Re: [PATCH] staging: Add rtl8821ce PCIe WiFi driver

[Greg KH]

On Wed, May 15, 2019 at 09:06:44PM +0800, Kai-Heng Feng wrote:
> at 20:33, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> 
> > On Wed, May 15, 2019 at 07:54:58PM +0800, Kai-Heng Feng wrote:
> > > at 19:40, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > > 
> > > > On Wed, May 15, 2019 at 07:24:01PM +0800, Kai-Heng Feng wrote:
> > > > > The rtl8821ce can be found on many HP and Lenovo laptops.
> > > > > Users have been using out-of-tree module for a while,
> > > > > 
> > > > > The new Realtek WiFi driver, rtw88, will support rtl8821ce in 2020 or
> > > > > later.
> > > > 
> > > > Where is that driver, and why is it going to take so long to get merged?
> > > 
> > > rtw88 is in 5.2 now, but it doesnât support 8821ce yet.
> > > 
> > > They plan to add the support in 2020.
> > 
> > Who is "they" and what is needed to support this device and why wait a
> > full year?
> 
> âTheyâ refers to Realtek.
> Itâs their plan so I canât really answer that on behalf of Realtek.

Where did they say that?  Any reason their developers are not on this
patch?

> > > > > 296 files changed, 206166 insertions(+)
> > > > 
> > > > Ugh, why do we keep having to add the whole mess for every single one of
> > > > these devices?
> > > 
> > > Because Realtek devices are unfortunately ubiquitous so the support is
> > > better come from kernel.
> > 
> > That's not the issue here.  The issue is that we keep adding the same
> > huge driver files to the kernel tree, over and over, with no real change
> > at all.  We have seen almost all of these files in other realtek
> > drivers, right?
> 
> Yes. They use one single driver to support different SoCs, different
> architectures and even different OSes.

Well, they try to, it doesn't always work :(

> Thatâs why itâs a mess.

Oh we all know why this is a mess.  But they have been saying for
_years_ they would clean up this mess.  So push back, I'm not going to
take another 200k lines for a simple wifi driver, again.

Along those lines, we should probably just delete the other old realtek
drivers that don't seem to be going anywhere from staging as well,
because those are just confusing people.

> > Why not use the ones we already have?
> 
> Itâs virtually impossible because Realtekâs mega wifi driver uses tons of
> #ifdefs, only one chip can be selected to be supported at compile time.

That's not what I asked.

I want to know why they can't just add support for their new devices to
one of the many existing realtek drivers we already have.  That is the
simpler way, and the correct way to do this.  We don't do this by adding
200k lines, again.

> > But better yet, why not add proper support for this hardware and not use
> > a staging driver?
> 
> Realtek plans to add the support in 2020, if everything goes well.

Device "goes well" please.  And when in 2020?  And why 2020?  Why not
2022?  2024?

> Meanwhile, many users of HP and Lenovo laptops are using out-of-tree driver,
> some of them are stuck to older kernels because they donât know how to fix
> the driver. So I strongly think having this in kernel is beneficial to many
> users, even itâs only for a year.

So who is going to be responsible for "fixing the driver" for all new
kernel api updates?  I'm tired of seeing new developers get lost in the
maze of yet-another realtek wifi driver.  We've been putting up with
this crud for years, and it has not gotten any better if you want to add
another 200k lines for some unknown amount of time with the hope that a
driver might magically show up one day.

thanks,

greg k-h