The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
This makes udev miss the device creation and prompted workarounds in userspace. See the linked util-linux/losetup bug.
Explicitly emit the events and drop the confusingly worded comments.
Link: https://github.com/util-linux/util-linux/issues/2434 Fixes: 498ef5c777d9 ("loop: suppress uevents while reconfiguring the device") Cc: stable@vger.kernel.org Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de --- Changes in v2: - Use correct Fixes tag - Rework commit message slightly - Rebase onto v6.15-rc1 - Link to v1: https://lore.kernel.org/r/20250317-loop-uevent-changed-v1-1-cb29cb91b62d@lin... --- drivers/block/loop.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 674527d770dc669e982a2b441af1171559aa427c..09a725710a21171e0adf5888f929ccaf94e98992 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -667,8 +667,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
error = 0; done: - /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0); + kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE); return error;
out_err: @@ -1129,8 +1129,8 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (partscan) clear_bit(GD_SUPPRESS_PART_SCAN, &lo->lo_disk->state);
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0); + kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE);
loop_global_unlock(lo, is_loop); if (partscan)
--- base-commit: 0af2f6be1b4281385b618cb86ad946eded089ac8 change-id: 20250307-loop-uevent-changed-aa3690f43e03
Best regards,
On Tue, Apr 15, 2025 at 10:51:47AM +0200, Thomas Weißschuh wrote:
The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
This makes udev miss the device creation and prompted workarounds in userspace. See the linked util-linux/losetup bug.
Explicitly emit the events and drop the confusingly worded comments.
Link: https://github.com/util-linux/util-linux/issues/2434 Fixes: 498ef5c777d9 ("loop: suppress uevents while reconfiguring the device") Cc: stable@vger.kernel.org Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de
Changes in v2:
- Use correct Fixes tag
- Rework commit message slightly
- Rebase onto v6.15-rc1
- Link to v1: https://lore.kernel.org/r/20250317-loop-uevent-changed-v1-1-cb29cb91b62d@lin...
drivers/block/loop.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 674527d770dc669e982a2b441af1171559aa427c..09a725710a21171e0adf5888f929ccaf94e98992 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -667,8 +667,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev, error = 0; done:
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE); return error;
out_err: @@ -1129,8 +1129,8 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (partscan) clear_bit(GD_SUPPRESS_PART_SCAN, &lo->lo_disk->state);
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE);
loop_global_unlock(lo, is_loop); if (partscan)
base-commit: 0af2f6be1b4281385b618cb86ad946eded089ac8 change-id: 20250307-loop-uevent-changed-aa3690f43e03
Best regards,
Thomas Weißschuh thomas.weissschuh@linutronix.de
Acked-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
On Tue 15-04-25 10:51:47, Thomas Weißschuh wrote:
The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
This makes udev miss the device creation and prompted workarounds in userspace. See the linked util-linux/losetup bug.
Explicitly emit the events and drop the confusingly worded comments.
Link: https://github.com/util-linux/util-linux/issues/2434 Fixes: 498ef5c777d9 ("loop: suppress uevents while reconfiguring the device") Cc: stable@vger.kernel.org Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de
Thanks for the fix! When reading the code, I'm a bit curious: What is actually generating events for partitions with loop_change_fd() call? Because there loop_reread_partitions() still happens with uevents supressed... I suspect event supressing there should be shorter.
Honza
Changes in v2:
- Use correct Fixes tag
- Rework commit message slightly
- Rebase onto v6.15-rc1
- Link to v1: https://lore.kernel.org/r/20250317-loop-uevent-changed-v1-1-cb29cb91b62d@lin...
drivers/block/loop.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 674527d770dc669e982a2b441af1171559aa427c..09a725710a21171e0adf5888f929ccaf94e98992 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -667,8 +667,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev, error = 0; done:
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE); return error;
out_err: @@ -1129,8 +1129,8 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (partscan) clear_bit(GD_SUPPRESS_PART_SCAN, &lo->lo_disk->state);
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE);
loop_global_unlock(lo, is_loop); if (partscan)
base-commit: 0af2f6be1b4281385b618cb86ad946eded089ac8 change-id: 20250307-loop-uevent-changed-aa3690f43e03
Best regards,
Thomas Weißschuh thomas.weissschuh@linutronix.de
On Tue, Apr 15, 2025 at 12:24:55PM +0200, Jan Kara wrote:
On Tue 15-04-25 10:51:47, Thomas Weißschuh wrote:
The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
This makes udev miss the device creation and prompted workarounds in userspace. See the linked util-linux/losetup bug.
