2017-10-30 22:53:06 |
Ryan Harper |
bug |
|
|
added bug |
2017-10-31 00:28:13 |
Nobuto Murata |
bug |
|
|
added subscriber Nobuto Murata |
2017-11-19 14:25:49 |
Dmitrii Shcherbakov |
bug task added |
|
bcache-tools |
|
2017-11-20 09:02:38 |
Ante Karamatić |
tags |
|
cpe-onsite |
|
2018-01-27 15:25:39 |
Dmitrii Shcherbakov |
bug |
|
|
added subscriber Dmitrii Shcherbakov |
2018-01-29 17:56:16 |
Ryan Harper |
attachment added |
|
fix_bcache_dname.patch https://bugs.launchpad.net/curtin/+bug/1728742/+attachment/5045099/+files/fix_bcache_dname.patch |
|
2018-01-30 21:32:38 |
Launchpad Janitor |
merge proposal linked |
|
https://code.launchpad.net/~raharper/curtin/+git/curtin/+merge/336871 |
|
2018-02-08 23:01:10 |
Ryan Harper |
attachment added |
|
bionic-lp172874-bcache-tools.debdiff https://bugs.launchpad.net/bcache-tools/+bug/1728742/+attachment/5051507/+files/bionic-lp172874-bcache-tools.debdiff |
|
2018-02-08 23:02:08 |
Ryan Harper |
attachment added |
|
xenial-lp172874-bcache-tools.debdiff https://bugs.launchpad.net/bcache-tools/+bug/1728742/+attachment/5051508/+files/xenial-lp172874-bcache-tools.debdiff |
|
2018-02-14 17:03:13 |
Scott Moser |
bug task added |
|
bcache-tools (Ubuntu) |
|
2018-02-14 17:03:28 |
Scott Moser |
bcache-tools (Ubuntu): status |
New |
Confirmed |
|
2018-02-14 17:03:50 |
Scott Moser |
bcache-tools (Ubuntu): importance |
Undecided |
Medium |
|
2018-02-14 17:03:53 |
Scott Moser |
bcache-tools (Ubuntu): status |
Confirmed |
In Progress |
|
2018-02-14 17:03:59 |
Scott Moser |
bcache-tools (Ubuntu): assignee |
|
Ryan Harper (raharper) |
|
2018-02-14 17:04:11 |
Scott Moser |
nominated for series |
|
Ubuntu Xenial |
|
2018-02-14 17:04:11 |
Scott Moser |
bug task added |
|
bcache-tools (Ubuntu Xenial) |
|
2018-02-14 17:04:11 |
Scott Moser |
nominated for series |
|
Ubuntu Trusty |
|
2018-02-14 17:04:11 |
Scott Moser |
bug task added |
|
bcache-tools (Ubuntu Trusty) |
|
2018-02-14 17:04:11 |
Scott Moser |
nominated for series |
|
Ubuntu Bionic |
|
2018-02-14 17:04:11 |
Scott Moser |
bug task added |
|
bcache-tools (Ubuntu Bionic) |
|
2018-02-14 17:04:11 |
Scott Moser |
nominated for series |
|
Ubuntu Artful |
|
2018-02-14 17:04:11 |
Scott Moser |
bug task added |
|
bcache-tools (Ubuntu Artful) |
|
2018-02-14 17:04:19 |
Scott Moser |
bcache-tools (Ubuntu Trusty): status |
New |
Confirmed |
|
2018-02-14 17:04:21 |
Scott Moser |
bcache-tools (Ubuntu Xenial): status |
New |
Confirmed |
|
2018-02-14 17:04:24 |
Scott Moser |
bcache-tools (Ubuntu Artful): status |
New |
Confirmed |
|
2018-02-14 17:04:27 |
Scott Moser |
bcache-tools (Ubuntu Trusty): importance |
Undecided |
Medium |
|
2018-02-14 17:04:30 |
Scott Moser |
bcache-tools (Ubuntu Xenial): importance |
Undecided |
Medium |
|
2018-02-14 17:04:33 |
Scott Moser |
bcache-tools (Ubuntu Artful): importance |
Undecided |
Medium |
|
2018-02-14 20:19:19 |
Ubuntu Foundations Team Bug Bot |
tags |
cpe-onsite |
cpe-onsite patch |
|
2018-02-14 20:19:27 |
Ubuntu Foundations Team Bug Bot |
bug |
|
|
added subscriber Ubuntu Review Team |
2018-02-14 22:04:15 |
Ryan Harper |
bug watch added |
|
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=890446 |
|
2018-02-14 22:04:15 |
Ryan Harper |
bug task added |
