linux-kernel - Re: [PATCH] Revert "usb: cdns3: core: quit if it uses role switch class"

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: <20201125003550.GB9929@b29397-desktop>
Date:   Wed, 25 Nov 2020 00:36:19 +0000
From:   Peter Chen <peter.chen@....com>
To:     Roger Quadros <rogerq@...com>
CC:     "heikki.krogerus@...ux.intel.com" <heikki.krogerus@...ux.intel.com>,
        "pawell@...ence.com" <pawell@...ence.com>,
        "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
        "balbi@...nel.org" <balbi@...nel.org>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Revert "usb: cdns3: core: quit if it uses role switch
 class"

On 20-11-24 14:22:25, Roger Quadros wrote:
> Peter,
> 
> On 24/11/2020 13:47, Peter Chen wrote:
> > On 20-11-24 12:33:34, Roger Quadros wrote:
> > > > > > 
> > > > > > I am sorry about that. Do you use role switch /sys entry, if you have
> > > > > > used, I prefer using "usb-role-switch" property at dts to judge if SoC
> > > > > > OTG signals or external signals for role switch. If you have not used
> > > > > > it, I prefer only setting cdns->role_sw for role switch use cases.
> > > > > > 
> > > > > 
> > > > > We use both hardware role switch and /sys entries for manually forcing a
> > > > > certain role.
> > > > > 
> > > > > We do not set any "usb-role-switch" property at DTS.
> > > > > 
> > > > > Currently cdns->role_sw is being always set by driver irrespective of any DT
> > > > > property, so this patch is clearly wrong and needs to be reverted.
> > > > > 
> > > > > What do you think?
> > > > > 
> > > > 
> > > > Could you accept below fix?
> > > > 
> > > > diff --git a/drivers/usb/cdns3/core.c b/drivers/usb/cdns3/core.c
> > > > index 2e469139769f..fdd52e87a7b2 100644
> > > > --- a/drivers/usb/cdns3/core.c
> > > > +++ b/drivers/usb/cdns3/core.c
> > > > @@ -280,8 +280,8 @@ int cdns3_hw_role_switch(struct cdns3 *cdns)
> > > >           enum usb_role real_role, current_role;
> > > >           int ret = 0;
> > > > 
> > > > -       /* Depends on role switch class */
> > > > -       if (cdns->role_sw)
> > > > +       /* quit if switch role through external signals */
> > > > +       if (device_property_read_bool(cdns->dev, "usb-role-switch"))
> > > >                   return 0;
> > > > 
> > > >           pm_runtime_get_sync(cdns->dev);
> > > 
> > > Although this will fix the issue I don't think this is making the driver to behave
> > > as expected with usb-role-switch property.
> > > 
> > > Now, even if usb-role-switch property is not present the driver will still register
> > > the role switch driver.
> > > 
> > > I think we need to register the role switch driver only if usb-role-switch property
> > > is present. We would also need to set the default role if role-switch-default-mode is present.
> > > 
> > > How about the following? It still doesn't handle role-switch-default-mode property though.
> > > 
> > 
> > Roger, you said you also use /sys entries (I suppose it means through role
> > switch class) to do role switch, with your change, there will be no /sys
> > entry for role switch.
> 
> Sorry for the confusion. Although we do need both features (SW role switch + HW role switch)
> I don't think it is required to operate simultaneously. If users need SW control they can set the DT flag.
> 

I see. I prefer embracing all things related to role switch under the
firmware entry condition. Besides, I find another issue that devm_request_irq
for wakeup_irq does not call usb_role_switch_unregister if it has
failed. So, probably, two patches are needed. I am OK you send the
patches to fix both.

diff --git a/drivers/usb/cdns3/core.c b/drivers/usb/cdns3/core.c
index 2e469139769f..fc6a8152406c 100644
--- a/drivers/usb/cdns3/core.c
+++ b/drivers/usb/cdns3/core.c
@@ -427,7 +427,6 @@ static irqreturn_t cdns3_wakeup_irq(int irq, void *data)
  */
 static int cdns3_probe(struct platform_device *pdev)
 {
-	struct usb_role_switch_desc sw_desc = { };
 	struct device *dev = &pdev->dev;
 	struct resource	*res;
 	struct cdns3 *cdns;
@@ -529,18 +528,21 @@ static int cdns3_probe(struct platform_device *pdev)
 	if (ret)
 		goto err2;
 
-	sw_desc.set = cdns3_role_set;
-	sw_desc.get = cdns3_role_get;
-	sw_desc.allow_userspace_control = true;
-	sw_desc.driver_data = cdns;
-	if (device_property_read_bool(dev, "usb-role-switch"))
+	if (device_property_read_bool(dev, "usb-role-switch")) {
+		struct usb_role_switch_desc sw_desc = { };
+
+		sw_desc.set = cdns3_role_set;
+		sw_desc.get = cdns3_role_get;
+		sw_desc.allow_userspace_control = true;
+		sw_desc.driver_data = cdns;
 		sw_desc.fwnode = dev->fwnode;
 
-	cdns->role_sw = usb_role_switch_register(dev, &sw_desc);
-	if (IS_ERR(cdns->role_sw)) {
-		ret = PTR_ERR(cdns->role_sw);
-		dev_warn(dev, "Unable to register Role Switch\n");
-		goto err3;
+		cdns->role_sw = usb_role_switch_register(dev, &sw_desc);
+		if (IS_ERR(cdns->role_sw)) {
+			ret = PTR_ERR(cdns->role_sw);
+			dev_warn(dev, "Unable to register Role Switch\n");
+			goto err3;
+		}
 	}
 
 	if (cdns->wakeup_irq) {
@@ -583,7 +585,8 @@ static int cdns3_probe(struct platform_device *pdev)
 	return 0;
 err4:
 	cdns3_drd_exit(cdns);
-	usb_role_switch_unregister(cdns->role_sw);
+	if (cdns->role_sw)
+		usb_role_switch_unregister(cdns->role_sw);
 err3:
 	set_phy_power_off(cdns);
 err2:

-- 

Thanks,
Peter Chen