[2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Adrian Bunk]

On Wed, Feb 23, 2005 at 01:42:33AM -0800, Andrew Morton wrote:
>...
> Changes since 2.6.11-rc3-mm1:
>...
>  bk-netdev.patch
>...


Some of the options that needlessly wrote in their help text which 
options they do selct (patch already sent) didn't obey the most 
important rule of select

  If you select something, you have to ensure that the dependencies
  of what you do select are fulfilled.

resulting in the following compile error:


<--  snip  -->

...
  LD      .tmp_vmlinux1
crypto/built-in.o(.init.text+0x31b): In function `aes_init':
: undefined reference to `crypto_register_alg'
crypto/built-in.o(.init.text+0x326): In function `michael_mic_init':
: undefined reference to `crypto_register_alg'
crypto/built-in.o(.exit.text+0x6): In function `aes_fini':
: undefined reference to `crypto_unregister_alg'
crypto/built-in.o(.exit.text+0x16): In function `michael_mic_exit':
: undefined reference to `crypto_unregister_alg'
net/built-in.o(.text+0x5ba52): In function `ieee80211_ccmp_init':
: undefined reference to `crypto_alloc_tfm'
net/built-in.o(.text+0x5ba94): In function `ieee80211_ccmp_init':
: undefined reference to `crypto_free_tfm'
net/built-in.o(.text+0x5bab7): In function `ieee80211_ccmp_deinit':
: undefined reference to `crypto_free_tfm'
net/built-in.o(.text+0x5c5c2): In function `ieee80211_tkip_init':
: undefined reference to `crypto_alloc_tfm'
net/built-in.o(.text+0x5c5d5): In function `ieee80211_tkip_init':
: undefined reference to `crypto_alloc_tfm'
net/built-in.o(.text+0x5c623): In function `ieee80211_tkip_init':
: undefined reference to `crypto_free_tfm'
net/built-in.o(.text+0x5c62a): In function `ieee80211_tkip_init':
: undefined reference to `crypto_free_tfm'
net/built-in.o(.text+0x5c65e): In function `ieee80211_tkip_deinit':
: undefined reference to `crypto_free_tfm'
net/built-in.o(.text+0x5c665): In function `ieee80211_tkip_deinit':
: undefined reference to `crypto_free_tfm'
make: *** [.tmp_vmlinux1] Error 1

<--  snip  -->


This patch adds the missing selects of CRYPTO.


--- linux-2.6.11-rc4-mm1-full/net/ieee80211/Kconfig.old	2005-02-26 12:12:44.000000000 +0100
+++ linux-2.6.11-rc4-mm1-full/net/ieee80211/Kconfig	2005-02-26 12:13:47.000000000 +0100
@@ -42,10 +42,11 @@
 	"ieee80211_crypt_wep".
 
 config IEEE80211_CRYPT_CCMP
 	tristate "IEEE 802.11i CCMP support"
 	depends on IEEE80211
+	select CRYPTO
 	select CRYPTO_AES
 	---help---
 	Include software based cipher suites in support of IEEE 802.11i 
 	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
 	networks.
@@ -54,10 +55,11 @@
 	"ieee80211_crypt_ccmp".
 
 config IEEE80211_CRYPT_TKIP
 	tristate "IEEE 802.11i TKIP encryption"
 	depends on IEEE80211
+	select CRYPTO
 	select CRYPTO_MICHAEL_MIC
 	---help---
 	Include software based cipher suites in support of IEEE 802.11i 
 	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with TKIP enabled 
 	networks.

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Jeff Garzik]

Adrian Bunk wrote:
> +	select CRYPTO
>  	select CRYPTO_AES
>  	---help---
>  	Include software based cipher suites in support of IEEE 802.11i 
>  	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
>  	networks.
> @@ -54,10 +55,11 @@
>  	"ieee80211_crypt_ccmp".
>  
>  config IEEE80211_CRYPT_TKIP
>  	tristate "IEEE 802.11i TKIP encryption"
>  	depends on IEEE80211
> +	select CRYPTO
>  	select CRYPTO_MICHAEL_MIC


'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.

	Jeff

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Adrian Bunk]

On Wed, Mar 02, 2005 at 01:43:04AM -0500, Jeff Garzik wrote:
> Adrian Bunk wrote:
> >+	select CRYPTO
> > 	select CRYPTO_AES
> > 	---help---
> > 	Include software based cipher suites in support of IEEE 802.11i 
> > 	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
> > 	networks.
> >@@ -54,10 +55,11 @@
> > 	"ieee80211_crypt_ccmp".
> > 
> > config IEEE80211_CRYPT_TKIP
> > 	tristate "IEEE 802.11i TKIP encryption"
> > 	depends on IEEE80211
> >+	select CRYPTO
> > 	select CRYPTO_MICHAEL_MIC
> 
> 
> 'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.

