linux-kernel - Re: [PATCH v2] usb: core: reduce power-on-good delay time of root hub

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] [thread-next>] [day] [month] [year] [list]

Message-ID: <13d55059-9f66-8599-54fc-46698bae41d1@linux.intel.com>
Date:   Wed, 3 Nov 2021 22:37:33 +0200
From:   Mathias Nyman <mathias.nyman@...ux.intel.com>
To:     "Walt Jr. Brake" <mr.yming81@...il.com>,
        Alan Stern <stern@...land.harvard.edu>,
        Mathias Nyman <mathias.nyman@...el.com>
Cc:     Chunfeng Yun <chunfeng.yun@...iatek.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Matthias Brugger <matthias.bgg@...il.com>,
        Nishad Kamdar <nishadkamdar@...il.com>,
        linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org,
        linux-mediatek@...ts.infradead.org,
        Eddie Hung <eddie.hung@...iatek.com>
Subject: Re: [PATCH v2] usb: core: reduce power-on-good delay time of root hub

On 2.11.2021 22.29, Walt Jr. Brake wrote:
> On 2/11/2021 17:05, Mathias Nyman wrote:
>> On 1.11.2021 16.06, Alan Stern wrote:
>>> On Sat, Oct 30, 2021 at 12:49:37PM +0800, Walt Jr. Brake wrote:
>>>> This patch make USB 3.1 device cannot be detected, and I report the bug [1]
>>>> to archlinux three month ago. Yesterday I try to fix it myself, and after I
>>>> revert this patch, compile the kernel and test, it works.
>>>>
>>>> [1] https://bugs.archlinux.org/task/71660?project=1&pagenum=2
>>>>
>>>>
>>>> diff --git a/drivers/usb/core/hub.h b/drivers/usb/core/hub.h
>>>> index 22ea1f4f2d66..73f4482d833a 100644
>>>> --- a/drivers/usb/core/hub.h
>>>> +++ b/drivers/usb/core/hub.h
>>>> @@ -148,10 +148,8 @@ static inline unsigned hub_power_on_good_delay(struct
>>>> usb_hub *hub)
>>>>   {
>>>>          unsigned delay = hub->descriptor->bPwrOn2PwrGood * 2;
>>>>
>>>> -       if (!hub->hdev->parent) /* root hub */
>>>> -               return delay;
>>>> -       else /* Wait at least 100 msec for power to become stable */
>>>> -               return max(delay, 100U);
>>>> +       /* Wait at least 100 msec for power to become stable */
>>>> +       return max(delay, 100U);
>>>>   }
>>> Mathias:
>>>
>>> It looks like the bPwrOn2PwrGood value in xhci-hcd's hub descriptor is
>>> too small for some USB 3.1 devices.
>>>
>>> Can you look into this?
>>>
>>> Alan Stern
>>>
>> At first glance the xhci roothub bPwrOn2PwrGood value looks ok.
>> xhci spec 5.4.8 states software should wait 20ms after asserting PP, before
>> attempting to change the state of the port.
>>
>> xhci driver sets desc->bPwrOn2PwrGood = 10; (2ms interval, so equals 20ms )
>>
>> We should probably get this working immediately, so maybe revert that patch
>> while looking into the rootcause.
>>
>> Walt Jr. Brake, instead of reverting that patch, could you test if changing the
>> xhci roothub bPwrOn2PwrGood value helps.
>>
>> diff --git a/drivers/usb/host/xhci-hub.c b/drivers/usb/host/xhci-hub.c
>> index a3f875eea751..756231a55602 100644
>> --- a/drivers/usb/host/xhci-hub.c
>> +++ b/drivers/usb/host/xhci-hub.c
>> @@ -257,7 +257,7 @@ static void xhci_common_hub_descriptor(struct xhci_hcd *xhci,
>>   {
>>          u16 temp;
>>   -       desc->bPwrOn2PwrGood = 10;      /* xhci section 5.4.9 says 20ms max */
>> +       desc->bPwrOn2PwrGood = 50;      /* The 20ms in xhci 5.4.8 isn't enough for USB 3.1 */
>>          desc->bHubContrCurrent = 0;
>>            desc->bNbrPorts = ports;
>>
>> Thanks
>> -Mathias
> 
> Mathias:
> 
> Sorry to reply lately. I test with your patch, it works.
> 
> I also test with setting bPwrOn2PwrGood to 45, and it not work.
> 
> Seems that the minimal value should be 50 for this case.
> 

Thanks for testing, and for checking that 90ms wait isn't enough

-Mathias