Skip to contentSkip to conversation

Remove ARM dependency for dm9000 driver

NET Franck 7 participants 18 messages Nov 30, 2005 – Dec 1, 2005 Network drivers

lore
Share & tools
lore

Franck

author
#0 lore raw

Hi,

What about this patch which removes ARM dependency for dm9000 ethernet controller driver ?

drivers/net/Kconfig +1 -1
index f15f909..4af63dd 100644
@@ -856,7 +856,7 @@ config SMC9194
 
 config DM9000
 	tristate "DM9000 support"
-	depends on ARM && NET_ETHERNET
+	depends on NET_ETHERNET
 	select CRC32
 	select MII
 	---help---

My platform based on MIPS cpu used it...

Thanks
--
Franck

#1 lore raw
Quoted 19 lines · Franck

Hi,

What about this patch which removes ARM dependency for dm9000 ethernet controller driver ?

drivers/net/Kconfig +1 -1
index f15f909..4af63dd 100644
@@ -856,7 +856,7 @@ config SMC9194
 
 config DM9000
 	tristate "DM9000 support"
-	depends on ARM && NET_ETHERNET
+	depends on NET_ETHERNET
 	select CRC32
 	select MII
 	---help---
Quoted 1 line

My platform based on MIPS cpu used it...

Maybe that should be (ARM || MIPS) && NET_ETHERNET ?

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
author
#2 lore raw

2005/11/30, Russell King <[email protected]>:

Quoted 8 lines · Franck
Quoted 5 lines

Hi,

What about this patch which removes ARM dependency for dm9000 ethernet controller driver ?

Maybe that should be (ARM || MIPS) && NET_ETHERNET ?

Well, if this dependency means "it has only be tested on ARM and MIPS", then you're probably right. But if it means "this controller must be run with an ARM or MIPS cpu" then I don't see why setting such restriction. What do you think ?

thanks
--
Franck

#3 lore raw
Quoted 7 lines · Franck

2005/11/30, Russell King <[email protected]>:

Quoted 6 lines · Franck
Quoted 5 lines

Hi,

What about this patch which removes ARM dependency for dm9000 ethernet controller driver ?

Quoted 7 lines
Quoted 2 lines

Maybe that should be (ARM || MIPS) && NET_ETHERNET ?

Well, if this dependency means "it has only be tested on ARM and MIPS", then you're probably right. But if it means "this controller must be run with an ARM or MIPS cpu" then I don't see why setting such restriction. What do you think ?

If other CPUs use this then fine, but I find that having config options needlessly available to all architectures is annoying - especially when they are never used.

Eg, would you ever expect to see a DM9000 ethernet device on an x86 machine? Probably not - there's far better PCI solutions now.

So until someone says "I want to use this on such and such arch" I think it's better to keep it dependent on those we know are likely to support it.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
author
#4 lore raw

2005/11/30, Russell King <[email protected]>:

Quoted 5 lines

If other CPUs use this then fine, but I find that having config options needlessly available to all architectures is annoying - especially when they are never used.

Quoted 7 lines

Eg, would you ever expect to see a DM9000 ethernet device on an x86 machine? Probably not - there's far better PCI solutions now.

So until someone says "I want to use this on such and such arch" I think it's better to keep it dependent on those we know are likely to support it.

That makes sens, specially if it's going to be used only by few arch. I'll resend a new patch.

Thanks
--
Franck

#5 lore raw
In reply to Russell King
Quoted 3 lines · Russell King

So until someone says "I want to use this on such and such arch" I think it's better to keep it dependent on those we know are likely to support it.

I disagree; unless there's a reason why it _shouldn't_ work on a given architecture, it should be possible to enable it there.

I believe I've seen FR-V boards with dm9000 too.

Signature · 1 line
dwmw2
#6 lore raw
Quoted 7 lines · David Woodhouse
Quoted 3 lines · Russell King

So until someone says "I want to use this on such and such arch" I think it's better to keep it dependent on those we know are likely to support it.

