#!/usr/local/cpanel/3rdparty/bin/perl

# cpanel - scripts/shrink_modsec_ip_database       Copyright 2022 cPanel, L.L.C.
#                                                           All rights reserved.
# copyright@cpanel.net                                         http://cpanel.net
# This code is subject to the cPanel license. Unauthorized copying is prohibited

package scripts::shrink_modsec_ip_database;

use strict;
use warnings;
use File::Temp                ();
use Cpanel::PwCache           ();
use Cpanel::FileUtils::Move   ();
use Cpanel::SafetyBits        ();
use Cpanel::SafetyBits::Chown ();
use Cpanel::AccessIds         ();
use Cpanel::SafeRun::Object   ();
use Cpanel::Imports;

our $MODSEC_SDBM_UTIL    = '/usr/sbin/modsec-sdbm-util';
our $DEFAULT_SECDATADIR  = '/var/cpanel/secdatadir';
our @DB_FILE_SUFFIXES    = qw( .pag .dir );                # Database file suffixes used by modsec-sdbm-util
our $NEW_DB_NAME         = 'new_db';                       # This name is hard-coded in modsec-sdbm-util
our $DB_PERMS            = 0640;                           # S_IRUSR | S_IWUSR | S_IRGRP
our $OTHER_EXECUTE_PERMS = 01;                             # S_IXOTH

sub new {
    my ( $pkg, $opts ) = @_;
    my $self = ref($opts) eq 'HASH' ? { %{$opts} } : {};
    bless $self, $pkg;
    return $self;
}

sub as_script {
    my $self = shift;
    logger->die('as_script() is a method call.') unless ref $self eq __PACKAGE__;

    if ( not $ARGV[0] or $ARGV[0] ne '-x' ) {
        my $msg = 'To execute, use the -x flag.';
        logger()->die($msg);
    }

    $self->run();

    return 1;
}

sub run {
    my $self = shift;
    logger->die('run() is a method call.') unless ref $self eq __PACKAGE__;

    return 0 unless $self->_bin_check;    # Bail out early and silently if the util is not installed

    my $databases = $self->_gather_databases();

    while ( my ( $db_path, $uid ) = each %{$databases} ) {
        if ( ( stat($MODSEC_SDBM_UTIL) )[2] & $OTHER_EXECUTE_PERMS ) {    # Can run util as "other" user?
            $self->_shrink_db_as_user( $uid, $db_path );
        }
        else {
            # Will have to settle for doing this as root.
            $self->_shrink_db( $uid, $db_path );
        }
    }
    return;
}

sub _bin_check {
    return -x $MODSEC_SDBM_UTIL ? 1 : 0;
}

sub _gather_databases {

    # All files that belong to the same database and that match @DB_FILE_SUFFIXES will need to have the same file owner or that database will not be in the final output
    my $self = shift;
    logger->die('_gather_databases() is a method call.') unless ref $self eq __PACKAGE__;

    return $self->{'databases'} if defined $self->{'databases'};

    my $secdatadir = $self->_secdatadir();

    my %databases;

    if ( opendir( my $dir_fh, $secdatadir ) ) {
      FILE: while ( my $filename = readdir($dir_fh) ) {
          SUFFIX: for my $suffix (@DB_FILE_SUFFIXES) {
                if ( $filename =~ m{ \A (.*) \Q$suffix\E \Z }xms ) {
                    my $short_name = $1;    # Filename without suffix

                    my $db_path = $secdatadir . '/' . $short_name;    # Database path name suitable for passing to modsec-sdbm-util
                    next FILE if exists $databases{$db_path};         # Move along if this belongs to a database already in the collection

                    my $owner = $self->_validate_database_files_owner($db_path);    # Check if there is a full set of files for this database path

                    if ( $self->_allowed_owner($owner) ) {
                        $databases{$db_path} = $owner;                              # Verified, add it to the collection
                        next FILE;
                    }

                }
            }
        }
        closedir($dir_fh);
    }
    return $self->{'databases'} = \%databases;
}

sub _shrink_db_as_user {
    my ( $self, $uid, $db_path ) = @_;
    logger->die('_shrink_db_as_user() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_shrink_db_as_user() called without expected arguments.') unless length $uid && length $db_path;
    return Cpanel::AccessIds::do_as_user( $uid, sub { $self->_shrink_db( $uid, $db_path ) } );
}

sub _shrink_db {
    my ( $self, $uid, $db_path ) = @_;
    logger->die('_shrink_db() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_shrink_db() called without expected arguments.') unless length $uid && length $db_path;

    my $secdatadir = $self->_secdatadir();

    my $workdir = File::Temp->newdir( CLEANUP => 1, TEMPLATE => 'shrink_modsec_db_XXXXXXXX', DIR => $secdatadir );
    Cpanel::SafetyBits::Chown::safe_chown_guess_gid( $uid, $workdir ) or logger->warn("Failed to chown $workdir to uid $uid");

    my @original_files = $self->_get_db_files($db_path);

