zfs/tests/zfs-tests/include/blkdev.shlib

595 lines
14 KiB
Plaintext

#
# 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 2009 Sun Microsystems, Inc. All rights reserved.
# Use is subject to license terms.
# Copyright (c) 2012, 2019 by Delphix. All rights reserved.
# Copyright 2016 Nexenta Systems, Inc.
# Copyright (c) 2016, 2017 by Intel Corporation. All rights reserved.
# Copyright (c) 2017 Lawrence Livermore National Security, LLC.
# Copyright (c) 2017 Datto Inc.
# Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
# Copyright 2019 Richard Elling
#
#
# Returns SCSI host number for the given disk
#
function get_scsi_host #disk
{
typeset disk=$1
ls /sys/block/${disk}/device/scsi_device | cut -d : -f 1
}
#
# Cause a scan of all scsi host adapters by default
#
# $1 optional host number
#
function scan_scsi_hosts
{
typeset hostnum=${1}
if is_linux; then
if [[ -z $hostnum ]]; then
for host in /sys/class/scsi_host/host*; do
log_must eval "echo '- - -' > $host/scan"
done
else
log_must eval \
"echo /sys/class/scsi_host/host$hostnum/scan" \
> /dev/null
log_must eval \
"echo '- - -' > /sys/class/scsi_host/host$hostnum/scan"
fi
fi
}
#
# Wait for newly created block devices to have their minors created.
# Additional arguments can be passed to udevadm trigger, with the expected
# arguments to typically be a block device pathname. This is useful when
# checking waiting on a specific device to settle rather than triggering
# all devices and waiting for them all to settle.
#
# The udevadm settle timeout can be 120 or 180 seconds by default for
# some distros. If a long delay is experienced, it could be due to some
# strangeness in a malfunctioning device that isn't related to the devices
# under test. To help debug this condition, a notice is given if settle takes
# too long.
#
# Note: there is no meaningful return code if udevadm fails. Consumers
# should not expect a return code (do not call as argument to log_must)
#
function block_device_wait
{
if is_linux; then
udevadm trigger $*
typeset local start=$SECONDS
udevadm settle
typeset local elapsed=$((SECONDS - start))
[[ $elapsed > 60 ]] && \
log_note udevadm settle time too long: $elapsed
elif is_freebsd; then
sleep 3
fi
}
#
# Check if the given device is physical device
#
function is_physical_device #device
{
typeset device=${1#$DEV_DSKDIR}
device=${device#$DEV_RDSKDIR}
if is_linux; then
is_disk_device "$DEV_DSKDIR/$device" && \
[[ -f /sys/module/loop/parameters/max_part ]]
return $?
elif is_freebsd; then
is_disk_device "$DEV_DSKDIR/$device" && \
echo $device | egrep -q \
-e '^a?da[0-9]+$' \
-e '^md[0-9]+$' \
-e '^mfid[0-9]+$' \
-e '^nda[0-9]+$' \
-e '^nvd[0-9]+$' \
-e '^vtbd[0-9]+$'
return $?
else
echo $device | egrep "^c[0-F]+([td][0-F]+)+$" > /dev/null 2>&1
return $?
fi
}
#
# Check if the given device is a real device (ie SCSI device)
#
function is_real_device #disk
{
typeset disk=$1
[[ -z $disk ]] && log_fail "No argument for disk given."
if is_linux; then
lsblk $DEV_RDSKDIR/$disk -o TYPE 2>/dev/null | \
egrep disk >/dev/null
return $?
fi
}
#
# Check if the given device is a loop device
#
function is_loop_device #disk
{
typeset disk=$1
[[ -z $disk ]] && log_fail "No argument for disk given."
if is_linux; then
lsblk $DEV_RDSKDIR/$disk -o TYPE 2>/dev/null | \
egrep loop >/dev/null
return $?
fi
}
#
# Linux:
# Check if the given device is a multipath device and if there is a symbolic
# link to a device mapper and to a disk
# Currently no support for dm devices alone without multipath
#
# FreeBSD:
# Check if the given device is a gmultipath device.
#
# Others:
# No multipath detection.
#
function is_mpath_device #disk
{
typeset disk=$1
[[ -z $disk ]] && log_fail "No argument for disk given."
