[PATCH] sdla non-module build fix.

[PATCH] sdla non-module build fix.

[Stephen Hemminger]

Last round of changes broke sdla build if not a module.

diff -Nru a/drivers/net/wan/sdla.c b/drivers/net/wan/sdla.c
--- a/drivers/net/wan/sdla.c	Mon Sep  8 09:49:56 2003
+++ b/drivers/net/wan/sdla.c	Mon Sep  8 09:49:56 2003
@@ -1719,9 +1719,9 @@
 	
 	unregister_frad(devname);
 }
-#endif
 
 MODULE_LICENSE("GPL");
 
 module_init(init_sdla);
 module_exit(exit_sdla);
+#endif

Re: [PATCH] sdla non-module build fix.

[Linus Torvalds]

On Mon, 8 Sep 2003, Stephen Hemminger wrote:
>
> Last round of changes broke sdla build if not a module.

This patch can't be right. It may fix the compile, but it looks like 
init/exit_sdla is never registered at all if it's built-in.

I suspect the whole #ifdef should go away. A driver that does things 
differently depending on whether it is a module or not is generally a 
_broken_ driver.

		Linus

Re: [PATCH] sdla non-module build fix.

[Stephen Hemminger]

On Mon, 8 Sep 2003 10:32:26 -0700 (PDT)
Linus Torvalds <torvalds@osdl.org> wrote:

> 
> On Mon, 8 Sep 2003, Stephen Hemminger wrote:
> >
> > Last round of changes broke sdla build if not a module.
> 
> This patch can't be right. It may fix the compile, but it looks like 
> init/exit_sdla is never registered at all if it's built-in.
> 
> I suspect the whole #ifdef should go away. A driver that does things 
> differently depending on whether it is a module or not is generally a 
> _broken_ driver.
> 
> 		Linus

The registration for non-modular takes place via drivers/net/Space.c 
calling sdla_init like it always did.

Re: [PATCH] sdla non-module build fix.

[Jeff Garzik]

On Mon, Sep 08, 2003 at 10:32:26AM -0700, Linus Torvalds wrote:
> 
> On Mon, 8 Sep 2003, Stephen Hemminger wrote:
> >
> > Last round of changes broke sdla build if not a module.
> 
> This patch can't be right. It may fix the compile, but it looks like 
> init/exit_sdla is never registered at all if it's built-in.
> 
> I suspect the whole #ifdef should go away. A driver that does things 
> differently depending on whether it is a module or not is generally a 
> _broken_ driver.

Nope, this is one of those weird two-ways-of-initialization things, like
drivers/net/Space.c or drivers/video/fbmem.c.  If you look at the sdla
code, just above where Stephen's patch's context ends, you see
"sdla_init" which is called by both modular and non-modular code.

So -- with the exception of #ifdef'ing out MODULE_LICENSE, Stephen's
patch is ok.

Now that said, this situation is only temporary, until we get all the
network device structs being allocated dynamically, and interface
ordering issues are resolved.  Once that happens, you are absolutely
right:  sdla's #ifdef MODULE should be removed.

	Jeff