Fwd: ebtables broken for kernel above 2.6.28

Fwd: ebtables broken for kernel above 2.6.28

[sachin sanap]

Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
works well with 2.6.27). Its mostly a structure alignment issue with
standard targets like DROP,ACCEPT etc. Message seen by user is
"eb_tables: standard target: invalid size 8 != 4"
Iam trying to fix it, this mail is just to check if its already fixed
by someone else?

Thanks
-Sachin

---------- Forwarded message ----------
From: Bart De Schuymer <bdschuym@pandora.be>
Date: Thu, May 28, 2009 at 9:40 PM
Subject: Re: ebtables broken for kernel above 2.6.28
To: sachin sanap <sachin.sanap.1gb@gmail.com>
Cc: nick@fedchik.org.ua, grzes@gnu.univ.gda.pl


You can ask around on netfilter-devel-owner@vger.kernel.org to see if
it has been fixed.

Regards,
Bart


sachin sanap schreef:
>
> Hi ,
> It seems there is some structure alignment problem with the kernel ,
> since the ebtables binary works correctly with 2.6.27 kernel.Iam
> digging into the problem. This mail is just to confirm that its not
> already fixed.
>
> Thanks
> -Sachin
>
>
>

Re: Fwd: ebtables broken for kernel above 2.6.28

[Jan Engelhardt]

On Friday 2009-05-29 07:51, sachin sanap wrote:

>Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
>works well with 2.6.27). Its mostly a structure alignment issue with
>standard targets like DROP,ACCEPT etc. Message seen by user is
>"eb_tables: standard target: invalid size 8 != 4"
>Iam trying to fix it, this mail is just to check if its already fixed
>by someone else?

Indeed this seems to happen, but only when using a mixed-bitness
combination:

/tmp/usr/sbin # LD_LIBRARY_PATH=../lib ./ebtables -A INPUT -j ACCEPT
The kernel doesn't support the ebtables 'filter' table.

/tmp/usr/sbin # uname -a
Linux sovereign 2.6.29.3-jen80-default #1 SMP 2009-04-06 18:10:58 +0200 
x86_64 x86_64 x86_64 GNU/Linux

/tmp/usr/sbin # file ebtables
ebtables: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), for 
GNU/Linux 2.6.4, dynamically linked (uses shared libs), stripped

ebtables does work when it is 64-bit too, so the immediate workaround —
assuming you are on x86 — is to not use a 32-bit userspace with a
64-bit kernel.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Fwd: ebtables broken for kernel above 2.6.28

[sachin sanap]

Sorry dint inform you earlier that iam on 32bit ARM machine.
The real problem is the kernel is computing the alignment on
__alignof__(struct _xt_align) and the userspace ebtables is computing
it on __alignof__(struct ebt_entry_target) , which are two different
structs.

I have the following patch for ebtables-v2.0.8-2 that fixes this.



--- ebtables-v2.0.8-2.orig/include/ebtables_u.h 2009-05-29
21:41:32.000000000 +0530
+++ ebtables-v2.0.8-2/include/ebtables_u.h      2009-05-29
21:43:32.000000000 +0530
@@ -25,6 +25,7 @@
 #define EBTABLES_U_H
 #include <netinet/in.h>
 #include <linux/netfilter_bridge/ebtables.h>
+#include <linux/netfilter/x_tables.h>

 #ifndef IPPROTO_SCTP
 #define IPPROTO_SCTP           132
@@ -37,7 +38,7 @@
 #define EXEC_STYLE_DAEMON      1

 #ifndef EBT_MIN_ALIGN
-#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
+#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
 #endif
 #define EBT_ALIGN(s) (((s) + (EBT_MIN_ALIGN-1)) & ~(EBT_MIN_ALIGN-1))
 #define ERRORMSG_MAXLEN 128


-Sachin




On Fri, May 29, 2009 at 3:32 PM, Jan Engelhardt <jengelh@medozas.de> wrote:
>
> On Friday 2009-05-29 07:51, sachin sanap wrote:
>
>>Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
>>works well with 2.6.27). Its mostly a structure alignment issue with
>>standard targets like DROP,ACCEPT etc. Message seen by user is
>>"eb_tables: standard target: invalid size 8 != 4"
>>Iam trying to fix it, this mail is just to check if its already fixed
>>by someone else?
>
> Indeed this seems to happen, but only when using a mixed-bitness
> combination:
>
> /tmp/usr/sbin # LD_LIBRARY_PATH=../lib ./ebtables -A INPUT -j ACCEPT
> The kernel doesn't support the ebtables 'filter' table.
>
> /tmp/usr/sbin # uname -a
> Linux sovereign 2.6.29.3-jen80-default #1 SMP 2009-04-06 18:10:58 +0200
> x86_64 x86_64 x86_64 GNU/Linux
>
> /tmp/usr/sbin # file ebtables
> ebtables: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), for
> GNU/Linux 2.6.4, dynamically linked (uses shared libs), stripped
>
> ebtables does work when it is 64-bit too, so the immediate workaround —
> assuming you are on x86 — is to not use a 32-bit userspace with a
> 64-bit kernel.
>
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Fwd: ebtables broken for kernel above 2.6.28

