vdev_id support for device link aliases

Add a vdev_id feature to map device names based on already defined
udev device links.  To increase the odds that vdev_id will run after
the rules it depends on, increase the vdev.rules rule number from 60
to 69.  With this change, vdev_id now provides functionality analogous
to zpool_id and zpool_layout, paving the way to retire those tools.

A defined alias takes precedence over a topology-derived name, but the
two naming methods can otherwise coexist. For example, one might name
drives in a JBOD with the sas_direct topology while naming an internal
L2ARC device with an alias.

For example, the following lines in vdev_id.conf will result in the
creation of links /dev/disk/by-vdev/{d1,d2}, each pointing to the same
target as the device link specified in the third field.

  #     by-vdev
  #     name     fully qualified or base name of device link
  alias d1       /dev/disk/by-id/wwn-0x5000c5002de3b9ca
  alias d2       wwn-0x5000c5002def789e

Also perform some minor vdev_id cleanup, such as removal of the unused
-s command line option.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #981
This commit is contained in:
Ned Bass 2012-11-29 11:08:37 -08:00 committed by Brian Behlendorf
parent e89260a1c8
commit 2957f38d78
7 changed files with 262 additions and 151 deletions

View File

@ -11,13 +11,17 @@
# default numbering is unsatisfactory. The drive aliases will be
# created as symbolic links in /dev/disk/by-vdev.
#
# The only currently supported topologies are sas_direct and
# sas_switch. A multipath mode is supported in which dm-mpath
# devices are handled by examining the first-listed running
# component disk. In multipath mode the configuration file
# should contain a channel definition with the same name for
# each path to a given enclosure.
# The currently supported topologies are sas_direct and sas_switch.
# A multipath mode is supported in which dm-mpath devices are
# handled by examining the first-listed running component disk. In
# multipath mode the configuration file should contain a channel
# definition with the same name for each path to a given enclosure.
#
# The alias keyword provides a simple way to map already-existing
# device symlinks to more convenient names. It is suitable for
# small, static configurations or for sites that have some automated
# way to generate the mapping file.
#
#
# Some example configuration files are given below.
@ -72,12 +76,19 @@
# channel 86:00.0 1 A
# channel 86:00.0 0 B
# #
# # Example vdev_id.conf - alias
# #
#
# # by-vdev
# # name fully qualified or base name of device link
# alias d1 /dev/disk/by-id/wwn-0x5000c5002de3b9ca
# alias d2 wwn-0x5000c5002def789e
PATH=/bin:/sbin:/usr/bin:/usr/sbin
CONFIG=/etc/zfs/vdev_id.conf
PHYS_PER_PORT=
DEV=
SLOT_MAP=
CHANNEL_MAP=
MULTIPATH=
TOPOLOGY=
@ -128,14 +139,188 @@ map_channel() {
printf "%s" ${MAPPED_CHAN}
}
while getopts 'c:s:d:g:mp:h' OPTION; do