if is_linux; then
lsblk $DEV_MPATHDIR/$disk -o TYPE 2>/dev/null | \
egrep mpath >/dev/null
if (($? == 0)); then
readlink $DEV_MPATHDIR/$disk > /dev/null 2>&1
return $?
else
return $?
fi
elif is_freebsd; then
is_disk_device $DEV_MPATHDIR/$disk
else
false
fi
}
#
# Check if the given path is the appropriate sort of device special node.
#
function is_disk_device #path
{
typeset path=$1
if is_freebsd; then
# FreeBSD doesn't have block devices, only character devices.
test -c $path
else
test -b $path
fi
}
# Set the slice prefix for disk partitioning depending
# on whether the device is a real, multipath, or loop device.
# Currently all disks have to be of the same type, so only
# checks first disk to determine slice prefix.
#
function set_slice_prefix
{
typeset disk
typeset -i i=0
if is_linux; then
while (( i < $DISK_ARRAY_NUM )); do
disk="$(echo $DISKS | nawk '{print $(i + 1)}')"
if ( is_mpath_device $disk ) && [[ -z $(echo $disk | awk 'substr($1,18,1)\
~ /^[[:digit:]]+$/') ]] || ( is_real_device $disk ); then
export SLICE_PREFIX=""
return 0
elif ( is_mpath_device $disk || is_loop_device \
$disk ); then
export SLICE_PREFIX="p"
return 0
else
log_fail "$disk not supported for partitioning."
fi
(( i = i + 1))
done
elif is_freebsd; then
export SLICE_PREFIX="p"
return 0
fi
}
#
# Set the directory path of the listed devices in $DISK_ARRAY_NUM
# Currently all disks have to be of the same type, so only
# checks first disk to determine device directory
# default = /dev (linux)
# real disk = /dev (linux)
# multipath device = /dev/mapper (linux)
#
function set_device_dir
{
typeset disk
typeset -i i=0
if is_linux; then
while (( i < $DISK_ARRAY_NUM )); do
disk="$(echo $DISKS | nawk '{print $(i + 1)}')"
if is_mpath_device $disk; then
export DEV_DSKDIR=$DEV_MPATHDIR
return 0
else
export DEV_DSKDIR=$DEV_RDSKDIR
return 0
fi
(( i = i + 1))
done
else
export DEV_DSKDIR=$DEV_RDSKDIR
fi
}
#
# Get the directory path of given device
#
function get_device_dir #device
{
typeset device=$1
if ! is_freebsd && ! is_physical_device $device; then
if [[ $device != "/" ]]; then
device=${device%/*}
fi
if is_disk_device "$DEV_DSKDIR/$device"; then
device="$DEV_DSKDIR"
fi
echo $device
else
echo "$DEV_DSKDIR"
fi
}
#
# Get persistent name for given disk
#
function get_persistent_disk_name #device
{
typeset device=$1
typeset dev_id
if is_linux; then
if is_real_device $device; then
dev_id="$(udevadm info -q all -n $DEV_DSKDIR/$device \
| egrep disk/by-id | nawk '{print $2; exit}' \
| nawk -F / '{print $3}')"
echo $dev_id
elif is_mpath_device $device; then
dev_id="$(udevadm info -q all -n $DEV_DSKDIR/$device \
| egrep disk/by-id/dm-uuid \
| nawk '{print $2; exit}' \
| nawk -F / '{print $3}')"
