The DEVHTL lookup in btrfs/003 is broken. It can only handle full LUNs and
not partitions on a disk.
Rather than returning 2:0:0:0 for /dev/sdc7, it returns
''block'' and we see:
./common/rc: line 2081: /sys/class/scsi_device/block/device/delete:
No such file or directory
If we look up the device by dev instead of by name, we can handle working
with full disks and partitions more easily and get the ability to use
any device name rather than just the ones that match sysfs.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
---
tests/btrfs/003 | 18 +++++++++++++++---
1 file changed, 15 insertions(+), 3 deletions(-)
--- a/tests/btrfs/003
+++ b/tests/btrfs/003
@@ -137,9 +137,21 @@ _test_replace()
#pick the 2nd last disk
ds=${devs[@]:$(($n-1)):1}
- # retrive the HTL for this scsi disk
- d=`echo $ds|cut -d"/" -f3`
- DEVHTL=`ls -l /sys/class/block/${d} | rev | cut -d "/" -f 3 | rev`
+ HEXMAJOR="$(stat -c "%t" "$ds")"
+ HEXMINOR="$(stat -c "%T" "$ds")"
+ if [ -z "$HEXMAJOR" -o -z "$HEXMINOR" ]; then
+ _fail "tr: HEXMAJOR and/or HEXMINOR is unset for $ds"
+ fi
+
+ DIR="/sys/dev/block/$(( 0x$HEXMAJOR )):$(( 0x$HEXMINOR ))"
+
+ if [ -L "$DIR/device" ]; then # whole disk
+ DEVHTL="$(basename $(readlink "$DIR/device"))"
+ elif [ -L "$DIR/../device" ]; then # partition
+ DEVHTL="$(basename $(readlink "$DIR/../device"))"
+ else
+ _fail "tr: Can''t locate device backing $ds"
+ fi
#fail disk
_devmgt_remove ${DEVHTL}
--
Jeff Mahoney
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Add xfs@oss.sgi.com to the cc list. On Tue, Sep 03, 2013 at 06:19:01PM -0400, Jeff Mahoney wrote:> The DEVHTL lookup in btrfs/003 is broken. It can only handle full LUNs and > not partitions on a disk. > > Rather than returning 2:0:0:0 for /dev/sdc7, it returns ''block'' and we see: > ./common/rc: line 2081: /sys/class/scsi_device/block/device/delete: > No such file or directory > > If we look up the device by dev instead of by name, we can handle working > with full disks and partitions more easily and get the ability to use > any device name rather than just the ones that match sysfs. > > Signed-off-by: Jeff Mahoney <jeffm@suse.com> > --- > tests/btrfs/003 | 18 +++++++++++++++--- > 1 file changed, 15 insertions(+), 3 deletions(-) > > --- a/tests/btrfs/003 > +++ b/tests/btrfs/003 > @@ -137,9 +137,21 @@ _test_replace() > #pick the 2nd last disk > ds=${devs[@]:$(($n-1)):1} > > - # retrive the HTL for this scsi disk > - d=`echo $ds|cut -d"/" -f3` > - DEVHTL=`ls -l /sys/class/block/${d} | rev | cut -d "/" -f 3 | rev` > + HEXMAJOR="$(stat -c "%t" "$ds")" > + HEXMINOR="$(stat -c "%T" "$ds")" > + if [ -z "$HEXMAJOR" -o -z "$HEXMINOR" ]; then > + _fail "tr: HEXMAJOR and/or HEXMINOR is unset for $ds" > + fi > + > + DIR="/sys/dev/block/$(( 0x$HEXMAJOR )):$(( 0x$HEXMINOR ))" > + > + if [ -L "$DIR/device" ]; then # whole disk > + DEVHTL="$(basename $(readlink "$DIR/device"))" > + elif [ -L "$DIR/../device" ]; then # partition > + DEVHTL="$(basename $(readlink "$DIR/../device"))" > + else > + _fail "tr: Can''t locate device backing $ds" > + fi > > #fail disk > _devmgt_remove ${DEVHTL} > > -- > Jeff Mahoney > SUSE Labs > -- > To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >-- Dave Chinner david@fromorbit.com _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs
