mirror_zfs/scripts/common.sh.in

749 lines
16 KiB
Bash
Raw Normal View History

#!/bin/bash
#
Support custom build directories and move includes One of the neat tricks an autoconf style project is capable of is allow configurion/building in a directory other than the source directory. The major advantage to this is that you can build the project various different ways while making changes in a single source tree. For example, this project is designed to work on various different Linux distributions each of which work slightly differently. This means that changes need to verified on each of those supported distributions perferably before the change is committed to the public git repo. Using nfs and custom build directories makes this much easier. I now have a single source tree in nfs mounted on several different systems each running a supported distribution. When I make a change to the source base I suspect may break things I can concurrently build from the same source on all the systems each in their own subdirectory. wget -c http://github.com/downloads/behlendorf/zfs/zfs-x.y.z.tar.gz tar -xzf zfs-x.y.z.tar.gz cd zfs-x-y-z ------------------------- run concurrently ---------------------- <ubuntu system> <fedora system> <debian system> <rhel6 system> mkdir ubuntu mkdir fedora mkdir debian mkdir rhel6 cd ubuntu cd fedora cd debian cd rhel6 ../configure ../configure ../configure ../configure make make make make make check make check make check make check This change also moves many of the include headers from individual incude/sys directories under the modules directory in to a single top level include directory. This has the advantage of making the build rules cleaner and logically it makes a bit more sense.
2010-09-05 00:26:23 +04:00
# Common support functions for testing scripts. If a script-config
# files is available it will be sourced so in-tree kernel modules and
Support custom build directories and move includes One of the neat tricks an autoconf style project is capable of is allow configurion/building in a directory other than the source directory. The major advantage to this is that you can build the project various different ways while making changes in a single source tree. For example, this project is designed to work on various different Linux distributions each of which work slightly differently. This means that changes need to verified on each of those supported distributions perferably before the change is committed to the public git repo. Using nfs and custom build directories makes this much easier. I now have a single source tree in nfs mounted on several different systems each running a supported distribution. When I make a change to the source base I suspect may break things I can concurrently build from the same source on all the systems each in their own subdirectory. wget -c http://github.com/downloads/behlendorf/zfs/zfs-x.y.z.tar.gz tar -xzf zfs-x.y.z.tar.gz cd zfs-x-y-z ------------------------- run concurrently ---------------------- <ubuntu system> <fedora system> <debian system> <rhel6 system> mkdir ubuntu mkdir fedora mkdir debian mkdir rhel6 cd ubuntu cd fedora cd debian cd rhel6 ../configure ../configure ../configure ../configure make make make make make check make check make check make check This change also moves many of the include headers from individual incude/sys directories under the modules directory in to a single top level include directory. This has the advantage of making the build rules cleaner and logically it makes a bit more sense.
2010-09-05 00:26:23 +04:00
# utilities will be used. If no script-config can be found then the
# installed kernel modules and utilities will be used.
basedir="$(dirname $0)"
Support custom build directories and move includes One of the neat tricks an autoconf style project is capable of is allow configurion/building in a directory other than the source directory. The major advantage to this is that you can build the project various different ways while making changes in a single source tree. For example, this project is designed to work on various different Linux distributions each of which work slightly differently. This means that changes need to verified on each of those supported distributions perferably before the change is committed to the public git repo. Using nfs and custom build directories makes this much easier. I now have a single source tree in nfs mounted on several different systems each running a supported distribution. When I make a change to the source base I suspect may break things I can concurrently build from the same source on all the systems each in their own subdirectory. wget -c http://github.com/downloads/behlendorf/zfs/zfs-x.y.z.tar.gz tar -xzf zfs-x.y.z.tar.gz cd zfs-x-y-z ------------------------- run concurrently ---------------------- <ubuntu system> <fedora system> <debian system> <rhel6 system> mkdir ubuntu mkdir fedora mkdir debian mkdir rhel6 cd ubuntu cd fedora cd debian cd rhel6 ../configure ../configure ../configure ../configure make make make make make check make check make check make check This change also moves many of the include headers from individual incude/sys directories under the modules directory in to a single top level include directory. This has the advantage of making the build rules cleaner and logically it makes a bit more sense.