echo $dev_id
else
echo $device
fi
else
echo $device
fi
}
#
# Online or offline a disk on the system
#
# First checks state of disk. Test will fail if disk is not properly onlined
# or offlined. Online is a full rescan of SCSI disks by echoing to every
# host entry.
#
function on_off_disk # disk state{online,offline} host
{
typeset disk=$1
typeset state=$2
typeset host=$3
[[ -z $disk ]] || [[ -z $state ]] && \
log_fail "Arguments invalid or missing"
if is_linux; then
if [[ $state == "offline" ]] && ( is_mpath_device $disk ); then
dm_name="$(readlink $DEV_DSKDIR/$disk \
| nawk -F / '{print $2}')"
slave="$(ls /sys/block/${dm_name}/slaves \
| nawk '{print $1}')"
while [[ -n $slave ]]; do
#check if disk is online
lsscsi | egrep $slave > /dev/null
if (($? == 0)); then
slave_dir="/sys/block/${dm_name}"
slave_dir+="/slaves/${slave}/device"
ss="${slave_dir}/state"
sd="${slave_dir}/delete"
log_must eval "echo 'offline' > ${ss}"
log_must eval "echo '1' > ${sd}"
lsscsi | egrep $slave > /dev/null
if (($? == 0)); then
log_fail "Offlining" \
"$disk failed"
fi
fi
slave="$(ls /sys/block/$dm_name/slaves \
2>/dev/null | nawk '{print $1}')"
done
elif [[ $state == "offline" ]] && ( is_real_device $disk ); then
#check if disk is online
lsscsi | egrep $disk > /dev/null
if (($? == 0)); then
dev_state="/sys/block/$disk/device/state"
dev_delete="/sys/block/$disk/device/delete"
log_must eval "echo 'offline' > ${dev_state}"
log_must eval "echo '1' > ${dev_delete}"
lsscsi | egrep $disk > /dev/null
if (($? == 0)); then
log_fail "Offlining $disk" \
"failed"
fi
else
log_note "$disk is already offline"
fi
elif [[ $state == "online" ]]; then
#force a full rescan
scan_scsi_hosts $host
block_device_wait
if is_mpath_device $disk; then
dm_name="$(readlink $DEV_DSKDIR/$disk \
| nawk -F / '{print $2}')"
slave="$(ls /sys/block/$dm_name/slaves \
| nawk '{print $1}')"
lsscsi | egrep $slave > /dev/null
if (($? != 0)); then
log_fail "Onlining $disk failed"
fi
elif is_real_device $disk; then
block_device_wait
typeset -i retries=0
while ! lsscsi | egrep -q $disk; do
if (( $retries > 2 )); then
log_fail "Onlining $disk failed"
break
fi
(( ++retries ))
sleep 1
done
else
log_fail "$disk is not a real dev"
fi
else
log_fail "$disk failed to $state"
fi
fi
}
#
# Simulate disk removal
#
function remove_disk #disk
{
typeset disk=$1
on_off_disk $disk "offline"
block_device_wait
}
#
# Simulate disk insertion for the given SCSI host
#
function insert_disk #disk scsi_host
{
typeset disk=$1
typeset scsi_host=$2
on_off_disk $disk "online" $scsi_host
block_device_wait
}
#
# Load scsi_debug module with specified parameters
# $blksz can be either one of: < 512b | 512e | 4Kn >
#
function load_scsi_debug # dev_size_mb add_host num_tgts max_luns blksz
{
typeset devsize=$1
typeset hosts=$2
typeset tgts=$3
typeset luns=$4
typeset blksz=$5
[[ -z $devsize ]] || [[ -z $hosts ]] || [[ -z $tgts ]] || \
[[ -z $luns ]] || [[ -z $blksz ]] && \
log_fail "Arguments invalid or missing"
case "$5" in
'512b')
typeset sector=512
typeset blkexp=0
;;
'512e')
typeset sector=512
typeset blkexp=3
;;
'4Kn')
typeset sector=4096
typeset blkexp=0
;;
*) log_fail "Unsupported blksz value: $5" ;;
esac
if is_linux; then
modprobe -n scsi_debug
if (($? != 0)); then
log_unsupported "Platform does not have scsi_debug"
"module"
fi
lsmod | egrep scsi_debug > /dev/null
if (($? == 0)); then
log_fail "scsi_debug module already installed"
else
log_must modprobe scsi_debug dev_size_mb=$devsize \
add_host=$hosts num_tgts=$tgts max_luns=$luns \
sector_size=$sector physblk_exp=$blkexp
block_device_wait
lsscsi | egrep scsi_debug > /dev/null
if (($? == 1)); then
log_fail "scsi_debug module install failed"
