Re: [PATCH 6/6] sched: Change task_struct::state

[Mathieu Desnoyers <mathieu.desnoyers-vg+e7yoeK/dWk0Htik3J/w@public.gmane.org>]

----- On Jun 2, 2021, at 9:12 AM, Peter Zijlstra peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org wrote:

> Change the type and name of task_struct::state. Drop the volatile and
> shrink it to an 'unsigned int'. Rename it in order to find all uses
> such that we can use READ_ONCE/WRITE_ONCE as appropriate.
> 
[...]
> 
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c

[...]
> @@ -1559,7 +1560,8 @@ static int fill_psinfo(struct elf_prpsin
> 	psinfo->pr_pgrp = task_pgrp_vnr(p);
> 	psinfo->pr_sid = task_session_vnr(p);
> 
> -	i = p->state ? ffz(~p->state) + 1 : 0;
> +	state = READ_ONCE(p->__state);
> +	i = state ? ffz(~state) + 1 : 0;
> 	psinfo->pr_state = i;
> 	psinfo->pr_sname = (i > 5) ? '.' : "RSDTZW"[i];
> 	psinfo->pr_zomb = psinfo->pr_sname == 'Z';

[...]

> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -113,13 +113,13 @@ struct task_group;
> 					 __TASK_TRACED | EXIT_DEAD | EXIT_ZOMBIE | \
> 					 TASK_PARKED)
> 
> -#define task_is_running(task)		(READ_ONCE((task)->state) == TASK_RUNNING)
> +#define task_is_running(task)		(READ_ONCE((task)->__state) == TASK_RUNNING)
> 
> -#define task_is_traced(task)		((task->state & __TASK_TRACED) != 0)
> +#define task_is_traced(task)		((READ_ONCE(task->__state) & __TASK_TRACED) != 0)
> 
> -#define task_is_stopped(task)		((task->state & __TASK_STOPPED) != 0)
> +#define task_is_stopped(task)		((READ_ONCE(task->__state) & __TASK_STOPPED) !=
> 0)
> 
> -#define task_is_stopped_or_traced(task)	((task->state & (__TASK_STOPPED |
> __TASK_TRACED)) != 0)
> +#define task_is_stopped_or_traced(task)	((READ_ONCE(task->__state) &
> (__TASK_STOPPED | __TASK_TRACED)) != 0)
> 
> #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
> 
> @@ -134,14 +134,14 @@ struct task_group;
> 	do {							\
> 		WARN_ON_ONCE(is_special_task_state(state_value));\
> 		current->task_state_change = _THIS_IP_;		\
> -		current->state = (state_value);			\
> +		WRITE_ONCE(current->__state, (state_value));	\
> 	} while (0)

Why not introduce set_task_state(p) and get_task_state(p) rather than sprinkle
READ_ONCE() and WRITE_ONCE() all over the kernel ?

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com