE disposition to ignore, and expects the script to handle it. In that case simply setting an SIGPIPE handler instead of IGNORE (even an empty handler) would work, since in that case the signal disposition will be reset to default upon executing external commands:
$SIG{PIPE} = sub { };
open my $fh, '-|', 'trap - PIPE; ... | tac';
When run as a standalone script it could be some setup bug (I've see it happen in questions related to containerization on Linux), or someone trying to exploit buggy programs running with elevated privileges not bothering to handle write(2) errors (EPIPE in this case).
my $pid = open3(my $in, my $fh, undef, "gunzip -c \"$fname\"");
...and then killing it just like I did above, I don't get any errors.
Where should you get the errors from, if you're redirecting its stderr to the same $fh you only read the first line from?
The thing is absolutely no different with open3:
$ (trap "" PIPE; perl -MIPC::Open3 -e 'my $pid = open3 my $in, my $out, my $err, "seq 100000 | tac 2>/dev/tty"; print scalar <$out>')
100000
$ tac: write error

Related

perl open() always returns the PID of the sh instead of the underlying program

I have to kill a program that I am opening via
$pid = open(FH, "program|")
or
$pid = or open(FH, "-|", "program")
However, the program (mosquittto_sub, to be specific) still lingers around in the background, because open is returning the PID of the sh that perl is using to run the program, so I am only killing the sh wrapper instead of the actual program.
Is there a way to get the programs real PID? What is the point of getting the sh's PID?
There are a few ways to deal with this.
First, you can use a list form to open a process and then no shell is involved so the child process (with pid returned by open) is precisely the one with the program you need to stop
my #cmd = ('progname', '-arg1', ...);
my $pid = open my $fh, '-|', #cmd // die "Can't open \"#cmd\": $!";
...
my $num_signaled = kill 15, $pid;
This sketch needs some checks added. Please see the linked documentation (look for "pipe").
If this isn't suitable for some reason -- perhaps you need the shell to run that program -- then you can find the program's pid, and Proc::ProcessTable module is good for this. A basic demo
use Proc::ProcessTable;
my $prog_name = ...
my $pid;
my $pt = Proc::ProcessTable->new();
foreach my $proc (#{$pt->table}) {
if ($proc->cmndline =~ /\Q$prog_name/) { # is this enough to identify it?
$pid = $proc->pid;
last;
}
}
my $num_signaled = kill 15, $pid;
Please be careful with identifying the program by its name -- on a modern system there may be all kinds of processes running that contain the name of the program you want to terminate. For more detail and discussion please see this post and this post, for starters.
Finally, you can use a module to run your external programs and then you'll be able to manage and control them far more nicely. Here I'd recommend IPC::Run.

How to get STDERR in Perl from a command executed in pipe with su -c

I'm trying to capture the output of the command executed as a different user using:
my $command = qq(sudo su - <username> -c '/usr/bin/whatever');
my $pid = open $cmdOutput, "-|", $command;
How can I capture the STDERR of /usr/bin/whatever?
I tried
$pid = open $cmdOutput, "-|", $command || die " something went wrong: $!";
but it looks like this is capturing the possible errors of "open" itself.
I also tried
my $command = qq(sudo su - <username> -c '/usr/bin/whatever' 2>/tmp/error.message);
which will redirect the STDERR to the file, which I can parse later, but I wanted some more straightforward solution.
Also, I only want to use core modules.
This is covered thoroughly in perlfaq8. Since you are using a piped open, the relevant examples are those that go by open3 from the core IPC::Open3 module.
Another option is to use IPC::Run for managing your processes, and the pump function will do what you need. The IPC::Open3 documentation says for IPC::Run
This is a CPAN module that has better error handling and more facilities than Open3.
With either of these you can manipulate STDOUT and STDERR separately or together, as needed. For convenient and complete output capture also see Capture::Tiny.
Other than 2>output redirection, there are no more elementary methods for the piped open.
If you don't mind mixing the streams or losing STDOUT altogether, another option is
my $command = 'cmd 2>&1 1>/dev/null' # Remove 1>/dev/null to have both
my $pid = open my $cmdOutput, "-|", $command;
while (<$cmdOutput>) { print } # STDERR only
The first redirection merges STDERR stream with STDOUT so you get them both, and mixed (with STDOUT subject to buffering, thus things may well come out of order). The second redirect sends the STDOUT away so with it in place you read only the command's STDERR from the handle.
The question is about running an external command using open but I'd like to mention that the canonical and simple qx (backticks) can be used in the same way. It returns the STDOUT so redirection just like above is needed to get STDERR. For completeness:
my $cmd = 'cmd_to_execute';
my $allout = qx($cmd 2>&1); # Both STDOUT and STDERR in $out, or
my $stderr = qx($cmd 2>&1 1>/dev/null); # Only STDERR
my $exit_status = $?;
The qx puts the child process exit code (status) in $?. This can then be inspected for failure modes; see a summary in the qx page or a very thorough discussion in I/O operators in perlop.
Note that the STDERR returned this way is from the command, if it ran. If the command itself couldn't be run (for a typo in command name, or fork failed for some reason) then $? will be -1 and the error will be in $!.
As suggested by zdim I used the IPC::Open3 module for the matter and I've got something like this doing the job for me
$instanceCommand = qq(sudo su - <username> -c '<command>');
my ($infh,$outfh,$errfh,$pid);
$errfh = gensym();
$pid = open3($infh, $outfh, $errfh, $instanceCommand);
my $sel = new IO::Select;
$sel->add($outfh,$errfh);
while (my #ready = $sel->can_read){
foreach my $fh (#ready){
my $line =<$fh>;
if (not defined $line){
$sel->remove($fh);
next;
}
if ($fh == $outfh){
chomp($line);
#<----- command output processing ----->
}
elsif ($fh == $errfh){
chomp $line;
#<----- command error processing ----->
}
else {
die "Reading from something else\n";
}
}
}
waitpid($pid, 0);
Maybe not completely bullet proof, but its working fine for me. Even whilst executing funny cascaded script as < command > .
The desired destination, opened for writing, could be dup()'ed to FD #2