|
bcache-tools (Debian) |
|
2018-02-14 22:44:44 |
Bug Watch Updater |
bcache-tools (Debian): status |
Unknown |
New |
|
2018-02-15 02:15:02 |
Ryan Harper |
merge proposal linked |
|
https://code.launchpad.net/~raharper/ubuntu/+source/bcache-tools/+git/bcache-tools/+merge/337751 |
|
2018-02-15 03:07:16 |
Ryan Harper |
description |
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time. |
[Impact]
* Current users of bcache devices may encounter unreliable device
numbering as the Linux kernel does not guarantee that bcache
minor numbers are assigned to the same devices at each boot.
Users who may have used /dev/bcacheN in paths to a specific
device could possible be pointing to a different dataset
altogether. bcache udev rules do provide some mechanism to
generate persistent symlinks in /dev/bcache/by-uuid or
/dev/bcache/by-label which is based on superblock data on
the underlying device. However, the Linux kernel does not
always generate an kernel uevent to trigger the udev rules
to create the symlink.
* The fix adds a udev program which will read bcache superblock
of slave devices and extract the UUID and LABEL, exporting them
to udev for use in the bcache rule files.
* This is affected in upstream bcache-tools, the owning package
of the udev rules. This affects all releases of bcache-tools
as the rules rely upon the kernel to trigger these events,
though that is not a requirement to resolve the lack of
persistent links.
[Test Case]
* Launch and Ubuntu Cloud Image with 3 unused disks
- apt install bcache-tools tree
- make-bcache -C /dev/vdb
- make-bcache -B /dev/vdc
- make-bcache -B /dev/vdd
- echo "vdc" > /sys/class/block/bcache0/bcache/label
- echo "vdd" > /sys/class/block/bcache1/bcache/label
- reboot
- Run this test:
#!/bin/bash
FAIL=0
[ ! -d /dev/bcache ] && {
echo "FAIL: /dev/bcache is not a directory";
exit 1
}
for label in /dev/bcache/by-label/*; do
LABEL_TARGET="$(ls -1 /sys/class/block/`basename $label`/holders/)"
DEVNAME=`readlink -f $label`;
KNAME="${DEVNAME#*/dev/}"
if [ "$LABEL_TARGET" != "$KNAME" ]; then
echo "FAIL: label points to $LABEL_TARGET but symlink points to $DEVNAME";
FAIL=1
fi;
done
if [ "$FAIL" == "0" ]; then
echo "PASS";
exit 0
fi
exit 1
[Regression Potential]
* As bcache minor numbers and these symlinks have been unreliable in
the past there may be code that makes assumptions about
/dev/bcache* expanded only to the block devices, versus
/dev/bcache which is a directory.
[Original Description]
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time. |
|
2018-02-15 16:53:01 |
Scott Moser |
description |
[Impact]
* Current users of bcache devices may encounter unreliable device
numbering as the Linux kernel does not guarantee that bcache
minor numbers are assigned to the same devices at each boot.
Users who may have used /dev/bcacheN in paths to a specific
device could possible be pointing to a different dataset
altogether. bcache udev rules do provide some mechanism to
generate persistent symlinks in /dev/bcache/by-uuid or
/dev/bcache/by-label which is based on superblock data on
the underlying device. However, the Linux kernel does not
always generate an kernel uevent to trigger the udev rules
to create the symlink.
