1
0
mirror of https://github.com/samba-team/samba.git synced 2025-01-25 06:04:04 +03:00
samba-mirror/source4/setup/tests/blackbox_start_backup.sh
Tim Beale 8e58954f2f tests: Add test that Samba cannot be started with a backup DB
We don't want users to take a backup file, and then simply untar it and
run Samba (Several modifications to the DB need to be made as part of
the restore process, so users should always run the 'backup restore'
command).

To enforce this, prime_ldb_databases() now refuses to start Samba if the
backupDate marker is present in the DB. This patch adds a test-case that
proves this basic behaviour works.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
2018-06-28 03:34:26 +02:00

80 lines
2.2 KiB
Bash
Executable File

#!/bin/sh
# Simple test that a DB from a backup file cannot be untarred and started
# manually (you have to run the samba-tool 'backup restore' command instead).
if [ $# -lt 1 ]; then
cat <<EOF
Usage: $0 PREFIX
EOF
exit 1;
fi
PREFIX="$1"
shift 1
DBPATH=$PREFIX/start-backup
mkdir -p $DBPATH
. `dirname $0`/../../../testprogs/blackbox/subunit.sh
do_provision()
{
$PYTHON $BINDIR/samba-tool domain provision \
--domain=FOO --realm=foo.example.com --use-ntvfs \
--targetdir=$DBPATH --option="pid directory = $DBPATH"
}
add_backup_marker()
{
# manually add the backup marker that the backup cmd usually adds
$BINDIR/ldbmodify \
-H tdb://$DBPATH/private/sam.ldb <<EOF
dn: @SAMBA_DSDB
changetype: modify
add: backupDate
backupDate: who-knows-when
-
EOF
}
start_backup()
{
# start samba in interactive mode (if we don't, samba daemonizes and so the
# command's exit status is always zero (success), regardless of whether
# samba actually starts up or not). However, this means if this assertion
# were ever to fail (i.e. samba DOES startup from a backup file), then the
# test case would just hang. So we use a max-run-time of 5 secs so that
# samba will self-destruct in the bad case (max_runtime_handler() returns
# zero/success in this case, which allows us to tell the good case from the
# bad case).
OPTS="--maximum-runtime=5 -i"
# redirect logs to stderr (which we'll then redirect to stdout so we can
# capture it in a bash variable)
OPTS="$OPTS --debug-stderr"
# start samba and capture the debug output
OUTPUT=$($BINDIR/samba -s $DBPATH/etc/smb.conf $OPTS 2>&1)
if [ $? -eq 0 ] ; then
echo "ERROR: Samba should not have started successfully"
return 1
fi
# check the reason we're failing is because prime_ldb_databases() is
# detecting that this is a backup DB (and not some other reason)
echo "$OUTPUT" | grep "failed to start: Database is a backup"
}
# setup a DB and manually mark it as being a "backup"
testit "provision" do_provision
testit "add-backup-marker" add_backup_marker
# check that Samba won't start using this DB (because it's a backup)
testit "start-samba-backup" start_backup
rm -rf $DBPATH
exit $failed