On Wed, Nov 14, 2012 at 06:52:18PM +0100, Johan Hovold wrote:

On Wed, Nov 14, 2012 at 12:12:01PM -0500, Richard Retanubun wrote:

quoted

On 10/11/12 09:30 AM, Johan Hovold wrote:
Hi Johan,

quoted

There was a reference-count fix for the probe error path that went in to
v3.5. Haven't read all the details on how you trigger your leak, but at
the face of it, it could be related.

Have a look at 0658a3366db7e27fa ("usb: use usb_serial_put in
usb_serial_probe errors). If related, you should be seeing "Ignoring
blacklisted interface #n" messages when you enable debug (e.g. #define
DEBUG) in the sierra driver.

That was it! Thanks so much for the research.
I can apply it cleanly to 3.0.22 and see usb_release_dev() being
called and thus no more kmemleak.

quoted

Greg, it seems to me that the fix referred to above should be backported
to the earlier stable trees either way.

I would vote "yes" for this also.

While my setup circumstances may be a corner case, (modem kept
resetting to re-establish PPP connection) it was leaking 1192 bytes
per occurrence.

The leak affects every failed probe, for example due to blacklisted
interfaces which is quite common, so commit 0658a3366db7 ("usb: use
usb_serial_put in usb_serial_probe errors) should be backported to the
<= 3.4 stable trees.

Thanks, now applied.

greg k-h