Patch for show_task

Patch for show_task

[Pete Zaitcev]

Andrew, bkbits says you changed the line above to be p->thread_info.
Unfortunately, there's another.

--- linux-2.5.66/kernel/sched.c	2003-03-24 14:01:16.000000000 -0800
+++ linux-2.5.66-sparc/kernel/sched.c	2003-04-03 22:33:29.000000000 -0800
@@ -2197,7 +2197,7 @@
 		unsigned long * n = (unsigned long *) (p->thread_info+1);
 		while (!*n)
 			n++;
-		free = (unsigned long) n - (unsigned long)(p+1);
+		free = (unsigned long) n - (unsigned long) (p->thread_info+1);
 	}
 	printk("%5lu %5d %6d ", free, p->pid, p->parent->pid);
 	if ((relative = eldest_child(p)))

-- Pete

Re: Patch for show_task

[Andrew Morton]

Pete Zaitcev <zaitcev@redhat.com> wrote:
>
> Andrew, bkbits says you changed the line above to be p->thread_info.
> Unfortunately, there's another.
> 
> --- linux-2.5.66/kernel/sched.c	2003-03-24 14:01:16.000000000 -0800
> +++ linux-2.5.66-sparc/kernel/sched.c	2003-04-03 22:33:29.000000000 -0800
> @@ -2197,7 +2197,7 @@
>  		unsigned long * n = (unsigned long *) (p->thread_info+1);
>  		while (!*n)
>  			n++;
> -		free = (unsigned long) n - (unsigned long)(p+1);
> +		free = (unsigned long) n - (unsigned long) (p->thread_info+1);
>  	}
>  	printk("%5lu %5d %6d ", free, p->pid, p->parent->pid);
>  	if ((relative = eldest_child(p)))
> 

Yup.  But the whole thing's dead anyway - we do not clear the kernel stack
when it is allocated hence the attempt to work out how much was used cannot
work.

That's what

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.66/2.5.66-mm3/broken-out/show_task-free-stack-fix.patch

is about, but it needs finishing off.

Re: Patch for show_task

[Pete Zaitcev]

> Date: Thu, 3 Apr 2003 22:43:46 -0800
> From: Andrew Morton <akpm@digeo.com>

> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.66/2.5.66-mm3/broken-out/show_task-free-stack-fix.patch

I see. What do you say to adding thread_saved_sp() in addition
to thread_saved_pc()? E.g. on sparc that would return
p->thread_info->ti_ksp, then you can calculate free stack.

-- Pete

Re: Patch for show_task

[Andrew Morton]

Pete Zaitcev <zaitcev@redhat.com> wrote:
>
> I see. What do you say to adding thread_saved_sp() in addition
> to thread_saved_pc()? E.g. on sparc that would return
> p->thread_info->ti_ksp, then you can calculate free stack.
> 

I don't think the current sp value is very interesting really.

What that code is trying to do is to tell you the minimum amount of stack
which was *ever* available to that process.  Now that is interesting.  But it
requires that the kernel stack be zeroed out when it is created, and there's
no code in there to do that.

It is probably a useful thing to retain though.  A new CONFIG_DEBUG_STACK
would suit.