sas_handler() {
if [ -z "$PHYS_PER_PORT" ] ; then
PHYS_PER_PORT=`awk "/^phys_per_port /{print \\$2;exit}" $CONFIG`
fi
PHYS_PER_PORT=${PHYS_PER_PORT:-4}
if ! echo $PHYS_PER_PORT | grep -q -E '^[0-9]+$' ; then
echo "Error: phys_per_port value $PHYS_PER_PORT is non-numeric"
exit 1
fi
if [ -z "$MULTIPATH_MODE" ] ; then
MULTIPATH_MODE=`awk "/^multipath /{print \\$2; exit}" $CONFIG`
fi
# Use first running component device if we're handling a dm-mpath device
if [ "$MULTIPATH_MODE" = "yes" ] ; then
# If udev didn't tell us the UUID via DM_NAME, check /dev/mapper
if [ -z "$DM_NAME" ] ; then
DM_NAME=`ls -l --full-time /dev/mapper |
awk "/\/$DEV$/{print \\$9}"`
fi
# For raw disks udev exports DEVTYPE=partition when
# handling partitions, and the rules can be written to
# take advantage of this to append a -part suffix. For
# dm devices we get DEVTYPE=disk even for partitions so
# we have to append the -part suffix directly in the
# helper.
if [ "$DEVTYPE" != "partition" ] ; then
PART=`echo $DM_NAME | awk -Fp '/p/{print "-part"$2}'`
fi
# Strip off partition information.
DM_NAME=`echo $DM_NAME | sed 's/p[0-9][0-9]*$//'`
if [ -z "$DM_NAME" ] ; then
return
fi
# Get the raw scsi device name from multipath -l.
DEV=`multipath -l $DM_NAME |awk '/running/{print $3 ; exit}'`
if [ -z "$DEV" ] ; then
return
fi
fi
if echo $DEV | grep -q ^/devices/ ; then
sys_path=$DEV
else
sys_path=`udevadm info -q path -p /sys/block/$DEV 2>/dev/null`
fi
# Use positional parameters as an ad-hoc array
set -- $(echo "$sys_path" | tr / ' ')
num_dirs=$#
scsi_host_dir="/sys"
# Get path up to /sys/.../hostX
i=1
while [ $i -le $num_dirs ] ; do
d=$(eval echo \${$i})
scsi_host_dir="$scsi_host_dir/$d"
echo $d | grep -q -E '^host[0-9]+$' && break
i=$(($i + 1))
done
if [ $i = $num_dirs ] ; then
return
fi
PCI_ID=$(eval echo \${$(($i -1))} | awk -F: '{print $2":"$3}')
# In sas_switch mode, the directory four levels beneath
# /sys/.../hostX contains symlinks to phy devices that reveal
# the switch port number. In sas_direct mode, the phy links one
# directory down reveal the HBA port.
port_dir=$scsi_host_dir
case $TOPOLOGY in
"sas_switch") j=$(($i + 4)) ;;
"sas_direct") j=$(($i + 1)) ;;
esac
i=$(($i + 1))
while [ $i -le $j ] ; do
port_dir="$port_dir/$(eval echo \${$i})"
i=$(($i + 1))
done
PHY=`ls -d $port_dir/phy* 2>/dev/null | head -1 | awk -F: '{print $NF}'`
if [ -z "$PHY" ] ; then
return
fi
PORT=$(( $PHY / $PHYS_PER_PORT ))
# Look in /sys/.../sas_device/end_device-X for the bay_identifier
# attribute.
end_device_dir=$port_dir
while [ $i -lt $num_dirs ] ; do
d=$(eval echo \${$i})
end_device_dir="$end_device_dir/$d"
if echo $d | grep -q '^end_device' ; then
end_device_dir="$end_device_dir/sas_device/$d"
break
fi
i=$(($i + 1))
done
SLOT=`cat $end_device_dir/bay_identifier 2>/dev/null`
if [ -z "$SLOT" ] ; then
return
fi
SLOT=`map_slot $SLOT`
CHAN=`map_channel $PCI_ID $PORT`
if [ -z "$CHAN" ] ; then
return
fi
echo ${CHAN}${SLOT}${PART}
