Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Feedback-ID: i51094778:Fastmail
User-Agent: Cyrus-JMAP/3.7.0-alpha0-1115-g8b801eadce-fm-20221102.001-g8b801ead
Mime-Version: 1.0
Message-Id: <0bc7b309-6f3b-4821-81d0-435671d64e2f@app.fastmail.com>
In-Reply-To: <Y2jrQgj53z/mhHmm@smile.fi.intel.com>
References: <20221106214804.2814-1-j@jannau.net>
 <Y2jrQgj53z/mhHmm@smile.fi.intel.com>
Date:   Mon, 07 Nov 2022 17:00:47 +0100
From:   "Sven Peter" <sven@svenpeter.dev>
To:     "Andy Shevchenko" <andriy.shevchenko@linux.intel.com>,
        "Janne Grunau" <j@jannau.net>
Cc:     linux-usb@vger.kernel.org, stable@kernel.org,
        "Thinh Nguyen" <Thinh.Nguyen@synopsys.com>,
        "Greg Kroah-Hartman" <gregkh@linuxfoundation.org>,
        "Andrey Smirnov" <andrew.smirnov@gmail.com>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/1] usb: dwc3: Do not get extcon device when usb-role-switch is
 used
Content-Type: text/plain
Precedence: bulk

On Mon, Nov 7, 2022, at 12:25, Andy Shevchenko wrote:
> On Sun, Nov 06, 2022 at 10:48:04PM +0100, Janne Grunau wrote:
>> The change breaks device tree based platforms with PHY device and use
>> usb-role-switch instead of an extcon switch. extcon_find_edev_by_node()
>> will return EPROBE_DEFER if it can not find a device so probing without
>> an extcon device will be deferred indefinitely. Fix this by
>> explicitly checking for usb-role-switch.
>> At least the out-of-tree USB3 support on Apple silicon based platforms
>> using dwc3 with tipd USB Type-C and PD controller is affected by this
>> issue.
>
> We don't care about out-of-tree modules, do we?
>
> OTOH, the problem you are trying to workaround is probably in a (mis)use of
> deferred probe somewhere.
>
> Btw, does it prevent the system boot or you just see the extcon in the list of
> deferred devices after booting?

Which extcon? The commit description already mentions that the issue is that there
is no extcon and that the dwc3 probe gets stuck with EPROBE_DEFER forever.

This happens because the code after Janne's new check looks for the PHY and then
just assumes that if the PHY has a "port" that the other end always is an extcon.
It then tries extcon_find_edev_by_node which will always fail with EPROBE_DEFER
if that node never registers an extcon.

If "usb-role-switch" is used and configured in the DT there is no extcon.
There actually cannot ever be a working extcon with "usb-role-switch" because
the very first thing dwc3_drd_init does is to look for a role switch partner
and then skip the entire extcon setup:

int dwc3_drd_init(struct dwc3 *dwc)
{
	int ret, irq;

	if (ROLE_SWITCH &&
	    device_property_read_bool(dwc->dev, "usb-role-switch"))
		return dwc3_setup_role_switch(dwc);
[....]


This entire issue was actually first fixed in ab7aa2866d29, then broken
again in 0f0101719138 due to a merge resolution, then fixed again with
7a84e7353e23 (where we actually had a brief discussion about this already
on the ML) and then broken again in d182c2e1bc92.

Janne's fix is much less subtle and should hopefully survive this time.

For the patch:

Reviewed-by: Sven Peter <sven@svenpeter.dev>


Best,

Sven