Virtualization - Jun 2016 - [PATCH] virtio-blk: Generate uevent after attribute available

Userspace listens to the KOBJ_ADD uevent generated in add_disk. At that
point we haven't created the serial attribute file, therefore depending
on how fast udev reacts, the /dev/disk/by-id/ entry doesn't always get
created.

This race condition can be easily reproduced by hot plugging a number of
virtio-blk disks.

Also in systemd, there used to be a related workaround in udev rules
called 'WAIT_FOR="serial"', but it is removed in later
versions.

Now let's generate a KOBJ_CHANGE event after the attributes are ready.

Signed-off-by: Fam Zheng <famz at redhat.com>
---
 drivers/block/virtio_blk.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 42758b5..5056007 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -567,6 +567,7 @@ static int virtblk_probe(struct virtio_device *vdev)
 {
 	struct virtio_blk *vblk;
 	struct request_queue *q;
+	struct device *ddev;
 	int err, index;
 
 	u64 cap;
@@ -746,6 +747,8 @@ static int virtblk_probe(struct virtio_device *vdev)
 					 &dev_attr_cache_type_ro);
 	if (err)
 		goto out_del_disk;
+	ddev = disk_to_dev(vblk->disk);
+	kobject_uevent(&ddev->kobj, KOBJ_CHANGE);
 	return 0;
 
 out_del_disk:
-- 
2.9.0

On Tue, Jun 28, 2016 at 10:39:15AM +0800, Fam Zheng
wrote:
> Userspace listens to the KOBJ_ADD uevent generated in add_disk. At that
> point we haven't created the serial attribute file, therefore depending
> on how fast udev reacts, the /dev/disk/by-id/ entry doesn't always get
> created.
> 
> This race condition can be easily reproduced by hot plugging a number of
> virtio-blk disks.
> 
> Also in systemd, there used to be a related workaround in udev rules
> called 'WAIT_FOR="serial"', but it is removed in later
versions.
> 
> Now let's generate a KOBJ_CHANGE event after the attributes are ready.
The same race is present in other drivers as well, e.g. nvme.  Please
find a way to make this work properly without needing to hack every
driver to send events manually.

On Tue, 06/28 04:45, Christoph Hellwig wrote:
> On Tue, Jun 28, 2016 at 10:39:15AM +0800, Fam Zheng wrote:
> > Userspace listens to the KOBJ_ADD uevent generated in add_disk. At
that
> > point we haven't created the serial attribute file, therefore
depending
> > on how fast udev reacts, the /dev/disk/by-id/ entry doesn't always
get
> > created.
> > 
> > This race condition can be easily reproduced by hot plugging a number
of
> > virtio-blk disks.
> > 
> > Also in systemd, there used to be a related workaround in udev rules
> > called 'WAIT_FOR="serial"', but it is removed in
later versions.
> > 
> > Now let's generate a KOBJ_CHANGE event after the attributes are
ready.
> 
> The same race is present in other drivers as well, e.g. nvme.  Please
> find a way to make this work properly without needing to hack every
> driver to send events manually.
OK, I'll take a look today!

Fam