* The fix adds a udev program which will read bcache superblock
of slave devices and extract the UUID and LABEL, exporting them
to udev for use in the bcache rule files.
* This is affected in upstream bcache-tools, the owning package
of the udev rules. This affects all releases of bcache-tools
as the rules rely upon the kernel to trigger these events,
though that is not a requirement to resolve the lack of
persistent links.
[Test Case]
* Launch and Ubuntu Cloud Image with 3 unused disks
- apt install bcache-tools tree
- make-bcache -C /dev/vdb
- make-bcache -B /dev/vdc
- make-bcache -B /dev/vdd
- echo "vdc" > /sys/class/block/bcache0/bcache/label
- echo "vdd" > /sys/class/block/bcache1/bcache/label
- reboot
- Run this test:
#!/bin/bash
FAIL=0
[ ! -d /dev/bcache ] && {
echo "FAIL: /dev/bcache is not a directory";
exit 1
}
for label in /dev/bcache/by-label/*; do
LABEL_TARGET="$(ls -1 /sys/class/block/`basename $label`/holders/)"
DEVNAME=`readlink -f $label`;
KNAME="${DEVNAME#*/dev/}"
if [ "$LABEL_TARGET" != "$KNAME" ]; then
echo "FAIL: label points to $LABEL_TARGET but symlink points to $DEVNAME";
FAIL=1
fi;
done
if [ "$FAIL" == "0" ]; then
echo "PASS";
exit 0
fi
exit 1
[Regression Potential]
* As bcache minor numbers and these symlinks have been unreliable in
the past there may be code that makes assumptions about
/dev/bcache* expanded only to the block devices, versus
/dev/bcache which is a directory.
[Original Description]
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time. |
[Impact]
* Current users of bcache devices may encounter unreliable device
numbering as the Linux kernel does not guarantee that bcache
minor numbers are assigned to the same devices at each boot.
Users who may have used /dev/bcacheN in paths to a specific
device could possible be pointing to a different dataset
altogether. bcache udev rules do provide some mechanism to
generate persistent symlinks in /dev/bcache/by-uuid or
/dev/bcache/by-label which is based on superblock data on
the underlying device. However, the Linux kernel does not
always generate an kernel uevent to trigger the udev rules
to create the symlink.
* The fix adds a udev program which will read bcache superblock
of slave devices and extract the UUID and LABEL, exporting them
to udev for use in the bcache rule files.
* This is affected in upstream bcache-tools, the owning package
of the udev rules. This affects all releases of bcache-tools
as the rules rely upon the kernel to trigger these events,
though that is not a requirement to resolve the lack of
persistent links.
[Test Case]
* Launch and Ubuntu Cloud Image with 3 unused disks
- apt install bcache-tools tree
- make-bcache -C /dev/vdb
- make-bcache -B /dev/vdc
- make-bcache -B /dev/vdd
- echo "vdc" > /sys/class/block/bcache0/bcache/label
- echo "vdd" > /sys/class/block/bcache1/bcache/label
- reboot
- Run this test:
#!/bin/bash
FAIL=0
[ ! -d /dev/bcache ] && {
echo "FAIL: /dev/bcache is not a directory";
exit 1
}
for label in /dev/bcache/by-label/*; do
LABEL_TARGET="$(ls -1 /sys/class/block/`basename $label`/holders/)"
DEVNAME=`readlink -f $label`;
KNAME="${DEVNAME#*/dev/}"
if [ "$LABEL_TARGET" != "$KNAME" ]; then
echo "FAIL: label points to $LABEL_TARGET but symlink points to $DEVNAME";
FAIL=1
fi;
done
if [ "$FAIL" == "0" ]; then
echo "PASS";
exit 0
fi
exit 1
[Regression Potential]
* As bcache minor numbers and these symlinks have been unreliable in
the past there may be code that makes assumptions about
/dev/bcache* expanded only to the block devices, versus
/dev/bcache which is a directory.
