current_is_kswapd is a function

current_is_kswapd is a function

[William Lee Irwin III]

current_is_kswapd() is a function, not an integer; this explains some
anomalies in reported statistics.

===== fs/inode.c 1.107 vs edited =====
--- 1.107/fs/inode.c	Sun Oct  5 01:07:55 2003
+++ edited/fs/inode.c	Sat Oct 11 19:15:48 2003
@@ -453,7 +453,7 @@
 	dispose_list(&freeable);
 	up(&iprune_sem);
 
-	if (current_is_kswapd)
+	if (current_is_kswapd())
 		mod_page_state(kswapd_inodesteal, reap);
 	else
 		mod_page_state(pginodesteal, reap);

Re: current_is_kswapd is a function

[Andrew Morton]

William Lee Irwin III <wli@holomorphy.com> wrote:
>
> -	if (current_is_kswapd)
>  +	if (current_is_kswapd())

Well damn, I must have looked at that a hundred times when wondering why
/proc/vmstat:pginodesteal was always zero.  Thanks.

It would probably be worthwhile teaching the compiler to generate a warning
in this case; I doubt if anyone is likely to want to find out at runtime
whether the linker happened to place a particular function at address zero.
I shall suggest that.

Re: current_is_kswapd is a function

[Jamie Lokier]

Andrew Morton wrote:
> It would probably be worthwhile teaching the compiler to generate a warning
> in this case; I doubt if anyone is likely to want to find out at runtime
> whether the linker happened to place a particular function at address zero.
> I shall suggest that.

I agree it would be a very useful warning.

Testing the address is useful occasionally, to ask whether the target
of a weak linkage reference was linked in, so a syntax is needed to
suppress the warning, such as "&function != 0".

-- Jamie