linux-kernel - [BUG] USB (EHCI/OHCI) broken on Allwinner SoCs

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>] [day] [month] [year] [list]

Message-ID: <20190318084536.GA27150@wens.csie.org>
Date:   Mon, 18 Mar 2019 16:45:36 +0800
From:   Chen-Yu Tsai <wens@...nel.org>
To:     Maxime Ripard <maxime.ripard@...tlin.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Grygorii Strashko <grygorii.strashko@...com>
Cc:     Kishon Vijay Abraham I <kishon@...com>, linux-usb@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: [BUG] USB (EHCI/OHCI) broken on Allwinner SoCs

Hi everyone,

USB EHCI/OHCI hosts are broken on Allwinner SoCs on v5.1-rc1. I believe
this is partially triggered by commit b97a31348379 ("usb: core: comply
to PHY framework"), and partially due to how the Allwinner USB PHY
driver handles phy_set_mode for non-OTG PHYs. Note this is not from a
bisect.

The following changes fix this.

Hunk 1 makes our PHY driver accept PHY_MODE_USB_HOST for the non-OTG
PHYs as a no-op.

diff --git a/drivers/phy/allwinner/phy-sun4i-usb.c b/drivers/phy/allwinner/phy-sun4i-usb.c
index 5163097b43df..4bbd9ede38c8 100644
--- a/drivers/phy/allwinner/phy-sun4i-usb.c
+++ b/drivers/phy/allwinner/phy-sun4i-usb.c
@@ -485,8 +485,11 @@ static int sun4i_usb_phy_set_mode(struct phy *_phy,
 	struct sun4i_usb_phy_data *data = to_sun4i_usb_phy_data(phy);
 	int new_mode;

-	if (phy->index != 0)
+	if (phy->index != 0) {
+		if (mode == PHY_MODE_USB_HOST)
+			return 0;
 		return -EINVAL;
+	}

 	switch (mode) {
 	case PHY_MODE_USB_HOST:
---

Hunk 2 makes the USB core call phy_set_mode with PHY_MODE_USB_HOST
instead of PHY_MODE_USB_HOST_SS. Not sure why PHY_MODE_USB_HOST_SS was
used here, but I believe only a small set of hardware actually support
SuperSpeed, so it might be better to use the more encompasing
PHY_MODE_USB_HOST. That or the core code would have to check which speed
class the port is for and pass in the appropriate mode. The Allwinner
PHYs do not support SuperSpeed.

diff --git a/drivers/usb/core/hcd.c b/drivers/usb/core/hcd.c
index 3189181bb628..7f25e384144d 100644
--- a/drivers/usb/core/hcd.c
+++ b/drivers/usb/core/hcd.c
@@ -2740,7 +2740,7 @@ int usb_add_hcd(struct usb_hcd *hcd,
 			return retval;

 		retval = usb_phy_roothub_set_mode(hcd->phy_roothub,
-						  PHY_MODE_USB_HOST_SS);
+						  PHY_MODE_USB_HOST);
 		if (retval)
 			goto err_usb_phy_roothub_power_on;

---

If the two changes make sense I'll send them out as actual patches.