From mboxrd@z Thu Jan  1 00:00:00 1970
From: peter.chen@freescale.com (Peter Chen)
Date: Mon, 28 Jan 2013 13:21:59 +0800
Subject: [RESEND PATCH v5 4/7] usb: chipidea: consolidate
 ci_role_driver's API for both roles
In-Reply-To: <87sj5pbgwr.fsf@ashishki-desk.ger.corp.intel.com>
References: <1358733418-17969-1-git-send-email-peter.chen@freescale.com>
 <1358733418-17969-5-git-send-email-peter.chen@freescale.com>
 <8738xqodhp.fsf@ashishki-desk.ger.corp.intel.com>
 <20130125033013.GB29795@nchen-desktop>
 <87sj5pbgwr.fsf@ashishki-desk.ger.corp.intel.com>
Message-ID: <20130128052158.GA8066@nchen-desktop>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Fri, Jan 25, 2013 at 02:12:20PM +0200, Alexander Shishkin wrote:
> Peter Chen <peter.chen@freescale.com> writes:
> 
> > On Thu, Jan 24, 2013 at 04:35:30PM +0200, Alexander Shishkin wrote:
> >> Peter Chen <peter.chen@freescale.com> writes:
> >> 
> >> > - Create init/destroy API for probe and remove
> >> > - start/stop API are only used otg id switch process
> >> > - Create the gadget at ci_hdrc_probe if the gadget is supported
> >> > at that port, the main purpose for this is to avoid gadget module
> >> > load fail at init.rc
> >> 
> @@ -402,6 +402,12 @@ static ssize_t store_role(struct device *dev, struct device_attribute *attr,
>  	if (ret)
>  		return ret;
>  
> +	/*
> +	 * there won't be an interrupt in case of manual switching,
> +	 * so we need to check vbus session manually
> +	 */
> +	ci_handle_vbus_change(ci);
> +
It may not be used as there will be a vbus interrupt.
>  	return count;
>  }
>  
>  	}
>  	dbg_remove_files(&ci->gadget.dev);
>  	device_unregister(&ci->gadget.dev);
> -	/* my kobject is dynamic, I swear! */
> -	memset(&ci->gadget, 0, sizeof(ci->gadget));
Any reason to delete it, another fix?
>  }
>  
>  /**
> @@ -1842,13 +1839,11 @@ int ci_hdrc_gadget_init(struct ci13xxx *ci)
>  	if (!rdrv)
>  		return -ENOMEM;
>  
> -	rdrv->init	= udc_start;
>  	rdrv->start	= udc_id_switch_for_device;
>  	rdrv->stop	= udc_id_switch_for_host;
> -	rdrv->destroy	= udc_stop;
Where we call udc_start and udc_stop? And the udc_start should only be called
one time.
> 
> Which is shorter (32 insertions(+), 53 deletions(-)) and makes the main
> probe easier on the eyes. What do you think?
OK, not matter how to change it, it needs to cover below things:
(Covers last email)
- Gadget needs to be only initialized one time.
- Host/device function should be OK when the device on it or
the usb cable connects to host during the boots up.
- When the OTG port is at host mode, it should not call any
register writing operations at gadget's API.

-- 

Best Regards,
Peter Chen