kernel 2.6.0, wrong Kconfig directives

kernel 2.6.0, wrong Kconfig directives

[Andrea Barisani]

Hi folks,

Installing 2.6.0 I've found that some kernel options directives are wrong,
in fact the option turns out to be always enabled. I don't think this is 
a desired behaviour.

Sorry for the format, yes I know it's ugly :) but I'll leave to you the proper 
solution :) so I can't make a proper patch.


- IPV6_SCTP___ option is always turned on

./net/sctp/Kconfig:

8:  config IPV6_SCTP__
9: 	    tristate
10:         default y if IPV6=n
11:	    default IPV6 if IPV6
12:
13: config IP_SCTP
14:	    tristate "The SCTP Protocol (EXPERIMENTAL)"
15:	    depends on IPV6_SCTP__


I think something is wrong here, why the 'default y if IPV6=n' ???

 
- INPUT_MOUSEDEV option is always turned on

./drivers/input/Kconfig:

27: config INPUT_MOUSEDEV
28:	    tristate "Mouse interface" if EMBEDDED
29:	    default y
30:	    depends on INPUT

43: config INPUT_MOUSEDEV_PSAUX
44:         bool "Provide legacy /dev/psaux device" if EMBEDDED
45:         default y
46:         depends on INPUT_MOUSEDEV


the tristate directive is ignored in most default configurations since EMBEDDED
is not set, however this doesn't allow to disable INPUT_MOUSEDEV and 
INPUT_MOUSEDEV_PSAUX. I don't suppose this is right.


- SOUND_GAMEPORT option is always turned on

./drivers/input/gameport/Kconfig

22: config SOUND_GAMEPORT
23:         tristate
24:         default y if GAMEPORT!=m
25:         default m if GAMEPORT=m

line 24 is definetly wrong, option is enabled if GAMEPORT=n.


Bye 

P.S.
I'm not subscribed to this list so CC me if it's needed

--
------------------------------------------------------------
INFIS Network Administrator & Security Officer         .*. 
Department of Physics       - University of Trieste     V 
lcars@infis.univ.trieste.it - GPG Key 0x864C9B9E      (   )
----------------------------------------------------  (   )
"How would you know I'm mad?" said Alice.             ^^-^^
"You must be,'said the Cat,'or you wouldn't have come here."
------------------------------------------------------------

Re: kernel 2.6.0, wrong Kconfig directives

[OGAWA Hirofumi]

> - SOUND_GAMEPORT option is always turned on
> 
> ./drivers/input/gameport/Kconfig
> 
> 22: config SOUND_GAMEPORT
> 23:         tristate
> 24:         default y if GAMEPORT!=m
> 25:         default m if GAMEPORT=m
> 
> line 24 is definetly wrong, option is enabled if GAMEPORT=n.

This patch uses "select" for the dependency of GAMEPORT.

Thanks.

 drivers/input/gameport/Kconfig |    5 -----
 sound/oss/Kconfig              |   18 ++++++++++++------
 2 files changed, 12 insertions(+), 11 deletions(-)

diff -puN drivers/input/gameport/Kconfig~fix-config-gameport drivers/input/gameport/Kconfig
--- linux-2.6.0/drivers/input/gameport/Kconfig~fix-config-gameport	2003-12-26 04:10:12.000000000 +0900
+++ linux-2.6.0-hirofumi/drivers/input/gameport/Kconfig	2003-12-26 04:10:12.000000000 +0900
@@ -19,11 +19,6 @@ config GAMEPORT
 	  To compile this driver as a module, choose M here: the
 	  module will be called gameport.
 
-config SOUND_GAMEPORT
-	tristate
-	default y if GAMEPORT!=m
-	default m if GAMEPORT=m
-
 config GAMEPORT_NS558
 	tristate "Classic ISA and PnP gameport support"
 	depends on GAMEPORT
diff -puN sound/oss/Kconfig~fix-config-gameport sound/oss/Kconfig
--- linux-2.6.0/sound/oss/Kconfig~fix-config-gameport	2003-12-26 04:10:12.000000000 +0900
+++ linux-2.6.0-hirofumi/sound/oss/Kconfig	2003-12-26 04:10:12.000000000 +0900
@@ -173,7 +173,8 @@ config SOUND_CS4281
 