}
alias_handler () {
# Special handling is needed to correctly append a -part suffix
# to partitions of device mapper devices. The DEVTYPE attribute
# is normally set to "disk" instead of "partition" in this case,
# so the udev rules won't handle that for us as they do for
# "plain" block devices.
#
# For example, we may have the following links for a device and its
# partitions,
#
# /dev/disk/by-id/dm-name-isw_dibgbfcije_ARRAY0 -> ../../dm-0
# /dev/disk/by-id/dm-name-isw_dibgbfcije_ARRAY0p1 -> ../../dm-1
# /dev/disk/by-id/dm-name-isw_dibgbfcije_ARRAY0p2 -> ../../dm-3
#
# and the following alias in vdev_id.conf.
#
# alias A0 dm-name-isw_dibgbfcije_ARRAY0
#
# The desired outcome is for the following links to be created
# without having explicitly defined aliases for the partitions.
#
# /dev/disk/by-vdev/A0 -> ../../dm-0
# /dev/disk/by-vdev/A0-part1 -> ../../dm-1
# /dev/disk/by-vdev/A0-part2 -> ../../dm-3
#
# Warning: The following grep pattern will misidentify whole-disk
# devices whose names end with 'p' followed by a string of
# digits as partitions, causing alias creation to fail. This
# ambiguity seems unavoidable, so devices using this facility
# must not use such names.
local DM_PART=
if echo $DM_NAME | grep -q -E 'p[0-9][0-9]*$' ; then
if [ "$DEVTYPE" != "partition" ] ; then
DM_PART=`echo $DM_NAME | awk -Fp '/p/{print "-part"$2}'`
fi
fi
# DEVLINKS attribute must have been populated by already-run udev rules.
for link in $DEVLINKS ; do
# Remove partition information to match key of top-level device.
if [ -n "$DM_PART" ] ; then
link=`echo $link | sed 's/p[0-9][0-9]*$//'`
fi
# Check both the fully qualified and the base name of link.
for l in $link `basename $link` ; do
alias=`awk "/^alias / && \\$3 == \"${l}\" \
{ print \\$2; exit }" $CONFIG`
if [ -n "$alias" ] ; then
echo ${alias}${DM_PART}
return
fi
done
done
}
while getopts 'c:d:g:mp:h' OPTION; do
case ${OPTION} in
c)
CONFIG=`readlink -e ${OPTARG}`
;;
CONFIG=${OPTARG}
;;
d)
DEV=${OPTARG}
;;
DEV=${OPTARG}
;;
g)
TOPOLOGY=$OPTARG
;;
@ -145,13 +330,6 @@ while getopts 'c:s:d:g:mp:h' OPTION; do
m)
MULTIPATH_MODE=yes
;;
s)
SLOT_MAP=`readlink -e ${OPTARG}`
if [ ! -r $SLOT_MAP ] ; then
echo "Error: $SLOT_MAP is nonexistant or unreadable"
exit 1
fi
;;
h)
usage
;;
@ -170,129 +348,24 @@ fi
if [ -z "$TOPOLOGY" ] ; then
TOPOLOGY=`awk "/^topology /{print \\$2; exit}" $CONFIG`
fi
TOPOLOGY=${TOPOLOGY:-sas_direct}
case $TOPOLOGY in
sas_direct|sas_switch)
;;
*)
echo "Error: unknown topology $TOPOLOGY"
exit 1
;;
esac
if [ -z "$PHYS_PER_PORT" ] ; then
PHYS_PER_PORT=`awk "/^phys_per_port /{print \\$2; exit}" $CONFIG`
fi
PHYS_PER_PORT=${PHYS_PER_PORT:-4}
if ! echo $PHYS_PER_PORT | grep -q -E '^[0-9]+$' ; then
echo "Error: phys_per_port value $PHYS_PER_PORT is non-numeric"
exit 1
# First check if an alias was defined for this device.
ID_VDEV=`alias_handler`
if [ -z "$ID_VDEV" ] ; then
TOPOLOGY=${TOPOLOGY:-sas_direct}
case $TOPOLOGY in
