lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20201224132944.5odphwxtnzhvkeg7@pali>
Date:   Thu, 24 Dec 2020 14:29:44 +0100
From:   Pali Rohár <pali@...nel.org>
To:     Peter Chen <peter.chen@....com>
Cc:     Mathias Nyman <mathias.nyman@...el.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Gregory CLEMENT <gregory.clement@...tlin.com>,
        Miquel Raynal <miquel.raynal@...tlin.com>,
        Tomasz Maciej Nowak <tmn505@...il.com>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] usb: host: xhci: mvebu: make USB 3.0 PHY optional for
 Armada 3720

On Thursday 24 December 2020 14:24:01 Pali Rohár wrote:
> On Thursday 24 December 2020 13:15:21 Peter Chen wrote:
> >  
> > > > > +	/* Old bindings miss the PHY handle */
> > > > > +	phy = of_phy_get(dev->of_node, "usb3-phy");
> > > > > +	if (IS_ERR(phy) && PTR_ERR(phy) == -EPROBE_DEFER)
> > > > > +		return -EPROBE_DEFER;
> > > >
> > > > Doesn't need to judge IS_ERR(phy).
> > > 
> > > Ok, I can remove it. I used same condition which is already in SATA and PCIe
> > > phy code.
> > > 
> > > > > +	else if (IS_ERR(phy))
> > > > > +		goto phy_out;
> > > > > +
> > > > > +	ret = phy_init(phy);
> > > > > +	if (ret)
> > > > > +		goto phy_put;
> > > > > +
> > > > > +	ret = phy_set_mode(phy, PHY_MODE_USB_HOST_SS);
> > > > > +	if (ret)
> > > > > +		goto phy_exit;
> > > > > +
> > > > > +	ret = phy_power_on(phy);
> > > > > +	if (ret == -EOPNOTSUPP) {
> > > > > +		/* Skip initializatin of XHCI PHY when it is unsupported by
> > > firmware */
> > > > > +		dev_warn(dev, "PHY unsupported by firmware\n");
> > > > > +		xhci->quirks |= XHCI_SKIP_PHY_INIT;
> > > > > +	}
> > > > > +	if (ret)
> > > > > +		goto phy_exit;
> > > > > +
> > > > > +	phy_power_off(phy);
> > > > > +phy_exit:
> > > > > +	phy_exit(phy);
> > > > > +phy_put:
> > > > > +	of_phy_put(phy);
> > > > > +phy_out:
> > > > > +
> > > >
> > > > You do power on and off again only want to know if PHY has already
> > > > powered at ATF, right?
> > > 
> > > I need to know if power on/off procedure is supported by ATF. And if not
> > > (indicated by -EOPNOTSUPP) then I need to ensure that usb hdc code would not
> > > try to call phy_power_on() as it would cause failure as described in the commit
> > > message. You can look at those other two commits for PCIe and SATA. Same
> > > thing is needed for USB.
> > 
> > If not supported by ATF, then where to power on and off PHY since no other place calls PHY APIs? Is it always on?
> 
> Yes, in this case (when -EOPNOTSUPP is returned) SMC API is not
> supported by ATF, and PHY is always on.

To make it clear, core/hcd.c function usb_add_hcd() when
hcd->skip_phy_initialization is false is calling
usb_phy_roothub_power_on() which calls phy_power_on(). If this call
fails then error is propagated back to the usb_add_hcd() and this
function fails too.

But on boards with older ATF (which do not support PHY power on SMC API)
is phy_power_on() returning error -EOPNOTSUPP and therefore whole USB
3.0 initialization fails.

This patch is adding init hook to detect if ATF supports PHY power
on/off functions and in case it does not support it, code sets
XHCI_SKIP_PHY_INIT flag to instruct xhci-plat code to set
hcd->skip_phy_initialization flag to instruct core/hcd.c to skip calling
usb_phy_roothub_power_on() function as it is know that it would fail.

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