compat-wireless with CONFIG_MAC80211disabled

compat-wireless with CONFIG_MAC80211disabled

[Vitja Makarov]

Hi!

I was trying to build compat wireless for my arm board.
First of all I disabled CONFIG_MAC80211 in my kernel, and successfully
built compat-wireless.

Then at runtime I get kernel panic.
Then I found that sk_buff in 2.6.28 depends on CONFIG_MAC80211, so
enabling it solves my problem.

Shouldn't compat-wireless Makefile check kernel dot config for this
option as well as for wireless extension?

Thanks,
vitja.

Re: compat-wireless with CONFIG_MAC80211disabled

[Luis R. Rodriguez]

On Fri, Feb 13, 2009 at 12:15 AM, Vitja Makarov <vitja.makarov@gmail.com> wrote:
> Hi!
>
> I was trying to build compat wireless for my arm board.
> First of all I disabled CONFIG_MAC80211 in my kernel, and successfully
> built compat-wireless.
>
> Then at runtime I get kernel panic.
> Then I found that sk_buff in 2.6.28 depends on CONFIG_MAC80211, so
> enabling it solves my problem.

Ah nice catch. Well you're the first to try that and report it.

> Shouldn't compat-wireless Makefile check kernel dot config for this
> option as well as for wireless extension?

Yes, Can you send a patch, put some nasty warn on config.mk if its not
enabled on 2.6.28 and error out.

  Luis

Re: compat-wireless with CONFIG_MAC80211disabled

[Vitja Makarov]

[-- Attachment #1: Type: text/plain, Size: 846 bytes --]

2009/2/13 Luis R. Rodriguez <mcgrof@gmail.com>:
> On Fri, Feb 13, 2009 at 12:15 AM, Vitja Makarov <vitja.makarov@gmail.com> wrote:
>> Hi!
>>
>> I was trying to build compat wireless for my arm board.
>> First of all I disabled CONFIG_MAC80211 in my kernel, and successfully
>> built compat-wireless.
>>
>> Then at runtime I get kernel panic.
>> Then I found that sk_buff in 2.6.28 depends on CONFIG_MAC80211, so
>> enabling it solves my problem.
>
> Ah nice catch. Well you're the first to try that and report it.
>
>> Shouldn't compat-wireless Makefile check kernel dot config for this
>> option as well as for wireless extension?
>
> Yes, Can you send a patch, put some nasty warn on config.mk if its not
> enabled on 2.6.28 and error out.
>
>  Luis
>

Here is patch, sorry for attachment, firewall here blocks all email
related ports.

vitja.

[-- Attachment #2: error-on-mac80211-disabled.diff --]
[-- Type: text/x-diff, Size: 1378 bytes --]

diff --git a/config.mk b/config.mk
index 49e82b3..f91f8f1 100644
--- a/config.mk
+++ b/config.mk
@@ -16,8 +16,6 @@ endif
 ifeq ($(CONFIG_MAC80211),y)
 $(error "ERROR: you have MAC80211 compiled into the kernel, CONFIG_MAC80211=y, as such you cannot replace its mac80211 driver. You need this set to CONFIG_MAC80211=m. If you are using Fedora upgrade your kernel as later version should this set as modular. For further information on Fedora see https://bugzilla.redhat.com/show_bug.cgi?id=470143. If you are using your own kernel recompile it and make mac80211 modular")
 endif
-# Wireless subsystem stuff
-CONFIG_MAC80211=m
 
 # We will warn when you don't have MQ support or NET_SCHED enabled.
 #
@@ -44,6 +42,12 @@ $(error "ERROR: Your 2.6.27 kernel has CONFIG_DYNAMIC_FTRACE, please upgrade you
 endif
 endif
 
+ifeq ($(shell test $(KERNEL_SUBLEVEL) -ge 28 && echo yes),yes)
+ifeq ($(CONFIG_MAC80211),)
+$(error "ERROR: Your >=2.6.28 kernel has CONFIG_MAC80211 disabled, you should have it CONFIG_MAC80211=m if you want to use this thing.")
+endif
+endif
+
 ifneq ($(KERNELRELEASE),) # This prevents a warning
 
 ifeq ($(CONFIG_NET_SCHED),)
@@ -59,6 +63,8 @@ endif
 endif # build check
 endif # kernel Makefile check
 
+# Wireless subsystem stuff
+CONFIG_MAC80211=m
 
 # choose between pid and minstrel as default rate control algorithm
 CONFIG_MAC80211_RC_DEFAULT=minstrel

Re: compat-wireless with CONFIG_MAC80211disabled

[Vitja Makarov]

This should be problem for 2.6.27 too.

vitja.

