linux-kernel - RE: [EXT] Re: [PATCH v1] usb: dwc3: remove the call trace of USBx

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: <87sgpmx1pn.fsf@gmail.com>
Date:   Tue, 27 Aug 2019 14:55:00 +0300
From:   Felipe Balbi <balbi@...nel.org>
To:     Ran Wang <ran.wang_1@....com>, Yinbo Zhu <yinbo.zhu@....com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "linux-usb\@vger.kernel.org" <linux-usb@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>
Cc:     Xiaobo Xie <xiaobo.xie@....com>, Jiafei Pan <jiafei.pan@....com>
Subject: RE: [EXT] Re: [PATCH v1] usb: dwc3: remove the call trace of USBx_GFLADJ


Hi,

Ran Wang <ran.wang_1@....com> writes:
>> Yinbo Zhu <yinbo.zhu@....com> writes:
>> >> Yinbo Zhu <yinbo.zhu@....com> writes:
>> >> >> > diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
>> >> >> > index
>> >> >> > 98bce85c29d0..a133d8490322 100644
>> >> >> > --- a/drivers/usb/dwc3/core.c
>> >> >> > +++ b/drivers/usb/dwc3/core.c
>> >> >> > @@ -300,8 +300,7 @@ static void
>> >> >> > dwc3_frame_length_adjustment(struct
>> >> >> > dwc3 *dwc)
>> >> >> >
>> >> >> >       reg = dwc3_readl(dwc->regs, DWC3_GFLADJ);
>> >> >> >       dft = reg & DWC3_GFLADJ_30MHZ_MASK;
>> >> >> > -     if (!dev_WARN_ONCE(dwc->dev, dft == dwc->fladj,
>> >> >> > -         "request value same as default, ignoring\n")) {
>> >> >> > +     if (dft != dwc->fladj) {
>> >> >>
>> >> >> if the value isn't different, why do you want to change it?
>> >> >>
>> >> >> --
>> >> >> Balbi
>> >> > Hi Balbi,
>> >> >
>> >> > I don't change any value. I was remove that call trace.
>> >>
>> >> Sure you do. The splat only shows when you request a FLADJ value
>> >> that's the same as the one already in the register. The reason you
>> >> see the splat is because your requested value is what's already in the HW.
>> >>
>> >> So, again, why are you adding this device tree property if the value
>> >> is already the correct one?
>> >>
>> >> > In addition that GFLADJ_30MHZ value intial value is 0, and it's
>> >> > value must be 0x20, if not, usb will not work.
>> >>
>> >> it's not zero, otherwise the splat wouldn't trigger. You're
>> >> requesting the value that's already in your register by default.
>> >>
>> >> --
>> >> Balbi
>> >
>> > Hi Balbi,
>> >
>> > According that rm doc that GFLADJ_30MHZ has a default value is 0x20,
>> > when GFLADJ_30MHZ_REG_SEL is 0, this 0x20 is a hard-coded value.
>> >
>> > But in fact, that default value is 0, please you note!
>> >
>> > Then according that xhci spec 5.2.4, that register the sixth bit if is
>> > 0, then that can support Frame Lenth Timing value.
>> >
>> > So set GFLADJ_30MHZ_REG_SEL to 1 for use FLADJ, then I find that it
>> > must use 0x20 usb will work well, even thoug xhci can permit
>> > GFLADJ_30MHZ use other value
>> 
>> You only get the splat because you try to sent GFLADJ to 0x20 and it's ALREADY
>> 0x20. This means that you don't need the property in DTS.
>> 
>> > In addition about what you said is about dts patch, and that patch had
>> > merged by upstream, patch owner isn't me,
>> 
>> Well, then remove the setting from DTS, since clearly it's not needed.
>
> Please considering below scenarios on the same board which needs GFLADJ property on kernel DTS:
>
> 1. Board boot to U-Boot first, then load kernel. In this case, we need kernel DTS
>     help to get GFLADJ setting right, everything is as expected.
>
> 2. Board boot to U-Boot console, then execute 'usb start' under U-Boot console to init
>     DWC3 controller, then load kernel. In this case, actually GFLADJ is correctly
>     configured already, and the GFLADJ config double-checking is fine (because kernel
>     cannot know if U-Boot has initialized it or not), but warning looks not necessary.
>
> 3. Board boot to kernel, GFLADJ get set from DTS, then system suspend & resume. In this case
>     when resuming, GFLADJ setting has been restored correctly, so here we might not need
>     send out the warning message (double-checking might be fine).
>
> So, what's your suggestion to remove this looks non-necessary warning message?

now this is well explained! So the value in the register is *NOT* 0x20
by default, however, u-boot _can_ use dwc3 if we're flashing, then it'll
result in the splat.

Okay, this is a valid scenario that the kernel should consider. I agree
that we should remove the WARN() from there.

Thanks

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (833 bytes)