On Tue, Sep 03, 2013 at 06:19:01PM -0400, Jeff Mahoney wrote:> The DEVHTL lookup in btrfs/003 is broken. It can only handle full LUNs and > not partitions on a disk. > > Rather than returning 2:0:0:0 for /dev/sdc7, it returns ''block'' and we see: > ./common/rc: line 2081: /sys/class/scsi_device/block/device/delete: > No such file or directory > > If we look up the device by dev instead of by name, we can handle working > with full disks and partitions more easily and get the ability to use > any device name rather than just the ones that match sysfs. > > @@ -137,9 +137,21 @@ _test_replace() > #pick the 2nd last disk > ds=${devs[@]:$(($n-1)):1} > > - # retrive the HTL for this scsi disk > - d=`echo $ds|cut -d"/" -f3` > - DEVHTL=`ls -l /sys/class/block/${d} | rev | cut -d "/" -f 3 | rev` > + HEXMAJOR="$(stat -c "%t" "$ds")" > + HEXMINOR="$(stat -c "%T" "$ds")" > + if [ -z "$HEXMAJOR" -o -z "$HEXMINOR" ]; then > + _fail "tr: HEXMAJOR and/or HEXMINOR is unset for $ds" > + fi > + > + DIR="/sys/dev/block/$(( 0x$HEXMAJOR )):$(( 0x$HEXMINOR ))" > + > + if [ -L "$DIR/device" ]; then # whole disk > + DEVHTL="$(basename $(readlink "$DIR/device"))" > + elif [ -L "$DIR/../device" ]; then # partition > + DEVHTL="$(basename $(readlink "$DIR/../device"))" > + else > + _fail "tr: Can''t locate device backing $ds" > + fi > > #fail disk > _devmgt_remove ${DEVHTL}The test assumes tha SCRATCH_DEV_POOL consists of standalone devices and more strongly that they can be removed from the system by $ echo 1 > /sys/class/scsi_device/sdx/device/delete via _devmgt_remove. Ordinary sdX devices are ok, but eg. MD devices are not because they do not have a corresponding "2:0:0:0"-like entry (I haven''t checked device-mapper devices). Looks like the tests have to do more fine grained checks of the devices in SCRATCH_DEV_POOL because eg. the ''device replace'' test btrfs/011 is just fine with partitions. david -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 9/6/13 10:03 AM, David Sterba wrote:> On Tue, Sep 03, 2013 at 06:19:01PM -0400, Jeff Mahoney wrote: >> The DEVHTL lookup in btrfs/003 is broken. It can only handle full LUNs and >> not partitions on a disk. >> >> Rather than returning 2:0:0:0 for /dev/sdc7, it returns ''block'' and we see: >> ./common/rc: line 2081: /sys/class/scsi_device/block/device/delete: >> No such file or directory >> >> If we look up the device by dev instead of by name, we can handle working >> with full disks and partitions more easily and get the ability to use >> any device name rather than just the ones that match sysfs. >> >> @@ -137,9 +137,21 @@ _test_replace() >> #pick the 2nd last disk >> ds=${devs[@]:$(($n-1)):1} >> >> - # retrive the HTL for this scsi disk >> - d=`echo $ds|cut -d"/" -f3` >> - DEVHTL=`ls -l /sys/class/block/${d} | rev | cut -d "/" -f 3 | rev` >> + HEXMAJOR="$(stat -c "%t" "$ds")" >> + HEXMINOR="$(stat -c "%T" "$ds")" >> + if [ -z "$HEXMAJOR" -o -z "$HEXMINOR" ]; then >> + _fail "tr: HEXMAJOR and/or HEXMINOR is unset for $ds" >> + fi >> + >> + DIR="/sys/dev/block/$(( 0x$HEXMAJOR )):$(( 0x$HEXMINOR ))" >> + >> + if [ -L "$DIR/device" ]; then # whole disk >> + DEVHTL="$(basename $(readlink "$DIR/device"))" >> + elif [ -L "$DIR/../device" ]; then # partition >> + DEVHTL="$(basename $(readlink "$DIR/../device"))" >> + else >> + _fail "tr: Can''t locate device backing $ds" >> + fi >> >> #fail disk >> _devmgt_remove ${DEVHTL} > > The test assumes tha SCRATCH_DEV_POOL consists of standalone devices > and more strongly that they can be removed from the system by > > $ echo 1 > /sys/class/scsi_device/sdx/device/delete > > via _devmgt_remove. Ordinary sdX devices are ok, but eg. MD devices are > not because they do not have a corresponding "2:0:0:0"-like entry (I > haven''t checked device-mapper devices). > > Looks like the tests have to do more fine grained checks of the devices > in SCRATCH_DEV_POOL because eg. the ''device replace'' test btrfs/011 is > just fine with partitions.Yeah. It doesn''t work with any type of device that doesn''t present a SCSI-style host interface with a delete sysfs entry. I wasn''t looking to make it work everywhere, just allow it to work with the types of devices it was originally targeting in a slightly less fragile way. -Jeff -- Jeff Mahoney SUSE Labs _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs