#
# This file and its contents are supplied under the terms of the
# Common Development and Distribution License ("CDDL"), version 1.0.
# You may only use this file in accordance with the terms of version
# 1.0 of the CDDL.
#
# A full copy of the text of the CDDL should have accompanied this
# source.  A copy of the CDDL is also available via the Internet at
# http://www.illumos.org/license/CDDL.
#

#
# Copyright (c) 2017, 2018 by Delphix. All rights reserved.
#

. $STF_SUITE/include/libtest.shlib
. $STF_SUITE/tests/functional/removal/removal.kshlib

#
# In general all the tests related to the pool checkpoint can
# be divided into two categories. TESTS that verify features
# provided by the checkpoint (e.g. checkpoint_rewind) and tests
# that stress-test the checkpoint (e.g. checkpoint_big_rewind).
#
# For the first group we don't really care about the size of
# the pool or the individual file sizes within the filesystems.
# This is why these tests run directly on pools that use a
# "real disk vdev" (meaning not a file based one). These tests
# use the $TESTPOOL pool that is created on top of $TESTDISK.
# This pool is referred to as the "test pool" and thus all
# the tests of this group use the testpool-related functions of
# this file (not the nested_pools ones).
#
# For the second group we generally try to bring the pool to its
# limits by increasing fragmentation, filling all allocatable
# space, attempting to use vdevs that the checkpoint spacemap
# cannot represent, etc. For these tests we need to control
# almost all parameters of the pool and the vdevs that back it
# so we create them based on file-based vdevs that we carefully
# create within the $TESTPOOL pool. So most of these tests, in
# order to create this nested pool sctructure, generally start
# like this:
# 1] We create the test pool ($TESTPOOL).
# 2] We create a filesystem and we populate it with files of
#    some predetermined size.
# 3] We use those files as vdevs for the pool that the test
#    will use ($NESTEDPOOL).
# 4] Go on and let the test run and operate on $NESTEDPOOL.
#

#
# These disks are used to back $TESTPOOL
#
TESTDISK="$(echo $DISKS | cut -d' ' -f1)"
EXTRATESTDISK="$(echo $DISKS | cut -d' ' -f2)"

FS0=$TESTPOOL/$TESTFS
FS1=$TESTPOOL/$TESTFS1
FS2=$TESTPOOL/$TESTFS2

FS0FILE=/$FS0/$TESTFILE0
FS1FILE=/$FS1/$TESTFILE1
FS2FILE=/$FS2/$TESTFILE2

#
# The following are created within $TESTPOOL and
# will be used to back $NESTEDPOOL
#
DISKFS=$TESTPOOL/disks
FILEDISKDIR=/$DISKFS
FILEDISK1=/$DISKFS/dsk1
FILEDISK2=/$DISKFS/dsk2
FILEDISKS="$FILEDISK1 $FILEDISK2"

#
# $NESTEDPOOL related variables
#
NESTEDPOOL=nestedpool
NESTEDFS0=$NESTEDPOOL/$TESTFS
NESTEDFS1=$NESTEDPOOL/$TESTFS1
NESTEDFS2=$NESTEDPOOL/$TESTFS2
NESTEDFS0FILE=/$NESTEDFS0/$TESTFILE0
NESTEDFS1FILE=/$NESTEDFS1/$TESTFILE1
NESTEDFS2FILE=/$NESTEDFS2/$TESTFILE2

#
# In the tests that stress-test the pool (second category
# mentioned above), there exist some that need to bring
# fragmentation at high percentages in a relatively short
# period of time. In order to do that we set the following
# parameters:
#
# * We use two disks of 1G each, to create a pool of size 2G.
#   The point is that 2G is not small nor large, and we also
#   want to have 2 disks to introduce indirect vdevs on our
#   setup.
# * We enable compression and set the record size of all
#   filesystems to 8K. The point of compression is to
#   ensure that we are not filling up the whole pool (that's
#   what checkpoint_capacity is for), and the specific
#   record size is set to match the block size of randwritecomp
#   which is used to increase fragmentation by writing on
#   files.
# * We always have 2 big files present of 512M each, which
#   should account for 40%~50% capacity by the end of each
#   test with fragmentation around 50~60%.
# * At each file we attempt to do enough random writes to
#   touch every offset twice on average.
#
# Note that the amount of random writes per files are based
# on the following calculation:
#
# ((512M / 8K) * 3) * 2 = ~400000
#
# Given that the file is 512M and one write is 8K, we would
# need (512M / 8K) writes to go through the whole file.
# Assuming though that each write has a compression ratio of
# 3, then we want 3 times that to cover the same amount of
# space. Finally, we multiply that by 2 since our goal is to
# touch each offset twice on average.
#
# Examples of those tests are checkpoint_big_rewind and
# checkpoint_discard_busy.
#
FILEDISKSIZE=1g
DISKSIZE=1g
BIGFILESIZE=512M
RANDOMWRITES=400000


