2002-01-18 13:43:07

by Roger W. Brown

Subject: Type-change of kdev_t

Hi,

I'm no kernel hacker so I am little hesitant to speak; however,
I'm looking at kdev_t.h from the linux-2.5.3-pre1 source.

The type of kdev_t has changed recently from a scalar type to a
structured type. Should macro definitions such as MINOR(dev) also
be revised to be consistent with the "new" kdev_t ?

Something like:

#define MINOR(dev) ((unsigned int) ((dev.value) & MINORMASK))

rather than

#define MINOR(dev) ((unsigned int) ((dev) & MINORMASK))

Then usage of the MINOR() macro remains unchanged.

Regards,

Roger Brown

2002-01-18 16:42:22

by Randy.Dunlap

Subject: Re: Type-change of kdev_t

On Sat, 19 Jan 2002, Roger W.Brown wrote:

|
| Hi,
|
| I'm no kernel hacker so I am little hesitant to speak; however,
| I'm looking at kdev_t.h from the linux-2.5.3-pre1 source.
|
| The type of kdev_t has changed recently from a scalar type to a
| structured type. Should macro definitions such as MINOR(dev) also
| be revised to be consistent with the "new" kdev_t ?

Macros to use kdev_t also changed.

| Something like:
|
| #define MINOR(dev) ((unsigned int) ((dev.value) & MINORMASK))
|
| rather than
|
| #define MINOR(dev) ((unsigned int) ((dev) & MINORMASK))
|
| Then usage of the MINOR() macro remains unchanged.

Nope, use major() and minor() instead [although I prefer
the kmajor() and kminor() patch].

See http://www.osdl.org/archive/rddunlap/linux-port-25x.html,
item 7, bullet 3, which points to an email from one Linus Torvalds
about the kdev_t change.

--
~Randy