 config SOUND_ES1370
 	tristate "Ensoniq AudioPCI (ES1370)"
-	depends on SOUND_PRIME!=n && SOUND && PCI && SOUND_GAMEPORT
+	depends on SOUND_PRIME!=n && SOUND && PCI
+	select GAMEPORT
 	help
 	  Say Y or M if you have a PCI sound card utilizing the Ensoniq
 	  ES1370 chipset, such as Ensoniq's AudioPCI (non-97). To find
@@ -186,7 +187,8 @@ config SOUND_ES1370
 
 config SOUND_ES1371
 	tristate "Creative Ensoniq AudioPCI 97 (ES1371)"
-	depends on SOUND_PRIME!=n && SOUND && PCI && SOUND_GAMEPORT
+	depends on SOUND_PRIME!=n && SOUND && PCI
+	select GAMEPORT
 	help
 	  Say Y or M if you have a PCI sound card utilizing the Ensoniq
 	  ES1371 chipset, such as Ensoniq's AudioPCI97. To find out if
@@ -199,7 +201,8 @@ config SOUND_ES1371
 
 config SOUND_ESSSOLO1
 	tristate "ESS Technology Solo1"
-	depends on SOUND_PRIME!=n && SOUND && SOUND_GAMEPORT
+	depends on SOUND_PRIME!=n && SOUND
+	select GAMEPORT
 	help
 	  Say Y or M if you have a PCI sound card utilizing the ESS Technology
 	  Solo1 chip. To find out if your sound card uses a
@@ -237,7 +240,8 @@ config SOUND_HARMONY
 
 config SOUND_SONICVIBES
 	tristate "S3 SonicVibes"
-	depends on SOUND_PRIME!=n && SOUND && SOUND_GAMEPORT
+	depends on SOUND_PRIME!=n && SOUND
+	select GAMEPORT
 	help
 	  Say Y or M if you have a PCI sound card utilizing the S3
 	  SonicVibes chipset. To find out if your sound card uses a
@@ -269,7 +273,8 @@ config SOUND_VRC5477
 
 config SOUND_TRIDENT
 	tristate "Trident 4DWave DX/NX, SiS 7018 or ALi 5451 PCI Audio Core"
-	depends on SOUND_PRIME!=n && SOUND && SOUND_GAMEPORT
+	depends on SOUND_PRIME!=n && SOUND
+	select GAMEPORT
 	---help---
 	  Say Y or M if you have a PCI sound card utilizing the Trident
 	  4DWave-DX/NX chipset or your mother board chipset has SiS 7018
@@ -782,7 +787,8 @@ config SOUND_NM256
 
 config SOUND_MAD16
 	tristate "OPTi MAD16 and/or Mozart based cards"
-	depends on SOUND_OSS && SOUND_GAMEPORT
+	depends on SOUND_OSS
+	select GAMEPORT
 	---help---
 	  Answer Y if your card has a Mozart (OAK OTI-601) or MAD16 (OPTi
 	  82C928 or 82C929 or 82C931) audio interface chip. These chips are

_

-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: kernel 2.6.0, wrong Kconfig directives

[Muli Ben-Yehuda]

[-- Attachment #1: Type: text/plain, Size: 715 bytes --]

On Tue, Dec 23, 2003 at 12:56:23AM +0100, Andrea Barisani wrote:

> - SOUND_GAMEPORT option is always turned on
> 
> ./drivers/input/gameport/Kconfig
> 
> 22: config SOUND_GAMEPORT
> 23:         tristate
> 24:         default y if GAMEPORT!=m
> 25:         default m if GAMEPORT=m

> line 24 is definetly wrong, option is enabled if GAMEPORT=n.

line 24 is correct. CONFIG_SOUND_GAMEPORT is an odd beast, see
e.g.
http://marc.theaimsgroup.com/?l=linux-kernel&m=106479206731633&w=2 for
details. Please leave poor CONFIG_GAMEPORT alone. 

Cheers, 
Muli 
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/

"the nucleus of linux oscillates my world" - gccbot@#offtopic


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: kernel 2.6.0, wrong Kconfig directives

[Muli Ben-Yehuda]

[-- Attachment #1: Type: text/plain, Size: 2036 bytes --]

On Fri, Dec 26, 2003 at 04:20:21AM +0900, OGAWA Hirofumi wrote:
> 
> > - SOUND_GAMEPORT option is always turned on
> > 
> > ./drivers/input/gameport/Kconfig
> > 
> > 22: config SOUND_GAMEPORT
> > 23:         tristate
> > 24:         default y if GAMEPORT!=m
> > 25:         default m if GAMEPORT=m
> > 
> > line 24 is definetly wrong, option is enabled if GAMEPORT=n.
> 
> This patch uses "select" for the dependency of GAMEPORT.