#
# Assumes create_test_pool has been called beforehand.
#
function setup_nested_pool
{
	log_must zfs create $DISKFS

	log_must truncate -s $DISKSIZE $FILEDISK1
	log_must truncate -s $DISKSIZE $FILEDISK2

	log_must zpool create -O sync=disabled $NESTEDPOOL $FILEDISKS
}

function setup_test_pool
{
	log_must zpool create -O sync=disabled $TESTPOOL "$TESTDISK"
}

function setup_nested_pools
{
	setup_test_pool
	setup_nested_pool
}

function cleanup_nested_pool
{
	if poolexists $NESTEDPOOL; then
		log_must zpool destroy $NESTEDPOOL
	fi

	log_must rm -f $FILEDISKS
}

function cleanup_test_pool
{
	if poolexists $TESTPOOL; then
		log_must zpool destroy $TESTPOOL
	fi

	#
	# We always clear the labels of all disks
	# between tests so imports from zpool or
	# or zdb do not get confused with leftover
	# data from old pools.
	#
	for disk in $DISKS; do
		zpool labelclear -f $disk
	done
}

function cleanup_nested_pools
{
	cleanup_nested_pool
	cleanup_test_pool
}

#
# Remove and re-add each vdev to ensure that data is
# moved between disks and indirect mappings are created
#
function introduce_indirection
{
	for disk in ${FILEDISKS[@]}; do
		log_must zpool remove $NESTEDPOOL $disk
		log_must wait_for_removal $NESTEDPOOL
		log_mustnot vdevs_in_pool $NESTEDPOOL $disk
		log_must zpool add $NESTEDPOOL $disk
	done
}

FILECONTENTS0="Can't wait to be checkpointed!"
FILECONTENTS1="Can't wait to be checkpointed too!"
NEWFILECONTENTS0="I survived after the checkpoint!"
NEWFILECONTENTS2="I was born after the checkpoint!"

function populate_test_pool
{
	log_must zfs create -o compression=lz4 -o recordsize=8k $FS0
	log_must zfs create -o compression=lz4 -o recordsize=8k $FS1

	echo $FILECONTENTS0 > $FS0FILE
	echo $FILECONTENTS1 > $FS1FILE
}

function populate_nested_pool
{
	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS0
	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS1

	echo $FILECONTENTS0 > $NESTEDFS0FILE
	echo $FILECONTENTS1 > $NESTEDFS1FILE
}

function test_verify_pre_checkpoint_state
{
	log_must zfs list $FS0
	log_must zfs list $FS1
	log_must [ "$(<$FS0FILE)" = "$FILECONTENTS0" ]
	log_must [ "$(<$FS1FILE)" = "$FILECONTENTS1" ]

	#
	# If we've opened the checkpointed state of the
	# pool as read-only without rewinding on-disk we
	# can't really use zdb on it.
	#
	if [[ "$1" != "ro-check" ]] ; then
		log_must zdb $TESTPOOL
	fi

	#
	# Ensure post-checkpoint state is not present
	#
	log_mustnot zfs list $FS2
	log_mustnot [ "$(<$FS0FILE)" = "$NEWFILECONTENTS0" ]
}

function nested_verify_pre_checkpoint_state
{
	log_must zfs list $NESTEDFS0
	log_must zfs list $NESTEDFS1
	log_must [ "$(<$NESTEDFS0FILE)" = "$FILECONTENTS0" ]
	log_must [ "$(<$NESTEDFS1FILE)" = "$FILECONTENTS1" ]

	#
	# If we've opened the checkpointed state of the
	# pool as read-only without rewinding on-disk we
	# can't really use zdb on it.
	#
	if [[ "$1" != "ro-check" ]] ; then
		log_must zdb $NESTEDPOOL
	fi