Explicitly emit the events and drop the confusingly worded comments.
Link: https://github.com/util-linux/util-linux/issues/2434 Fixes: 498ef5c777d9 ("loop: suppress uevents while reconfiguring the device") Cc: stable@vger.kernel.org Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de
Thanks for the fix! When reading the code, I'm a bit curious: What is actually generating events for partitions with loop_change_fd() call? Because there loop_reread_partitions() still happens with uevents supressed... I suspect event supressing there should be shorter.
Makes sense. For loop_configure() this was fixed in commit bb430b694226 ("loop: LOOP_CONFIGURE: send uevents for partitions"). I guess we need the same for loop_change_fd().
I'm not entirely sure on how to order the commits or if they should be folded together. My current preference is to first have the current patch under discussion and then the fix for loop_change_fd().
Thomas
Changes in v2:
- Use correct Fixes tag
- Rework commit message slightly
- Rebase onto v6.15-rc1
- Link to v1: https://lore.kernel.org/r/20250317-loop-uevent-changed-v1-1-cb29cb91b62d@lin...
drivers/block/loop.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 674527d770dc669e982a2b441af1171559aa427c..09a725710a21171e0adf5888f929ccaf94e98992 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -667,8 +667,8 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev, error = 0; done:
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE); return error;
out_err: @@ -1129,8 +1129,8 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (partscan) clear_bit(GD_SUPPRESS_PART_SCAN, &lo->lo_disk->state);
- /* enable and uncork uevent now that we are done */ dev_set_uevent_suppress(disk_to_dev(lo->lo_disk), 0);
- kobject_uevent(&disk_to_dev(lo->lo_disk)->kobj, KOBJ_CHANGE);
loop_global_unlock(lo, is_loop); if (partscan)
base-commit: 0af2f6be1b4281385b618cb86ad946eded089ac8 change-id: 20250307-loop-uevent-changed-aa3690f43e03
Best regards,
Thomas Weißschuh thomas.weissschuh@linutronix.de
-- Jan Kara jack@suse.com SUSE Labs, CR
On Tue 15-04-25 13:07:48, Thomas Weißschuh wrote:
On Tue, Apr 15, 2025 at 12:24:55PM +0200, Jan Kara wrote:
On Tue 15-04-25 10:51:47, Thomas Weißschuh wrote:
The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
This makes udev miss the device creation and prompted workarounds in userspace. See the linked util-linux/losetup bug.
Explicitly emit the events and drop the confusingly worded comments.
Link: https://github.com/util-linux/util-linux/issues/2434 Fixes: 498ef5c777d9 ("loop: suppress uevents while reconfiguring the device") Cc: stable@vger.kernel.org Signed-off-by: Thomas Weißschuh thomas.weissschuh@linutronix.de
Thanks for the fix! When reading the code, I'm a bit curious: What is actually generating events for partitions with loop_change_fd() call? Because there loop_reread_partitions() still happens with uevents supressed... I suspect event supressing there should be shorter.
Makes sense. For loop_configure() this was fixed in commit bb430b694226 ("loop: LOOP_CONFIGURE: send uevents for partitions"). I guess we need the same for loop_change_fd().
I'm not entirely sure on how to order the commits or if they should be folded together. My current preference is to first have the current patch under discussion and then the fix for loop_change_fd().
Yeah, whatever. I was asking mainly to make sure my understanding of the current code is correct :). With this one feel free to add:
Reviewed-by: Jan Kara jack@suse.cz
Honza
On Tue, 15 Apr 2025 10:51:47 +0200, Thomas Weißschuh wrote:
The original commit message and the wording "uncork" in the code comment indicate that it is expected that the suppressed event instances are automatically sent after unsuppressing. This is not the case, instead they are discarded. In effect this means that no "changed" events are emitted on the device itself by default. While each discovered partition does trigger a changed event on the device, devices without partitions don't have any event emitted.
[...]
Applied, thanks!
[1/1] loop: properly send KOBJ_CHANGED uevent for disk device commit: 7ed2a771b5fb3edee9c4608181235c30b40bb042
Best regards,
linux-stable-mirror@lists.linaro.org
-
Greg KH -
Jan Kara -
Jens Axboe -
Thomas Weißschuh