[Original Description]
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time.
Related bugs:
* bug 1729145: /dev/bcache/by-uuid links not created after reboot |
|
2018-02-15 23:15:41 |
Bug Watch Updater |
bcache-tools (Debian): status |
New |
Confirmed |
|
2018-02-16 14:38:00 |
Scott Moser |
description |
[Impact]
* Current users of bcache devices may encounter unreliable device
numbering as the Linux kernel does not guarantee that bcache
minor numbers are assigned to the same devices at each boot.
Users who may have used /dev/bcacheN in paths to a specific
device could possible be pointing to a different dataset
altogether. bcache udev rules do provide some mechanism to
generate persistent symlinks in /dev/bcache/by-uuid or
/dev/bcache/by-label which is based on superblock data on
the underlying device. However, the Linux kernel does not
always generate an kernel uevent to trigger the udev rules
to create the symlink.
* The fix adds a udev program which will read bcache superblock
of slave devices and extract the UUID and LABEL, exporting them
to udev for use in the bcache rule files.
* This is affected in upstream bcache-tools, the owning package
of the udev rules. This affects all releases of bcache-tools
as the rules rely upon the kernel to trigger these events,
though that is not a requirement to resolve the lack of
persistent links.
[Test Case]
* Launch and Ubuntu Cloud Image with 3 unused disks
- apt install bcache-tools tree
- make-bcache -C /dev/vdb
- make-bcache -B /dev/vdc
- make-bcache -B /dev/vdd
- echo "vdc" > /sys/class/block/bcache0/bcache/label
- echo "vdd" > /sys/class/block/bcache1/bcache/label
- reboot
- Run this test:
#!/bin/bash
FAIL=0
[ ! -d /dev/bcache ] && {
echo "FAIL: /dev/bcache is not a directory";
exit 1
}
for label in /dev/bcache/by-label/*; do
LABEL_TARGET="$(ls -1 /sys/class/block/`basename $label`/holders/)"
DEVNAME=`readlink -f $label`;
KNAME="${DEVNAME#*/dev/}"
if [ "$LABEL_TARGET" != "$KNAME" ]; then
echo "FAIL: label points to $LABEL_TARGET but symlink points to $DEVNAME";
FAIL=1
fi;
done
if [ "$FAIL" == "0" ]; then
echo "PASS";
exit 0
fi
exit 1
[Regression Potential]
* As bcache minor numbers and these symlinks have been unreliable in
the past there may be code that makes assumptions about
/dev/bcache* expanded only to the block devices, versus
/dev/bcache which is a directory.
[Original Description]
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time.
Related bugs:
* bug 1729145: /dev/bcache/by-uuid links not created after reboot |
[Impact]
* Current users of bcache devices may encounter unreliable device
numbering as the Linux kernel does not guarantee that bcache
minor numbers are assigned to the same devices at each boot.
Users who may have used /dev/bcacheN in paths to a specific
device could possible be pointing to a different dataset
altogether. bcache udev rules do provide some mechanism to
generate persistent symlinks in /dev/bcache/by-uuid or
/dev/bcache/by-label which is based on superblock data on
the underlying device. However, the Linux kernel does not
always generate an kernel uevent to trigger the udev rules
to create the symlink.
* The fix adds a udev program which will read bcache superblock
of slave devices and extract the UUID and LABEL, exporting them
to udev for use in the bcache rule files.
* This is affected in upstream bcache-tools, the owning package
of the udev rules. This affects all releases of bcache-tools
as the rules rely upon the kernel to trigger these events,
though that is not a requirement to resolve the lack of
persistent links.