2010-09-05 00:26:23 +04:00
SCRIPT_CONFIG=zfs-script-config.sh
if [ -f "${basedir}/../${SCRIPT_CONFIG}" ]; then
. "${basedir}/../${SCRIPT_CONFIG}"
else
KERNEL_MODULES=(zlib_deflate zlib_inflate)
MODULES=(spl splat zavl znvpair zunicode zcommon zfs)
fi
PROG="<define PROG>"
CLEANUP=
VERBOSE=
VERBOSE_FLAG=
FORCE=
FORCE_FLAG=
DUMP_LOG=
ERROR=
RAID0S=()
RAID10S=()
RAIDZS=()
RAIDZ2S=()
TESTS_RUN=${TESTS_RUN:-'*'}
TESTS_SKIP=${TESTS_SKIP:-}
prefix=@prefix@
exec_prefix=@exec_prefix@
libexecdir=@libexecdir@
pkglibexecdir=${libexecdir}/@PACKAGE@
bindir=@bindir@
sbindir=@sbindir@
udevdir=@udevdir@
udevruledir=@udevruledir@
sysconfdir=@sysconfdir@
ETCDIR=${ETCDIR:-/etc}
DEVDIR=${DEVDIR:-/dev/disk/zpool}
ZPOOLDIR=${ZPOOLDIR:-${pkglibexecdir}/zpool-config}
ZPIOSDIR=${ZPIOSDIR:-${pkglibexecdir}/zpios-test}
ZPIOSPROFILEDIR=${ZPIOSPROFILEDIR:-${pkglibexecdir}/zpios-profile}
ZDB=${ZDB:-${sbindir}/zdb}
ZFS=${ZFS:-${sbindir}/zfs}
ZINJECT=${ZINJECT:-${sbindir}/zinject}
ZPOOL=${ZPOOL:-${sbindir}/zpool}
ZPOOL_ID=${ZPOOL_ID:-${bindir}/zpool_id}
ZTEST=${ZTEST:-${sbindir}/ztest}
ZPIOS=${ZPIOS:-${sbindir}/zpios}
COMMON_SH=${COMMON_SH:-${pkglibexecdir}/common.sh}
ZFS_SH=${ZFS_SH:-${pkglibexecdir}/zfs.sh}
ZPOOL_CREATE_SH=${ZPOOL_CREATE_SH:-${pkglibexecdir}/zpool-create.sh}
ZPIOS_SH=${ZPIOS_SH:-${pkglibexecdir}/zpios.sh}
ZPIOS_SURVEY_SH=${ZPIOS_SURVEY_SH:-${pkglibexecdir}/zpios-survey.sh}
LDMOD=${LDMOD:-/sbin/modprobe}
LSMOD=${LSMOD:-/sbin/lsmod}
RMMOD=${RMMOD:-/sbin/rmmod}
INFOMOD=${INFOMOD:-/sbin/modinfo}
LOSETUP=${LOSETUP:-/sbin/losetup}
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
MDADM=${MDADM:-/sbin/mdadm}
PARTED=${PARTED:-/sbin/parted}
BLOCKDEV=${BLOCKDEV:-/sbin/blockdev}
LSSCSI=${LSSCSI:-/usr/bin/lsscsi}
SCSIRESCAN=${SCSIRESCAN:-/usr/bin/scsi-rescan}
SYSCTL=${SYSCTL:-/sbin/sysctl}
UDEVADM=${UDEVADM:-/sbin/udevadm}
AWK=${AWK:-/usr/bin/awk}
COLOR_BLACK="\033[0;30m"
COLOR_DK_GRAY="\033[1;30m"
COLOR_BLUE="\033[0;34m"
COLOR_LT_BLUE="\033[1;34m"
COLOR_GREEN="\033[0;32m"
COLOR_LT_GREEN="\033[1;32m"
COLOR_CYAN="\033[0;36m"
COLOR_LT_CYAN="\033[1;36m"
COLOR_RED="\033[0;31m"
COLOR_LT_RED="\033[1;31m"
COLOR_PURPLE="\033[0;35m"
COLOR_LT_PURPLE="\033[1;35m"
COLOR_BROWN="\033[0;33m"
COLOR_YELLOW="\033[1;33m"
COLOR_LT_GRAY="\033[0;37m"
COLOR_WHITE="\033[1;37m"
COLOR_RESET="\033[0m"
die() {
echo -e "${PROG}: $1" >&2
exit 1
}
msg() {
if [ ${VERBOSE} ]; then
echo "$@"
fi
}
pass() {
echo -e "${COLOR_GREEN}Pass${COLOR_RESET}"
}
fail() {
echo -e "${COLOR_RED}Fail${COLOR_RESET} ($1)"
exit $1
}
skip() {
echo -e "${COLOR_BROWN}Skip${COLOR_RESET}"
}
populate() {
local ROOT=$1
local MAX_DIR_SIZE=$2
local MAX_FILE_SIZE=$3
mkdir -p $ROOT/{a,b,c,d,e,f,g}/{h,i}
DIRS=`find $ROOT`
for DIR in $DIRS; do
COUNT=$(($RANDOM % $MAX_DIR_SIZE))
for i in `seq $COUNT`; do
FILE=`mktemp -p ${DIR}`
SIZE=$(($RANDOM % $MAX_FILE_SIZE))
dd if=/dev/urandom of=$FILE bs=1k count=$SIZE &>/dev/null
done