fi
fi
fi
}
#
# Unload scsi_debug module, if needed.
#
function unload_scsi_debug
{
log_must_retry "in use" 5 modprobe -r scsi_debug
}
#
# Get scsi_debug device name.
# Returns basename of scsi_debug device (for example "sdb").
#
function get_debug_device
{
for i in {1..10} ; do
val=$(lsscsi | nawk '/scsi_debug/ {print $6; exit}' | cut -d / -f3)
# lsscsi can take time to settle
if [ "$val" != "-" ] ; then
break
fi
sleep 1
done
echo "$val"
}
#
# Get actual devices used by the pool (i.e. linux sdb1 not sdb).
#
function get_pool_devices #testpool #devdir
{
typeset testpool=$1
typeset devdir=$2
typeset out=""
if is_linux || is_freebsd; then
out=$(zpool status -P $testpool |grep ${devdir} | awk '{print $1}')
out=$(echo $out | sed -e "s|${devdir}/||g" | tr '\n' ' ')
fi
echo $out
}
#
# Write to standard out giving the level, device name, offset and length
# of all blocks in an input file. The offset and length are in units of
# 512 byte blocks. In the case of mirrored vdevs, only the first
# device is listed, as the levels, blocks and offsets will be the same
# on other devices. Note that this function only works with mirrored
# or non-redundant pools, not raidz.
#
# The output of this function can be used to introduce corruption at
# varying levels of indirection.
#
function list_file_blocks # input_file
{
typeset input_file=$1
[[ -f $input_file ]] || log_fail "Couldn't find $input_file"
typeset ds="$(zfs list -H -o name $input_file)"
typeset pool="${ds%%/*}"
typeset inum="$(stat -c '%i' $input_file)"
#
# Establish a mapping between vdev ids as shown in a DVA and the
# pathnames they correspond to in ${VDEV_MAP[]}.
#
eval $(zdb -C $pool | awk '
BEGIN {
printf("typeset VDEV_MAP\n");
looking = 0;
}
/^ children/ {
id = $1;
looking = 1;
}
/path: / && looking == 1 {
print id" "$2;
looking = 0;
}
' | sed -n 's/^children\[\([0-9]\)\]: \(.*\)$/VDEV_MAP[\1]=\2/p')
#
# The awk below parses the output of zdb, printing out the level
# of each block along with vdev id, offset and length. The last
# two are converted to decimal in the while loop. 4M is added to
# the offset to compensate for the first two labels and boot
# block. Lastly, the offset and length are printed in units of
# 512b blocks for ease of use with dd.
#
log_must zpool sync -f
typeset level path offset length
zdb -ddddd $ds $inum | awk -F: '
BEGIN { looking = 0 }
/^Indirect blocks:/ { looking = 1}
/^\t\tsegment / { looking = 0}
/L[0-8]/ && looking == 1 { print $0}
' | sed -n 's/^.*\(L[0-9]\) \([0-9]*\):\([0-9a-f]*\):\([0-9a-f]*\) .*$/\1 \2 \3 \4/p' | \
while read level path offset length; do
offset=$((16#$offset)) # Conversion from hex
length=$((16#$length))
offset="$(((offset + 4 * 1024 * 1024) / 512))"
length="$((length / 512))"
echo "$level ${VDEV_MAP[$path]} $offset $length"
done 2>/dev/null
}
function corrupt_blocks_at_level # input_file corrupt_level
{
typeset input_file=$1
typeset corrupt_level="L${2:-0}"
typeset level path offset length
[[ -f $input_file ]] || log_fail "Couldn't find $input_file"
log_must list_file_blocks $input_file | \
while read level path offset length; do
if [[ $level = $corrupt_level ]]; then
log_must dd if=/dev/urandom of=$path bs=512 \
count=$length seek=$offset conv=notrunc
fi
done
# This is necessary for pools made of loop devices.
sync
}