[PATCH] psmouse: touchpad doesn't reconnect after resume: Synaptics ps2 Bug: 551234

[Andy Whitcroft]

On Mon, Mar 29, 2010 at 05:18:35PM -0400, Peter M. Petrakis wrote:
> Hi All,
> 
> This is my first stab at correcting some PS/2 misbehavior when returning
> from S3. Fire away :). Thanks.
> 
> Peter
> 
> P.S. Collaborated with Colin King on this one.

> From 006e55bacfe64f5fe6b8c12af071f7b2bf528c3a Mon Sep 17 00:00:00 2001
> From: Peter M. Petrakis <peter.petrakis at canonical.com>
> Date: Fri, 26 Mar 2010 17:23:43 -0400
> Subject: [PATCH] Perform a reset should psmouse_probe fail for any reason. This is necessary for some PS/2 devices who can't even be identified when returning from sleep (S3).
> 
> ---
>  psmouse-base.c |   28 +++++++++++++++++++++++++---
>  1 files changed, 25 insertions(+), 3 deletions(-)
> 
> diff --git a/psmouse-base.c b/psmouse-base.c
> index b407b35..3b8e63c 100644
> --- a/psmouse-base.c
> +++ b/psmouse-base.c
> @@ -856,12 +856,12 @@ static const struct psmouse_protocol *psmouse_protocol_by_name(const char *name,
>  	return NULL;
>  }
>  
> -
>  /*
> - * psmouse_probe() probes for a PS/2 mouse.
> + * __psmouse_probe() probes for a PS/2 mouse.
> + * Wrapped by psmouse_probe() for clean reset code.
>   */
>  
> -static int psmouse_probe(struct psmouse *psmouse)
> +static int __psmouse_probe(struct psmouse *psmouse)
>  {
>  	struct ps2dev *ps2dev = &psmouse->ps2dev;
>  	unsigned char param[2];
> @@ -892,6 +892,28 @@ static int psmouse_probe(struct psmouse *psmouse)
>  }
>  
>  /*
> + * Wrapper for probe routine to cleanly reset the device should
> + * the initial probe fail for any reason.
> + */
> +static int psmouse_probe(struct psmouse *psmouse) {
> +	struct ps2dev *ps2dev = &psmouse->ps2dev;
> +	int tries = 3;
> +	int ret = -1;
> +
> +retry:
> +	if (tries > 0 && ((ret = __psmouse_probe(psmouse)) != 0)) {
> +		printk(KERN_ERR "psmouse.c: Error encountered while probing PS/2 device on %s "
> +			"reseting...\n", ps2dev->serio->phys);
> +		psmouse_reset(psmouse);
> +		msleep(500);
> +		tries--;
> +		goto retry;
> +	}

How does this behave if the machine does not have a psmouse on this
port?  Do we still call here?  If so that sounds like it might add 1.5s
to boot/resume?

Is 500ms a number picked out of the air or does it have some basis in an
errata?  It seems like a long time in the real world.

Also what machines have this issue, what devices?  Can we quirk this on
only for the machines which are broken? 

Finally, what releases is this aimed at?

> +
> +	return ret;
> +}
> +
> +/*
>   * Here we set the mouse resolution.
>   */

-apw