> On Fri, Nov 12, 2010 at 08:55:36PM +0100, Felix Fietkau wrote:
> > On 2010-11-12 8:51 PM, Wojciech Dubowik wrote:
> > >> Having both in the kernel is completely pointless. There are no
> > >> devices
> > >> that can support both, nor will there ever be.
> > >
> > > I agree. The only advantage would be that some of the typos and
> missing
> > > functions in either of the files would be detected during compile
> time.
>
> Compile-testing coverage is a fine reason to allow people to compile
> both modules. Ok, I accept that it's a bit of a pain to create
> separate
> modules, but can we at least do it like:
>
> ath5k-$(CONFIG_ATH5K_PCI) += pci.o
> ath5k-$(CONFIG_ATH5K_AHB) += ahb.o
>
> and add the appropriate selects and mutual exclusion to the Kconfig?
> Of course, that presupposes a Kconfig entry for CONFIG_ATHEROS_AR2316
> where there is not one currently.
>
> --
> Bob Copeland %% http://www.bobcopeland.com
What about such changes compared to original patch? It's then possible to
compile both modules at the same time. Of course linking won't work because
I would need to separate functions like in ath9k.
---
diff --git a/drivers/net/wireless/ath/ath5k/Kconfig b/drivers/net/wireless/ath/ath5k/Kconfig
index eb83b7b..eba97f8 100644
--- a/drivers/net/wireless/ath/ath5k/Kconfig
+++ b/drivers/net/wireless/ath/ath5k/Kconfig
@@ -1,9 +1,11 @@
config ATH5K
tristate "Atheros 5xxx wireless cards support"
- depends on PCI && MAC80211
+ depends on (PCI || ATHEROS_AR231X) && MAC80211
select MAC80211_LEDS
select LEDS_CLASS
select NEW_LEDS
+ select ATH5K_AHB if ATHEROS_AR231X
+ select ATH5K_PCI if PCI
---help---
This module adds support for wireless adapters based on
Atheros 5xxx chipset.
@@ -37,3 +39,14 @@ config ATH5K_DEBUG
modprobe ath5k debug=0x00000400
+config ATH5K_AHB
+ bool "Atheros 5xxx AHB bus support"
+ depends on ATHEROS_AR231X
+ ---help---
+ This adds support for WiSoC type chipsets of the 5xxx family
+
+config ATH5K_PCI
+ bool "Atheros 5xxx PCI bus support"
+ depends on PCI
+ ---help---
+ This adds support for PCI type chipsets of the 5xxx family
diff --git a/drivers/net/wireless/ath/ath5k/Makefile b/drivers/net/wireless/ath/ath5k/Makefile
index 50d5e08..a16b178 100644
--- a/drivers/net/wireless/ath/ath5k/Makefile
+++ b/drivers/net/wireless/ath/ath5k/Makefile
@@ -15,5 +15,6 @@ ath5k-y += rfkill.o
ath5k-y += ani.o
ath5k-y += sysfs.o
ath5k-$(CONFIG_ATH5K_DEBUG) += debug.o
-ath5k-y += $(if $(CONFIG_ATHEROS_AR231X),ahb,pci).o
+ath5k-$(CONFIG_ATH5K_PCI) += pci.o
+ath5k-$(CONFIG_ATH5K_AHB) += ahb.o
obj-$(CONFIG_ATH5K) += ath5k.o
On Mon, Nov 15, 2010 at 02:16:06PM +0100, Wojciech Dubowik wrote:
> What about such changes compared to original patch? It's then possible to
> compile both modules at the same time. Of course linking won't work because
> I would need to separate functions like in ath9k.
Is there a plan to add a config entry for ATHEROS_AR231X (or did
I miss it?) Someone will one day run their Kconfig checker and
report it missing.
> --- a/drivers/net/wireless/ath/ath5k/Kconfig
> +++ b/drivers/net/wireless/ath/ath5k/Kconfig
> @@ -1,9 +1,11 @@
> config ATH5K
> tristate "Atheros 5xxx wireless cards support"
> - depends on PCI && MAC80211
> + depends on (PCI || ATHEROS_AR231X) && MAC80211
> select MAC80211_LEDS
> select LEDS_CLASS
> select NEW_LEDS
> + select ATH5K_AHB if ATHEROS_AR231X
> + select ATH5K_PCI if PCI
I like this better, but there probably needs to be a mutual
exclusion so that randconfig builds don't break. Then you don't
get to build both still, but the resulting Makefile is a bit more
normal. And then if someone shows up saying that they really do
have some AR231X device with PCI, then it's more obvious what they
need to do to fix it (i.e. separate bus interfaces from core).
Something like?
select ATH5K_PCI if (PCI && !ATHEROS_AR231X)
or
depends on ((PCI && !ATHEROS_AR231X) || (!PCI && ATHEROS_AR231X)) ...
--
Bob Copeland %% http://www.bobcopeland.com