Skip to contentSkip to conversation

Best way of making some drivers common across kirkwood and orion5x SoCs?

U-Boot Prafulla Wadaskar 2 participants 3 messages Nov 9, 2009 – Nov 10, 2009 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

Dear Albert

Quoted 49 lines

-----Original Message-----
From: Albert ARIBAUD [mailto:albert.aribaud at free.fr]
Sent: Monday, November 02, 2009 5:46 PM
To: Prafulla Wadaskar
Subject: Re: [U-Boot] Best way of making some drivers common across kirkwood and orion5x SoCs?

Prafulla Wadaskar a ?crit :

Quoted 4 lines
Quoted 3 lines

Still, the drivers would be full of 'KWxxx" and "kwxxx" symbols of which many are not kirkwood-specific actually.

Quoted 8 lines

Any way, those are not even orion specific nor Marvell specific. Those are related to the functionality supported by SOCs that may be customized by each SOC

Quoted 2 lines

In order for these drivers to compile with an orion5x SoC, I would have to adopt kirkwood names in the

What harm in this?

I would say it harms maintenability and reuseability of the code. If those drivers are neither kirkwood- nor even marvell-specific, then they
should not lead readers to believe they are, otherwise people might not
realize they can ruse these drivers in their own SoCs.

Quoted 8 lines
Quoted 2 lines

orion5x code, which I don't like as much as I would like fixing the

If you see kernel code even there you can see kirkwood call from Orion drivers and vice versa.

Quoted 1 line

symbols to make them marvell-, not kirkwood-, specific.

This will not solve the root problem, what about some non

marvell SoC

Quoted 3 lines

have same h/w and want to reuse the code? Do we again change the suffixes? (kirkwood re-used external serial driver adopting external definitions).

Point valid and taken--see my suggestion below.

Quoted 1 line

I suggest here to adopt kw symbols in Orion. This would

make it clear

Quoted 1 line

for anybody that kirkwood code is reused by orion. With

this kirkwood

Quoted 1 line

drives will be untouched.

Quoted 9 lines

This does not solve the root problem any better than switching to Marvell prefixes, as you rightly point out. I thus suggest removing any
Marvell, Kirkwood or Orion prefixes from symbols in these drivers altogether. For instance, egiga symbols would take EGIGA_ as a prefix.
Then each SoC (kirkwood, orion5x, any other) header file would provide
adequate definitions (#define EGIGA_xxxx yyyy).

I understands your concerns-
1. For SOC support Orion should be implemented similar to Kirkwood (you agreed on this)
2. For common drivers for (ex: kirkwood_egiga.c) the common code must be abstracted to egiga_core.c and architecture specific code must be supported through kirkwood_egiga.c/h and orion_egiga.c/h, this way it would make it more clean and structured.

Quoted 7 lines
Quoted 4 lines

While doing this reuse activity, if we find something blocking, certainly we should address that, but let's avoid changing identity of drivers.

I personally believe that having the Orion5x SoC dependent of the kirkwood one is blocking enough from a design standpoint.

What is your opinion with #2 suggestion?

Regards..
Prafulla . .

Quoted 7 lines
Quoted 2 lines

Regards.. Prafulla . .

Amicalement,

Signature · 1 line
Albert.
#2 lore raw

Hi PRafulla,

Prafulla Wadaskar a ?crit :

Quoted 2 lines

1. For SOC support Orion should be implemented similar to Kirkwood (you agreed on this)

I still do. :)

Quoted 4 lines

2. For common drivers for (ex: kirkwood_egiga.c) the common code must be abstracted to egiga_core.c and architecture specific code must be supported through kirkwood_egiga.c/h and orion_egiga.c/h, this way it would make it more clean and structured.

Agreed.

Quoted 1 line

What is your opinion with #2 suggestion?

Sounds fine! I'll start submitting patches for both common drivers and orion support based on this proposal.

Regarding the arch-specific parts of current driver code, should they end up in cpu/arm926ejs/{kirkwood,orion5x} or stay in drivers/*? I tend like cpu/arm926ejs/* better as it puts everything arch-specific in a single place.

Amicalement,

Signature · 1 line
Albert.
#3 lore raw
Quoted 35 lines

-----Original Message-----
From: Albert ARIBAUD [mailto:albert.aribaud at free.fr]
Sent: Tuesday, November 10, 2009 12:27 AM
To: Prafulla Wadaskar
Cc: u-boot at lists.denx.de; Ashish Karkare; Prabhanjan Sarnaik
Subject: Re: [U-Boot] Best way of making some drivers common across kirkwood and orion5x SoCs?

Hi PRafulla,

Prafulla Wadaskar a ?crit :

Quoted 2 lines

1. For SOC support Orion should be implemented similar to Kirkwood (you agreed on this)

I still do. :)

Quoted 1 line

2. For common drivers for (ex: kirkwood_egiga.c) the

common code must

Quoted 1 line

be abstracted to egiga_core.c and architecture specific

code must be

Quoted 1 line

supported through kirkwood_egiga.c/h and orion_egiga.c/h,

this way it

Quoted 1 line

would make it more clean and structured.

Agreed.

Quoted 1 line

What is your opinion with #2 suggestion?

Sounds fine! I'll start submitting patches for both common drivers and
orion support based on this proposal.

Quoted 5 lines

Regarding the arch-specific parts of current driver code, should they end up in cpu/arm926ejs/{kirkwood,orion5x} or stay in drivers/*? I tend
like cpu/arm926ejs/* better as it puts everything arch-specific in a single place.

I agree for this. Good luck :-)

Regards...
Prafulla . . .

Quoted 5 lines

Amicalement,

Signature · 1 line
Albert.