refine xhci-plat-Fix-xhci_plat-shutdown-sequence

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, Apr 18, 2018 at 01:41:45PM +0800, leiwan@codeaurora.org wrote:
> On 2018-04-17 18:32, Greg KH wrote:
> > On Tue, Apr 17, 2018 at 11:32:42AM +0800, leiwan@codeaurora.org wrote:
> > > 
> > > xhci-plat Shutdown callback should check HCD_FLAG_HW_ACCESSIBLE
> > > before accessing any register. This should avoid hung with access
> > > controllers which support runtime suspend
> > > 
> > > This can fix for issue of https://patchwork.kernel.org/patch/10339317/
> > > corresponding upload in CAF:
> > > https://source.codeaurora.org/quic/la/kernel/msm-4.4/commit/?h=LV.HB.1.1.5-03810-8x96.0&id=a7a5307ee04ad349d365ad50f304605a9cd9bd0a
> > > 
> > > full patch refer attachment.
> > > diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
> > > index 9b27798..bdf914d 100644
> > > --- a/drivers/usb/host/xhci.c
> > > +++ b/drivers/usb/host/xhci.c
> > > @@ -702,6 +702,10 @@ static void xhci_shutdown(struct usb_hcd *hcd)
> > >                 usb_disable_xhci_ports(to_pci_dev(hcd->self.sysdev));
> > > 
> > >         spin_lock_irq(&xhci->lock);
> > > +       if (!HCD_HW_ACCESSIBLE(hcd)) {
> > > +               spin_unlock_irq(&xhci->lock);
> > > +               return;
> > > +       }
> > >         xhci_halt(xhci);
> > 
> > A blank line after the if statement?
> > > [lei]I checked other code in this file, seems we do not neeed a
> > > blank line after if here.
> > 
> > What about all of the other places in this driver that you should also
> > check for this?  Look at the other host controllers, shouldn't you
> > mirror what they are doing?
> > > [lei] The issue here is when shutdown called, xhci-host driver will
> > > try to access
> > > usb register which may be in runtime suspend state, this lead to a
> > > hang up.
> > > Similar as what xhci-host suspend workflow has done, we need check
> > > HW_ACCESS.
> > And this needs a Fixes: tag, along with a cc: stable so as to properly
> > get backported as this is broken in some stable kernels right now.
> > > [lei] Added Cc and Fixes part in commit message
> > 
> > thanks,
> > 
> > greg k-h
> 
> 
> From c03697fa259ab38d1002598ec2ccfac37607ca0b Mon Sep 17 00:00:00 2001
> From: Lei wang <leiwan@codeaurora.org>
> Date: Tue, 17 Apr 2018 10:55:35 +0800
> Subject: [PATCH v2] xhci: plat: Fix xhci_plat shutdown hung
> 
> xhci-plat Shutdown callback should check HCD_FLAG_HW_ACCESSIBLE
> before accessing any register. This should avoid hung with access
> controllers which support runtime suspend
> 
> Fixes: b07c12517f2a ("xhci: plat: Register shutdown for xhci_plat")
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Lei wang <leiwan@codeaurora.org>
> ---
>  drivers/usb/host/xhci.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
> index 9b27798..bdf914d 100644
> --- a/drivers/usb/host/xhci.c
> +++ b/drivers/usb/host/xhci.c
> @@ -702,6 +702,10 @@ static void xhci_shutdown(struct usb_hcd *hcd)
>                 usb_disable_xhci_ports(to_pci_dev(hcd->self.sysdev));
> 
>         spin_lock_irq(&xhci->lock);
> +       if (!HCD_HW_ACCESSIBLE(hcd)) {
> +               spin_unlock_irq(&xhci->lock);
> +               return;
> +       }
>         xhci_halt(xhci);
>         /* Workaround for spurious wakeups at shutdown with HSW */
>         if (xhci->quirks & XHCI_SPURIOUS_WAKEUP)
> --
> 1.9.1

Please step back, go find someone at Qualcomm that knows how to submit
patches upstream, and have them read this email and point out what you
did incorrectly.

Also, you did not address any of the patch review comments I made on the
first version for some unknown reason, making it less likely that anyone
wants to review this patch again as it's obviously not correct...

thanks,

greg k-h
---
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html