sas_direct|sas_switch)
ID_VDEV=`sas_handler`
;;
*)
echo "Error: unknown topology $TOPOLOGY"
exit 1
;;
esac
fi
if [ -z "$MULTIPATH_MODE" ] ; then
MULTIPATH_MODE=`awk "/^multipath /{print \\$2; exit}" $CONFIG`
if [ -n "$ID_VDEV" ] ; then
echo "ID_VDEV=${ID_VDEV}"
echo "ID_VDEV_PATH=disk/by-vdev/${ID_VDEV}"
fi
# Use first running component device if we're handling a dm-mpath device.
if [ "$MULTIPATH_MODE" = "yes" ] ; then
# If udev didn't tell us the UUID via DM_NAME, find it in /dev/mapper
if [ -z "$DM_NAME" ] ; then
DM_NAME=`ls -l --full-time /dev/mapper |
awk "/\/$DEV$/{print \\$9}"`
fi
# For raw disks udev exports DEVTYPE=partition when handling partitions,
# and the rules can be written to take advantage of this to append a
# -part suffix. For dm devices we get DEVTYPE=disk even for partitions
# so we have to append the -part suffix directly in the helper.
if [ "$DEVTYPE" != "partition" ] ; then
PART=`echo $DM_NAME | awk -Fp '/p/{print "-part"$2}'`
fi
# Strip off partition information.
DM_NAME=`echo $DM_NAME | sed 's/p[0-9][0-9]*$//'`
if [ -z "$DM_NAME" ] ; then
exit 0
fi
# Get the raw scsi device name from multipath -l.
DEV=`multipath -l $DM_NAME |awk '/running/{print $3 ; exit}'`
if [ -z "$DEV" ] ; then
exit 0
fi
fi
if echo $DEV | grep -q ^/devices/ ; then
sys_path=$DEV
else
sys_path=`udevadm info -q path -p /sys/block/$DEV 2>/dev/null`
fi
# Use positional parameters as an ad-hoc array
set -- $(echo "$sys_path" | tr / ' ')
num_dirs=$#
scsi_host_dir="/sys"
# Get path up to /sys/.../hostX
i=1
while [ $i -le $num_dirs ] ; do
d=$(eval echo \${$i})
scsi_host_dir="$scsi_host_dir/$d"
echo $d | grep -q -E '^host[0-9]+$' && break
i=$(($i + 1))
done
if [ $i = $num_dirs ] ; then
exit 0
fi
PCI_ID=$(eval echo \${$(($i -1))} | awk -F: '{print $2":"$3}')
# In sas_switch mode, the directory four levels beneath /sys/.../hostX
# contains symlinks to phy devices that reveal the switch port number. In
# sas_direct mode, the phy links one directory down reveal the HBA port.
port_dir=$scsi_host_dir
case $TOPOLOGY in
"sas_switch") j=$(($i + 4)) ;;
"sas_direct") j=$(($i + 1)) ;;
esac
i=$(($i + 1))
while [ $i -le $j ] ; do
port_dir="$port_dir/$(eval echo \${$i})"
i=$(($i + 1))
done
PHY=`ls -d $port_dir/phy* 2>/dev/null | head -1 | awk -F: '{print $NF}'`
if [ -z "$PHY" ] ; then
exit 0
fi
PORT=$(( $PHY / $PHYS_PER_PORT ))
# Look in /sys/.../sas_device/end_device-X for the bay_identifier
# attribute.
end_device_dir=$port_dir
while [ $i -lt $num_dirs ] ; do
d=$(eval echo \${$i})
end_device_dir="$end_device_dir/$d"
if echo $d | grep -q '^end_device' ; then
end_device_dir="$end_device_dir/sas_device/$d"
break
fi
i=$(($i + 1))
done
SLOT=`cat $end_device_dir/bay_identifier 2>/dev/null`
if [ -z "$SLOT" ] ; then
exit 0
fi
SLOT=`map_slot $SLOT`
CHAN=`map_channel $PCI_ID $PORT`
if [ -z "$CHAN" ] ; then
exit 0
fi
ID_VDEV=${CHAN}${SLOT}${PART}
echo "ID_VDEV=${ID_VDEV}"
echo "ID_VDEV_PATH=disk/by-vdev/${ID_VDEV}"