[Bart De Schuymer]

Hi,

I think it's best to fix this in the kernel if it's possible, since we 
don't want to break compatibility.
If the fix has to be in userspace, then it must be a fix that still 
works for older kernels.

cheers,
Bart

sachin sanap schreef:
> Sorry dint inform you earlier that iam on 32bit ARM machine.
> The real problem is the kernel is computing the alignment on
> __alignof__(struct _xt_align) and the userspace ebtables is computing
> it on __alignof__(struct ebt_entry_target) , which are two different
> structs.
>
> I have the following patch for ebtables-v2.0.8-2 that fixes this.
>
>
>
> --- ebtables-v2.0.8-2.orig/include/ebtables_u.h 2009-05-29
> 21:41:32.000000000 +0530
> +++ ebtables-v2.0.8-2/include/ebtables_u.h      2009-05-29
> 21:43:32.000000000 +0530
> @@ -25,6 +25,7 @@
>  #define EBTABLES_U_H
>  #include <netinet/in.h>
>  #include <linux/netfilter_bridge/ebtables.h>
> +#include <linux/netfilter/x_tables.h>
>
>  #ifndef IPPROTO_SCTP
>  #define IPPROTO_SCTP           132
> @@ -37,7 +38,7 @@
>  #define EXEC_STYLE_DAEMON      1
>
>  #ifndef EBT_MIN_ALIGN
> -#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
> +#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
>  #endif
>  #define EBT_ALIGN(s) (((s) + (EBT_MIN_ALIGN-1)) & ~(EBT_MIN_ALIGN-1))
>  #define ERRORMSG_MAXLEN 128
>
>
> -Sachin
>
>
>
>
> On Fri, May 29, 2009 at 3:32 PM, Jan Engelhardt <jengelh@medozas.de> wrote:
>   
>> On Friday 2009-05-29 07:51, sachin sanap wrote:
>>
>>     
>>> Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
>>> works well with 2.6.27). Its mostly a structure alignment issue with
>>> standard targets like DROP,ACCEPT etc. Message seen by user is
>>> "eb_tables: standard target: invalid size 8 != 4"
>>> Iam trying to fix it, this mail is just to check if its already fixed
>>> by someone else?
>>>       
>> Indeed this seems to happen, but only when using a mixed-bitness
>> combination:
>>
>> /tmp/usr/sbin # LD_LIBRARY_PATH=../lib ./ebtables -A INPUT -j ACCEPT
>> The kernel doesn't support the ebtables 'filter' table.
>>
>> /tmp/usr/sbin # uname -a
>> Linux sovereign 2.6.29.3-jen80-default #1 SMP 2009-04-06 18:10:58 +0200
>> x86_64 x86_64 x86_64 GNU/Linux
>>
>> /tmp/usr/sbin # file ebtables
>> ebtables: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), for
>> GNU/Linux 2.6.4, dynamically linked (uses shared libs), stripped
>>
>> ebtables does work when it is 64-bit too, so the immediate workaround —
>> assuming you are on x86 — is to not use a 32-bit userspace with a
>> 64-bit kernel.
>>
>>     
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>
>   

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Fwd: ebtables broken for kernel above 2.6.28

[Jan Engelhardt]

On Friday 2009-05-29 18:20, Bart De Schuymer wrote:

> Hi,
>
> I think it's best to fix this in the kernel if it's possible, since we don't
> want to break compatibility.
> If the fix has to be in userspace, then it must be a fix that still works for
> older kernels.

Since the kernel uses xt_align already, it's best for userspace to do the same.

