sched.h - INIT_TASK(tsk) - kernel programming/syntax question

[MoToR]

Hi, everyone,

I've got an assignment to add some functionality to a linux kernel in educational purposes.

I need to add a new processes group management functionality to the kernel. In order to do so, among some other things, I need to add a field to the task_struct structure in the file sched.h and therefore I need to update INIT_TASK (sched.h file) definition to initialize my added fields during system initialization.

I added new fields to the task_struct and went to update the INIT_TASK macro. Unfortunately I've met a C syntax there that I found somewhat difficult to comprehend:


482 #define INIT_TASK(tsk) \
483 { \
484 state: 0, \
485 flags: 0, \
486 sigpending: 0, \
487 addr_limit: KERNEL_DS, \
488 exec_domain: &default_exec_domain, \
489 lock_depth: -1, \
490 counter: DEF_COUNTER, \
491 nice: DEF_NICE, \
492 policy: SCHED_OTHER, \
493 mm: NULL, \
494 active_mm: &init_mm, \
495 cpus_runnable: ~0UL, \
496 cpus_allowed: ~0UL, \
497 run_list: LIST_HEAD_INIT(tsk.run_list), \
498 next_task: &tsk, \
499 prev_task: &tsk, \
500 p_opptr: &tsk, \
501 p_pptr: &tsk, \
502 thread_group: LIST_HEAD_INIT(tsk.thread_group), \
503 wait_chldexit: __WAIT_QUEUE_HEAD_INITIALIZER(tsk.wait_chldexit),\
504 real_timer: { \
505 function: it_real_fn \
506 }, \
507 cap_effective: CAP_INIT_EFF_SET, \
508 cap_inheritable: CAP_INIT_INH_SET, \
509 cap_permitted: CAP_FULL_SET, \
510 keep_capabilities: 0, \
511 rlim: INIT_RLIMITS, \
512 user: INIT_USER, \
513 comm: "swapper", \
514 thread: INIT_THREAD, \
515 fs: &init_fs, \
516 files: &init_files, \
517 sigmask_lock: SPIN_LOCK_UNLOCKED, \
518 sig: &init_signals, \
519 pending: { NULL, &tsk.pending.head, {{0}}}, \
520 blocked: {{0}}, \
521 alloc_lock: SPIN_LOCK_UNLOCKED, \
522 journal_info: NULL, \
523 }

(the numbers on the left are just a line numbering from the TOMOYO Linux cross reference, see links below)


I would be very thankful if someone could clarify how exactly does INIT_TASK(tsk) work, what does its syntax mean and where can I learn more about this.

As far as I was able to understand, these are designated initializers, but why colons? Why not .field syntax? What am I missing?

If we look inside init_task.c we'll find something like:

__asm__(".section \".text\",#alloc\n");
union task_union init_task_union =
{ INIT_TASK(init_task_union.task) };

Why do we initialize a union with so many fields? Isn't only the first field used for that?

Any explanation of how exactly these lines work will be most appreciated.

Thank you!



P.S. The next cross reference might be helpful if you don't have time or opportunity to look inside your distribution source files:

http://tomoyo.sourceforge.jp/cgi-bin/lxr/source/include/linux/sched.h?v=linux-2.4.37.1-ccs-1.6.7#L482

http://tomoyo.sourceforge.jp/cgi-bin/lxr/source/arch/sparc/kernel/init_task.c?v=linux-2.4.37.1-ccs-1.6.7#L19

http://tomoyo.sourceforge.jp/cgi-bin/lxr/ident?v=linux-2.4.37.1-ccs-1.6.7&i=INIT_TASK

Thanks to the guys that provided the reference.

[MoToR]

* bump *

By the way explanation in any level of complexity below the original will be welcome. I'm not a beginner in C, so to speak.

[MoToR]

No replies in three days.

That's just sad.

Well I figured out the second question, and I understood what the first code does, I'd just like to know what syntax is used in the first piece of code. The colon thing is undocumented.

[MoToR]

* bump *

[hod139]

No replies in three days.

That's just sad.

Well I figured out the second question, and I understood what the first code does, I'd just like to know what syntax is used in the first piece of code. The colon thing is undocumented.

The "colon thing" is a gcc extension to the C language (C89 standard anyways) that allows you to use element labels to initialize a union (http://gcc.gnu.org/onlinedocs/gcc-4.4.0/gcc/Designated-Inits.html#Designated-Inits)

[MoToR]

Awesome! Thank you for your reply! I've found tons of useful info at that page.

To sum up the thread, here's the explanation I was looking for of the syntax, as can be found on the page you guided me to:


...
In a structure initializer, specify the name of a field to initialize with `fieldname:' before the element value. For example, given the following structure,


struct point { int x, y; };


the following initialization


struct point p = { y: yvalue, x: xvalue };


is equivalent to


struct point p = { xvalue, yvalue };


Another syntax which has the same meaning is `.fieldname ='., as shown here:


struct point p = { .y = yvalue, .x = xvalue };


You can also use an element label (with either the colon syntax or the period-equal syntax) when initializing a union, to specify which element of the union should be used. For example,


union foo { int i; double d; };


union foo f = { d: 4 };


will convert 4 to a double to store it in the union using the second element. By contrast, casting 4 to type union foo would store it into the union as the integer i, since it is an integer.
...


Thank you again! That was very helpful.