Question

i found this macro in one kernel code..
http://lxr.free-electrons.com/source/arch/alpha/include/asm/io.h?v=3.0;a=arm#L140

#define IO_CONCAT(a,b)  _IO_CONCAT(a,b)
#define _IO_CONCAT(a,b) a ## _ ## b

i am unable to understand the meaning of this.Does anybody know this?

Edit :

Then here what it will return

  return IO_CONCAT(__IO_PREFIX,readl)(addr);

Answer 1

Double hash is used to concatenate two tokens together:

#define CONCAT(a,b) a ## b

CONCAT(x, y) # Gives 'xy'

However, such naive implementation doesn’t work in case when one of the arguments being passed is a macro itself:

#define Z y
CONCAT(x, Z) # Gives 'xZ', not 'xy' as one might expect

That is why macro indirection is used in your question:

#define CONCAT(a,b)  __CONCAT(a,b)
#define __CONCAT(a,b) a ## b

#define Z y
CONCAT(x, Z) # Gives 'xy' 

UPD.

Now consider the concrete example you’re asking about:

return IO_CONCAT(__IO_PREFIX,readl)(addr);

Here __IO_PREFIX is obviously a macro (uppercase identifiers in Linux kernel are often macros). It is defined in several places, one of them is:

#define __IO_PREFIX             generic

Now let’s see which steps are taken to expand the original statement:

1. Expand __IO_PREFIX:

   return IO_CONCAT(generic,readl)(addr);
   

2. Expand IO_CONCAT(...):

   return _IO_CONCAT(generic,readl)(addr);
   

3. Expand _IO_CONCAT(...):

   return generic_readl(addr);