This would result in a recursive dependency.

> 	Jeff

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: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Jeff Garzik]

Adrian Bunk wrote:
> On Wed, Mar 02, 2005 at 01:43:04AM -0500, Jeff Garzik wrote:
> 
>>Adrian Bunk wrote:
>>
>>>+	select CRYPTO
>>>	select CRYPTO_AES
>>>	---help---
>>>	Include software based cipher suites in support of IEEE 802.11i 
>>>	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
>>>	networks.
>>>@@ -54,10 +55,11 @@
>>>	"ieee80211_crypt_ccmp".
>>>
>>>config IEEE80211_CRYPT_TKIP
>>>	tristate "IEEE 802.11i TKIP encryption"
>>>	depends on IEEE80211
>>>+	select CRYPTO
>>>	select CRYPTO_MICHAEL_MIC
>>
>>
>>'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.
> 
> 
> This would result in a recursive dependency.

No, it wouldn't.  CRYPTO_AES depends on CRYPTO, which depends on nothing.

	Jeff

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Andrew Morton]

Jeff Garzik <jgarzik@pobox.com> wrote:
>
> Adrian Bunk wrote:
> > On Wed, Mar 02, 2005 at 01:43:04AM -0500, Jeff Garzik wrote:
> > 
> >>Adrian Bunk wrote:
> >>
> >>>+	select CRYPTO
> >>>	select CRYPTO_AES
> >>>	---help---
> >>>	Include software based cipher suites in support of IEEE 802.11i 
> >>>	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
> >>>	networks.
> >>>@@ -54,10 +55,11 @@
> >>>	"ieee80211_crypt_ccmp".
> >>>
> >>>config IEEE80211_CRYPT_TKIP
> >>>	tristate "IEEE 802.11i TKIP encryption"
> >>>	depends on IEEE80211
> >>>+	select CRYPTO
> >>>	select CRYPTO_MICHAEL_MIC
> >>
> >>
> >>'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.
> > 
> > 
> > This would result in a recursive dependency.
> 
> No, it wouldn't.  CRYPTO_AES depends on CRYPTO, which depends on nothing.
> 

Thing is, CRYPTO_AES on only selectable on x86.

So really, IEEE80211_CRYPT_CCMP should depend upon CRYPTO_AES rather than
selecting it.  But that confuses users.

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Jeff Garzik]

Andrew Morton wrote:
> Jeff Garzik <jgarzik@pobox.com> wrote:
> 
>>Adrian Bunk wrote:
>>
>>>On Wed, Mar 02, 2005 at 01:43:04AM -0500, Jeff Garzik wrote:
>>>
>>>
>>>>Adrian Bunk wrote:
>>>>
>>>>
>>>>>+	select CRYPTO
>>>>>	select CRYPTO_AES
>>>>>	---help---
>>>>>	Include software based cipher suites in support of IEEE 802.11i 
>>>>>	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
>>>>>	networks.
>>>>>@@ -54,10 +55,11 @@
>>>>>	"ieee80211_crypt_ccmp".
>>>>>
>>>>>config IEEE80211_CRYPT_TKIP
>>>>>	tristate "IEEE 802.11i TKIP encryption"
>>>>>	depends on IEEE80211
>>>>>+	select CRYPTO
>>>>>	select CRYPTO_MICHAEL_MIC
>>>>
>>>>
>>>>'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.
>>>
>>>
>>>This would result in a recursive dependency.
>>
>>No, it wouldn't.  CRYPTO_AES depends on CRYPTO, which depends on nothing.
>>
> 
> 
> Thing is, CRYPTO_AES on only selectable on x86.

You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
the dependencies are a bit weird:

config CRYPTO_AES
         tristate "AES cipher algorithms"
         depends on CRYPTO && !(X86 && !X86_64)
config CRYPTO_AES_586
         tristate "AES cipher algorithms (i586)"
         depends on CRYPTO && (X86 && !X86_64)

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Andrew Morton]

Jeff Garzik <jgarzik@pobox.com> wrote:
>
> > Thing is, CRYPTO_AES on only selectable on x86.
> 
>  You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
>  the dependencies are a bit weird:
> 
>  config CRYPTO_AES
>           tristate "AES cipher algorithms"
>           depends on CRYPTO && !(X86 && !X86_64)
>  config CRYPTO_AES_586
>           tristate "AES cipher algorithms (i586)"
>           depends on CRYPTO && (X86 && !X86_64)

That's pretty broken, isn't it?

