[PATCH net-next] mdio: Export mdio.h to userland

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Signed-off-by: Ben Hutchings <redacted>
---
 include/linux/Kbuild |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/include/linux/Kbuild b/include/linux/Kbuild
index c94e717..4b0b7ed 100644
--- a/include/linux/Kbuild
+++ b/include/linux/Kbuild

@@ -238,6 +238,7 @@ header-y += magic.h
 header-y += major.h
 header-y += map_to_7segment.h
 header-y += matroxfb.h
+header-y += mdio.h
 header-y += media.h
 header-y += mempolicy.h
 header-y += meye.h

-- 
1.7.7.6


-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

On Wed, 2012-02-29 at 19:02 +0000, Ben Hutchings wrote:

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Minor issue with this: mdio_phy_id_is_c45() returns bool, and that's not
defined by default.  Should I add #include <stdbool.h> or change the
return type to int for maximum compatibility?

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 19:12:34 +0000

On Wed, 2012-02-29 at 19:02 +0000, Ben Hutchings wrote:

quoted

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Minor issue with this: mdio_phy_id_is_c45() returns bool, and that's not
defined by default.  Should I add #include <stdbool.h> or change the
return type to int for maximum compatibility?

I do not see any precedence for including stdbool.h in a kernel header
file.  And I sure as hell am not going to compromise using a correct
type just to make this usable by userspace.

I think this is of dubious value, and combined with the bool
complication I'm not going to apply this, sorry.

On Wed, 2012-02-29 at 15:42 -0500, David Miller wrote:

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 19:12:34 +0000

quoted

On Wed, 2012-02-29 at 19:02 +0000, Ben Hutchings wrote:

quoted

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Minor issue with this: mdio_phy_id_is_c45() returns bool, and that's not
defined by default.  Should I add #include <stdbool.h> or change the
return type to int for maximum compatibility?

I do not see any precedence for including stdbool.h in a kernel header
file.  And I sure as hell am not going to compromise using a correct
type just to make this usable by userspace.

I think this is of dubious value, and combined with the bool
complication I'm not going to apply this, sorry.

Without this, userland doing clause 45 MDIO has to duplicate the ID
packing definitions.   I intended to export these to userland in the
first place; that's why the header already has an #ifdef __KERNEL__.
I was about to get rid of the duplicate definitions in our utilities
when I realised I hadn't exported them.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 22:43:48 +0000

On Wed, 2012-02-29 at 15:42 -0500, David Miller wrote:

quoted

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 19:12:34 +0000

quoted

On Wed, 2012-02-29 at 19:02 +0000, Ben Hutchings wrote:

quoted

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Minor issue with this: mdio_phy_id_is_c45() returns bool, and that's not
defined by default.  Should I add #include <stdbool.h> or change the
return type to int for maximum compatibility?

I do not see any precedence for including stdbool.h in a kernel header
file.  And I sure as hell am not going to compromise using a correct
type just to make this usable by userspace.

I think this is of dubious value, and combined with the bool
complication I'm not going to apply this, sorry.

Without this, userland doing clause 45 MDIO has to duplicate the ID
packing definitions.

Then come up with a real solution to the bool issue.

But note that changing the bool to an int is not a solution, neither
is #include <stdbool.h> unless we can reach consensus on lkml that
this is OK.

On Wed, 2012-02-29 at 18:02 -0500, David Miller wrote:

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 22:43:48 +0000

quoted

On Wed, 2012-02-29 at 15:42 -0500, David Miller wrote:

quoted

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 19:12:34 +0000

quoted

On Wed, 2012-02-29 at 19:02 +0000, Ben Hutchings wrote:

quoted

The ID packing/unpacking functions are needed by userland and the
register definitions may also be useful there.

Minor issue with this: mdio_phy_id_is_c45() returns bool, and that's not
defined by default.  Should I add #include <stdbool.h> or change the
return type to int for maximum compatibility?

I do not see any precedence for including stdbool.h in a kernel header
file.  And I sure as hell am not going to compromise using a correct
type just to make this usable by userspace.

I think this is of dubious value, and combined with the bool
complication I'm not going to apply this, sorry.

Without this, userland doing clause 45 MDIO has to duplicate the ID
packing definitions.

Then come up with a real solution to the bool issue.

But note that changing the bool to an int is not a solution, neither
is #include <stdbool.h> unless we can reach consensus on lkml that
this is OK.

I could just move the mdio_phy_id_is_c45(), mdio_phy_id_prtad(),
mdio_phy_id_devad() functions inside #ifdef __KERNEL__ since userland
shouldn't need to unpack PHY IDs.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

From: Ben Hutchings <redacted>
Date: Wed, 29 Feb 2012 23:09:52 +0000

I could just move the mdio_phy_id_is_c45(), mdio_phy_id_prtad(),
mdio_phy_id_devad() functions inside #ifdef __KERNEL__ since userland
shouldn't need to unpack PHY IDs.

That would work too.