Re: [PATCH v4 2/6] can: etas_es58x: add devlink support

From: Vincent MAILHOL <mailhol.vincent@wanadoo.fr>
To: Alan Stern <stern@rowland.harvard.edu>
Cc: Andrew Lunn <andrew@lunn.ch>,
	linux-can@vger.kernel.org, Marc Kleine-Budde <mkl@pengutronix.de>,
	linux-kernel@vger.kernel.org,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	netdev@vger.kernel.org, linux-usb@vger.kernel.org,
	Saeed Mahameed <saeed@kernel.org>, Jiri Pirko <jiri@nvidia.com>,
	Lukas Magel <lukas.magel@posteo.net>
Subject: Re: [PATCH v4 2/6] can: etas_es58x: add devlink support
Date: Mon, 28 Nov 2022 14:32:23 +0900	[thread overview]
Message-ID: <CAMZ6RqKS0sUFZWQfmRU6q2ivWEUFD06uiQekDr=u94L3uij3yQ@mail.gmail.com> (raw)
In-Reply-To: <CAMZ6Rq+Gi+rcLqSj2-kug7c1G_nNuj6peh5nH1DNoo8B3aSxzw@mail.gmail.com>

On Mon. 28 Nov. 2022 at 10:34, Vincent MAILHOL
<mailhol.vincent@wanadoo.fr> wrote:
> On Mon. 28 Nov. 2022 at 00:41, Alan Stern <stern@rowland.harvard.edu> wrote:
> > On Sun, Nov 27, 2022 at 02:10:32PM +0900, Vincent MAILHOL wrote:
> > > > Should devlink_free() be after usb_set_inftdata()?
> > >
> > > A look at
> > >   $ git grep -W "usb_set_intfdata(.*NULL)"
> > >
> > > shows that the two patterns (freeing before or after
> > > usb_set_intfdata()) coexist.
> > >
> > > You are raising an important question here. usb_set_intfdata() does
> > > not have documentation that freeing before it is risky. And the
> > > documentation of usb_driver::disconnect says that:
> > >   "@disconnect: Called when the interface is no longer accessible,
> > >    usually because its device has been (or is being) disconnected
> > >    or the driver module is being unloaded."
> > >   Ref: https://elixir.bootlin.com/linux/v6.1-rc6/source/include/linux/usb.h#L1130
> > >
> > > So the interface no longer being accessible makes me assume that the
> > > order does not matter. If it indeed matters, then this is a foot gun
> > > and there is some clean-up work waiting for us on many drivers.
> > >
> > > @Greg, any thoughts on whether or not the order of usb_set_intfdata()
> > > and resource freeing matters or not?
> >
> > In fact, drivers don't have to call usb_set_intfdata(NULL) at all; the
> > USB core does it for them after the ->disconnect() callback returns.
>
> Interesting. This fact is widely unknown, cf:
>   $ git grep "usb_set_intfdata(.*NULL)" | wc -l
>   215
>
> I will do some clean-up later on, at least for the CAN USB drivers.
>
> > But if a driver does make the call, it should be careful to ensure that
> > the call happens _after_ the driver is finished using the interface-data
> > pointer.  For example, after all outstanding URBs have completed, if the
> > completion handlers will need to call usb_get_intfdata().
>
> ACK. I understand that it should be called *after* the completion of
> any ongoing task.
>
> My question was more on:
>
>         devlink_free(priv_to_devlink(es58x_dev));
>         usb_set_intfdata(intf, NULL);
>
> VS.
>
>         usb_set_intfdata(intf, NULL);
>         devlink_free(priv_to_devlink(es58x_dev));
>
> From your comments, I understand that both are fine.

Do we agree that the usb-skeleton is doing it wrong?
  https://elixir.bootlin.com/linux/latest/source/drivers/usb/usb-skeleton.c#L567
usb_set_intfdata(interface, NULL) is called before deregistering the
interface and terminating the outstanding URBs!

> > Remember, the interface-data pointer is owned by the driver.  Nothing
> > else in the kernel uses it.  So the driver merely has to be careful not
> > to clear the pointer while it is still using it.
>
> Thanks for your comments!
>
>
> Yours sincerely,
> Vincent Mailhol