done
return 0
}
init() {
# Disable the udev rule 90-zfs.rules to prevent the zfs module
# stack from being loaded due to the detection of a zfs device.
# This is important because the test scripts require full control
# over when and how the modules are loaded/unloaded. A trap is
# set to ensure the udev rule is correctly replaced on exit.
local RULE=${udevruledir}/90-zfs.rules
if test -e ${RULE}; then
trap "mv ${RULE}.disabled ${RULE}" INT TERM EXIT
mv ${RULE} ${RULE}.disabled
fi
# Create a random directory tree of files and sub-directories to
# to act as a copy source for the various regression tests.
SRC_DIR=`mktemp -d -p /var/tmp/ zfs.src.XXXXXXXX`
trap "rm -Rf $SRC_DIR" INT TERM EXIT
populate $SRC_DIR 10 100
}
spl_dump_log() {
${SYSCTL} -w kernel.spl.debug.dump=1 &>/dev/null
local NAME=`dmesg | tail -n 1 | cut -f5 -d' '`
${SPLBUILD}/cmd/spl ${NAME} >${NAME}.log
echo
echo "Dumped debug log: ${NAME}.log"
tail -n1 ${NAME}.log
echo
return 0
}
check_modules() {
local LOADED_MODULES=()
local MISSING_MODULES=()
for MOD in ${MODULES[*]}; do
local NAME=`basename $MOD .ko`
if ${LSMOD} | egrep -q "^${NAME}"; then
LOADED_MODULES=(${NAME} ${LOADED_MODULES[*]})
fi
if [ ${INFOMOD} ${MOD} 2>/dev/null ]; then
MISSING_MODULES=("\t${MOD}\n" ${MISSING_MODULES[*]})
fi
done
if [ ${#LOADED_MODULES[*]} -gt 0 ]; then
ERROR="Unload these modules with '${PROG} -u':\n"
ERROR="${ERROR}${LOADED_MODULES[*]}"
return 1
fi
if [ ${#MISSING_MODULES[*]} -gt 0 ]; then
ERROR="The following modules can not be found,"
ERROR="${ERROR} ensure your source trees are built:\n"
ERROR="${ERROR}${MISSING_MODULES[*]}"
return 1
fi
return 0
}
load_module() {
local NAME=`basename $1 .ko`
if [ ${VERBOSE} ]; then
echo "Loading ${NAME} ($@)"
fi
${LDMOD} $* &>/dev/null || ERROR="Failed to load $1" return 1
return 0
}
load_modules() {
mkdir -p /etc/zfs
for MOD in ${KERNEL_MODULES[*]}; do
load_module ${MOD}
done
for MOD in ${MODULES[*]}; do
local NAME=`basename ${MOD} .ko`
local VALUE=
for OPT in "$@"; do
OPT_NAME=`echo ${OPT} | cut -f1 -d'='`
if [ ${NAME} = "${OPT_NAME}" ]; then
VALUE=`echo ${OPT} | cut -f2- -d'='`
fi
done
load_module ${MOD} ${VALUE} || return 1
done
if [ ${VERBOSE} ]; then
echo "Successfully loaded ZFS module stack"
fi
return 0
}
unload_module() {
local NAME=`basename $1 .ko`
if [ ${VERBOSE} ]; then
echo "Unloading ${NAME} ($@)"
fi
${RMMOD} ${NAME} || ERROR="Failed to unload ${NAME}" return 1
return 0
}
unload_modules() {
local MODULES_REVERSE=( $(echo ${MODULES[@]} |
${AWK} '{for (i=NF;i>=1;i--) printf $i" "} END{print ""}') )
for MOD in ${MODULES_REVERSE[*]}; do
local NAME=`basename ${MOD} .ko`
local USE_COUNT=`${LSMOD} |
egrep "^${NAME} "| ${AWK} '{print $3}'`
if [ "${USE_COUNT}" = 0 ] ; then
if [ "${DUMP_LOG}" -a ${NAME} = "spl" ]; then
spl_dump_log
fi
unload_module ${MOD} || return 1
fi
done
if [ ${VERBOSE} ]; then
echo "Successfully unloaded ZFS module stack"
