Re: [PATCH] mmc: sdhci-pci-gli: Set SDR104's clock to 205MHz and enable SSC for GL975x

[Greg KH <gregkh@linuxfoundation.org>]

On Wed, Oct 21, 2020 at 11:27:42AM +0800, Ben Chuang wrote:
> Hi Greg,
> 
> On Thu, Oct 15, 2020 at 9:17 AM Ben Chuang <benchuanggli@gmail.com> wrote:
> >
> > Hi, Greg,
> >
> > On Wed, Oct 14, 2020 at 7:37 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > >
> > > On Wed, Oct 14, 2020 at 06:42:06PM +0800, Ben Chuang wrote:
> > > > Hi Greg,
> > > >
> > > > On Wed, Oct 14, 2020 at 4:10 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > > > >
> > > > > On Wed, Oct 14, 2020 at 04:00:49PM +0800, Ben Chuang wrote:
> > > > > > Hi Greg,
> > > > > >
> > > > > > On Tue, Oct 13, 2020 at 9:48 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > > > > > >
> > > > > > > On Tue, Oct 13, 2020 at 07:11:13PM +0800, Ben Chuang wrote:
> > > > > > > > On Tue, Oct 13, 2020 at 4:57 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > > > > > > > >
> > > > > > > > > On Tue, Oct 13, 2020 at 04:33:38PM +0800, Ben Chuang wrote:
> > > > > > > > > > On Tue, Oct 13, 2020 at 4:00 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> > > > > > > > > > >
> > > > > > > > > > > On Tue, Oct 13, 2020 at 03:46:00PM +0800, Ben Chuang wrote:
> > > > > > > > > > > > From: Ben Chuang <ben.chuang@genesyslogic.com.tw>
> > > > > > > > > > > >
> > > > > > > > > > > > commit 786d33c887e15061ff95942db68fe5c6ca98e5fc upstream.
> > > > > > > > > > > >
> > > > > > > > > > > > Set SDR104's clock to 205MHz and enable SSC for GL9750 and GL9755
> > > > > > > > > > > >
> > > > > > > > > > > > Signed-off-by: Ben Chuang <ben.chuang@genesyslogic.com.tw>
> > > > > > > > > > > > Link: https://lore.kernel.org/r/20200717033350.13006-1-benchuanggli@gmail.com
> > > > > > > > > > > > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > > > > > > > > > > Cc: <stable@vger.kernel.org> # 5.4.x
> > > > > > > > > > > > ---
> > > > > > > > > > > > Hi Greg and Sasha,
> > > > > > > > > > > >
> > > > > > > > > > > > The patch is to improve the EMI of the hardware.
> > > > > > > > > > > > So it should be also required for some hardware devices using the v5.4.
> > > > > > > > > > > > Please tell me if have other questions.
> > > > > > > > > > >
> > > > > > > > > > > This looks like a "add support for new hardware" type of patch, right?
> > > > > > > > > >
> > > > > > > > > > No, this is for a mass production hardware.
> > > > > > > > >
> > > > > > > > > That does not make sense, sorry.
> > > > > > > > >
> > > > > > > > > Is this a bug that is being fixed, did the hardware work properly before
> > > > > > > > > 5.4 and now it does not?  Or has it never worked properly and 5.9 is the
> > > > > > > > > first kernel that it now works on?
> > > > > > > >
> > > > > > > > It seems there is misunderstanding regarding “hardware” means.
> > > > > > > > I originally thought that the "hardware" refers to GL975x chips.
> > > > > > > >
> > > > > > > > This Genesys patch is to fix the EMI problem for GL975x controller on a system.
> > > > > > >
> > > > > > > Did it work on the 4.19 kernel?  Another older kernel?  Or is 5.9 the
> > > > > > > first kernel release where it works?
> > > > > >
> > > > > > The patch works on after v5.4.
> > > > >
> > > > > You are not answering the question I am trying to ask.
> > > > >
> > > > > My question is:
> > > > >         Did this hardware ever work properly before the 5.9 kernel
> > > > >         release.
> > > >
> > > > Yes.
> > >
> > > It did?  What kernel release from kernel.org did it work, and what
> > > kernel release did it break on?
> >
> > Sorry, I don't know if  "this Hardware" refers to a single GL975x
> > controller or the whole system.
> > "Yes", for those existing systems built in GL975x controller. These
> > systems are no EMI issue on v5.4~v5.9.
> >
> > "No", for some systems with developing stage build in GL975x
> > controller because there may be an EMI issue.
> > For those systems, they need the v5.9 to pass the EMI test.
> >
> > >
> > > > > > > In other words, is this fixing a regression, or just enabling hardware
> > > > > > > support for something that has never worked before for this hardware?
> > > > > >
> > > > > > This patch is to reduce the EMI at SDR104 mode for GL975x.
> > > > > > It changes the preset frequency of SDR104 to 205Mhz and sets the SSC value.
> > > > > > So I think it is fixing a regression.
> > > > >
> > > > > A regression is when an older kernel works fine, but a newer kernel does
> > > > > not.  When that happens, you can point to a specific commit and say,
> > > > > "this commit here broke this previously working hardware".
> > > > >
> > > > > Is that the case here?  If not, this is not a regression.
> > > >
> > > > With this definition, no.
> > >
> > > Wait, I do not understand, this is the exact oposite of what you said
> > > above.
> > >
> > > Which is true?
> >
> > These existing systems built with GL975x have no EMI problems.
> > The patch can reduce the impact of EMI, so these existing systems are still ok.
> >
> > For these existing systems, on an older kernel (v5.4) they work fine
> > (no EMI issue), on a newer kernel (v5.9) they work fine (no EMI
> > issue).
> > For those developing systems, on an older kernel (v5.4) they don't
> > work fine (EMI issue), on a newer kernel (v5.9) they work fine(no EMI
> > issue).
> >
> 
> Sorry, I forgot to consider some existing machines with pre-installed windows.
> These machines fit the definition of regression.

Again, the definition of regression, for Linux is:
	The hardware worked properly on an older version of the kernel,
	and does not work properly on a newer version of the kernel.

We can not "regress" compared to any other operating system, sorry, that
is just crazy to think so.  What would Microsoft say if you reported a
regression to them where the hardware worked properly on Linux but does
not work on Windows?

Just have your users move to the 5.9 kernel, and all should be fine.

thanks,

greg k-h