View File

@ -1,6 +1,7 @@
pkgsysconfdir = $(sysconfdir)/zfs
pkgsysconf_DATA = \
vdev_id.conf.alias.example \
vdev_id.conf.sas_direct.example \
vdev_id.conf.sas_switch.example \
vdev_id.conf.multipath.example \

View File

@ -0,0 +1,4 @@
# by-vdev
# name fully qualified or base name of device link
alias d1 /dev/disk/by-id/wwn-0x5000c5002de3b9ca
alias d2 wwn-0x5000c5002def789e

View File

@ -17,6 +17,22 @@ keyword is ignored. Comments may optionally begin with a hash
character.
The following keywords and values are used.
.TP
\fIalias\fR <name> <devlink>
Maps a device link in the /dev directory hierarchy to a new device
name. The udev rule defining the device link must have run prior to
.BR vdev_id (8).
A defined alias takes precedence over a topology-derived name, but the
two naming methods can otherwise coexist. For example, one might name
drives in a JBOD with the sas_direct topology while naming an internal
L2ARC device with an alias.
\fIname\fR - the name of the link to the device that will by created in
/dev/disk/by-vdev.
\fIdevlink\fR - the name of the device link that has already been
defined by udev. This may be an absolute path or the base filename.
.TP
\fIchannel\fR [pci_slot] <port> <name>
Maps a physical path to a channel name (typically representing a single
@ -59,7 +75,7 @@ a SAS switch port number
.TP
\fIphys_per_port\fR <num>
Specifies the number of PHY devices are associated with a SAS HBA port or SAS
Specifies the number of PHY devices associated with a SAS HBA port or SAS
switch port.
.BR vdev_id (8)
internally uses this value to determine which HBA or switch port a
@ -150,6 +166,18 @@ definitions - one per physical path.
channel 86:00.0 0 B
.br
.P
A configuration using device link aliases.
.P
.br
# by-vdev
.br
# name fully qualified or base name of device link
.br
alias d1 /dev/disk/by-id/wwn-0x5000c5002de3b9ca
.br
alias d2 wwn-0x5000c5002def789e
.br
.P
.SH FILES
.TP

View File

@ -19,15 +19,20 @@ drives. Slot numbers may also be re-mapped in case the default
numbering is unsatisfactory. The drive aliases will be created as
symbolic links in /dev/disk/by-vdev.
The only currently supported topologies are sas_direct and
sas_switch. A multipath mode is supported in which dm-mpath
devices are handled by examining the first-listed running
component disk as reported by the
The currently supported topologies are sas_direct and sas_switch. A
multipath mode is supported in which dm-mpath devices are handled by
examining the first-listed running component disk as reported by the
.BR multipath (8)
command. In multipath mode the configuration file should contain a
channel definition with the same name for each path to a given
enclosure.
.BR vdev_id
also supports creating aliases based on existing udev links in the /dev
hierarchy using the \fIalias\fR configuration file keyword. See the
.BR vdev_id.conf (5)
man page for details.
.SH OPTIONS
.TP
\fB\-c\fR <config_file>

View File

@ -1,5 +1,5 @@
#
# @udevdir@/rules.d/60-vdev.rules
# @udevdir@/rules.d/69-vdev.rules
#
ENV{DEVTYPE}=="disk", IMPORT{program}="@udevdir@/vdev_id -d %k"

View File

@ -1,11 +1,11 @@
udevrule_DATA = \
$(top_srcdir)/udev/rules.d/60-vdev.rules \
$(top_srcdir)/udev/rules.d/69-vdev.rules \
$(top_srcdir)/udev/rules.d/60-zpool.rules \
$(top_srcdir)/udev/rules.d/60-zvol.rules \
$(top_srcdir)/udev/rules.d/90-zfs.rules
EXTRA_DIST = \
$(top_srcdir)/udev/rules.d/60-vdev.rules.in \
$(top_srcdir)/udev/rules.d/69-vdev.rules.in \
$(top_srcdir)/udev/rules.d/60-zpool.rules.in \
$(top_srcdir)/udev/rules.d/60-zvol.rules.in \
$(top_srcdir)/udev/rules.d/90-zfs.rules.in