fi
return 0
}
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
#
# Check that the mdadm utilities are installed.
#
check_loop_utils() {
test -f ${LOSETUP} || die "${LOSETUP} utility must be installed"
}
#
# Find and return an unused loopback device.
#
unused_loop_device() {
for DEVICE in `ls -1 /dev/loop[0-9]* 2>/dev/null`; do
${LOSETUP} ${DEVICE} &>/dev/null
if [ $? -ne 0 ]; then
echo ${DEVICE}
return
fi
done
die "Error: Unable to find unused loopback device"
}
#
# This can be slightly dangerous because the loop devices we are
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
# cleaning up may not be ours. However, if the devices are currently
# in use we will not be able to remove them, and we only remove
# devices which include 'zpool' in the name. So any damage we might
# do should be limited to other zfs related testing.
#
cleanup_loop_devices() {
local TMP_FILE=`mktemp`
${LOSETUP} -a | tr -d '()' >${TMP_FILE}
${AWK} -F":" -v losetup="$LOSETUP" \
'/zpool/ { system("losetup -d "$1) }' ${TMP_FILE}
${AWK} -F" " '/zpool/ { system("rm -f "$3) }' ${TMP_FILE}
rm -f ${TMP_FILE}
}
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
#
# Destroy the passed loopback devices, this is used when you know
# the names of the loopback devices.
#
destroy_loop_devices() {
local LODEVICES="$1"
msg "Destroying ${LODEVICES}"
${LOSETUP} -d ${LODEVICES} || \
die "Error $? destroying ${FILE} -> ${DEVICE} loopback"
