Re: [RFC PATCH] net: stmicro: eQOS IP Core

Re: [RFC PATCH] net: stmicro: eQOS IP Core

[Joao Pinto]

Às 4:20 PM de 1/20/2017, David Miller escreveu:
> From: Joao Pinto <Joao.Pinto@synopsys.com>
> Date: Fri, 20 Jan 2017 12:27:33 +0000
> 
>> For historical reasons the Designware eQOS IP Core started in version 4.x,
>> instead of starting in 1.x. This caused some misunderstanding, which resulted
>> in calling this IP as GMAC4. If you go to Synopsys website and check the
>> Ethernet portfolio there is no GMAC4 IP:
>> https://www.synopsys.com/designware-ip/interface-ip/ethernet.html
>>
>> There is a new version of eQOS comming out (5.x) that will have the same
>> mac, dma registers and flow, with an extra set of registers to enable
>> features like TSN. The creation of a dwmac5 or gmac5 would be wrong, and
>> including those new features related to eQOS 5.x in dwmac4 would also not
>> be very correct.
>>
>> This patch suggests to rename dwmac4_* driver to eqos_*, turning it the
>> official driver for eQOS IP. The registers definitions are also now
>> called QOS_ instead of GMAC4_.
>>
>> For more information about eQOS IP please check:
>> https://www.synopsys.com/dw/ipdir.php?ds=dwc_ether_qos
>>
>> Signed-off-by: Joao Pinto <jpinto@synopsys.com>
> 
> Sorry, you're not renaming every function in the driver that meets this
> qualification.
> 
> I understand you want to work hard on the driver, but part of that is
> _accepting_ things how they have been for years and making code
> improvements and additions rather than messing around playing
> rename-this/rename-that games until it looks how you want it to.
> 
> Please stop this charade of renaming, I'm getting really tired of it and
> I think you really have no concept what kind of effect it has on people
> who have to review, apply, and backport this code.
> 
> Kconfig text changes and clarifications, that's fine.  But all of these
> function and variable name changes to the driver, and comment sweeps, no
> it is not acceptable to me.
> 

Hello David.

Charade of renames? It is a bit strong, don't you agree?

So you suggest to keep the driver called as dwmac4 when the IP is in fact eQOS?
Where do you suggest to put the eQOS 5.x features?

Thanks,
Joao

Re: [RFC PATCH] net: stmicro: eQOS IP Core

[David Miller]

From: Joao Pinto <Joao.Pinto@synopsys.com>
Date: Fri, 20 Jan 2017 16:30:41 +0000

> Charade of renames? It is a bit strong, don't you agree?

Not at all.

> So you suggest to keep the driver called as dwmac4 when the IP is in fact eQOS?
> Where do you suggest to put the eQOS 5.x features?

Since you are adding the new code, you can name the functions that implement
eQOS 5.x support however you like.  You don't have to rename the dwmac4 code
in order to add support for new chip families.

Re: [RFC PATCH] net: stmicro: eQOS IP Core

[Joao Pinto]

Às 4:34 PM de 1/20/2017, David Miller escreveu:
> From: Joao Pinto <Joao.Pinto@synopsys.com>
> Date: Fri, 20 Jan 2017 16:30:41 +0000
> 
>> Charade of renames? It is a bit strong, don't you agree?
> 
> Not at all.

To avoid this kind of "charade" comments, I sent an RFC, not an official patch
to be applied. Just a suggestion to be analysed by network/ developers.

> 
>> So you suggest to keep the driver called as dwmac4 when the IP is in fact eQOS?
>> Where do you suggest to put the eQOS 5.x features?
> 
> Since you are adding the new code, you can name the functions that implement
> eQOS 5.x support however you like.  You don't have to rename the dwmac4 code
> in order to add support for new chip families.
> 

The rename suggestion of the dwmac4_* files was because it is stating that they
are for a GMAC4 IP Core that does not exist. The controller is eQOS that started
in version 4.x and now is going to release a 5.x. My sugestion intention is to
put things clear and well structured to everyone that wish to use the driver in
the future.

I understand that you maintain a busy subsystem, and backport is a pain, but it
shouldn't avoid you improving its structure. You have duplicated drivers
targeting the same IPs and I just trying to puts things clear to avoid that in
the future.

I am going to focus developing the new features for the eQOS IP, and if sometime
you think about organizing or change the backport workflow I am available to
help you.

Joao Pinto