On Thu, Mar 22, 2012 at 7:47 AM, James Bottomley
[off-list ref] wrote:

On Wed, 2012-03-21 at 23:32 -0700, Dan Williams wrote:

quoted

In scsi at least two cases of the parent device being deleted before the
child is added have been observed.

1/ scsi is performing async scans and the device is removed prior to the
   async can thread running.

This doesn't sound right.  We do an explicit get on the sdkp (and the
sdkp holds the sdp) before we schedule an async scan.  That's only
removed after the async scan has completed, so it should be impossible
for the parent to vanish.

Right, the parent does not get freed because we have a reference, but
it still gets device_del()'d which leads to:

  device_del()->kobject_del()->sysfs_remove_dir()

  kobj->sd = NULL;

...and then sysfs_create_dir() (without this fix) goes ahead and
de-references parent_sd via sysfs_ns_type():

  return (sd->s_flags & SYSFS_NS_TYPE_MASK) >> SYSFS_NS_TYPE_SHIFT;

--
Dan