>>
>> #ifndef IPPROTO_SCTP
>> #define IPPROTO_SCTP           132
>> @@ -37,7 +38,7 @@
>> #define EXEC_STYLE_DAEMON      1
>>
>> #ifndef EBT_MIN_ALIGN
>> -#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
>> +#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
>> #endif
>> #define EBT_ALIGN(s) (((s) + (EBT_MIN_ALIGN-1)) & ~(EBT_MIN_ALIGN-1))
>> #define ERRORMSG_MAXLEN 128
>>

Re: Fwd: ebtables broken for kernel above 2.6.28

[Patrick McHardy]

Jan Engelhardt wrote:
> On Friday 2009-05-29 18:20, Bart De Schuymer wrote:
> 
>> I think it's best to fix this in the kernel if it's possible, since we don't
>> want to break compatibility.
>> If the fix has to be in userspace, then it must be a fix that still works for
>> older kernels.
> 
> Since the kernel uses xt_align already, it's best for userspace to do the same.

But that doesn't work for older kernels. Please don't dismiss
compatibility issues that easily. Sometimes things unfortunately
do slip through, but I expect people to do their best to fix the
problem properly when this happens.

Re: ebtables broken for kernel above 2.6.28

[Jan Engelhardt]

On 2009-06-02 12:20:45, Patrick McHardy wrote:
>On Friday 2009-05-29 19:16, Jan Engelhardt wrote:
>>On Friday 2009-05-29 18:20, Bart De Schuymer wrote:
>>>
>>> I think it's best to fix this in the kernel if it's possible,
>>> since we don't want to break compatibility. If the fix has to be
>>> in userspace, then it must be a fix that still works for older
>>> kernels.
>>>
>>> -#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
>>> +#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
>>
>>Since the kernel uses xt_align already, it's best for userspace to
>>do the same.
>
>But that doesn't work for older kernels. Please don't dismiss
>compatibility issues that easily. Sometimes things unfortunately do
>slip through, but I expect people to do their best to fix the
>problem properly when this happens.

But if it gets changed back again to ebt_entry_target, it might
stop working for the combinations it currently works on. :/

Re: ebtables broken for kernel above 2.6.28

[Patrick McHardy]

Jan Engelhardt wrote:
> On 2009-06-02 12:20:45, Patrick McHardy wrote:
>   
>> On Friday 2009-05-29 19:16, Jan Engelhardt wrote:
>>     
>>> On Friday 2009-05-29 18:20, Bart De Schuymer wrote:
>>>       
>>>> I think it's best to fix this in the kernel if it's possible,
>>>> since we don't want to break compatibility. If the fix has to be
>>>> in userspace, then it must be a fix that still works for older
>>>> kernels.
>>>>
>>>> -#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
>>>> +#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
>>>>         
>>> Since the kernel uses xt_align already, it's best for userspace to
>>> do the same.
>>>       
>> But that doesn't work for older kernels. Please don't dismiss
>> compatibility issues that easily. Sometimes things unfortunately do
>> slip through, but I expect people to do their best to fix the
>> problem properly when this happens.
>>     
>
> But if it gets changed back again to ebt_entry_target, it might
> stop working for the combinations it currently works on. :/
>   

Why, ebtables userspace didn't change, which implies it will still work?

Re: Fwd: ebtables broken for kernel above 2.6.28

[Jan Engelhardt]

On Friday 2009-05-29 13:14, sachin sanap wrote:

>Sorry dint inform you earlier that iam on 32bit ARM machine.
>The real problem is the kernel is computing the alignment on
>__alignof__(struct _xt_align) and the userspace ebtables is computing
>it on __alignof__(struct ebt_entry_target) , which are two different
>structs.
>
>I have the following patch for ebtables-v2.0.8-2 that fixes this.
>@@ -37,7 +38,7 @@
> #define EXEC_STYLE_DAEMON      1
>
> #ifndef EBT_MIN_ALIGN
>-#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
>+#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
> #endif
> #define EBT_ALIGN(s) (((s) + (EBT_MIN_ALIGN-1)) & ~(EBT_MIN_ALIGN-1))
> #define ERRORMSG_MAXLEN 128

Can you run this, just to give me a better understanding of ARM32?


#include <stdio.h>
#include <stdint.h>
#define p(t) \
	printf(#t ": %zu\n", __alignof__(t));

int main(void)
{
	p(uint8_t);
	p(uint16_t);
	p(uint32_t);
	p(uint64_t);
	p(float);
	p(double);
	p(void *);
	p(void (*)(void));
}

Re: Fwd: ebtables broken for kernel above 2.6.28

[Sachin Sanap]