    # modsec-sdbm-util will drop $NEW_DB_NAME * @DB_FILE_SUFFIXES files into $tempdir
    return 0 unless $self->_call_modsec_sdbm_util( $workdir, $db_path );

    # Verify new files exist and adjust perms
    my $new_db_path = $workdir . '/' . $NEW_DB_NAME;
    my @new_files   = map { $new_db_path . $_ } @DB_FILE_SUFFIXES;
    if ( !defined $self->_validate_database_files_owner($new_db_path) ) {    # root owned files = 0
        logger->warn("Failed to verify the database files generated by modsec-sdbm-util in the working directory");
        return 0;
    }
    $self->_set_default_perms( $uid, \@new_files );

    # Move the existing files to the workdir so we can revert if the new-file move fails
    my @revert_files = map { $workdir . '/original' . $_ } @DB_FILE_SUFFIXES;
    my $can_revert   = $self->_move_files( \@original_files, \@revert_files ) or logger->warn("Failed to move original files for $db_path into working dir");

    # Move new files into place
    if ( !$self->_move_files( \@new_files, \@original_files ) ) {
        logger->warn("Failed to move new files into place for $db_path");
        if ($can_revert) {
            $self->_move_files( \@revert_files, \@original_files ) or logger->warn("Failed to move backup files for $db_path from working dir to original location");
            $self->_set_default_perms( $uid, \@original_files );
        }
        else {
            logger->warn("Not able to restore original files for db_path");
        }
        return 0;
    }

    # Fix up final database permissions
    return 0 unless $self->_set_default_perms( $uid, \@original_files );

    return 1;
}

sub _call_modsec_sdbm_util {
    my ( $self, $tempdir, $db_path ) = @_;
    logger->die('_call_modsec_sdbm_util() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_call_modsec_sdbm_util() called without expected arguments.') unless length $tempdir && length $db_path;

    my $run = Cpanel::SafeRun::Object->new(
        program => $MODSEC_SDBM_UTIL,
        args    => [ '-D', $tempdir, '-v', '-n', $db_path ],
    );

    # For whatever reason, if the util fails to open the specified db it doesn't exit with an error code, so parse out the error message.
    # It will fail to open if the file is immutable -- which is a crazy thing to do on purpose -- but it doesn't make that obvious.
    if ( $run->stdout() =~ m{ ^ Failed \s to \s open \s sdbm: \s (.*) $ }xms ) {
        logger()->warn("$MODSEC_SDBM_UTIL failed to open database (try checking all file/dir attributes): $1");
        return 0;
    }

    if ( $run->CHILD_ERROR() ) {
        logger()->warn( "$MODSEC_SDBM_UTIL exited with non-zero status: " . join( q{ }, map { $run->$_() // () } qw( autopsy stdout stderr ) ) );
        return 0;
    }

    return 1;
}

sub _validate_database_files_owner {

    # Expects a database path such as "$secdatadir/$db_name" without a suffix
    # Returns owner (uid) of a full set of database files if they exist, undef otherwise
    # Remember that root has uid 0!
    my ( $self, $db_path ) = @_;
    logger->die('_validate_database_files_owner() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_validate_database_files_owner() called without expected arguments.') unless length $db_path;

    my $owner;
    for my $file ( $self->_get_db_files($db_path) ) {
        return unless -f $file;           # All generated filenames must exist
        my $seen = ( stat(_) )[4];
        $owner //= $seen;                 # Record owner of the first file we see
        return unless $owner == $seen;    # Validation fails if any file doesn't match recorded owner
    }
    return $owner;
}

sub _move_files {

    # Move a new set of files in place.  The indexes of the source and dest lists of files are expected to correlate directly for the rename.
    # For example, $source_files->[0] will be renamed to $dest_files->[0].
    my ( $self, $source_files, $dest_files ) = @_;
    logger->die('_move_files() is a method call.')                         unless ref $self eq __PACKAGE__;
    logger->die('_move_files() called without expected arguments.')        unless ref($source_files) eq 'ARRAY' && ref($dest_files) eq 'ARRAY';
    logger->die('_move_files() called without file lists of equal count.') unless scalar @$source_files == scalar @$dest_files;