Would be better to just do:

config CRYPTO_AES
	select CRYPTO_AES_586 if (X86 && !X86_64)
	select CRYPTO_AES_OTHER if !(X86 && !X86_64)

and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Adrian Bunk]

On Wed, Mar 02, 2005 at 01:18:17PM -0800, Andrew Morton wrote:
> Jeff Garzik <jgarzik@pobox.com> wrote:
> >
> > > Thing is, CRYPTO_AES on only selectable on x86.
> > 
> >  You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
> >  the dependencies are a bit weird:
> > 
> >  config CRYPTO_AES
> >           tristate "AES cipher algorithms"
> >           depends on CRYPTO && !(X86 && !X86_64)
> >  config CRYPTO_AES_586
> >           tristate "AES cipher algorithms (i586)"
> >           depends on CRYPTO && (X86 && !X86_64)
> 
> That's pretty broken, isn't it?
> 
> Would be better to just do:
> 
> config CRYPTO_AES
> 	select CRYPTO_AES_586 if (X86 && !X86_64)
> 	select CRYPTO_AES_OTHER if !(X86 && !X86_64)
> 
> and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.


  http://www.ussg.iu.edu/hypermail/linux/kernel/0502.3/0518.html
  http://www.ussg.iu.edu/hypermail/linux/kernel/0502.3/0523.html
  

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: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Adrian Bunk]

On Wed, Mar 02, 2005 at 02:12:04PM -0500, Jeff Garzik wrote:
> Adrian Bunk wrote:
> >On Wed, Mar 02, 2005 at 01:43:04AM -0500, Jeff Garzik wrote:
> >
> >>Adrian Bunk wrote:
> >>
> >>>+	select CRYPTO
> >>>	select CRYPTO_AES
> >>>	---help---
> >>>	Include software based cipher suites in support of IEEE 802.11i 
> >>>	(aka TGi, WPA, WPA2, WPA-PSK, etc.) for use with CCMP enabled 
> >>>	networks.
> >>>@@ -54,10 +55,11 @@
> >>>	"ieee80211_crypt_ccmp".
> >>>
> >>>config IEEE80211_CRYPT_TKIP
> >>>	tristate "IEEE 802.11i TKIP encryption"
> >>>	depends on IEEE80211
> >>>+	select CRYPTO
> >>>	select CRYPTO_MICHAEL_MIC
> >>
> >>
> >>'select CRYPTO_AES' should 'select CRYPTO' automatically, I would hope.
> >
> >
> >This would result in a recursive dependency.
> 
> No, it wouldn't.  CRYPTO_AES depends on CRYPTO, which depends on nothing.

Exactly.

And if CRYPTO_AES would select CRYPTO, you'd have a recursive 
dependency.

The only possible thing would be to change all dependencies on CRYPTO to 
selects. This wouldn't be unlogical since the whole crypto subsystem is 
only a helper for other subsystems.

James, any opinions on this issue?

> 	Jeff

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: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Andrew Morton]

Adrian Bunk <bunk@stusta.de> wrote:
>
> > Would be better to just do:
> > 
> > config CRYPTO_AES
> > 	select CRYPTO_AES_586 if (X86 && !X86_64)
> > 	select CRYPTO_AES_OTHER if !(X86 && !X86_64)
> > 
> > and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.
> 
> 
>   http://www.ussg.iu.edu/hypermail/linux/kernel/0502.3/0518.html

Please resubmit.

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Jeff Garzik]

Andrew Morton wrote:
> Jeff Garzik <jgarzik@pobox.com> wrote:
> 
>>>Thing is, CRYPTO_AES on only selectable on x86.
>>
>> You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
>> the dependencies are a bit weird:
>>
>> config CRYPTO_AES
>>          tristate "AES cipher algorithms"
>>          depends on CRYPTO && !(X86 && !X86_64)
>> config CRYPTO_AES_586
>>          tristate "AES cipher algorithms (i586)"
>>          depends on CRYPTO && (X86 && !X86_64)
> 
> 
> That's pretty broken, isn't it?
> 
> Would be better to just do:
> 
> config CRYPTO_AES
> 	select CRYPTO_AES_586 if (X86 && !X86_64)
> 	select CRYPTO_AES_OTHER if !(X86 && !X86_64)
> 
> and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.

