linux-kernel - Re: [PATCH v8 6/6] usb: dwc3: qcom: Keep power domain on to support wakeup

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]

Date:   Wed, 18 Aug 2021 12:56:56 +0300
From:   Felipe Balbi <balbi@...nel.org>
To:     Sandeep Maheswaram <sanm@...eaurora.org>
Cc:     Matthias Kaehlcke <mka@...omium.org>,
        Andy Gross <agross@...nel.org>,
        Bjorn Andersson <bjorn.andersson@...aro.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Stephen Boyd <swboyd@...omium.org>,
        Doug Anderson <dianders@...omium.org>,
        Mathias Nyman <mathias.nyman@...el.com>,
        linux-arm-msm@...r.kernel.org, linux-usb@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Pratham Pratap <prathampratap@...eaurora.org>
Subject: Re: [PATCH v8 6/6] usb: dwc3: qcom: Keep power domain on to support
 wakeup


Hi,

Sandeep Maheswaram <sanm@...eaurora.org> writes:
>> This means that in order for glue_suspend() to run, dwc3 has to suspend
>> first and xhci has to suspend before dwc3.
>>
>> For example, in the suspend call above, qcom (the glue) is directly
>> accessing dwc3 core data, which is incorrect. It looks like we want to
>> know if the PHY is not powered off and if it isn't, then we want to
>> change the power domain ACTIVE_WAKEUP flag. Now, phy_power_off is false
>> whenever any of xHCI's children enable USB wakeup.
>>
>> It seems like we need to way to generically propagate that knowledge up
>> the parent tree. I.e., a parent needs to know if its child is wakeup
>> capable, then dwc3 could, in its suspend routine:
>>
>> static int dwc3_suspend(struct device *dev)
>> {
>> 	/* ... */
>>
>> 	if (device_children_wakeup_capable(dev))
>>          	device_enable_wakeup(dev);
>>
>> 	/* ... */
>> }
>
> Can we use like  this device_may_wakeup(&dwc->xhci->dev) to check if
> children is wakeup capable like below ?

that really doesn't sound like a good idea, IMHO. We're still passing
through layers of abstraction without anyone's knowledge :-)

It looks to me like we're missing some infrastructure in the wakeup code
so parents can make decisions based on the state of their children.

>> and similarly for qcom glue:
>>
>> static int dwc3_qcom_suspend(struct device *dev)
>> {
>> 	/* ... */
>>
>>
>> 	if (device_children_wakeup_capable(dev)) {
>>          	device_enable_wakeup(dev);
>> 		genpd->flags |= GENPD_FLAG_ACTIVE_WAKEUP;
>>          }
>>
>> 	/* ... */
>> }
>>
>> It also seems plausible that this could be done at driver core and
>> completely hidden away from drivers.
>
> And in qcom glue like this
>
> static int dwc3_qcom_suspend(struct dwc3_qcom *qcom)
> {
>
> /* ... */
>
>     struct dwc3 *dwc = platform_get_drvdata(qcom->dwc3);

you see, here there's an assumption that the platform data is still
valid and not some bogus dangling pointer. There's also an assumption
that the type is struct dwc3 (which is unlikely to change, but still).

-- 
balbi