This is wrong. It forces the joystick (GAMEPORT) in even when it's not
needed, whereas SOUND_GAMEPORT handles all cases fine. That way lies
kernel bloat. Please apply this documentation patch instead: 

Index: drivers/input/gameport/Kconfig
===================================================================
RCS file: /home/muli/kernel/cvsroot/linux-2.5/drivers/input/gameport/Kconfig,v
retrieving revision 1.4
diff -u -u -r1.4 Kconfig
--- drivers/input/gameport/Kconfig	26 Sep 2003 00:23:18 -0000	1.4
+++ drivers/input/gameport/Kconfig	25 Dec 2003 19:48:49 -0000
@@ -19,6 +19,17 @@
 	  To compile this driver as a module, choose M here: the
 	  module will be called gameport.
 
+
+# Yes, this looks a bit odd. Yes, it ends up being turned on in lots
+# of cases. Please don't touch it. It is here to handle the case where
+# a sound driver can be either a module or compiled in if GAMEPORT is
+# not selected, but must be a module if the joystick is selected as a 
+# module. The sound driver calls GAMEPORT functions. If GAMEPORT is
+# not selected, stubs are provided. If GAMEPORT is built in,
+# everything is fine. If GAMEPORT is a module, however, it would need
+# to be loaded for the sound driver to be able to link
+# properly. Therefore, the sound driver must be a module as well in
+# that case (and the GAMEPORT module must be loaded first). 
 config SOUND_GAMEPORT
 	tristate
 	default y if GAMEPORT!=m
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/

"the nucleus of linux oscillates my world" - gccbot@#offtopic


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: kernel 2.6.0, wrong Kconfig directives

[OGAWA Hirofumi]

Muli Ben-Yehuda <mulix@mulix.org> writes:

> On Fri, Dec 26, 2003 at 04:20:21AM +0900, OGAWA Hirofumi wrote:
> > 
> > > - SOUND_GAMEPORT option is always turned on
> > > 
> > > ./drivers/input/gameport/Kconfig
> > > 
> > > 22: config SOUND_GAMEPORT
> > > 23:         tristate
> > > 24:         default y if GAMEPORT!=m
> > > 25:         default m if GAMEPORT=m
> > > 
> > > line 24 is definetly wrong, option is enabled if GAMEPORT=n.
> > 
> > This patch uses "select" for the dependency of GAMEPORT.
> 
> This is wrong. It forces the joystick (GAMEPORT) in even when it's not
> needed, whereas SOUND_GAMEPORT handles all cases fine. That way lies
> kernel bloat. Please apply this documentation patch instead: 
> 
> Index: drivers/input/gameport/Kconfig
> ===================================================================
> RCS file: /home/muli/kernel/cvsroot/linux-2.5/drivers/input/gameport/Kconfig,v
> retrieving revision 1.4
> diff -u -u -r1.4 Kconfig
> --- drivers/input/gameport/Kconfig	26 Sep 2003 00:23:18 -0000	1.4
> +++ drivers/input/gameport/Kconfig	25 Dec 2003 19:48:49 -0000
> @@ -19,6 +19,17 @@
>  	  To compile this driver as a module, choose M here: the
>  	  module will be called gameport.
>  
> +
> +# Yes, this looks a bit odd. Yes, it ends up being turned on in lots
> +# of cases. Please don't touch it. It is here to handle the case where
> +# a sound driver can be either a module or compiled in if GAMEPORT is
> +# not selected, but must be a module if the joystick is selected as a 
> +# module. The sound driver calls GAMEPORT functions. If GAMEPORT is
> +# not selected, stubs are provided. If GAMEPORT is built in,
> +# everything is fine. If GAMEPORT is a module, however, it would need
> +# to be loaded for the sound driver to be able to link
> +# properly. Therefore, the sound driver must be a module as well in
> +# that case (and the GAMEPORT module must be loaded first). 
>  config SOUND_GAMEPORT
>  	tristate
>  	default y if GAMEPORT!=m