Not really that easy.  For x86 we have

	aes
	aes-586
	aes-via

And my own personal custom-kernel preference is to use the C version of 
the code on my x86 and x86-64 boxes.

	Jeff

Re: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Adrian Bunk]

On Wed, Mar 02, 2005 at 05:41:50PM -0500, Jeff Garzik wrote:
> Andrew Morton wrote:
> >Jeff Garzik <jgarzik@pobox.com> wrote:
> >
> >>>Thing is, CRYPTO_AES on only selectable on x86.
> >>
> >>You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
> >>the dependencies are a bit weird:
> >>
> >>config CRYPTO_AES
> >>         tristate "AES cipher algorithms"
> >>         depends on CRYPTO && !(X86 && !X86_64)
> >>config CRYPTO_AES_586
> >>         tristate "AES cipher algorithms (i586)"
> >>         depends on CRYPTO && (X86 && !X86_64)
> >
> >
> >That's pretty broken, isn't it?
> >
> >Would be better to just do:
> >
> >config CRYPTO_AES
> >	select CRYPTO_AES_586 if (X86 && !X86_64)
> >	select CRYPTO_AES_OTHER if !(X86 && !X86_64)
> >
> >and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.
> 
> Not really that easy.  For x86 we have
> 
> 	aes
> 	aes-586
> 	aes-via

Where is aes-via?

> And my own personal custom-kernel preference is to use the C version of 
> the code on my x86 and x86-64 boxes.

That's already not possible today.

> 	Jeff

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: [2.6.11-rc4-mm1 patch] fix buggy IEEE80211_CRYPT_* selects

[Jeff Garzik]

Adrian Bunk wrote:
> On Wed, Mar 02, 2005 at 05:41:50PM -0500, Jeff Garzik wrote:
> 
>>Andrew Morton wrote:
>>
>>>Jeff Garzik <jgarzik@pobox.com> wrote:
>>>
>>>
>>>>>Thing is, CRYPTO_AES on only selectable on x86.
>>>>
>>>>You're thinking about CRYPTO_AES_586.  But looking at crypto/Kconfig, 
>>>>the dependencies are a bit weird:
>>>>
>>>>config CRYPTO_AES
>>>>        tristate "AES cipher algorithms"
>>>>        depends on CRYPTO && !(X86 && !X86_64)
>>>>config CRYPTO_AES_586
>>>>        tristate "AES cipher algorithms (i586)"
>>>>        depends on CRYPTO && (X86 && !X86_64)
>>>
>>>
>>>That's pretty broken, isn't it?
>>>
>>>Would be better to just do:
>>>
>>>config CRYPTO_AES
>>>	select CRYPTO_AES_586 if (X86 && !X86_64)
>>>	select CRYPTO_AES_OTHER if !(X86 && !X86_64)
>>>
>>>and hide CRYPTO_AES_586 and CRYPTO_AES_OTHER from the outside world.
>>
>>Not really that easy.  For x86 we have
>>
>>	aes
>>	aes-586
>>	aes-via
> 
> 
> Where is aes-via?

drivers/crypto


>>And my own personal custom-kernel preference is to use the C version of 
>>the code on my x86 and x86-64 boxes.
> 
> 
> That's already not possible today.

It should be.

	Jeff

How to handle the multiple aes variants on i386?

[Adrian Bunk]

On Wed, Mar 02, 2005 at 05:49:26PM -0500, Jeff Garzik wrote:
> Adrian Bunk wrote:
> >On Wed, Mar 02, 2005 at 05:41:50PM -0500, Jeff Garzik wrote:
>...
> >>Not really that easy.  For x86 we have
> >>
> >>	aes
> >>	aes-586
> >>	aes-via
> >
> >
> >Where is aes-via?
> 
> drivers/crypto
> 
> 
> >>And my own personal custom-kernel preference is to use the C version of 
> >>the code on my x86 and x86-64 boxes.
> >
> >
> >That's already not possible today.
> 
> It should be.

OK, rethinking about it, your arguments sound reasonable.

Could anyone explain, what exactly happens if multiple "aes" algorithms 
are compiled into the kernel?

Choosing between the i386 asm and the generic versions seems easy, bug 
the VIA Padlock case sounds more tricky since it works only on a subset 
of the i386 architecture.

> 	Jeff

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