rm -f ${FILES}
return 0
}
#
# Create a device label.
#
label() {
local DEVICE=$1
local LABEL=$2
${PARTED} ${DEVICE} --script -- mklabel ${LABEL} || return 1
return 0
}
#
# Create a primary partition on a block device.
#
partition() {
local DEVICE=$1
local TYPE=$2
local START=$3
local END=$4
${PARTED} --align optimal ${DEVICE} --script -- \
mkpart ${TYPE} ${START} ${END} || return 1
udev_trigger
return 0
}
#
# Create a filesystem on the block device
#
format() {
local DEVICE=$1
local FSTYPE=$2
# Force 4K blocksize, else mkfs.ext2 tries to use 8K, which
# won't mount
/sbin/mkfs.${FSTYPE} -b 4096 -F -q ${DEVICE} || return 1
return 0
}
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
#
# Check that the mdadm utilities are installed.
#
check_md_utils() {
test -f ${MDADM} || die "${MDADM} utility must be installed"
test -f ${PARTED} || die "${PARTED} utility must be installed"
}
check_md_partitionable() {
local LOFILE=`mktemp -p /tmp zpool-lo.XXXXXXXX`
local LODEVICE=`unused_loop_device`
local MDDEVICE=`unused_md_device`
local RESULT=1
check_md_utils
rm -f ${LOFILE}
dd if=/dev/zero of=${LOFILE} bs=1M count=0 seek=16 \
&>/dev/null || return ${RESULT}
msg "Creating ${LODEVICE} using ${LOFILE}"
${LOSETUP} ${LODEVICE} ${LOFILE}
if [ $? -ne 0 ]; then
rm -f ${LOFILE}
return ${RESULT}
fi
msg "Creating ${MDDEVICE} using ${LODEVICE}"
${MDADM} --build ${MDDEVICE} --level=faulty \
--raid-devices=1 ${LODEVICE} &>/dev/null
if [ $? -ne 0 ]; then
destroy_loop_devices ${LODEVICE}
rm -f ${LOFILE}
return ${RESULT}
fi
wait_udev ${MDDEVICE} 30
${BLOCKDEV} --rereadpt ${MDDEVICE} 2>/dev/null
RESULT=$?
destroy_md_devices ${MDDEVICE}
destroy_loop_devices ${LODEVICE}
rm -f ${LOFILE}
return ${RESULT}
}
#
# Find and return an unused md device.
#
unused_md_device() {
for (( i=0; i<32; i++ )); do
MDDEVICE=md${i}
# Skip active devicesudo in /proc/mdstat.
grep -q "${MDDEVICE} " /proc/mdstat && continue
# Device doesn't exist, use it.
if [ ! -e $/dev/{MDDEVICE} ]; then
echo /dev/${MDDEVICE}
return
fi
# Device exists but may not be in use.
if [ -b /dev/${MDDEVICE} ]; then
${MDADM} --detail /dev/${MDDEVICE} &>/dev/null
if [ $? -eq 1 ]; then
echo /dev/${MDDEVICE}
return
fi
fi
done
die "Error: Unable to find unused md device"
}
#
# This can be slightly dangerous because it is possible the md devices
# we are cleaning up may not be ours. However, if the devices are
# currently in use we will not be able to remove them, and even if
# we remove devices which were not out we do not zero the super block
# so you should be able to reconstruct them.
#
cleanup_md_devices() {
destroy_md_devices "`ls /dev/md* 2>/dev/null | grep -v p`"
udev_trigger
}
#
# Destroy the passed md devices, this is used when you know
# the names of the md devices.
#
destroy_md_devices() {
local MDDEVICES="$1"
msg "Destroying ${MDDEVICES}"
for MDDEVICE in ${MDDEVICES}; do
${MDADM} --stop ${MDDEVICE} &>/dev/null
${MDADM} --remove ${MDDEVICE} &>/dev/null
${MDADM} --detail ${MDDEVICE} &>/dev/null
done
return 0
}
#
# Check that the scsi utilities are installed.
#
check_sd_utils() {
${INFOMOD} scsi_debug &>/dev/null || die "scsi_debug module required"
test -f ${LSSCSI} || die "${LSSCSI} utility must be installed"
}
#
# Rescan the scsi bus for scsi_debug devices. It is preferable to use the
# scsi-rescan tool if it is installed, but if it's not we can fall back to
# removing and readding the device manually. This rescan will only effect
# the first scsi_debug device if scsi-rescan is missing.
#
scsi_rescan() {
local AWK_SCRIPT="/scsi_debug/ { print \$1; exit }"
if [ -f ${SCSIRESCAN} ]; then
${SCSIRESCAN} --forcerescan --remove &>/dev/null
else
local SCSIID=`${LSSCSI} | ${AWK} "${AWK_SCRIPT}" | tr -d '[]'`
local SCSIHOST=`echo ${SCSIID} | cut -f1 -d':'`
echo 1 >"/sys/class/scsi_device/${SCSIID}/device/delete"