[Test Case]
* Launch and Ubuntu Cloud Image with 3 unused disks
- apt install bcache-tools tree
- make-bcache -C /dev/vdb
- make-bcache -B /dev/vdc
- make-bcache -B /dev/vdd
- echo "vdc" > /sys/class/block/bcache0/bcache/label
- echo "vdd" > /sys/class/block/bcache1/bcache/label
- reboot
- Run this test:
#!/bin/bash
FAIL=0
[ ! -d /dev/bcache ] && {
echo "FAIL: /dev/bcache is not a directory";
exit 1
}
for label in /dev/bcache/by-label/*; do
LABEL_TARGET="$(ls -1 /sys/class/block/`basename $label`/holders/)"
DEVNAME=`readlink -f $label`;
KNAME="${DEVNAME#*/dev/}"
if [ "$LABEL_TARGET" != "$KNAME" ]; then
echo "FAIL: label points to $LABEL_TARGET but symlink points to $DEVNAME";
FAIL=1
fi;
done
if [ "$FAIL" == "0" ]; then
echo "PASS";
exit 0
fi
exit 1
[Regression Potential]
* As bcache minor numbers and these symlinks have been unreliable in
the past there may be code that makes assumptions about
/dev/bcache* expanded only to the block devices, versus
/dev/bcache which is a directory.
[Original Description]
Bcache device names like /dev/bcache0 are unstable. Bcache does not use any predictable ordering when assembling bcache devices, so on systems with multiple bcache devices, a symlink to /dev/bcache0 may end up pointing do a different device.
the bcache dname symlink should point to the /dev/bcache/by-uuid/<UUID> which matches the backing device UUID that's set at creation time.
Related bugs:
* bug 1729145: [kernel] /dev/bcache/by-uuid links not created after reboot |
|
2018-03-22 15:51:36 |
Launchpad Janitor |
merge proposal linked |
|
https://code.launchpad.net/~raharper/curtin/+git/curtin/+merge/341915 |
|
2018-03-22 19:05:28 |
Scott Moser |
curtin: status |
New |
Fix Committed |
|
2018-03-22 19:18:21 |
Launchpad Janitor |
merge proposal unlinked |
https://code.launchpad.net/~raharper/curtin/+git/curtin/+merge/341915 |
|
|
2018-05-09 06:56:41 |
Peter Sabaini |
bug |
|
|
added subscriber Peter Sabaini |
2018-05-09 14:14:18 |
David Britton |
bug task deleted |
bcache-tools (Ubuntu) |
|
|
2018-05-09 14:17:19 |
David Britton |
bcache-tools (Ubuntu Bionic): status |
In Progress |
Invalid |
|
2018-05-09 14:17:28 |
David Britton |
bcache-tools (Ubuntu Artful): status |
Confirmed |
Invalid |
|
2018-05-09 14:17:33 |
David Britton |
bcache-tools (Ubuntu Xenial): status |
Confirmed |
Invalid |
|
2018-05-09 14:18:02 |
David Britton |
bcache-tools (Ubuntu Trusty): status |
Confirmed |
Invalid |
|
2018-05-09 14:18:12 |
David Britton |
bcache-tools (Ubuntu Bionic): assignee |
Ryan Harper (raharper) |
|
|
2018-05-09 14:20:27 |
David Britton |
bug task added |
|
curtin (Ubuntu) |
|
2018-05-09 14:26:31 |
David Britton |
curtin (Ubuntu Trusty): status |
New |
Won't Fix |
|
2018-05-09 14:27:40 |
David Britton |
curtin (Ubuntu Bionic): status |
New |
Fix Released |
|
2018-05-09 14:28:13 |
David Britton |
curtin (Ubuntu): status |
New |
Fix Released |
|
2018-05-10 08:26:36 |
Gábor Mészáros |
bug |
|
|
added subscriber Gábor Mészáros |
2018-05-29 17:35:41 |
Aymen Frikha |
bug |
|
|
added subscriber Aymen Frikha |
2018-06-01 14:02:20 |
Ryan Harper |
curtin: status |
Fix Committed |
Fix Released |
|