Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

[Kees Cook]

On Thu, Aug 17, 2023 at 01:03:20PM +0800, kernel test robot wrote:
> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next.git testing
> head:   015e2d9101d3713c7bee16dccad171df04a3bbd5
> commit: 61b9e6bd48a6317c0a44ee4f3fecdec9de5baa9e [2/9] netfilter: ebtables: replace zero-length array members
> config: i386-buildonly-randconfig-r004-20230817 (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/config)
> compiler: clang version 16.0.4 (https://github.com/llvm/llvm-project.git ae42196bc493ffe877a7e3dff8be32035dea4d07)
> reproduce: (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/reproduce)
> 
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> | Reported-by: kernel test robot <lkp@intel.com>
> | Closes: https://lore.kernel.org/oe-kbuild-all/202308171249.g1ywxhII-lkp@intel.com/
> 
> All warnings (new ones prefixed by >>):
> 
>    In file included from <built-in>:1:
> >> ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
>            struct ebt_entry_target target;
>                                    ^
>    1 warning generated.

Eww, it looks like "struct ebt_entry_target" is used _within_ another
struct:

struct ebt_standard_target {
        struct ebt_entry_target target;
        int verdict;
};

So "verdict" overlaps with the "data" FAM:

struct ebt_entry_target {
        union { 
                struct {
                        char name[EBT_EXTENSION_MAXNAMELEN];
                        __u8 revision;
                };
                struct xt_target *target;
        } u;
        /* size of data */
        unsigned int target_size;
        unsigned char data[0] __attribute__ ((aligned (__alignof__(struct ebt_replace))));
};  

These have been fixed in the past in a variety of ways -- it all depends
on how userspace is using them. In looking at Debian Code Search:
https://codesearch.debian.net/search?q=struct+ebt_standard_target&literal=1

It is exclusively doing casts and looking at the "verdict" member. So
the easiest conversion might be this:

 struct ebt_standard_target {
-       struct ebt_entry_target target;
+       unsigned char hdr[sizeof(struct ebt_entry_target)];
        int verdict;
 };

Or this might work (not tested):

diff --git a/include/uapi/linux/netfilter_bridge/ebtables.h b/include/uapi/linux/netfilter_bridge/ebtables.h
index a494cf43a755..d6f10163b14a 100644
--- a/include/uapi/linux/netfilter_bridge/ebtables.h
+++ b/include/uapi/linux/netfilter_bridge/ebtables.h
@@ -146,23 +146,25 @@ struct ebt_entry_watcher {
 };
 
 struct ebt_entry_target {
-	union {
-		struct {
-			char name[EBT_EXTENSION_MAXNAMELEN];
-			__u8 revision;
-		};
-		struct xt_target *target;
-	} u;
-	/* size of data */
-	unsigned int target_size;
-	unsigned char data[0] __attribute__ ((aligned (__alignof__(struct ebt_replace))));
+	struct ebt_entry_target_hdr {
+		union {
+			struct {
+				char name[EBT_EXTENSION_MAXNAMELEN];
+				__u8 revision;
+			};
+			struct xt_target *target;
+		} u;
+		/* size of data */
+		unsigned int target_size;
+	};
+	unsigned char data[] __attribute__ ((aligned (__alignof__(struct ebt_replace))));
 };
 
 #define EBT_STANDARD_TARGET "standard"
 struct ebt_standard_target {
-	struct ebt_entry_target target;
-	int verdict;
+	struct ebt_entry_target_hdr target;
+	int verdict __attribute__ ((aligned (__alignof__(struct ebt_replace))));
 };
 
 /* one entry */


-- 
Kees Cook

Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

[Florian Westphal]

Kees Cook <keescook@chromium.org> wrote:
> On Thu, Aug 17, 2023 at 01:03:20PM +0800, kernel test robot wrote:
> > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next.git testing
> > head:   015e2d9101d3713c7bee16dccad171df04a3bbd5
> > commit: 61b9e6bd48a6317c0a44ee4f3fecdec9de5baa9e [2/9] netfilter: ebtables: replace zero-length array members
> > config: i386-buildonly-randconfig-r004-20230817 (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/config)
> > compiler: clang version 16.0.4 (https://github.com/llvm/llvm-project.git ae42196bc493ffe877a7e3dff8be32035dea4d07)
> > reproduce: (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/reproduce)
> > 
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > the same patch/commit), kindly add following tags
> > | Reported-by: kernel test robot <lkp@intel.com>
> > | Closes: https://lore.kernel.org/oe-kbuild-all/202308171249.g1ywxhII-lkp@intel.com/
> > 
> > All warnings (new ones prefixed by >>):
> > 
> >    In file included from <built-in>:1:
> > >> ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
> >            struct ebt_entry_target target;
> >                                    ^
> >    1 warning generated.
> 
> Eww, it looks like "struct ebt_entry_target" is used _within_ another
> struct:
> 
> struct ebt_standard_target {
>         struct ebt_entry_target target;
>         int verdict;
> };

Yes, same as xt_standard_target.