udev_trigger
echo "- - -" >/sys/class/scsi_host/host${SCSIHOST}/scan
udev_trigger
fi
}
#
# Trigger udev and wait for it to settle.
#
udev_trigger() {
if [ -f ${UDEVADM} ]; then
${UDEVADM} trigger --action=change --subsystem-match=block
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
${UDEVADM} settle
else
/sbin/udevtrigger
/sbin/udevsettle
fi
}
#
# The following udev helper functions assume that the provided
# udev rules file will create a /dev/disk/zpool/<CHANNEL><RANK>
# disk mapping. In this mapping each CHANNEL is represented by
# the letters a-z, and the RANK is represented by the numbers
# 1-n. A CHANNEL should identify a group of RANKS which are all
# attached to a single controller, each RANK represents a disk.
# This provides a simply mechanism to locate a specific drive
# given a known hardware configuration.
#
udev_setup() {
local SRC_PATH=$1
# When running in tree manually contruct symlinks in tree to
# the proper devices. Symlinks are installed for all entires
# in the config file regardless of if that device actually
# exists. When installed as a package udev can be relied on for
# this and it will only create links for devices which exist.
if [ ${INTREE} ]; then
PWD=`pwd`
mkdir -p ${DEVDIR}/
cd ${DEVDIR}/
${AWK} '!/^#/ && /./ { system( \
"ln -f -s /dev/disk/by-path/"$2" "$1";" \
"ln -f -s /dev/disk/by-path/"$2"-part1 "$1"p1;" \
"ln -f -s /dev/disk/by-path/"$2"-part9 "$1"p9;" \
) }' $SRC_PATH
cd ${PWD}
else
DST_FILE=`basename ${SRC_PATH} | cut -f1-2 -d'.'`
DST_PATH=/etc/zfs/${DST_FILE}
if [ -e ${DST_PATH} ]; then
die "Error: Config ${DST_PATH} already exists"
fi
cp ${SRC_PATH} ${DST_PATH}
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
udev_trigger
fi
return 0
}
udev_cleanup() {
local SRC_PATH=$1
if [ ${INTREE} ]; then
PWD=`pwd`
cd ${DEVDIR}/
${AWK} '!/^#/ && /./ { system( \
"rm -f "$1" "$1"p1 "$1"p9") }' $SRC_PATH
cd ${PWD}
fi
return 0
}
udev_cr2d() {
local CHANNEL=`echo "obase=16; $1+96" | bc`
local RANK=$2
printf "\x${CHANNEL}${RANK}"
}
udev_raid0_setup() {
local RANKS=$1
local CHANNELS=$2
local IDX=0
RAID0S=()
for RANK in `seq 1 ${RANKS}`; do
for CHANNEL in `seq 1 ${CHANNELS}`; do
DISK=`udev_cr2d ${CHANNEL} ${RANK}`
RAID0S[${IDX}]="${DEVDIR}/${DISK}"
let IDX=IDX+1
done
done
return 0
}
udev_raid10_setup() {
local RANKS=$1
local CHANNELS=$2
local IDX=0
RAID10S=()
for RANK in `seq 1 ${RANKS}`; do
for CHANNEL1 in `seq 1 2 ${CHANNELS}`; do
let CHANNEL2=CHANNEL1+1
DISK1=`udev_cr2d ${CHANNEL1} ${RANK}`
DISK2=`udev_cr2d ${CHANNEL2} ${RANK}`
GROUP="${DEVDIR}/${DISK1} ${DEVDIR}/${DISK2}"
RAID10S[${IDX}]="mirror ${GROUP}"
let IDX=IDX+1
done
done
return 0
}
udev_raidz_setup() {
local RANKS=$1
local CHANNELS=$2
RAIDZS=()
for RANK in `seq 1 ${RANKS}`; do
RAIDZ=("raidz")
for CHANNEL in `seq 1 ${CHANNELS}`; do
DISK=`udev_cr2d ${CHANNEL} ${RANK}`
RAIDZ[${CHANNEL}]="${DEVDIR}/${DISK}"
done
RAIDZS[${RANK}]="${RAIDZ[*]}"
done
return 0
}
udev_raidz2_setup() {
local RANKS=$1
local CHANNELS=$2
RAIDZ2S=()
for RANK in `seq 1 ${RANKS}`; do
RAIDZ2=("raidz2")
for CHANNEL in `seq 1 ${CHANNELS}`; do
DISK=`udev_cr2d ${CHANNEL} ${RANK}`
RAIDZ2[${CHANNEL}]="${DEVDIR}/${DISK}"
done
RAIDZ2S[${RANK}]="${RAIDZ2[*]}"
done
return 0
}
run_one_test() {
local TEST_NUM=$1
local TEST_NAME=$2
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
printf "%-4d %-34s " ${TEST_NUM} "${TEST_NAME}"
test_${TEST_NUM}
}
skip_one_test() {
local TEST_NUM=$1
local TEST_NAME=$2
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
printf "%-4d %-34s " ${TEST_NUM} "${TEST_NAME}"
skip
}
run_test() {
local TEST_NUM=$1
local TEST_NAME=$2
for i in ${TESTS_SKIP[@]}; do
if [[ $i == ${TEST_NUM} ]] ; then
skip_one_test ${TEST_NUM} "${TEST_NAME}"
return 0
fi
done
if [ "${TESTS_RUN[0]}" = "*" ]; then
run_one_test ${TEST_NUM} "${TEST_NAME}"