I disagree; unless there's a reason why it _shouldn't_ work on a given architecture, it should be possible to enable it there.

We agree to disagree. For example, in all probability, ARM will never see a TPM chip, yet it's offered for selection. Given that, does it really make sense to offer it for ARM?

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
#7 lore raw
Quoted 3 lines · Russell King

We agree to disagree. For example, in all probability, ARM will never see a TPM chip, yet it's offered for selection. Given that, does it really make sense to offer it for ARM?

You speak of _probability_. Yes, it makes sense to offer it as an _option_ for ARM. It just doesn't make sense to put it in the defconfig for any of the existing platforms.

Nobody expects 'allyesconfig' to be something you'd actually want to _use_.

Signature · 1 line
dwmw2
#8 lore raw
Quoted 11 lines · David Woodhouse
Quoted 3 lines · Russell King

We agree to disagree. For example, in all probability, ARM will never see a TPM chip, yet it's offered for selection. Given that, does it really make sense to offer it for ARM?

You speak of _probability_. Yes, it makes sense to offer it as an _option_ for ARM. It just doesn't make sense to put it in the defconfig for any of the existing platforms.

Nobody expects 'allyesconfig' to be something you'd actually want to _use_.

In which case why do we restrict floppy to only those machines which could have floppy? Why do we restrict IDE to only those platforms which may have IDE?

Hint: there's already a precident established for *not* offering configuration options which are pointless.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
maintainer
#9 lore raw

From: Russell King <[email protected]>
Date: Thu, 1 Dec 2005 09:41:11 +0000

Quoted 3 lines

In which case why do we restrict floppy to only those machines which could have floppy? Why do we restrict IDE to only those platforms which may have IDE?

These two examples require platform level support via an asm/*.h header file.

Whereas the driver's we are talking about use portable interfaces that should be available across the board.

So, bad example.

#10 lore raw
Quoted 7 lines · David S. Miller

From: Russell King <[email protected]>
Date: Thu, 1 Dec 2005 09:41:11 +0000

Quoted 3 lines

In which case why do we restrict floppy to only those machines which could have floppy? Why do we restrict IDE to only those platforms which may have IDE?

Quoted 7 lines

These two examples require platform level support via an asm/*.h header file.

Whereas the driver's we are talking about use portable interfaces that should be available across the board.

So, bad example.

Not in the IDE case. Bart restricted IDE to a smaller number of ARM platforms, plus any that had PCMCIA. There is no such restriction in the asm-arm/*.h header files.

if PCMCIA || ARCH_CLPS7500 || ARCH_IOP3XX || ARCH_IXP4XX \
        || ARCH_L7200 || ARCH_LH7A40X || ARCH_PXA || ARCH_RPC \
        || ARCH_S3C2410 || ARCH_SA1100 || ARCH_SHARK || FOOTBRIDGE

source "drivers/ide/Kconfig"
endif

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
#11 lore raw
Quoted 11 lines · Russell King
Quoted 10 lines · David S. Miller

From: Russell King <[email protected]>
Date: Thu, 1 Dec 2005 09:41:11 +0000

Quoted 3 lines

In which case why do we restrict floppy to only those machines which could have floppy? Why do we restrict IDE to only those platforms which may have IDE?

These two examples require platform level support via an asm/*.h header file.

Quoted 8 lines
Quoted 4 lines

Whereas the driver's we are talking about use portable interfaces that should be available across the board.

So, bad example.

Not in the IDE case. Bart restricted IDE to a smaller number of ARM platforms, plus any that had PCMCIA. There is no such restriction in the asm-arm/*.h header files.

When I did this change there was such restriction in asm-arm/mach-*/ide.h files (some platforms just lacked ide.h making IDE build break for them).

IDE is a bad example anyway because of legacy ordering issues etc etc.