> These have been fixed in the past in a variety of ways -- it all depends
> on how userspace is using them. In looking at Debian Code Search:
> https://codesearch.debian.net/search?q=struct+ebt_standard_target&literal=1
> 
> It is exclusively doing casts and looking at the "verdict" member. So
> the easiest conversion might be this:
> 
>  struct ebt_standard_target {
> -       struct ebt_entry_target target;
> +       unsigned char hdr[sizeof(struct ebt_entry_target)];
>         int verdict;

I don't think its worth doing all of this.

Can't we just keep it as-is and drop the relevant hunk from the patch?

Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

[Kees Cook]

On Thu, Aug 17, 2023 at 10:26:18AM +0200, Florian Westphal wrote:
> Kees Cook <keescook@chromium.org> wrote:
> > On Thu, Aug 17, 2023 at 01:03:20PM +0800, kernel test robot wrote:
> > > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next.git testing
> > > head:   015e2d9101d3713c7bee16dccad171df04a3bbd5
> > > commit: 61b9e6bd48a6317c0a44ee4f3fecdec9de5baa9e [2/9] netfilter: ebtables: replace zero-length array members
> > > config: i386-buildonly-randconfig-r004-20230817 (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/config)
> > > compiler: clang version 16.0.4 (https://github.com/llvm/llvm-project.git ae42196bc493ffe877a7e3dff8be32035dea4d07)
> > > reproduce: (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/reproduce)
> > > 
> > > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > > the same patch/commit), kindly add following tags
> > > | Reported-by: kernel test robot <lkp@intel.com>
> > > | Closes: https://lore.kernel.org/oe-kbuild-all/202308171249.g1ywxhII-lkp@intel.com/
> > > 
> > > All warnings (new ones prefixed by >>):
> > > 
> > >    In file included from <built-in>:1:
> > > >> ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
> > >            struct ebt_entry_target target;
> > >                                    ^
> > >    1 warning generated.
> > 
> > Eww, it looks like "struct ebt_entry_target" is used _within_ another
> > struct:
> > 
> > struct ebt_standard_target {
> >         struct ebt_entry_target target;
> >         int verdict;
> > };
> 
> Yes, same as xt_standard_target.
> 
> > These have been fixed in the past in a variety of ways -- it all depends
> > on how userspace is using them. In looking at Debian Code Search:
> > https://codesearch.debian.net/search?q=struct+ebt_standard_target&literal=1
> > 
> > It is exclusively doing casts and looking at the "verdict" member. So
> > the easiest conversion might be this:
> > 
> >  struct ebt_standard_target {
> > -       struct ebt_entry_target target;
> > +       unsigned char hdr[sizeof(struct ebt_entry_target)];
> >         int verdict;
> 
> I don't think its worth doing all of this.
> 
> Can't we just keep it as-is and drop the relevant hunk from the patch?

For now, yeah. But we'll need to find a solution as flex-array
structures overlapping variables is considered deprecated (it can lead
to ambiguous sizing results). But as long as the kernel itself doesn't
use struct ebt_standard_target, we can kick the can down the road a bit
more. :)

-- 
Kees Cook

Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

[GONG, Ruiqi]

On 2023/08/18 1:02, Kees Cook wrote:
> On Thu, Aug 17, 2023 at 10:26:18AM +0200, Florian Westphal wrote:
>> Kees Cook <keescook@chromium.org> wrote:
>>> On Thu, Aug 17, 2023 at 01:03:20PM +0800, kernel test robot wrote:
>>>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next.git testing
>>>> head:   015e2d9101d3713c7bee16dccad171df04a3bbd5
>>>> commit: 61b9e6bd48a6317c0a44ee4f3fecdec9de5baa9e [2/9] netfilter: ebtables: replace zero-length array members
>>>> config: i386-buildonly-randconfig-r004-20230817 (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/config)
>>>> compiler: clang version 16.0.4 (https://github.com/llvm/llvm-project.git ae42196bc493ffe877a7e3dff8be32035dea4d07)
>>>> reproduce: (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/reproduce)
>>>>
>>>> If you fix the issue in a separate patch/commit (i.e. not just a new version of
>>>> the same patch/commit), kindly add following tags
>>>> | Reported-by: kernel test robot <lkp@intel.com>
>>>> | Closes: https://lore.kernel.org/oe-kbuild-all/202308171249.g1ywxhII-lkp@intel.com/
>>>>
>>>> All warnings (new ones prefixed by >>):
>>>>
>>>>    In file included from <built-in>:1:
>>>>>> ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
>>>>            struct ebt_entry_target target;
>>>>                                    ^
>>>>    1 warning generated.
>>>
>>> Eww, it looks like "struct ebt_entry_target" is used _within_ another
>>> struct:
>>>
>>> struct ebt_standard_target {
>>>         struct ebt_entry_target target;
>>>         int verdict;
>>> };
>>
>> Yes, same as xt_standard_target.
>>
>>> These have been fixed in the past in a variety of ways -- it all depends
>>> on how userspace is using them. In looking at Debian Code Search:
>>> https://codesearch.debian.net/search?q=struct+ebt_standard_target&literal=1
>>>
>>> It is exclusively doing casts and looking at the "verdict" member. So
>>> the easiest conversion might be this:
>>>
>>>  struct ebt_standard_target {
>>> -       struct ebt_entry_target target;
>>> +       unsigned char hdr[sizeof(struct ebt_entry_target)];
>>>         int verdict;
>>
>> I don't think its worth doing all of this.
>>
>> Can't we just keep it as-is and drop the relevant hunk from the patch?
> 
> For now, yeah. But we'll need to find a solution as flex-array
> structures overlapping variables is considered deprecated (it can lead
> to ambiguous sizing results). But as long as the kernel itself doesn't
> use struct ebt_standard_target, we can kick the can down the road a bit
> more. :)
> 