else
for i in ${TESTS_RUN[@]}; do
if [[ $i == ${TEST_NUM} ]] ; then
run_one_test ${TEST_NUM} "${TEST_NAME}"
return 0
fi
done
skip_one_test ${TEST_NUM} "${TEST_NAME}"
fi
}
wait_udev() {
local DEVICE=$1
local DELAY=$2
local COUNT=0
Add zfault zpool configurations and tests Eleven new zpool configurations were added to allow testing of various failure cases. The first 5 zpool configurations leverage the 'faulty' md device type which allow us to simuluate IO errors at the block layer. The last 6 zpool configurations leverage the scsi_debug module provided by modern kernels. This device allows you to create virtual scsi devices which are backed by a ram disk. With this setup we can verify the full IO stack by injecting faults at the lowest layer. Both methods of fault injection are important to verifying the IO stack. The zfs code itself also provides a mechanism for error injection via the zinject command line tool. While we should also take advantage of this appraoch to validate the code it does not address any of the Linux integration issues which are the most concerning. For the moment we're trusting that the upstream Solaris guys are running zinject and would have caught internal zfs logic errors. Currently, there are 6 r/w test cases layered on top of the 'faulty' md devices. They include 3 writes tests for soft/transient errors, hard/permenant errors, and all writes error to the device. There are 3 matching read tests for soft/transient errors, hard/permenant errors, and fixable read error with a write. Although for this last case zfs doesn't do anything special. The seventh test case verifies zfs detects and corrects checksum errors. In this case one of the drives is extensively damaged and by dd'ing over large sections of it. We then ensure zfs logs the issue and correctly rebuilds the damage. The next test cases use the scsi_debug configuration to injects error at the bottom of the scsi stack. This ensures we find any flaws in the scsi midlayer or our usage of it. Plus it stresses the device specific retry, timeout, and error handling outside of zfs's control. The eighth test case is to verify that the system correctly handles an intermittent device timeout. Here the scsi_debug device drops 1 in N requests resulting in a retry either at the block level. The ZFS code does specify the FAILFAST option but it turns out that for this case the Linux IO stack with still retry the command. The FAILFAST logic located in scsi_noretry_cmd() does no seem to apply to the simply timeout case. It appears to be more targeted to specific device or transport errors from the lower layers. The ninth test case handles a persistent failure in which the device is removed from the system by Linux. The test verifies that the failure is detected, the device is made unavailable, and then can be successfully re-add when brought back online. Additionally, it ensures that errors and events are logged to the correct places and the no data corruption has occured due to the failure.
2010-09-29 03:32:12 +04:00
udev_trigger
while [ ! -e ${DEVICE} ]; do
if [ ${COUNT} -gt ${DELAY} ]; then
return 1
fi
let COUNT=${COUNT}+1
sleep 1
done
return 0
}
stack_clear() {
local STACK_MAX_SIZE=/sys/kernel/debug/tracing/stack_max_size
local STACK_TRACER_ENABLED=/proc/sys/kernel/stack_tracer_enabled
if [ -e $STACK_MAX_SIZE ]; then
echo 1 >$STACK_TRACER_ENABLED
echo 0 >$STACK_MAX_SIZE
fi
}
stack_check() {
local STACK_MAX_SIZE=/sys/kernel/debug/tracing/stack_max_size
local STACK_TRACE=/sys/kernel/debug/tracing/stack_trace
local STACK_LIMIT=7000
if [ -e $STACK_MAX_SIZE ]; then
STACK_SIZE=`cat $STACK_MAX_SIZE`
if [ $STACK_SIZE -ge $STACK_LIMIT ]; then
echo
echo "Warning: max stack size $STACK_SIZE bytes"
cat $STACK_TRACE
fi
fi
}