linux-kernel - Re: [PATCH v2 2/2] PCI: Fix runtime PM race with PME polling

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: Tue, 23 Jan 2024 11:45:19 +0100
From: Lukas Wunner <lukas@...ner.de>
To: Alex Williamson <alex.williamson@...hat.com>
Cc: linux-pci@...r.kernel.org, bhelgaas@...gle.com,
	linux-kernel@...r.kernel.org, eric.auger@...hat.com,
	mika.westerberg@...ux.intel.com, rafael.j.wysocki@...el.com,
	Sanath.S@....com
Subject: Re: [PATCH v2 2/2] PCI: Fix runtime PM race with PME polling

On Mon, Jan 22, 2024 at 03:50:03PM -0700, Alex Williamson wrote:
> On Mon, 22 Jan 2024 23:17:30 +0100 Lukas Wunner <lukas@...ner.de> wrote:
> > On Thu, Jan 18, 2024 at 11:50:49AM -0700, Alex Williamson wrote:
> > > To do that I used pm_runtime_get_if_active(), but in retrospect this
> > > requires the device to be in RPM_ACTIVE so we end up skipping anything
> > > suspended or transitioning.  
> > 
> > How about dropping the calls to pm_runtime_get_if_active() and
> > pm_runtime_put() and instead simply do:
> > 
> > 			if (pm_runtime_suspended(&pdev->dev) &&
> > 			    pdev->current_state != PCI_D3cold)
> > 				pci_pme_wakeup(pdev, NULL);
> 
> Do we require that the polled device is in the RPM_SUSPENDED state?

If the device is RPM_SUSPENDING, why immediately resume it for polling?
It's sufficient to poll it the next time around, i.e. 1 second later.

Likewise, if it's already RPM_RESUMING or RPM_ACTIVE anyway, no need
to poll PME.

This leaves RPM_SUSPENDED as the only state in which it makes sense to
poll.

> Also pm_runtime_suspended() can also only be trusted while holding the
> device power.lock, we need a usage count reference to maintain that
> state.

Why?  Let's say there's a race and the device resumes immediately after
we call pm_runtime_suspended() here.  So we might call pci_pme_wakeup()
gratuitouly.  So what?  No biggie.

> +			if (bdev) {
> +				spin_lock_irq(&bdev->power.lock);

Hm, I'd expect that lock to be internal to the PM core,
although there *are* a few stray users outside of it.

Thanks,

Lukas