    unlink @$dest_files;    # Though they would be overwritten by safemv, there's less chance for a mixture of old and new files if we remove all now and then something goes wrong later
    my $result = 1;
    while ( my ( $index, $source_file ) = each @$source_files ) {
        my $dest_file = $dest_files->[$index];
        if ( !Cpanel::FileUtils::Move::safemv( '-f', $source_file, $dest_file ) ) {
            logger->warn("Failed to move $source_file to $dest_file");
            $result = 0;    # Overall fail if any file doesn't move
        }
    }

    return $result;
}

sub _set_default_perms {
    my ( $self, $uid, $files ) = @_;
    logger->die('_set_default_perms() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_set_default_perms() called without expected arguments.') unless length $uid && ref($files) eq 'ARRAY' && scalar @$files;
    for my $file (@$files) {
        if ( !-f $file ) {
            logger->warn("Missing expected file $file while trying to update permissions");
            return 0;    # Must bail out if all of the expected files don't exist.
        }

        Cpanel::SafetyBits::safe_chmod( $DB_PERMS, $uid, $file )       or logger->warn("Failed to chmod $file");
        Cpanel::SafetyBits::Chown::safe_chown_guess_gid( $uid, $file ) or logger->warn("Failed to chown $file to uid $uid");
    }
    return 1;
}

sub _get_db_files {

    # Expects a database path (i.e. "$secdatadir/$shortname") without a suffix
    # Generates list of files with known suffixes appended to database path (does not verify existence)
    my ( $self, $path ) = @_;
    logger->die('_get_db_files() is a method call.')                  unless ref $self eq __PACKAGE__;
    logger->die('_get_db_files() called without expected arguments.') unless length $path;

    return map { $path . $_ } @DB_FILE_SUFFIXES;
}

sub _allowed_owner {

    # If this is expanded to allow any user, ensure that $owner and its gid exists in Cpanel::PwCache to avoid death by Cpanel::SafetyBits::Chown::safe_chown_guess_gid
    my ( $self, $owner ) = @_;
    logger->die('_allowed_owner() is a method call.') unless ref $self eq __PACKAGE__;

    # undef $owner is not an implementation error here, it simply means the owner couldn't be determined or is intentionally being skipped.
    return unless defined $owner;

    my $nobody_uid = $self->{'nobody_uid'} //= ( Cpanel::PwCache::getpwnam('nobody') )[2];
    return unless defined $nobody_uid;

    return 1 if $owner == $nobody_uid;

    return 0;
}

sub _secdatadir {
    my $self = shift;
    logger->die('_secdatadir() is a method call.') unless ref $self eq __PACKAGE__;
    $self->{'secdatadir'} //= $DEFAULT_SECDATADIR;
    logger->die('Unable to determine secdatadir.') unless length $self->{'secdatadir'};
    return $self->{'secdatadir'};
}

if ( not caller() ) {
    my $shrink = scripts::shrink_modsec_ip_database->new();
    $shrink->as_script;
    exit 0;
}

1;

__END__

=head1 NAME

/scripts/shrink_modsec_ip_database

=head1 USAGE AS A SCRIPT

  /scripts/shrink_modsec_ip_database -x

=head2 AS A LIBRARY

This script is internally written as a modulino, which means it can be C<require>'d:

  use strict;
  require q{/scripts/shrink_modsec_ip_database};
  my $shrink = scripts::shrink_modsec_ip_database->new();
  $shrink->run();

=head1 REQUIRED ARGUMENTS

None

=head1 OPTIONS

=over 4

=item -x

Use this option to actually run the script, otherwise it will warn and return
without doing anything.

=back

=head1 DESCRIPTION

This script is called by C<scripts/maintenance>, and its purpose is to shrink
ModSecurity database files by removing expired entries.

=head1 DIAGNOSTICS

None

=head1 EXIT STATUS

Exit status is 0 (success) unless an unexpected error occurs.

=head1 DEPENDENCIES

This script relies on C</usr/sbin/modsec-sdbm-util> to be installed, and in order to be useful,
C<ModSecurity> must be installed and be enabled.

=head1 INCOMPATIBILITIES

None

=head1 BUGS AND LIMITATIONS

None

=head1 LICENSE AND COPYRIGHT

   Copyright 2022 cPanel, L.L.C.