Re: [PATCH 31/34] move virtrng_remove to .devexit.text

[Uwe Kleine-König]

Hello Michael,

On Thu, Oct 01, 2009 at 08:05:00PM +0200, Michael S. Tsirkin wrote:
> On Thu, Oct 01, 2009 at 07:41:16PM +0200, Uwe Kleine-König wrote:
> > Hello Michael,
> > 
> > > > But note it's not an error in general to use a .text function as remove
> > > > callback.  E.g. take drivers/gpio/twl4030-gpio.c.  gpio_twl4030_remove
> > > > is used in gpio_twl4030_probe which is defined using __devinit.  So
> > > > using __devexit for gpio_twl4030_remove is wrong.  (So there is a bug,
> > > > as gpio_twl4030_remove uses __devexit.)  I didn't try, but as far as I
> > > > understand this will result in a compile error if the driver is built-in
> > > > with HOTPLUG=n.
> > > 
> > > Wait a second.
> > > As far as I understand, __devexit makes it possible to remove code if
> > > hotplug is off.
> > right.
> >  
> > > At least for static functions, it's enough to mark their only use
> > > as _devexit_p, and compiler will remove the text as it's unused.
> > > 
> > > Isn't that right?
> > hmm, I don't know.  I'll try, one moment.  OK, you're right.  The
> > function is discarded with a compiler warning.
> >  
> > > If so, what, again, was the motivation for the patches that added
> > > __devexit to functions that were already used with __devexit_p?
> > I thought it saves some memory, but as it looks now it only fixes a
> > compiler warning.
> 
> We can redefine __devexit_p(x) to something like
> #define __devexit_p(x) ((typeof(x) *)NULL)
> and this will shut down the warning without need to fix the code.
Mmmmh, don't know.  A definitive advantage is that there is only a
single point in the code that defines if a function is discarded or not.
Nothing that needs to be consistent.
For me it feels somehow wrong anyhow, but that might be only because I'm
used to the current model.
 
> > Note there are two types of errors fixed in this series.  One is:
> > 
> > 	-static int func(void arg)
> > 	+static int __devexit func(void arg)
> > 
> > if the only usage of func() is wraped by __devexit_p.  This is (as
> > seen above) not that critical, there is only a warning fixed.
> > 
> > The other type results in a build failure:
> > 
> > 	-remove = __devexit_p(another_func),
> > 	+remove = __exit_p(another_func),
> > 
> > with another_func being defined using __exit.  In the case 
> > defined(MODULE) && defined(CONFIG_HOTPLUG) another_func is discarded,
> > but __devexit_p(another_func) evaluates to another_func and thus the
> > module doesn't link.
> 
> Yes, calling __exit function from non- __exit is always a bug.
> I think there's a make flag to warn about this, not sure why it's
> not the default.
I think it was disabled because there were too many warnings :-)

Best regards
Uwe

-- 
Pengutronix e.K.                              | Uwe Kleine-König            |
Industrial Linux Solutions                    | http://www.pengutronix.de/  |
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/