Quoted 5 lines
if PCMCIA || ARCH_CLPS7500 || ARCH_IOP3XX || ARCH_IXP4XX \
        || ARCH_L7200 || ARCH_LH7A40X || ARCH_PXA || ARCH_RPC \
        || ARCH_S3C2410 || ARCH_SA1100 || ARCH_SHARK || FOOTBRIDGE

source "drivers/ide/Kconfig"
endif

#12 lore raw
Quoted 4 lines · Bartlomiej Zolnierkiewicz
Quoted 3 lines · Russell King
Quoted 1 line · David S. Miller

So, bad example.

Quoted 8 lines
Quoted 3 lines

Not in the IDE case. Bart restricted IDE to a smaller number of ARM platforms, plus any that had PCMCIA. There is no such restriction in the asm-arm/*.h header files.

When I did this change there was such restriction in asm-arm/mach-*/ide.h files (some platforms just lacked ide.h making IDE build break for them).

IDE is a bad example anyway because of legacy ordering issues etc etc.

Okay. Given the general concensus in this thread, can this be removed now?

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
#13 lore raw
Quoted 9 lines · Russell King
Quoted 8 lines · Bartlomiej Zolnierkiewicz
Quoted 6 lines · Russell King
Quoted 1 line · David S. Miller

So, bad example.

Not in the IDE case. Bart restricted IDE to a smaller number of ARM platforms, plus any that had PCMCIA. There is no such restriction in the asm-arm/*.h header files.

Quoted 7 lines
Quoted 4 lines

When I did this change there was such restriction in asm-arm/mach-*/ide.h files (some platforms just lacked ide.h making IDE build break for them).

IDE is a bad example anyway because of legacy ordering issues etc etc.

Okay. Given the general concensus in this thread, can this be removed now?

No, I didn't say that I agree with DaveM. :)

For IDE keeping restriction makes it much easier to maintain (i.e. to answer questions like "why is this ugly hack needed?").

Bartlomiej

#14 lore raw
In reply to Russell King
Quoted 2 lines · Russell King

In which case why do we restrict floppy to only those machines which could have floppy?

Because the floopy drivers needs asm/floppy.h and wouldn't build on others.

Quoted 2 lines

Why do we restrict IDE to only those platforms
which may have IDE?

Dito with asm/ide.h

these are very old drivers where people weren't used to abstractions that would allow to write them platform-independent. For any driver that doesn't have an actual platform depency making it conditional is a bad idea.

maintainer
#15 lore raw
In reply to Russell King

From: Russell King <[email protected]>
Date: Wed, 30 Nov 2005 16:55:47 +0000

Quoted 6 lines

If other CPUs use this then fine, but I find that having config options needlessly available to all architectures is annoying - especially when they are never used.

Eg, would you ever expect to see a DM9000 ethernet device on an x86 machine? Probably not - there's far better PCI solutions now.

If I, for example, make changes across the tree to SKB handling, I'd like to be able to build as many drivers as possible and fix up the compile warnings and build failures before _you_ get to see them.

That's why it's a good idea to make drivers available to as many platforms as possible, even if the hardware isn't necessarily used there.

author
#16 lore raw

2005/11/30, David S. Miller <[email protected]>:

Quoted 10 lines

From: Russell King <[email protected]>
Date: Wed, 30 Nov 2005 16:55:47 +0000

Quoted 6 lines

If other CPUs use this then fine, but I find that having config options needlessly available to all architectures is annoying - especially when they are never used.

Eg, would you ever expect to see a DM9000 ethernet device on an x86 machine? Probably not - there's far better PCI solutions now.

Quoted 8 lines

If I, for example, make changes across the tree to SKB handling, I'd like to be able to build as many drivers as possible and fix up the compile warnings and build failures before _you_ get to see them.

That's why it's a good idea to make drivers available to as many platforms as possible, even if the hardware isn't necessarily used there.

this would mean that we should change almost all ethernet driver deps ?

thanks
--
Franck

#17 lore raw
In reply to Russell King
Quoted 3 lines · Russell King

So until someone says "I want to use this on such and such arch" I think it's better to keep it dependent on those we know are likely to support it.

I use DM9000 on SH-4.

Signature · 1 line
Matan Ziv-Av.                         [email protected]