What is the meaning of leading and trailing underscores in Linux kernel identifiers?

Question

I keep running across little conventions like __KERNEL__.

Are the __ in this case a naming convention used by kernel developers or is it a syntax specific reason for naming a macro this way?

There are many examples of this throughout the code.

For example some functions and variables begin with an _ or even __.

Is there a specific reason for this?

It seems pretty widely used and I just need some clarification as to whether these things have a syntactical purpose or is it simply a naming convention.

Furthermore I see lots of user declared types such as uid_t. Again I assume this is a naming convention telling the reader that it is a user-defined type?

Cheers for the response. Sorry, to clear up an ambiguity, when I say user-defined types I indeed mean those defined by typedef. — A.Smith, Feb 24 '12 at 14:16
Sure. As for the underscores, my guess is to prevent a macro being used where it was not intended when it has an ambiguous name with a function or class. — Linuxios, Feb 24 '12 at 14:25
http://stackoverflow.com/questions/3325815/what-does-double-underscore-mean-in-variable-name-in-c-language — Ciro Santilli OurBigBook.com, Oct 24 '15 at 13:58

score 7 · Accepted Answer · answered Feb 24 '12 at 15:28

There are several cases:

In public facing headers, i.e. anything that libc will be taking over and putting under /usr/include/linux, the standards specify which symbols should be defined and any other symbols specific to the system shall start with underscore and capital letter or two underscores. That's the reason for __KERNEL__ in particular, because it is used in headers that are included both in kernel and in libc and some declarations are different.
In internal code, the convention usually is that symbol __something is workhorse for something excluding some management, often locking. That is a reason for things like __d_lookup . Similar convention for system calls is that sys_something is the system call entry point that handles context switch to and from kernel and calls do_something to do the actual work.
The _t suffix is standard library convention for typedefs. E.g. size_t, ptrdiff_t, foff_t and such. Kernel code follows this convention for it's internal types too.

Thanks for your answer. With all of the coding style documentation about the kernel, I'm surprised that I don't find a lot of documentation on the naming conventions in the kernel. Did you just pick up these things by doing kernel hacking? — KarateSnowMachine, Nov 12 '13 at 22:19
@KarateSnowMachine: Well, the first and last point come from C standard. The middle point, mostly yes. I didn't do any serious kernel hacking, but I had an assignment in kernel back at school, so I had to read quite a bit of the code (especially to double-check the locking). — Jan Hudec, Nov 13 '13 at 07:23

Angelo Dureghello · Answer 2 · 2020-11-08T11:53:40.467

0

There are several __functions, as i.e. __alloc_pages_nodemask() that seems also be exported. Also, __functions are static, static inline, or globals too. Btw, observing __alloc_pages_nodemask() is called only from mm code, and in no other place, such functions may be meant "internal" of some kernel framekwork.

edited Nov 08 '20 at 11:53

answered Nov 08 '20 at 08:23

Angelo Dureghello

154
1
6

What is the meaning of leading and trailing underscores in Linux kernel identifiers?

2 Answers2

Linked