linux-kernel - Re: [PATCH 6/6 v2] arm: omap: usb: global Suspend and resume support of ehci and ohci

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: <20110704093056.GD25295@legolas.emea.dhcp.ti.com>
Date:	Mon, 4 Jul 2011 12:30:57 +0300
From:	Felipe Balbi <balbi@...com>
To:	Partha Basak <p-basak2@...com>
Cc:	balbi@...com, Alan Stern <stern@...land.harvard.edu>,
	Keshava Munegowda <keshava_mgowda@...com>,
	linux-usb@...r.kernel.org, linux-omap@...r.kernel.org,
	linux-kernel@...r.kernel.org, Anand Gadiyar <gadiyar@...com>,
	sameo@...ux.intel.com, parthab@...ia.ti.com, tony@...mide.com,
	Kevin Hilman <khilman@...com>,
	Benoit Cousson <b-cousson@...com>, paul@...an.com,
	johnstul@...ibm.com, Vishwanath Sripathy <vishwanath.bs@...com>
Subject: Re: [PATCH 6/6 v2] arm: omap: usb: global Suspend and resume support
 of ehci and ohci

Hi,

On Mon, Jul 04, 2011 at 02:56:30PM +0530, Partha Basak wrote:
> >> Both for EHCI & OHCI, the clocks are owned by the parent (uhh-tll).
> >>
> >> Calling pm_runtime_put_sync(dev->parent) within omap_ehci_suspend will
> >> turn-off the parent clocks in the Suspend path.
> >>
> >> Similarly, calling pm_runtime_get_sync(dev->parent) within
> >> omap_ehci_resume will turn-on the parent clocks in the resume path.
> >>
> >> This way, all reference counting are implicit within the Runtime PM
> >> layer and takes care of all combinations of only EHCI insmoded, OHCI
> >> insmoded, both insmoded etc.
> >>
> >> When both EHCI & OHCI are suspended, parent clocks will actually be
> >> turned OFF and vice-versa.
> >
> >not sure this is necessary. I would expect:
> >
> >pm_runtime_get_sync(dev) to propagate up the parent tree and enable all
> >necessary resources to get the child in a working state. IOW, you
> >shouldn't need to manuall access the parent device.
> >
> Refer to the description in Patch(5/6)
> <snip>
> In fact, the runtime framework takes care the get sync and put sync of the
> child
> in turn call the get sync and put sync of parent too; but calling get sync
> and
> put sync of parent is by ASYNC mode;
> This mode queues the work item in runtime pm work queue,
> which not getting scheduled in case of global suspend path.
> <snip>
> This approach was tried, but did not work in the Suspend path

sounds to me like a bug on pm runtime ? If you're calling
pm_runtime_*_sync() family, shouldn't all calls be _sync() too ?

> static int rpm_suspend(struct device *dev, int rpmflags)
> 	__releases(&dev->power.lock) __acquires(&dev->power.lock)
> {
> .
> .
> .
> no_callback:
> .
> .
> .
> 	/* Maybe the parent is now able to suspend. */
> 	if (parent && !parent->power.ignore_children &&
> !dev->power.irq_safe) {
> 		spin_unlock(&dev->power.lock);
> 
> 		spin_lock(&parent->power.lock);
> 		rpm_idle(parent, RPM_ASYNC);

to me this is bogus, if you called pm_runtime_put_sync() should should
be sync too. Shouldn't it ?

> 		spin_unlock(&parent->power.lock);
> 
> 		spin_lock(&dev->power.lock);
> 	}
> This is the reason of directly calling the parent Runtime PM calls from
> the children.
> If directly calling Runtime PM APIs with parent dev-pointer isn't
> acceptable,
> this can be achieved by exporting wrapper APIs from the
> parent and calling them from the chidren .suspend/.resume routines.

Still no good, IMHO.

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (491 bytes)