Skip to contentSkip to conversation

PNP: export pnp_bus_type

PATCH In review David Brownell 2 participants 5 messages Dec 27, 2006 – Dec 28, 2006 Pnp support

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PNP: export pnp_bus_type

The PNP framework doesn't export "pnp_bus_type", which is an unfortunate exception to the policy followed by pretty much every other bus. I noticed this when I had to find a device in order to provide its platform_data.

drivers/pnp/base.h +0 -1
Index: g26/drivers/pnp/base.h
===================================================================
@@ -1,4 +1,3 @@
-extern struct bus_type pnp_bus_type;
 extern spinlock_t pnp_lock;
 void *pnp_alloc(long size);
 int pnp_interface_attach_device(struct pnp_dev *dev);
drivers/pnp/driver.c +1 -0
Index: g26/drivers/pnp/driver.c
===================================================================
@@ -199,6 +199,7 @@ struct bus_type pnp_bus_type = {
 	.suspend = pnp_bus_suspend,
 	.resume = pnp_bus_resume,
 };
+EXPORT_SYMBOL(pnp_bus_type);
 
 int pnp_register_driver(struct pnp_driver *drv)
 {
include/linux/pnp.h +2 -0
Index: g26/include/linux/pnp.h
===================================================================
@@ -352,6 +352,8 @@ struct pnp_protocol {
 	(dev) = protocol_to_pnp_dev((dev)->protocol_list.next))
 
 
+extern struct bus_type pnp_bus_type;
+
 #if defined(CONFIG_PNP)
 
 /* device management */
#1 lore raw
Quoted 3 lines · David Brownell

The PNP framework doesn't export "pnp_bus_type", which is an unfortunate exception to the policy followed by pretty much every other bus. I noticed this when I had to find a device in order to provide its platform_data.

can you please merge the export together with the driver? We already have way too many unused exports, and the only sane way is to merge the export with the user..... (and yes exports are not free, they take up 100 to 150 bytes of kernel size for example)

#2 lore raw
Quoted 6 lines · Arjan van de Ven
Quoted 3 lines · David Brownell

The PNP framework doesn't export "pnp_bus_type", which is an unfortunate exception to the policy followed by pretty much every other bus. I noticed this when I had to find a device in order to provide its platform_data.

can you please merge the export together with the driver?

I'll send that stuff along; providing the platform data is actually an update to ACPI glue, not the driver, so the driver won't need to become needlessly coupled to ACPI. (Driver = rtc_cmos, I'll resend it in a few days.)

Quoted 4 lines

We already
have way too many unused exports, and the only sane way is to merge the export with the user..... (and yes exports are not free, they take up 100 to 150 bytes of kernel size for example)

Hmm, then maybe it'd be worth updating that patch I just sent so that the only change is to switch #includes for the extern decl ... i.e. to "export" it only to other statically linked kernel code, rather than to modules. I'll do that.

My own question about that EXPORT_SYMBOL was whether it instead be an EXPORT_SYMBOL_GPL, but if either one costs bytes ... I'm happy to avoid that cost!

- Dave

#3 lore raw
Quoted 9 lines

Hmm, then maybe it'd be worth updating that patch I just sent so that the only change is to switch #includes for the extern decl ... i.e. to "export" it only to other statically linked kernel code, rather than to modules. I'll do that.

My own question about that EXPORT_SYMBOL was whether it instead be an EXPORT_SYMBOL_GPL, but if either one costs bytes ... I'm happy to avoid that cost!

no export if it's not *really* need is obviously superior to either of those so yes I like the patch you're talking about already without even having seen it ;)

Quoted 1 line
Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
#4 lore raw
Quoted 6 lines · Arjan van de Ven
Quoted 6 lines

Hmm, then maybe it'd be worth updating that patch I just sent so that the only change is to switch #includes for the extern decl ... i.e. to "export" it only to other statically linked kernel code, rather than to modules. I'll do that.

Quoted 7 lines
Quoted 3 lines

My own question about that EXPORT_SYMBOL was whether it instead be an EXPORT_SYMBOL_GPL, but if either one costs bytes ... I'm happy to avoid that cost!

no export if it's not *really* need is obviously superior to either of those so yes I like the patch you're talking about already without even having seen it ;)

The change is trivial ... in pnp/driver.c strike the EXPORT_SYMBOL. So I won't bother posting it to this list; I hate wasting electrons.

I sent the updated patch to Andrew, but evidently just missed his 2.6.20-rc2-mm1 cutoff.

- Dave