On Sun, Jun 14, 2009 at 11:12 PM, Jan Engelhardt<jengelh@medozas.de> wrote:
> On Friday 2009-05-29 13:14, sachin sanap wrote:
>
>>Sorry dint inform you earlier that iam on 32bit ARM machine.
>>The real problem is the kernel is computing the alignment on
>>__alignof__(struct _xt_align) and the userspace ebtables is computing
>>it on __alignof__(struct ebt_entry_target) , which are two different
>>structs.
>>
>>I have the following patch for ebtables-v2.0.8-2 that fixes this.
>>@@ -37,7 +38,7 @@
>> #define EXEC_STYLE_DAEMON      1
>>
>> #ifndef EBT_MIN_ALIGN
>>-#define EBT_MIN_ALIGN (__alignof__(struct ebt_entry_target))
>>+#define EBT_MIN_ALIGN (__alignof__(struct _xt_align))
>> #endif
>> #define EBT_ALIGN(s) (((s) + (EBT_MIN_ALIGN-1)) & ~(EBT_MIN_ALIGN-1))
>> #define ERRORMSG_MAXLEN 128
>
> Can you run this, just to give me a better understanding of ARM32?
>
>
> #include <stdio.h>
> #include <stdint.h>
> #define p(t) \
>        printf(#t ": %zu\n", __alignof__(t));
>
> int main(void)
> {
>        p(uint8_t);
>        p(uint16_t);
>        p(uint32_t);
>        p(uint64_t);
>        p(float);
>        p(double);
>        p(void *);
>        p(void (*)(void));
> }
>
Here is the output from ARM

uint8_t: 1
uint16_t: 2
uint32_t: 4
uint64_t: 8
float: 4
double: 8
void *: 4
void (*)(void): 4

I have CONFIG_AEABI=y

-Sachin Sanap

Re: Fwd: ebtables broken for kernel above 2.6.28

[Peter Volkov]

Could anybody suggest correct fix for this problem? After reading full
thread I'm unsure what to do... BTW, where ebtables (userspace)
development now flows? Does it still uses the sourceforge's CVS server?

Thanks,
--
Peter.

В Птн, 29/05/2009 в 11:21 +0530, sachin sanap пишет:
> Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
> works well with 2.6.27). Its mostly a structure alignment issue with
> standard targets like DROP,ACCEPT etc. Message seen by user is
> "eb_tables: standard target: invalid size 8 != 4"
> Iam trying to fix it, this mail is just to check if its already fixed
> by someone else?
> 
> Thanks
> -Sachin
> 
> ---------- Forwarded message ----------
> From: Bart De Schuymer <bdschuym@pandora.be>
> Date: Thu, May 28, 2009 at 9:40 PM
> Subject: Re: ebtables broken for kernel above 2.6.28
> To: sachin sanap <sachin.sanap.1gb@gmail.com>
> Cc: nick@fedchik.org.ua, grzes@gnu.univ.gda.pl
> 
> 
> You can ask around on netfilter-devel-owner@vger.kernel.org to see if
> it has been fixed.
> 
> Regards,
> Bart
	
-- 
Peter.

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Fwd: ebtables broken for kernel above 2.6.28

[Bart De Schuymer]

The sourceforge page is still the site to obtain the userspace tool. 
Some unreleased code is in the cvs. Kernel changes seem to get in these 
days without checking for compatibility with userspace.
Please fix things up in the kernel code. I sure hope that's possible. I 
won't accept a patch to the userspace code that makes the application no 
longer work on older kernels.

cheers,
Bart

Peter Volkov schreef:
> Could anybody suggest correct fix for this problem? After reading full
> thread I'm unsure what to do... BTW, where ebtables (userspace)
> development now flows? Does it still uses the sourceforge's CVS server?
>
> Thanks,
> --
> Peter.
>
> В Птн, 29/05/2009 в 11:21 +0530, sachin sanap пишет:
>   
>> Iam facing problem with ebtables binay on 2.6.28 kernel (same binary
>> works well with 2.6.27). Its mostly a structure alignment issue with
>> standard targets like DROP,ACCEPT etc. Message seen by user is
>> "eb_tables: standard target: invalid size 8 != 4"
>> Iam trying to fix it, this mail is just to check if its already fixed
>> by someone else?
>>
>> Thanks
>> -Sachin
>>
>> ---------- Forwarded message ----------
>> From: Bart De Schuymer <bdschuym@pandora.be>
>> Date: Thu, May 28, 2009 at 9:40 PM
>> Subject: Re: ebtables broken for kernel above 2.6.28
>> To: sachin sanap <sachin.sanap.1gb@gmail.com>
>> Cc: nick@fedchik.org.ua, grzes@gnu.univ.gda.pl
>>
>>
>> You can ask around on netfilter-devel-owner@vger.kernel.org to see if
>> it has been fixed.
>>
>> Regards,
>> Bart
>>     
> 	
>   

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html