Date: Fri, 8 Apr 2016 15:45:40 +0800
From: Peter Chen <hzpeterchen@gmail.com>
To: Roger Quadros <rogerq@ti.com>
Cc: Felipe Balbi <balbi@kernel.org>, stern@rowland.harvard.edu,
        gregkh@linuxfoundation.org, peter.chen@freescale.com,
        dan.j.williams@intel.com, jun.li@freescale.com,
        mathias.nyman@linux.intel.com, tony@atomide.com,
        Joao.Pinto@synopsys.com, abrestic@chromium.org,
        linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-omap@vger.kernel.org
Subject: Re: [PATCH v6 01/12] usb: hcd: Initialize hcd->flags to 0
Message-ID: <20160408074540.GB9070@shlinux2.ap.freescale.net>
References: <1459865117-7032-1-git-send-email-rogerq@ti.com>
 <1459865117-7032-2-git-send-email-rogerq@ti.com>
 <87zit72rqz.fsf@intel.com>
 <5704AD76.9080806@ti.com>
 <20160407094246.GA9963@shlinux2.ap.freescale.net>
 <57063915.7000700@ti.com>
 <20160408010130.GA9070@shlinux2.ap.freescale.net>
 <57075ACE.1010702@ti.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <57075ACE.1010702@ti.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3487
Lines: 88

On Fri, Apr 08, 2016 at 10:16:30AM +0300, Roger Quadros wrote:
> On 08/04/16 04:01, Peter Chen wrote:
> > On Thu, Apr 07, 2016 at 01:40:21PM +0300, Roger Quadros wrote:
> >> On 07/04/16 12:42, Peter Chen wrote:
> >>> On Wed, Apr 06, 2016 at 09:32:22AM +0300, Roger Quadros wrote:
> >>>> On 06/04/16 09:09, Felipe Balbi wrote:
> >>>>>
> >>>>> Hi,
> >>>>>
> >>>>> Roger Quadros <rogerq@ti.com> writes:
> >>>>>> diff --git a/drivers/usb/core/hcd.c b/drivers/usb/core/hcd.c
> >>>>>> index 2ca2cef..6b1930d 100644
> >>>>>> --- a/drivers/usb/core/hcd.c
> >>>>>> +++ b/drivers/usb/core/hcd.c
> >>>>>> @@ -2706,6 +2706,7 @@ int usb_add_hcd(struct usb_hcd *hcd,
> >>>>>>  	int retval;
> >>>>>>  	struct usb_device *rhdev;
> >>>>>>  
> >>>>>> +	hcd->flags = 0;
> >>>>>
> >>>
> >>> I am not sure if this usb_add(remove)_hcd pair is safe and clean enough
> >>> for start/stop host role. From my point, we may need to do like
> >>> .probe/.remove host platform driver interface. In that case, we can make
> >>
> >> probe and remove are meant to be called from bus layer.
> >> I do not see a way how OTG framework can call probe/remove of HCD driver.
> >> Some HCDs may be platform devices, some PCI, so different entities are calling
> >> the HCD .probe hook.
> >>
> >>> sure the clocks and regulators are off, and hcd will be zero-initialized
> >>
> >> why can't we make that sure that is taken care of within the hcd_ops?
> >> Why should some driver keep its regulators and clocks enabled when hcd is stopped?
> >> It doesn't need to. If it is doing so now, it needs to be fixed.
> >>
> > 
> > Well, you may misunderstand me. I mean your hcd_ops->start or ->stop
> > is hard to be a general one which only calls usb_hcd_add or
> > usb_hcd_remove. It needs to implement like .probe or .remove at platform
> > driver, some example code like host_start and host_stop at
> > drivers/usb/chipidea/host.c.
> > 
> The only extra thing the host_start/stop() of that driver is doing is
> enabling/disabling the VBUS regulator.
> In the OTG/dual role scope VBUS regulator handling has to be done via the
> OTG driver using the otg ops otg_drv_vbus() and not at HCD level. Do you agree?
> 

I agree with you that vbus handling should be handled by otg_drv_vbus
for OTG case.

> I don't want to complicate the OTG to HCD interface by adding new hooks there.
> If HCD driver wants to do something special for OTG case it can always do that
> within the struct hc_driver interface. But ideally OTG specific handling must
> be done in the OTG driver. All that the HCD driver should care about is making sure
> all used resources are disabled once usb_remove_hcd() is called.
> 

In this patch, we set hcd->flag as 0 to fix one problem, but if any
problems are still existed due to non-zero hcd structure. We have
tested a lot for insmod/rmmod procedure for host driver, it doesn't
show many problems. In a word, once the host role is stopped do we
need to remove all things belong to host, and re-initialized all things
belong to host when the host role is re-started? From my point is 'yes'.

host_stop()
{
	usb_remove_hcd(hcd);
	usb_put_hcd(hcd);
	platform_deinit(); /* clock, gpio, etc */
}

host_start()
{
	platform_init(); /* clock, gpio, etc */
	usb_create_hcd();
	... /* initialize hcd structure from platform information */
	usb_add_hcd(hcd);
}

Maybe you have already done something, but there is no user now.
Things will be more clear after your post your dwc3 changes :)

-- 

Best Regards,
Peter Chen