Re: compat-wireless with CONFIG_MAC80211disabled

[Luis R. Rodriguez]

On Fri, Feb 13, 2009 at 4:54 AM, Vitja Makarov <vitja.makarov@gmail.com> wrote:
> This should be problem for 2.6.27 too.

modified for 27 and applied, thanks

Re: compat-wireless with CONFIG_MAC80211disabled

[pat-lkml]

Luis R. Rodriguez wrote:
> On Fri, Feb 13, 2009 at 4:54 AM, Vitja Makarov <vitja.makarov@gmail.com> wrote:
>> This should be problem for 2.6.27 too.
> 
> modified for 27 and applied, thanks

I didn't run into this problem with 2.6.28-gentoo-r1... I've been running 
compat-wireless from git since the 4th on this kernel, with no mac80211
enabled and had no problems.  Is there interest in a patch allowing you to
set something like FORCE=y on the CLI to override these patches?

Pat Erley

Re: compat-wireless with CONFIG_MAC80211disabled

[Luis R. Rodriguez]

On Fri, Feb 13, 2009 at 6:43 PM, pat-lkml <pat-lkml@erley.org> wrote:
> Luis R. Rodriguez wrote:
>> On Fri, Feb 13, 2009 at 4:54 AM, Vitja Makarov <vitja.makarov@gmail.com> wrote:
>>> This should be problem for 2.6.27 too.
>>
>> modified for 27 and applied, thanks
>
> I didn't run into this problem with 2.6.28-gentoo-r1... I've been running
> compat-wireless from git since the 4th on this kernel, with no mac80211
> enabled and had no problems.

Oh interesting, because skb->do_not_encrypt was there on 27, and
skb->requeue was added as of 28. Since config.mk was enforcing it to
be enabled and since compat-wireless uses its own
include/linux/compat_autoconf.h with values derived from config.mk
when building compat-wireless you were getting

#define CONFIG_MAC80211 1

and therefore compat-wireless did build correctly assuming your
kernel's skb_buff had skb->do_not_encrypt and skb->requeue although
the reality of it is that it didn't therefore probably forcing
mac80211 to mess around with other nearby fields in the skb_buff like
the skb->dma_cookie, skb->secmark or skb->mark, depending on whether
CONFIG_NET_DMA, CONFIG_NETWORK_SECMARK or neither were enabled
respectively.

So it seems it was working by chance as your kernel's skbs were being
raped by mac80211 in their sleep.

> Is there interest in a patch allowing you to
> set something like FORCE=y on the CLI to override these patches?

Do you have CONFIG_NET_DMA?
If not do you have CONFIG_NETWORK_SECMARK?

It'd be interesting to see what net core is doing due to the rape and
it'd be even more interesting if we can determine the rape is allowed
during certain circumstances, therefore allowing us to hack config.mk
to allow for such atrocities to go on.

 Luis

Re: compat-wireless with CONFIG_MAC80211disabled

[pat-lkml]

Luis R. Rodriguez wrote:
 >> Is there interest in a patch allowing you to
>> set something like FORCE=y on the CLI to override these patches?
> 
> Do you have CONFIG_NET_DMA?
> If not do you have CONFIG_NETWORK_SECMARK?
> 
> It'd be interesting to see what net core is doing due to the rape and
> it'd be even more interesting if we can determine the rape is allowed
> during certain circumstances, therefore allowing us to hack config.mk
> to allow for such atrocities to go on.

Yes to CONFIG_NETWORK_SECMARK, No to CONFIG_NET_DMA.

Pat

Re: compat-wireless with CONFIG_MAC80211disabled

[Luis R. Rodriguez]

On Fri, Feb 13, 2009 at 7:16 PM, pat-lkml <pat-lkml@erley.org> wrote:
> Luis R. Rodriguez wrote:
>  >> Is there interest in a patch allowing you to
>>> set something like FORCE=y on the CLI to override these patches?
>>
>> Do you have CONFIG_NET_DMA?
>> If not do you have CONFIG_NETWORK_SECMARK?
>>
>> It'd be interesting to see what net core is doing due to the rape and
>> it'd be even more interesting if we can determine the rape is allowed
>> during certain circumstances, therefore allowing us to hack config.mk
>> to allow for such atrocities to go on.
>
> Yes to CONFIG_NETWORK_SECMARK, No to CONFIG_NET_DMA.

Then I take it you haven't yet used this feature:


config NETWORK_SECMARK
        bool "Security Marking"
        help
          This enables security marking of network packets, similar
          to nfmark, but designated for security purposes.
          If you are unsure how to answer this question, answer N.

Which you seem to be able to use if you enable and use:


config NF_CONNTRACK_SECMARK
        bool  'Connection tracking security mark support'
        depends on NETWORK_SECMARK
        default m if NETFILTER_ADVANCED=n
        help
          This option enables security markings to be applied to
          connections.  Typically they are copied to connections from
          packets using the CONNSECMARK target and copied back from
          connections to packets with the same target, with the packets
          being originally labeled via SECMARK.

          If unsure, say 'N'.

or:

config NETFILTER_XT_TARGET_SECMARK
        tristate '"SECMARK" target support'
        depends on NETWORK_SECMARK
        default m if NETFILTER_ADVANCED=n
        help
          The SECMARK target allows security marking of network
          packets, for use with security subsystems.

          To compile it as a module, choose M here.  If unsure, say N.

Not like I have any clue what the hell this is used for despite the explanation.

  Luis