Re: [PATCH] libata: add refcounting to ata_host

[Tejun Heo]

On Fri, Mar 09, 2018 at 08:34:41AM +0000, Taras Kondratiuk wrote:
> After commit 9a6d6a2ddabb ("ata: make ata port as parent device of scsi
> host") manual driver unbind/remove causes use-after-free.
> 
> Unbind unconditionally invokes devres_release_all() which calls
> ata_host_release() and frees ata_host/ata_port memory while it is still
> being referenced as a parent of SCSI host. When SCSI host is finally
> released scsi_host_dev_release() calls put_device(parent) and accesses
> freed ata_port memory.
> 
> Add reference counting to make sure that ata_host lives long enough.
> 
> Bug report: https://lkml.org/lkml/2017/11/1/945
> Fixes: 9a6d6a2ddabb ("ata: make ata port as parent device of scsi host")
> Cc: Tejun Heo <tj@xxxxxxxxxx>
> Cc: Lin Ming <minggr@xxxxxxxxx>
> Cc: linux-ide@xxxxxxxxxxxxxxx
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Signed-off-by: Taras Kondratiuk <takondra@xxxxxxxxx>

Applied to libata/for-4.17.

Thanks.

-- 
tejun