From: Alan Stern <stern@rowland.harvard.edu>
To: Mathias Nyman <mathias.nyman@linux.intel.com>
Cc: Evan Green <evgreen@chromium.org>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Mathias Nyman <mathias.nyman@intel.com>,
Rajat Jain <rajatja@chromium.org>,
Thomas Gleixner <tglx@linutronix.de>,
Bjorn Helgaas <bhelgaas@google.com>,
"Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
Youngjin Jang <yj84.jang@samsung.com>,
LKML <linux-kernel@vger.kernel.org>,
linux-usb@vger.kernel.org
Subject: Re: [PATCH] USB: hcd-pci: Fully suspend across freeze/thaw cycle
Date: Thu, 14 Apr 2022 10:21:37 -0400 [thread overview]
Message-ID: <Ylgt8Y7Mz4nOAhtv@rowland.harvard.edu> (raw)
In-Reply-To: <b1df80e4-af6a-e84f-f49d-c74500bdec05@linux.intel.com>
On Thu, Apr 14, 2022 at 05:00:12PM +0300, Mathias Nyman wrote:
> On 12.4.2022 18.40, Alan Stern wrote:
> > On Tue, Apr 12, 2022 at 05:56:42PM +0300, Mathias Nyman wrote:
> >> On 11.4.2022 17.50, Alan Stern wrote:
> >>> For example, what would happen if the user unplugs a device right in the
> >>> middle of the freeze transition, after the root hub has been frozen but
> >>> before the controller is frozen? We don't want such an unplug event to
> >>> prevent the system from going into hibernation -- especially if the root
> >>> hub was not enabled for wakeup.
> >>
> >> We should be able to let system go to hibernate even if we get a disconnect
> >> interrupt between roothub and host controller freeze.
> >> Host is not yet suspended so no PME# wake is generated, only an interrupt.
> >>
> >> From Linux PM point of view it should be ok as well as the actual xhci
> >> device that is generating the interrupt is hasnt completer freeze()
> >>
> >> The xhci interrupt handler just needs to make sure that the disconnect
> >> isn't propagated if roothub is suspended and wake on disconnect
> >> is not set. And definitely make sure xhci doesn't start roothub polling.
> >>
> >> When freeze() is called for the host we should prevent the host from
> >> generating interrupts.
> >
> > I guess that means adding a new callback. Or we could just suspend the
> > controller, like Evan proposed originally
>
> Suspending the host in freeze should work.
> It will do an extra xhci controller state save stage, but that should be harmless.
>
> But is there really a need for the suggested noirq part?
>
> + .freeze_noirq = hcd_pci_suspend_noirq,
>
> That will try to set the host to PCI D3 state.
> It seems a bit unnecessary for freeze.
Agreed.
> >>> (If the root hub _is_ enabled for wakeup then it's questionable.
> >>> Unplugging a device would be a wakeup event, so you could easily argue
> >>> that it _should_ prevent the system from going into hibernation. After
> >>> all, if the unplug happened a few milliseconds later, after the system
> >>> had fully gone into hibernation, then it would cause the system to wake
> >>> up.)
> >>>
> >>>> Would it make sense prevent xHCI interrupt generation in the host
> >>>> freeze() stage, clearing the xHCI EINT bit in addition to calling
> >>>> check_roothub_suspend()?
> >>>> Then enable it back in thaw()
> >>>
> >>> That won't fully eliminate the problem mentioned in the preceding
> >>> paragraphs, although I guess it would help somewhat.
> >>
> >> Would the following steps solve this?
> >>
> >> 1. Disable device initiated resume for connected usb devices in freeze()
> >>
> >> 2. Don't propagate connect or OC changes if roothub is suspended and port wake
> >> flags are disabled. I.E don't kick roothub polling in xhci interrupt
> >> handler here.
> >
> > I guess you can't just halt the entire host controller when only one of
> > the root hubs is suspended with wakeup disabled. That does complicate
> > things. But you could halt it as soon as both of the root hubs are
> > frozen. Wouldn't that prevent interrupt generation?
>
> True, but probably easier to just suspend host in freeze() as you stated above.
Okay.
Evan, this discussion suggests that you rewrite your patch as a series
of three:
1. Change choose_wakeup() so that for PM_EVENT_FREEZE, wakeup is
always disabled.
2. Change the xhci-hcd interrupt handler so that port-status
changes are ignored if the port's root hub is suspended with
wakeup disabled.
3. As in the original patch, make the .freeze and .thaw callbacks
in hcd-pci.c call the appropriate suspend and resume routines,
but don't do anything for .freeze_noirq and .thaw_noirq.
How does that sound?
Alan Stern
next prev parent reply other threads:[~2022-04-14 15:40 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-04-07 18:59 Evan Green
2022-04-08 14:29 ` Alan Stern
2022-04-08 21:52 ` Evan Green
2022-04-09 1:58 ` Alan Stern
2022-04-11 10:43 ` Mathias Nyman
2022-04-11 14:50 ` Alan Stern
2022-04-12 14:56 ` Mathias Nyman
2022-04-12 15:40 ` Alan Stern
2022-04-14 14:00 ` Mathias Nyman
2022-04-14 14:21 ` Alan Stern [this message]
2022-04-14 16:30 ` Evan Green
2022-04-14 17:06 ` Mathias Nyman
2022-04-14 20:16 ` Alan Stern
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Ylgt8Y7Mz4nOAhtv@rowland.harvard.edu \
--to=stern@rowland.harvard.edu \
--cc=bhelgaas@google.com \
--cc=evgreen@chromium.org \
--cc=gregkh@linuxfoundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-usb@vger.kernel.org \
--cc=mathias.nyman@intel.com \
--cc=mathias.nyman@linux.intel.com \
--cc=rafael.j.wysocki@intel.com \
--cc=rajatja@chromium.org \
--cc=tglx@linutronix.de \
--cc=yj84.jang@samsung.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®