I agree that we can complete the whole transformation step-by-step. I
will make another patch without the problematic replacement.

Re: [netfilter-nf-next:testing 2/9] ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension

[GONG, Ruiqi]

On 2023/08/18 11:24, GONG, Ruiqi wrote:
> 
> 
> On 2023/08/18 1:02, Kees Cook wrote:
>> On Thu, Aug 17, 2023 at 10:26:18AM +0200, Florian Westphal wrote:
>>> Kees Cook <keescook@chromium.org> wrote:
>>>> On Thu, Aug 17, 2023 at 01:03:20PM +0800, kernel test robot wrote:
>>>>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next.git testing
>>>>> head:   015e2d9101d3713c7bee16dccad171df04a3bbd5
>>>>> commit: 61b9e6bd48a6317c0a44ee4f3fecdec9de5baa9e [2/9] netfilter: ebtables: replace zero-length array members
>>>>> config: i386-buildonly-randconfig-r004-20230817 (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/config)
>>>>> compiler: clang version 16.0.4 (https://github.com/llvm/llvm-project.git ae42196bc493ffe877a7e3dff8be32035dea4d07)
>>>>> reproduce: (https://download.01.org/0day-ci/archive/20230817/202308171249.g1ywxhII-lkp@intel.com/reproduce)
>>>>>
>>>>> If you fix the issue in a separate patch/commit (i.e. not just a new version of
>>>>> the same patch/commit), kindly add following tags
>>>>> | Reported-by: kernel test robot <lkp@intel.com>
>>>>> | Closes: https://lore.kernel.org/oe-kbuild-all/202308171249.g1ywxhII-lkp@intel.com/
>>>>>
>>>>> All warnings (new ones prefixed by >>):
>>>>>
>>>>>    In file included from <built-in>:1:
>>>>>>> ./usr/include/linux/netfilter_bridge/ebtables.h:163:26: warning: field 'target' with variable sized type 'struct ebt_entry_target' not at the end of a struct or class is a GNU extension [-Wgnu-variable-sized-type-not-at-end]
>>>>>            struct ebt_entry_target target;
>>>>>                                    ^
>>>>>    1 warning generated.
>>>>
>>>> Eww, it looks like "struct ebt_entry_target" is used _within_ another
>>>> struct:
>>>>
>>>> struct ebt_standard_target {
>>>>         struct ebt_entry_target target;
>>>>         int verdict;
>>>> };
>>>
>>> Yes, same as xt_standard_target.
>>>
>>>> These have been fixed in the past in a variety of ways -- it all depends
>>>> on how userspace is using them. In looking at Debian Code Search:
>>>> https://codesearch.debian.net/search?q=struct+ebt_standard_target&literal=1
>>>>
>>>> It is exclusively doing casts and looking at the "verdict" member. So
>>>> the easiest conversion might be this:
>>>>
>>>>  struct ebt_standard_target {
>>>> -       struct ebt_entry_target target;
>>>> +       unsigned char hdr[sizeof(struct ebt_entry_target)];
>>>>         int verdict;
>>>
>>> I don't think its worth doing all of this.
>>>
>>> Can't we just keep it as-is and drop the relevant hunk from the patch?
>>
>> For now, yeah. But we'll need to find a solution as flex-array
>> structures overlapping variables is considered deprecated (it can lead
>> to ambiguous sizing results). But as long as the kernel itself doesn't
>> use struct ebt_standard_target, we can kick the can down the road a bit
>> more. :)
>>
> 
> I agree that we can complete the whole transformation step-by-step. I
> will make another patch without the problematic replacement.

I didn't notice that the patch had been modified and merged into
nf-next's testing branch. Let's use that and I won't send another one.
Thanks Florian! XD