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;
From:   =?utf-8?B?7KCV7J6s7ZuI?= <jh0801.jung@samsung.com>
To:     "'Linyu Yuan'" <quic_linyyuan@quicinc.com>,
        "'Felipe Balbi'" <balbi@kernel.org>,
        "'Greg Kroah-Hartman'" <gregkh@linuxfoundation.org>,
        "'Thinh Nguyen'" <Thinh.Nguyen@synopsys.com>
Cc:     "'open list:USB XHCI DRIVER'" <linux-usb@vger.kernel.org>,
        "'open list'" <linux-kernel@vger.kernel.org>,
        "'Seungchull Suh'" <sc.suh@samsung.com>,
        "'Daehwan Jung'" <dh10.jung@samsung.com>
In-Reply-To: <c4e01a0a-1c98-3103-2b91-2fe0ba8c3118@quicinc.com>
Subject: RE: [PATCH] usb: dwc3: Clear DWC3_EVENT_PENDING when count is 0
Date:   Thu, 5 Jan 2023 18:54:55 +0900
Message-ID: <000201d920eb$c3715c50$4a5414f0$@samsung.com>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Thread-Index: AQHoaYWzl+91vPuYA9QqPKuxNdj5mgIxEgVOAi/I+n8Cu+PL3q44GoEQ
Content-Language: ko
Content-Type: text/plain; charset="utf-8"
CMS-TYPE: 102P
DLP-Filter: Pass
References: <CGME20230102050839epcas2p4b9d09d926f9a14c3b8e8df2574d334c3@epcas2p4.samsung.com>
        <20230102050831.105499-1-jh0801.jung@samsung.com>
        <af32e2f5-de45-38af-2b4f-47c1ac21ea9e@quicinc.com>
        <c4e01a0a-1c98-3103-2b91-2fe0ba8c3118@quicinc.com>
Precedence: bulk


> -----Original Message-----
> From: Linyu Yuan =5Bmailto:quic_linyyuan=40quicinc.com=5D
> Sent: Thursday, January 5, 2023 12:35 PM
> To: JaeHun Jung; Felipe Balbi; Greg Kroah-Hartman; Thinh Nguyen
> Cc: open list:USB XHCI DRIVER; open list; Seungchull Suh; Daehwan Jung
> Subject: Re: =5BPATCH=5D usb: dwc3: Clear DWC3_EVENT_PENDING when count i=
s 0
>=20
>=20
> On 1/5/2023 11:29 AM, Linyu Yuan wrote:
> > On 1/2/2023 1:08 PM, JaeHun Jung wrote:
> >> Sometimes very rarely, The count is 0 and the DWC3 flag is set has
> >> status.
> >> It must not have these status. Because, It can make happen interrupt
> >> storming status.
> >
> > could you help explain without clear the flag, how interrupt storming
> > happen ?
> >
> > as your change didn't touch any hardware register, i don't know how it
> > fix storming.
> >
H/W interrupts are still occur on IP.
But. ISR doesn't handle it. Because of this
=22if (evt->flags & DWC3_EVENT_PENDING)=22

This is event values.
(struct dwc3_event_buffer *) ev_buf =3D 0xFFFFFF883DBF1180 =3D kernel_size_=
le_lo32+0xFFFFFF883B391180 -> (
        (void *) buf =3D 0xFFFFFFC00DBDD000 =3D end+0x337D000 -> ,
        (void *) cache =3D 0xFFFFFF8839F54080 =3D kernel_size_le_lo32+0xFFF=
FFF88376F4080 -> ,
        (unsigned int) length =3D 0x1000,
        (unsigned int) lpos =3D 0x0,
        (unsigned int) count =3D 0x0,
        (unsigned int) flags =3D 0x00100001,
        (dma_addr_t) dma =3D 0x00000008BD7D7000,

*((struct dwc3_event_type *) 0xFFFFFF8839F54080) =3D (
    is_devspec =3D 1,
    type =3D 0,
    reserved8_31 =3D 774)

*((struct dwc3_event_devt  *) 0xFFFFFF8839F54080) =3D (
    one_bit =3D 1,
    device_event =3D 0,
    type =3D 6, << DWC3_DEVICE_EVENT_SUSPEND
    reserved15_12 =3D 0,
    event_info =3D 3,
    reserved31_25 =3D 0)

Occurred interrupts are =22DWC3_DEVICE_EVENT_SUSPEND=22.
If when =22count has 0=22 and DWC3_EVENT_PENDING are set at the same time t=
han
 ISR and bottom thread couldn't handle next interrupt.
We guessing connected with =22dwc3_gadget_process_pending_events()=22 funct=
ion.
But, We could not find reproduce way.


