Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Thinh Nguyen <thinh.nguyen@synopsys.com>
To:     John Stultz <john.stultz@linaro.org>,
        Felipe Balbi <balbi@kernel.org>,
        Zeng Tao <prime.zeng@hisilicon.com>,
        Jack Pham <jackp@codeaurora.org>,
        Thinh Nguyen <thinh.nguyen@synopsys.com>,
        Chen Yu <chenyu56@huawei.com>
CC:     lkml <linux-kernel@vger.kernel.org>,
        Linux USB List <linux-usb@vger.kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Subject: Re: Frequent dwc3 crashes on suspend or reboot since 5.0-rc1
Thread-Topic: Frequent dwc3 crashes on suspend or reboot since 5.0-rc1
Thread-Index: AQHUuozT6kyyipjNNU6b/TjVOcoaGw==
Date:   Sat, 2 Feb 2019 00:29:38 +0000
Message-ID: <30102591E157244384E984126FC3CB4F639BF445@us01wembx1.internal.synopsys.com>
References: <CALAqxLXy1zSTSoDEfYN0VJBbf=Q0N6kH+S_1hTTgEb6dQi0XrQ@mail.gmail.com>
Accept-Language: en-US
Content-Language: en-US
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi John,=0A=
=0A=
John Stultz wrote:=0A=
> Hey all,=0A=
>   Since the 5.0 merge window opened, I've been tripping on frequent=0A=
> dwc3 crashes on reboot and suspend, which I've added an example to the=0A=
> bottom of this mail.=0A=
>=0A=
> I've dug in a little bit and sort of have a sense of whats going on.=0A=
>=0A=
> In ffs_epfile_io():=0A=
> https://urldefense.proofpoint.com/v2/url?u=3Dhttps-3A__git.kernel.org_pub=
_scm_linux_kernel_git_torvalds_linux.git_tree_drivers_usb_gadget_function_f=
-5Ffs.c-23n1065&d=3DDwIBaQ&c=3DDPL6_X_6JkXFx7AXWqB0tg&r=3Du9FYoxKtyhjrGFcyi=
xFYqTjw1ZX0VsG2d8FCmzkTY-w&m=3Da8TU-itM8GBG_EARYf2yM-kVfCzmaPkKDNAUFQHTe3Q&=
s=3DBQiVAFiViSlxVg5_LemED0x_47FLVUD43M7R6h6T8qk&e=3D=0A=
>=0A=
> The completion done is setup on the stack:=0A=
>   DECLARE_COMPLETION_ONSTACK(done);=0A=
>=0A=
> Then later we setup a request and queue it:=0A=
>   req->context  =3D &done;=0A=
>   ...=0A=
>   ret =3D usb_ep_queue(ep->ep, req, GFP_ATOMIC);=0A=
>=0A=
> Then wait for it:=0A=
>   if (unlikely(wait_for_completion_interruptible(&done))) {=0A=
>     /*=0A=
>     * To avoid race condition with ffs_epfile_io_complete,=0A=
>     * dequeue the request first then check=0A=
>     * status. usb_ep_dequeue API should guarantee no race=0A=
>     * condition with req->complete callback.=0A=
>     */=0A=
>     usb_ep_dequeue(ep->ep, req);=0A=
>     interrupted =3D ep->status < 0;=0A=
>   }=0A=
>=0A=
> The problem is, that we end up being interrupted, supposedly dequeue=0A=
> the request, and exit.=0A=
>=0A=
> But then (or in parallel) the irq triggers and we try calling=0A=
> complete() on the context pointer which points to now random stack=0A=
> space, which results in the panic.=0A=
>=0A=
> It seems like something is wrong with usb_ep_dequeue not really=0A=
> stopping the irq from happening?=0A=
>=0A=
> If I revert all the changes to dwc3 back to 4.20, I don't see the issue.=
=0A=
>=0A=
> I'll do some bisection to try to narrow things down, but I wanted to=0A=
> see if this was a known issue or if anyone had immediate ideas as to=0A=
> what might be wrong.=0A=
>=0A=
=0A=
I'm not sure if this is related, but can you try to test using Felipe's=0A=
testing/next branch? There is a fix to a race condition when the gadget=0A=
driver tries to dequeue requests.=0A=
=0A=
See if you run into this issue again.=0A=
=0A=
Thanks,=0A=
Thinh=0A=