	#
	# Ensure post-checkpoint state is not present
	#
	log_mustnot zfs list $NESTEDFS2
	log_mustnot [ "$(<$NESTEDFS0FILE)" = "$NEWFILECONTENTS0" ]
}

function test_change_state_after_checkpoint
{
	log_must zfs destroy $FS1
	log_must zfs create -o compression=lz4 -o recordsize=8k $FS2

	echo $NEWFILECONTENTS0 > $FS0FILE
	echo $NEWFILECONTENTS2 > $FS2FILE
}

function nested_change_state_after_checkpoint
{
	log_must zfs destroy $NESTEDFS1
	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS2

	echo $NEWFILECONTENTS0 > $NESTEDFS0FILE
	echo $NEWFILECONTENTS2 > $NESTEDFS2FILE
}

function test_verify_post_checkpoint_state
{
	log_must zfs list $FS0
	log_must zfs list $FS2
	log_must [ "$(<$FS0FILE)" = "$NEWFILECONTENTS0" ]
	log_must [ "$(<$FS2FILE)" = "$NEWFILECONTENTS2" ]

	log_must zdb $TESTPOOL

	#
	# Ensure pre-checkpointed state that was removed post-checkpoint
	# is not present
	#
	log_mustnot zfs list $FS1
	log_mustnot [ "$(<$FS0FILE)" = "$FILECONTENTS0" ]
}

function fragment_before_checkpoint
{
	populate_nested_pool
	log_must mkfile -n $BIGFILESIZE $NESTEDFS0FILE
	log_must mkfile -n $BIGFILESIZE $NESTEDFS1FILE
	log_must randwritecomp $NESTEDFS0FILE $RANDOMWRITES
	log_must randwritecomp $NESTEDFS1FILE $RANDOMWRITES

	#
	# Display fragmentation on test log
	#
	log_must zpool list -v
}

function fragment_after_checkpoint_and_verify
{
	log_must zfs destroy $NESTEDFS1
	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS2
	log_must mkfile -n $BIGFILESIZE $NESTEDFS2FILE
	log_must randwritecomp $NESTEDFS0FILE $RANDOMWRITES
	log_must randwritecomp $NESTEDFS2FILE $RANDOMWRITES

	#
	# Display fragmentation on test log
	#
	log_must zpool list -v

	#
	# Typically we would just run zdb at this point and things
	# would be fine. Unfortunately, if there is still any
	# background I/O in the pool the zdb command can fail with
	# checksum errors temporarily.
	#
	# Export the pool when running zdb so the pool is idle and
	# the verification results are consistent.
	#
	log_must zpool export $NESTEDPOOL
	log_must zdb -e -p $FILEDISKDIR $NESTEDPOOL
	log_must zdb -e -p $FILEDISKDIR -kc $NESTEDPOOL
	log_must zpool import -d $FILEDISKDIR $NESTEDPOOL
}

function wait_discard_finish
{
	typeset pool="$1"

	typeset status
	status=$(zpool status $pool | grep "checkpoint:")
	while [ "" != "$status" ]; do
		sleep 5
		status=$(zpool status $pool | grep "checkpoint:")
	done
}

function test_wait_discard_finish
{
	wait_discard_finish $TESTPOOL
}

function nested_wait_discard_finish
{
	wait_discard_finish $NESTEDPOOL
}

#
# Creating the setup for the second group of tests mentioned in
# block comment of this file can take some time as we are doing
# random writes to raise capacity and fragmentation before taking
# the checkpoint. Thus we create this setup once and save the
# disks of the nested pool in a temporary directory where we can
# reuse it for each test that requires that setup.
#
SAVEDPOOLDIR="$TEST_BASE_DIR/ckpoint_saved_pool"

function test_group_premake_nested_pools
{
	setup_nested_pools

	#
	# Populate and fragment the pool.
	#
	fragment_before_checkpoint

	#
	# Export and save the pool for other tests.
	#
	log_must zpool export $NESTEDPOOL
	log_must mkdir $SAVEDPOOLDIR
	log_must cp $FILEDISKS $SAVEDPOOLDIR

	#
	# Reimport pool to be destroyed by
	# cleanup_nested_pools function
	#
	log_must zpool import -d $FILEDISKDIR $NESTEDPOOL
}

function test_group_destroy_saved_pool
{
	log_must rm -rf $SAVEDPOOLDIR
}

#
# Recreate nested pool setup from saved pool.
#
function setup_nested_pool_state
{
	setup_test_pool

	log_must zfs create $DISKFS
	log_must cp $SAVEDPOOLDIR/* $FILEDISKDIR

	log_must zpool import -d $FILEDISKDIR $NESTEDPOOL
}