> > storming from software layer ?
> >
> >> So, It have to clean up DWC3_EVENT_PENDING flags set when count is 0.
> >> It means =22There are no interrupts to handle.=22.
> >>
> >> (struct dwc3_event_buffer *) ev_buf =3D 0xFFFFFF883DBF1180 (
> >>     (void *) buf =3D 0xFFFFFFC00DBDD000 =3D end+0x337D000,
> >>     (void *) cache =3D 0xFFFFFF8839F54080,
> >>     (unsigned int) length =3D 0x1000,
> >>     (unsigned int) lpos =3D 0x0,
> >>     (unsigned int) count =3D 0x0,
> >>     (unsigned int) flags =3D 0x00000001,
> >>     (dma_addr_t) dma =3D 0x00000008BD7D7000,
> >>     (struct dwc3 *) dwc =3D 0xFFFFFF8839CBC880,
> >>     (u64) android_kabi_reserved1 =3D 0x0),
> >>
> >> (time =3D 47557628930999, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 1), (time =3D 47557628931268, irq =3D 165, fn =3D dwc3_interrup=
t,
> >> latency =3D 0, en =3D 3), (time =3D 47557628932383, irq =3D 165, fn =
=3D
> >> dwc3_interrupt, latency =3D 0, en =3D 1), (time =3D 47557628932652, ir=
q =3D
> >> 165, fn =3D dwc3_interrupt, latency =3D 0, en =3D 3), (time =3D
> >> 47557628933768, irq =3D 165, fn =3D dwc3_interrupt, latency =3D 0, en =
=3D 1),
> >> (time =3D 47557628934037, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 3), (time =3D 47557628935152, irq =3D 165, fn =3D dwc3_interrup=
t,
> >> latency =3D 0, en =3D 1), (time =3D 47557628935460, irq =3D 165, fn =
=3D
> >> dwc3_interrupt, latency =3D 0, en =3D 3), (time =3D 47557628936575, ir=
q =3D
> >> 165, fn =3D dwc3_interrupt, latency =3D 0, en =3D 1), (time =3D
> >> 47557628936845, irq =3D 165, fn =3D dwc3_interrupt, latency =3D 0, en =
=3D 3),
> >> (time =3D 47557628937960, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 1), (time =3D 47557628938229, irq =3D 165, fn =3D dwc3_interrup=
t,
> >> latency =3D 0, en =3D 3), (time =3D 47557628939345, irq =3D 165, fn =
=3D
> >> dwc3_interrupt, latency =3D 0, en =3D 1), (time =3D 47557628939652, ir=
q =3D
> >> 165, fn =3D dwc3_interrupt, latency =3D 0, en =3D 3), (time =3D
> >> 47557628940768, irq =3D 165, fn =3D dwc3_interrupt, latency =3D 0, en =
=3D 1),
> >> (time =3D 47557628941037, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 3), (time =3D 47557628942152, irq =3D 165, fn =3D dwc3_interrup=
t,
> >> latency =3D 0, en =3D 1), (time =3D 47557628942422, irq =3D 165, fn =
=3D
> >> dwc3_interrupt, latency =3D 0, en =3D 3), (time =3D 47557628943537, ir=
q =3D
> >> 165, fn =3D dwc3_interrupt, latency =3D 0, en =3D 1), (time =3D
> >> 47557628943806, irq =3D 165, fn =3D dwc3_interrupt, latency =3D 0, en =
=3D 3),
> >> (time =3D 47557628944922, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 1), (time =3D 47557628945229, irq =3D 165, fn =3D dwc3_interrup=
t,
> >> latency =3D 0, en =3D 3), (time =3D 47557628946345, irq =3D 165, fn =
=3D
> >> dwc3_interrupt, latency =3D 0, en =3D 1), (time =3D 47557628946614, ir=
q =3D
> >> 165, fn =3D dwc3_interrupt, latency =3D 0, en =3D 3), (time =3D
> >> 47557628947729, irq =3D 165, fn =3D dwc3_interrupt, latency =3D 0, en =
=3D 1),
> >> (time =3D 47557628947999, irq =3D 165, fn =3D dwc3_interrupt, latency =
=3D 0,
> >> en =3D 3),
> >>
> >> Signed-off-by: JaeHun Jung <jh0801.jung=40samsung.com>
> >> ---
> >>   drivers/usb/dwc3/gadget.c =7C 5 ++++-
> >>   1 file changed, 4 insertions(+), 1 deletion(-) diff --git
> >> a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c index
> >> 789976567f9f..5d2d5a9b9915 100644
> >> --- a/drivers/usb/dwc3/gadget.c
> >> +++ b/drivers/usb/dwc3/gadget.c
> >> =40=40 -4355,8 +4355,11 =40=40 static irqreturn_t dwc3_check_event_buf=
(struct
> >> dwc3_event_buffer *evt)
> >>        * irq event handler completes before caching new event to
> >> prevent
> >>        * losing events.
> >>        */
> >> -    if (evt->flags & DWC3_EVENT_PENDING)
> >> +    if (evt->flags & DWC3_EVENT_PENDING) =7B
> >> +        if (=21evt->count)
> >> +            evt->flags &=3D =7EDWC3_EVENT_PENDING;
> >>           return IRQ_HANDLED;
> >> +    =7D
> >>         count =3D dwc3_readl(dwc->regs, DWC3_GEVNTCOUNT(0));
> >>       count &=3D DWC3_GEVNTCOUNT_MASK;
> >
> > how about below change ?
> >
> >         count =3D dwc3_readl(dwc->regs, DWC3_GEVNTCOUNT(0));
> >         count &=3D DWC3_GEVNTCOUNT_MASK;
> > -       if (=21count)
> > +       if (=21count) =7B
> > +               evt->flags &=3D =7EDWC3_EVENT_PENDING;
> >                 return IRQ_NONE;
> > +       =7D
> ignore my suggestion, add Thinh to comment.
> >
> >         evt->count =3D count;
> >         evt->flags =7C=3D DWC3_EVENT_PENDING;
> >