Re: [PATCH] task name handling in proc fs

[Andrew Morton <akpm@osdl.org>]

Mike Kravetz <kravetz@us.ibm.com> wrote:
>
> +void set_task_comm(struct task_struct *tsk, char *name)
> +{
> +	int i, ch;
> +
> +	task_lock(tsk);
> +	for (i=0; (ch = *(name++)) != '\0';) {
> +		if (ch == '/')
> +			i = 0;
> +		else
> +			if (i < (sizeof(tsk->comm) - 1))
> +				tsk->comm[i++] = ch;
> +	}
> +	tsk->comm[i] = '\0';
> +	task_unlock(tsk);
> +}

I don't think the basename logic should be in this function.  Only one
caller needs it, and if we later try to use this function to set
current->comm for per-cpu kernel threads, it will mangle the text.

root         2  0.0  0.0     0    0 ?        SW   Jul06   0:00 [migration/0]
root         3  0.0  0.0     0    0 ?        SWN  Jul06   0:00 [ksoftirqd/0]
root         4  0.0  0.0     0    0 ?        SW   Jul06   0:00 [migration/1]
root         5  0.0  0.0     0    0 ?        SWN  Jul06   0:00 [ksoftirqd/1]
root         6  0.0  0.0     0    0 ?        SW   Jul06   0:00 [migration/2]
root         7  0.0  0.0     0    0 ?        SWN  Jul06   0:00 [ksoftirqd/2]
root         8  0.0  0.0     0    0 ?        SW   Jul06   0:00 [migration/3]
root         9  0.0  0.0     0    0 ?        SWN  Jul06   0:00 [ksoftirqd/3]

We probably won't actually _do_ that, since kthread_create() uses vsnprintf(),
but pushing code which is specific to one caller into a library function
seems wrong...