linux-kernel - Re: [PATCH] usb: dwc3: xilinx: Prevent spike in reset signal

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <a8be82ee-b6c8-4308-bf3a-d18050dd043a@amd.com>
Date: Tue, 18 Mar 2025 08:36:22 +0100
From: Michal Simek <michal.simek@....com>
To: Mike Looijmans <mike.looijmans@...ic.nl>,
 Thinh Nguyen <Thinh.Nguyen@...opsys.com>,
 Radhey Shyam Pandey <radhey.shyam.pandey@....com>
Cc: "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
 Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
 "linux-arm-kernel@...ts.infradead.org"
 <linux-arm-kernel@...ts.infradead.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] usb: dwc3: xilinx: Prevent spike in reset signal

+Radhey

On 3/18/25 07:21, Mike Looijmans wrote:
> On 18-03-2025 01:12, Thinh Nguyen wrote:
>> On Mon, Mar 17, 2025, Mike Looijmans wrote:
>>> On 14-03-2025 22:14, Thinh Nguyen wrote:
>>>> On Fri, Mar 14, 2025, Mike Looijmans wrote:
>>>>> Set the gpio to "high" on acquisition, instead of acquiring it in low
>>>>> state and then immediately setting it high again. This prevents a
>>>>> short "spike" on the reset signal.
>>>> How does this affect the current programming flow beside preventing a
>>>> spike to the reset signal?
>>> I don't understand your question. What "programming flow" are you referring
>>> to?
>> It's not obvious to me if this is an error in Xilinx documentation, the
>> driver issue, or whether this is found through experiment. Since I don't
>> have the info of this platform, it would help to know where the source
>> of error is so we can document this in the code or change-log.
> 
> It's a bug in the driver, found through code inspection.
> 
> The reset GPIO here is to control the reset signal to an external, usually ULPI 
> PHY, chip. This external chip is not part of the Xilinx hardware.
> 
>>> The reset sequence was just plain wrong, the issue is almost the same as the
>> Do we need a fix tag and add to stable then?
> 
> That would be appropriate I think.
> 
> 
>>
>>> one described in this commit:
>>> e0183b974d30 "net: mdiobus: Prevent spike on MDIO bus reset signal"
>>>
>>> Note that I see this high-low-high-low double reset toggle in many Xilinx
>>> software drivers, they seem to teach that at the Xilinx academy or so.
>>>
>>>
>>>>> Signed-off-by: Mike Looijmans <mike.looijmans@...ic.nl>
>>>>> ---
>>>>>
>>>>>    drivers/usb/dwc3/dwc3-xilinx.c | 3 +--
>>>>>    1 file changed, 1 insertion(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/drivers/usb/dwc3/dwc3-xilinx.c b/drivers/usb/dwc3/dwc3-xilinx.c
>>>>> index a33a42ba0249..a159a511483b 100644
>>>>> --- a/drivers/usb/dwc3/dwc3-xilinx.c
>>>>> +++ b/drivers/usb/dwc3/dwc3-xilinx.c
>>>>> @@ -207,7 +207,7 @@ static int dwc3_xlnx_init_zynqmp(struct dwc3_xlnx 
>>>>> *priv_data)
>>>>>    skip_usb3_phy:
>>>>>        /* ulpi reset via gpio-modepin or gpio-framework driver */
>>>>> -    reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
>>>>> +    reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_HIGH);
>>>>>        if (IS_ERR(reset_gpio)) {
>>>>>            return dev_err_probe(dev, PTR_ERR(reset_gpio),
>>>>>                         "Failed to request reset GPIO\n");
>>>>> @@ -215,7 +215,6 @@ static int dwc3_xlnx_init_zynqmp(struct dwc3_xlnx 
>>>>> *priv_data)
>>>>>        if (reset_gpio) {
>>>>>            /* Toggle ulpi to reset the phy. */
>>>> Does the comment above still apply?
>>> Now you mention it, the comment never made any sense anyway.
>>>
>> Then can we remove it?
> 
> Removing would be better, yes. I'll make a v2 patch.
> 
> 
>>>>> -        gpiod_set_value_cansleep(reset_gpio, 1);
>>>>>            usleep_range(5000, 10000);
>>>> Do we still need this usleep_range here?
>>> Yes, this is the "reset active" time.
>>>
>> But why do we need 2 calls to usleep_range? From just looking at this
>> here, it appears that the first was intended for the removed
>> gpiod_set_value_cansleep(reset_gpio, 1). If this "reset active" time is
>> needed irrespective of the existent reset_gpio, then shouldn't it be set
>> outside of this if statement?
> 
> It helps to see the whole thing instead of just the patch.
> 
> If I omit error handling and comments then the original code reads:
> 
>          reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
>          if (reset_gpio) {
>                  gpiod_set_value_cansleep(reset_gpio, 1);
>                  usleep_range(5000, 10000);
>                  gpiod_set_value_cansleep(reset_gpio, 0);
>                  usleep_range(5000, 10000);
>          }
> 
> So the gpio is acquired in a LOW state and then, without delay, is set to a high 
> state again. This causes the "spike" I'm mentioning here. The correct procedure 
> is to acquire it in the HIGH state immediately, so the sequence becomes:
> 
>          reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_HIGH);
>          if (reset_gpio) {
>                  usleep_range(5000, 10000);
>                  gpiod_set_value_cansleep(reset_gpio, 0);
>                  usleep_range(5000, 10000);
>          }
> 
> This patch does exactly that.

Please keep Radhey in loop. He will take a look at it from our side.

Thanks,
Michal