linux-kernel - Re: [PATCH v2] usb: host: xhci: Avoid XHCI resume delay if SSUSB device is not present

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: <8dd86cf5-6337-b8f5-34d5-dcd290dc2d38@linux.intel.com>
Date:   Mon, 11 Sep 2023 16:50:02 +0300
From:   Mathias Nyman <mathias.nyman@...ux.intel.com>
To:     Wesley Cheng <quic_wcheng@...cinc.com>, mathias.nyman@...el.com,
        gregkh@...uxfoundation.org
Cc:     linux-kernel@...r.kernel.org, linux-usb@...r.kernel.org,
        quic_jackp@...cinc.com
Subject: Re: [PATCH v2] usb: host: xhci: Avoid XHCI resume delay if SSUSB
 device is not present

On 1.9.2023 3.15, Wesley Cheng wrote:
> There is a 120ms delay implemented for allowing the XHCI host controller to
> detect a U3 wakeup pulse.  The intention is to wait for the device to retry
> the wakeup event if the USB3 PORTSC doesn't reflect the RESUME link status
> by the time it is checked.  As per the USB3 specification:
> 
>    tU3WakeupRetryDelay ("Table 7-12. LTSSM State Transition Timeouts")
> 
> This would allow the XHCI resume sequence to determine if the root hub
> needs to be also resumed.  However, in case there is no device connected,
> or if there is only a HSUSB device connected, this delay would still affect
> the overall resume timing.
> 
> Since this delay is solely for detecting U3 wake events (USB3 specific)
> then ignore this delay for the disconnected case and the HSUSB connected
> only case.

Thanks, makes sense

> 
> Signed-off-by: Wesley Cheng <quic_wcheng@...cinc.com>
> ---
>   drivers/usb/host/xhci.c | 26 +++++++++++++++++++++++++-
>   1 file changed, 25 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
> index e1b1b64a0723..640db6a4e686 100644
> --- a/drivers/usb/host/xhci.c
> +++ b/drivers/usb/host/xhci.c
> @@ -805,6 +805,24 @@ static void xhci_disable_hub_port_wake(struct xhci_hcd *xhci,
>   	spin_unlock_irqrestore(&xhci->lock, flags);
>   }
>   
> +static bool xhci_usb3_dev_connected(struct xhci_hcd *xhci)
> +{
> +	struct xhci_port	**ports;
> +	int			port_index;
> +	u32			portsc;
> +
> +	port_index = xhci->usb3_rhub.num_ports;
> +	ports = xhci->usb3_rhub.ports;
> +
> +	while (port_index--) {
> +		portsc = readl(ports[port_index]->addr);
> +		if (portsc & (portsc & PORT_CONNECT))
> +			return true;
> +	}
> +
> +	return false;
> +}
> +

This is one way, but we can probably avoid re-reading all the usb3 portsc registers
by checking if any bit is set in either:

  // bitfield, set if xhci usb3 port neatly set to U3 with a hub request
xhci->usb3_rhub.bus_state.suspended_ports

// bitfield, set if xhci usb3 port is forced to U3 during xhci suspend.
xhci->usb3_rhub.bus_state.bus_suspended

But haven't checked this works in all corner cases.

Thanks
Mathias