I see. So why did we need the SOUND_GAMEPORT?
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: kernel 2.6.0, wrong Kconfig directives

[Muli Ben-Yehuda]

[-- Attachment #1: Type: text/plain, Size: 1485 bytes --]

On Fri, Dec 26, 2003 at 05:33:06AM +0900, OGAWA Hirofumi wrote:

> > +# Yes, this looks a bit odd. Yes, it ends up being turned on in lots
> > +# of cases. Please don't touch it. It is here to handle the case where
> > +# a sound driver can be either a module or compiled in if GAMEPORT is
> > +# not selected, but must be a module if the joystick is selected as a 
> > +# module. The sound driver calls GAMEPORT functions. If GAMEPORT is
> > +# not selected, stubs are provided. If GAMEPORT is built in,
> > +# everything is fine. If GAMEPORT is a module, however, it would need
> > +# to be loaded for the sound driver to be able to link
> > +# properly. Therefore, the sound driver must be a module as well in
> > +# that case (and the GAMEPORT module must be loaded first). 
> >  config SOUND_GAMEPORT
> >  	tristate
> >  	default y if GAMEPORT!=m
> 
> I see. So why did we need the SOUND_GAMEPORT?

I thought I explained this above, quite verbosely :-) 
Rather than make the sound drivers depend directly on GAMEPORT, which
is troublesome because Kconfig has no provisions for this type of
dependancy, we create an artificial dependency, SOUND_GAMEPORT, which
the sound driver depends on. SOUND_GAMEPORT depends on GAMEPORT, and
Kconfig ends up doing the right thing. I hope that was clearer. 

Cheers, 
Muli 
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/

"the nucleus of linux oscillates my world" - gccbot@#offtopic


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: kernel 2.6.0, wrong Kconfig directives

[OGAWA Hirofumi]

Muli Ben-Yehuda <mulix@mulix.org> writes:

> > I see. So why did we need the SOUND_GAMEPORT?
> 
> I thought I explained this above, quite verbosely :-) 
> Rather than make the sound drivers depend directly on GAMEPORT, which
> is troublesome because Kconfig has no provisions for this type of
> dependancy, we create an artificial dependency, SOUND_GAMEPORT, which
> the sound driver depends on. SOUND_GAMEPORT depends on GAMEPORT, and
> Kconfig ends up doing the right thing. I hope that was clearer. 

Umm... SOUND_GAMEPORT is always true, so GAMEPORT is just option, not
dependency.

It may be good to be documented as the help of Kconfig instead of
SOUND_GAMEPORT...
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: kernel 2.6.0, wrong Kconfig directives

[OGAWA Hirofumi]

OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> writes:

> Muli Ben-Yehuda <mulix@mulix.org> writes:
> 
> > > I see. So why did we need the SOUND_GAMEPORT?
> > 
> > I thought I explained this above, quite verbosely :-) 
> > Rather than make the sound drivers depend directly on GAMEPORT, which
> > is troublesome because Kconfig has no provisions for this type of
> > dependancy, we create an artificial dependency, SOUND_GAMEPORT, which
> > the sound driver depends on. SOUND_GAMEPORT depends on GAMEPORT, and
> > Kconfig ends up doing the right thing. I hope that was clearer. 
> 
> Umm... SOUND_GAMEPORT is always true, so GAMEPORT is just option, not
> dependency.
> 
> It may be good to be documented as the help of Kconfig instead of
> SOUND_GAMEPORT...

Ah, sorry. You are right. I understanded it now.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: kernel 2.6.0, wrong Kconfig directives

[Adrian Bunk]

On Tue, Dec 23, 2003 at 12:56:23AM +0100, Andrea Barisani wrote:
> 
> Hi folks,

Hi Andrea,

> Installing 2.6.0 I've found that some kernel options directives are wrong,
> in fact the option turns out to be always enabled. I don't think this is 
> a desired behaviour.
> 
> Sorry for the format, yes I know it's ugly :) but I'll leave to you the proper 
> solution :) so I can't make a proper patch.
> 
> 
> - IPV6_SCTP___ option is always turned on
> 
> ./net/sctp/Kconfig:
> 
> 8:  config IPV6_SCTP__
> 9: 	    tristate
> 10:         default y if IPV6=n
> 11:	    default IPV6 if IPV6
> 12:
> 13: config IP_SCTP
> 14:	    tristate "The SCTP Protocol (EXPERIMENTAL)"
> 15:	    depends on IPV6_SCTP__
> 
> 
> I think something is wrong here, why the 'default y if IPV6=n' ???


