Re: [PATCH] usb: core: Unregister device on component_add() failure

[Fabio M. De Francesco]

On mercoled? 9 febbraio 2022 15:00:09 CET Heikki Krogerus wrote:
> On Tue, Feb 08, 2022 at 06:00:48PM +0100, Fabio M. De Francesco wrote:
> > Commit 8c67d06f3fd9 ("usb: Link the ports to the connectors they are
> > attached to") creates a link to the USB Type-C connector for every new
> > port that is added when possible. If component_add() fails,
> > usb_hub_create_port_device() prints a warning but does not unregister
> > the device and does not return errors to the callers.
> > 
> > Syzbot reported a "WARNING in component_del()".
> > 
> > Fix this issue in usb_hub_create_port_device by calling device_unregister()
> > and returning the errors from component_add().
> > 
> > Reported-by: syzbot+60df062e1c41940cae0f@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Fixes: 8c67d06f3fd9 ("usb: Link the ports to the connectors they are attached to")
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@xxxxxxxxx>
> > ---
> >  drivers/usb/core/port.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > [...]
> >
> You didn't remove the peer links. Either remove them here separately,
> or alternatively you can also just shuffle the code so that you only
> create those links after the component_add() call:
>
> [...]
> 

Hello Heikki,

Thanks for your review and suggestion. I think that I'll use the second of the
two possible solutions (shuffle the code).

I had to spend some time to understand the code of usb_hub_create_port_device(),
component_add() and component_del(). Unfortunately, the USB core is very far from
the usual things I look at or care of. Therefore I missed that find_and_link_peer()
does some work that must be either unwound or simply postponed. I agree with you 
that the latter is the best solution.

I need some minutes to submit v2.

Again, thanks,

Fabio

> thanks,
> 
> -- 
> heikki
>