Why system() returns 0 even though the program it executes dies

I'm trying to test a piece of code ($code) that should make sure that only one instance of the program is running at a time:
#!/usr/bin/perl
# test_lock
use strict;
use warnings;
( my $code = <<'CODE') =~ s/^\s+//gm;
#!/usr/bin/perl
use strict;
use warnings;
use Fcntl qw(:flock);
# Make sure only one instance of the program is running at a time.
open our $Lock, '<', $0 or die "Can't lock myself $0: $!";
flock $Lock, LOCK_EX | LOCK_NB
or die "Another instance of $0 is already running. Exiting ...\n";
sleep(2);
CODE
my $progfile = '/tmp/x';
open my $fh, '>', $progfile or die $!;
print $fh $code;
close $fh;
$|++;
my $ex1 = system("perl $progfile &");
print "First system(): $ex1\n";
my $ex2 = system("perl $progfile");
print "Second system(): $ex2\n";
I expected that the second call to system() would return a non-zero value ($ex2) as it can't get the lock and dies. However I get:
$ perl test_lock
First system(): 0
Another instance of /tmp/x is already running. Exiting ...
Second system(): 0
What is wrong with my assumption? (Is there a better way to test the $code?)
I think it likely because you have a race condition. How do you know that error is actually coming from your second process?
Because if you for example, run:
perl /tmp/x & perl /tmp/x ; echo $?
You may get a zero return, because the 'winner' of the race may well be the latter process (which return code you're catching). (Try it a few times, and you'll see different results)
You also do have slight difference is what the shell is doing between the two commands - from the docs:
If there is only one scalar argument, the argument is checked for shell metacharacters, and if there are any, the entire argument is passed to the system's command shell for parsing (this is /bin/sh -c on Unix platforms, but varies on other platforms). If there are no shell metacharacters in the argument, it is split into words and passed directly to execvp , which is more efficient.
So actually you should see invocation of sh before perl in your first, which means it's actually more likely to take longer to get to the lock point.
That means your command is more like:
sh -c "perl /tmp/x"& perl /tmp/x; echo $?
Run that a few times and see how many times you get non-zero error codes. It's not often, because usually the 'delay' of the shell start up is enough to ensure that the second instance wins the race most of the time!
If you've linux - try strace -fTt yourscript which will trace the execution flow. Or you can make judicious use of $$ to report the process-pid when running.
In both cases, you are obtaining the exit code of the shell you launch. Roughly speaking, the shell returns the exit code of the last program it ran.
Since the shell created by system("perl $progfile &") doesn't wait for the child to end, it will virtually always return 0 since launching perl in the background is unlikely to result in an error.
So if the second instance of perl managed to obtain the lock first, you'll get the outcome you got. This race condition can be seem more clearly if you identify the source of the exception.
#!/usr/bin/perl
# test_lock
use strict;
use warnings;
( my $code = <<'CODE') =~ s/^\s+//gm;
#!/usr/bin/perl
use strict;
use warnings;
use Fcntl qw(:flock);
# Make sure only one instance of the program is running at a time.
open our $Lock, '<', $0 or die "Can't lock myself $0: $!";
flock $Lock, LOCK_EX | LOCK_NB
or die "$ARGV[0]: Another instance of $0 is already running. Exiting ...\n";
sleep(2);
CODE
my $progfile = 'b.pl';
open my $fh, '>', $progfile or die $!;
print $fh $code;
close $fh;
$|++;
my $ex1 = system("perl $progfile 1 &");
print "First system(): $ex1\n";
my $ex2 = system("perl $progfile 2");
print "Second system(): $ex2\n";
Output:
$ perl a.pl
First system(): 0
1: Another instance of b.pl is already running. Exiting ...
Second system(): 0
$ perl a.pl
First system(): 0
2: Another instance of b.pl is already running. Exiting ...
Second system(): 2816

How can I redirect the output of Perl's system() to a filehandle?

