On Wed, Oct 24, 2012 at 07:24:15PM +0200, Johan Hovold wrote:
> This is an updated series of fixes rebased on the interface-data memory leak
> discovered by Lennart.
>
> One of the reasons that I missed it was that the interface data was allocated
> in probe rather than attach, which in itself is a bug as the memory would not
> get released if usb-serial probe failes before attach returns successfully. The
> third patch fixes this.
>
> I also split out the fix for the port-data memory leak in the error path of
> attach that was part of my previous patch. This way the first three patches can
> be backported to most (all?) stable trees, whereas the last patch applies to
> v3.6 (which contains the (in)famous 0998d063100 ("device-core: Ensure drvdata
> = NULL when no driver is bound" commit).
>
> I included Lennart's patch with an updated commit message for completeness.
>
> Lennart, if you care to test this series before it is applied that would be
> much appreciated.
I backported 1, 2 and 3 for the 3.0.23 kernel I am running, and so far it
looks to be working. I think our USB HCD driver may have a leak still,
but someone else is investigating that right now. At least the sierra
driver appears to have stopped leaking.
--
Len Sorensen
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
