On Wed, Dec 06, 2023 at 05:51:34PM +0000, Russell King (Oracle) wrote:
> On Wed, Dec 06, 2023 at 01:44:17AM +0000, Daniel Golle wrote:
> > +struct phylink_pcs *mtk_pcs_lynxi_select_pcs(struct device_node *np, phy_interface_t mode)
> > +{
> > + struct platform_device *pdev;
> > + struct mtk_pcs_lynxi *mpcs;
> > +
> > + if (!np)
> > + return NULL;
> > +
> > + if (!of_device_is_available(np))
> > + return ERR_PTR(-ENODEV);
> > +
> > + if (!of_match_node(mtk_pcs_lynxi_of_match, np))
> > + return ERR_PTR(-EINVAL);
> > +
> > + pdev = of_find_device_by_node(np);
> > + if (!pdev || !platform_get_drvdata(pdev)) {
> > + if (pdev)
> > + put_device(&pdev->dev);
> > + return ERR_PTR(-EPROBE_DEFER);
> > + }
> > +
> > + mpcs = platform_get_drvdata(pdev);
> > + put_device(&pdev->dev);
> > +
> > + return &mpcs->pcs;
> > +}
> > +EXPORT_SYMBOL(mtk_pcs_lynxi_select_pcs);
>
> If you're going to play games like this, then you must mark the driver
> with .suppress_bind_attrs = true to remove the bind/unbind attributes
> in userspace that could wreak havoc with the above - because there is
> _nothing_ that guarantees that the memory you're returning from this
> function will remain intact. Basically, it's racy.
Ack, I've set .suppress_bind_attrs = true in the usxgmii driver but
forgot to add it here.
> Also, I'm not sure I approve of using the "select_pcs" suffix (I
> haven't spotted _where_ you use this, but returning EPROBE_DEFER to
> phylink's mac_select_pcs() method doesn't do anything to defer any
> probe, so that's an entirely misleading error code.
EPROBE_DEFER is handled when the function is called by mtk_add_mac()
during probe of the Ethernet driver -- which we do want to postpone
in case the PCS hasn't been probed yet as at this point that's the
best we can do without adding lots of intrastructure to dynamically
attach the PCS later on...
But true, later the function is being called by mac_select_pcs() and
what ever it returns is returned to the caller of mac_select_pcs().
If you think it's better to return ENODEV (or EAGAIN?) I can change
that -- from what I could tell, the only error which receives special
handling by phylink is -EOPNOTSUPP, everything else just gets passed-
through to the callers.
> If we are going to have device drivers for PCS, then we need to
> seriously think about how we look up PCS and return the phylink_pcs
> pointer - and also how we handle the PCS device going away. None of
> that should be coded into _any_ PCS driver.
I agree -- just wasn't up to design and implement all that at once.