With the open command in Perl, you can use a filehandle. However I have trouble getting back the exit code with the open command in Perl.
With the system command in Perl, I can get back the exit code of the program I'm running. However I want to just redirect the STDOUT to some filehandle (no stderr).
My stdout is going to be a line-by-line output of key-value pairs that I want to insert into a mao in perl. That is why I want to redirect only my stdout from my Java program in perl. Is that possible?
Note: If I get errors, the errors get printed to stderr. One possibility is to check if anything gets printed to stderr so that I can quite the Perl script.
Canonically, if you're trying to get at the text output of a forked process, my understanding is that's what the backticks are for. If you need the exit status as well, you can check it with the $? special variable afterward, e.g.:
open my $fh, '>', "output.txt" or die $!;
print {$fh} `echo "Hello!"`;
print "Return code: $?\n";
Output to STDERR from the command in backticks will not be captured, but will instead be written directly to STDERR in the Perl program it's called from.
You may want to check out IPC::System::Simple -- it gives you many options for executing external commands, capturing its output and return value, and optionally dying if a bad result is returned.
This is one of the ways to do it.
open my $fh, '>', $file;
defined(my $pid = fork) or die "fork: $!";
if (!$pid) {
open STDOUT, '>&', $fh;
exec($command, #args);
}
waitpid $pid, 0;
print $? == 0 ? "ok\n" : "nok\n";
Use open in -| mode. When you close the filehandle, the exit status will be in $?.
open my $fh, '-|', "$command"; # older version: open my $fh, "$command |";
my #command_output = <$fh>;
close $fh;
my $command_status = $?;
From perldoc -f close
If the file handle came from a piped open, "close" will
additionally return false if one of the other system calls
involved fails, or if the program exits with non-zero status.
(If the only problem was that the program exited non-zero, $!
will be set to 0.) Closing a pipe also waits for the process
executing on the pipe to complete, in case you want to look at
the output of the pipe afterwards, and implicitly puts the exit
status value of that command into $? and
"${^CHILD_ERROR_NATIVE}".

How can I run a system command in Perl asynchronously?

I currently have a Perl script that runs an external command on the system, gathers the output, and performs some action based on what was returned. Right now, here is how I run this (where $cmd is a string with the command setup):
#output = `$cmd`;
I'd like to change this so if the command hangs and does not return a value after so much time then I kill the command. How would I go about running this asynchronously?
There's a LOT of ways to do this:
You can do this with a fork (perldoc -f fork)
or using threads (perldoc threads). Both of these make passing the returned information back to the main program difficult.
On systems that support it, you can set an alarm (perldoc -f alarm) and then clean up in the signal handler.
You can use an event loop like POE or Coro.
Instead of the backticks, you can use open() or respectively open2 or open3 (cf. IPC::Open2, IPC::Open3) to start a program while getting its STDOUT/STDERR via a file handle. Run non-blocking read operations on it. (perldoc -f select and probably google "perl nonblocking read")
As a more powerful variant of the openX()'s, check out IPC::Run/IPC::Cmd.
Probably tons I can't think of in the middle of the night.
If you really just need to put a timeout on a given system call that is a much simpler problem than asynchronous programming.
All you need is alarm() inside of an eval() block.
Here is a sample code block that puts these into a subroutine that you could drop into your code. The example calls sleep so isn't exciting for output, but does show you the timeout functionality you were interested in.
Output of running it is:
/bin/sleep 2 failure: timeout at
./time-out line 15.
$ cat time-out
#!/usr/bin/perl
use warnings;
use strict;
my $timeout = 1;
my #cmd = qw(/bin/sleep 2);
my $response = timeout_command($timeout, #cmd);
print "$response\n" if (defined $response);
sub timeout_command {
my $timeout = (shift);
my #command = #_;
undef $#;
my $return = eval {
local($SIG{ALRM}) = sub {die "timeout";};
alarm($timeout);
my $response;
open(CMD, '-|', #command) || die "couldn't run #command: $!\n";
while(<CMD>) {
$response .= $_;
}
close(CMD) || die "Couldn't close execution of #command: $!\n";
$response;
};
alarm(0);
if ($#) {
warn "#cmd failure: $#\n";
}
return $return;
}
If your external program doesn't take any input, look for the following words in the perlipc manpage:
Here's a safe backtick or pipe open for read:
Use the example code and guard it with an alarm (which is also explained in perlipc).
I coded below to run rsync on 20 directories simultaneously (in parallel instead of sequentially requiring me to wait hours for it to complete):
use threads;
for my $user ( keys %users ) {
my $host = $users{$user};
async {
system <<~ "SHELL";
ssh $host \\
rsync_user $user
SHELL
}
}
$ pgrep -lf rsync | wc -l
20
Not sure if it's best or even a good solution, but I was glad that it worked for my use case.
With this you get a mixed output on screen (what I ignored anyway), but it does its job successfully.
threads pragma exports the (very useful) async function by default.
rsync_user is my Perl script that wraps rsync command with options, and source and target directories set.
Ran on FreeBSD 13.1 with Perl 5.32.1