It's ___ugly___ but designed this way...

The whole purpose of IPV6_SCTP__ is to disallow static IP_SCTP with
modular IPV6.


> - INPUT_MOUSEDEV option is always turned on
> 
> ./drivers/input/Kconfig:
> 
> 27: config INPUT_MOUSEDEV
> 28:	    tristate "Mouse interface" if EMBEDDED
> 29:	    default y
> 30:	    depends on INPUT
> 
> 43: config INPUT_MOUSEDEV_PSAUX
> 44:         bool "Provide legacy /dev/psaux device" if EMBEDDED
> 45:         default y
> 46:         depends on INPUT_MOUSEDEV
> 
> 
> the tristate directive is ignored in most default configurations since EMBEDDED
> is not set, however this doesn't allow to disable INPUT_MOUSEDEV and 
> INPUT_MOUSEDEV_PSAUX. I don't suppose this is right.
>...

These EMBEDDED are there to help people not to accidentially disable 
these options although they might require them.

> Bye 
>...

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: kernel 2.6.0, wrong Kconfig directives

[Andrea Barisani]

On Mon, Jan 05, 2004 at 11:17:32PM +0100, Adrian Bunk wrote:
> On Tue, Dec 23, 2003 at 12:56:23AM +0100, Andrea Barisani wrote:
> > 
> > Hi folks,
> 
> Hi Andrea,

Hi!

> 
> > Installing 2.6.0 I've found that some kernel options directives are wrong,
> > in fact the option turns out to be always enabled. I don't think this is 
> > a desired behaviour.
> > 
> > Sorry for the format, yes I know it's ugly :) but I'll leave to you the proper 
> > solution :) so I can't make a proper patch.
> > 
> > 
> > - IPV6_SCTP___ option is always turned on
> > 
> > ./net/sctp/Kconfig:
> > 
> > 8:  config IPV6_SCTP__
> > 9: 	    tristate
> > 10:         default y if IPV6=n
> > 11:	    default IPV6 if IPV6
> > 12:
> > 13: config IP_SCTP
> > 14:	    tristate "The SCTP Protocol (EXPERIMENTAL)"
> > 15:	    depends on IPV6_SCTP__
> > 
> > 
> > I think something is wrong here, why the 'default y if IPV6=n' ???
> 
> 
> It's ___ugly___ but designed this way...
> 
> The whole purpose of IPV6_SCTP__ is to disallow static IP_SCTP with
> modular IPV6.

Ok, so it's actually a "dummy" config with not linked with real code. A
simple grep suggest that. If so I won't complain again, now it's clear ;)

However it's really ___ugly___ ;)

> 
> These EMBEDDED are there to help people not to accidentially disable 
> these options although they might require them.

Ok that's fine, however I personally think this should be documented in
EMBEDDED related options to let people know how they can disable them.

And I really think that MOUSEDEV should not be considered a 'not-standard'
option. It's quite common disabling MOUSE support on servers, terminals and
so on. :)

> cu
> Adrian

Bye and thanks

--
------------------------------------------------------------
INFIS Network Administrator & Security Officer         .*. 
Department of Physics       - University of Trieste     V 
lcars@infis.univ.trieste.it - GPG Key 0x864C9B9E      (   )
----------------------------------------------------  (   )
"How would you know I'm mad?" said Alice.             ^^-^^
"You must be,'said the Cat,'or you wouldn't have come here."
------------------------------------------------------------

Re: kernel 2.6.0, wrong Kconfig directives

[Adrian Bunk]

On Mon, Jan 05, 2004 at 11:59:54PM +0100, Andrea Barisani wrote:
> Hi!

Hi Andrea!

>...
> And I really think that MOUSEDEV should not be considered a 'not-standard'
> option. It's quite common disabling MOUSE support on servers, terminals and
> so on. :)

If the only effect of an unnecessaryly enabled option are a few kB 
wasted space, this isn't a big problem in most situations.

If it's easy to accidentially disable an important option, it's good to 
protect users against doing so.

People who _really_ need the space simply enable EMBEDDED to disable 
such an option.

> Bye and thanks

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed