[PATCH ethtool 0/3] Build fixes for older kernel headers and musl

[PATCH ethtool 0/3] Build fixes for older kernel headers and musl

[Florian Fainelli]

Hi Michal,

These 3 patches fix build issues encountered in the 6.1 release with
either older kernel headers (pre v4.11) or with musl-libc.

In case you want to add a prebuilt toolchain with your release procedure
you can use those binaries:

https://github.com/Broadcom/stbgcc-8.3/releases/tag/stbgcc-8.3-0.4

Florian Fainelli (3):
  misc: Fix build with kernel headers < v4.11
  netlink: Fix maybe uninitialized 'meters' variable
  marvell.c: Fix build with musl-libc

 internal.h       | 1 +
 marvell.c        | 2 +-
 netlink/parser.c | 2 +-
 3 files changed, 3 insertions(+), 2 deletions(-)

-- 
2.34.1

[PATCH ethtool 1/3] misc: Fix build with kernel headers < v4.11

[Florian Fainelli]

Not all toolchain kernel headers may contain upstream commit
2618be7dccf8739b89e1906b64bd8d551af351e6 ("uapi: fix linux/if.h
userspace compilation errors") which is included in v4.11 and onwards.
Err on the side of caution by including sys/socket.h ahead of including
linux/if.h.

Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
Reported-by: Markus Mayer <mmayer@broadcom.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 internal.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/internal.h b/internal.h
index b80f77afa4c0..f7aaaf5229f4 100644
--- a/internal.h
+++ b/internal.h
@@ -21,6 +21,7 @@
 #include <unistd.h>
 #include <endian.h>
 #include <sys/ioctl.h>
+#include <sys/socket.h>
 #include <linux/if.h>
 
 #include "json_writer.h"
-- 
2.34.1

[PATCH ethtool 2/3] netlink: Fix maybe uninitialized 'meters' variable

[Florian Fainelli]

GCC12 warns that 'meters' may be uninitialized, initialize it
accordingly.

Fixes: 9561db9b76f4 ("Add cable test TDR support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 netlink/parser.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/netlink/parser.c b/netlink/parser.c
index f982f229a040..6f863610a490 100644
--- a/netlink/parser.c
+++ b/netlink/parser.c
@@ -237,7 +237,7 @@ int nl_parse_direct_m2cm(struct nl_context *nlctx, uint16_t type,
 			 struct nl_msg_buff *msgbuff, void *dest)
 {
 	const char *arg = *nlctx->argp;
-	float meters;
+	float meters = 0.0;
 	uint32_t cm;
 	int ret;
 
-- 
2.34.1

[PATCH ethtool 3/3] marvell.c: Fix build with musl-libc

[Florian Fainelli]

After commit 1fa60003a8b8 ("misc: header includes cleanup") we stopped
including net/if.h which resolved the proper defines to pull in
sys/types.h and provide a definition for u_int32_t. With musl-libc we
need to define _GNU_SOURCE to ensure that sys/types.h does provide a
definition for u_int32_t.

Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 marvell.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/marvell.c b/marvell.c
index d3d570e4d4ad..be2fc36b8fc5 100644
--- a/marvell.c
+++ b/marvell.c
@@ -6,7 +6,7 @@
  */
 
 #include <stdio.h>
-
+#define _GNU_SOURCE
 #include "internal.h"
 
 static void dump_addr(int n, const u8 *a)
-- 
2.34.1

Re: [PATCH ethtool 1/3] misc: Fix build with kernel headers < v4.11

[Michal Kubecek]

[-- Attachment #1: Type: text/plain, Size: 1127 bytes --]

On Fri, Jan 13, 2023 at 03:31:46PM -0800, Florian Fainelli wrote:
> Not all toolchain kernel headers may contain upstream commit
> 2618be7dccf8739b89e1906b64bd8d551af351e6 ("uapi: fix linux/if.h
> userspace compilation errors") which is included in v4.11 and onwards.
> Err on the side of caution by including sys/socket.h ahead of including
> linux/if.h.
> 
> Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
> Reported-by: Markus Mayer <mmayer@broadcom.com>
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  internal.h | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/internal.h b/internal.h
> index b80f77afa4c0..f7aaaf5229f4 100644
> --- a/internal.h
> +++ b/internal.h
> @@ -21,6 +21,7 @@
>  #include <unistd.h>
>  #include <endian.h>
>  #include <sys/ioctl.h>
> +#include <sys/socket.h>
>  #include <linux/if.h>
>  
>  #include "json_writer.h"

No objection but I wonder if it wouldn't make sense to add linux/if.h to
the header copies in uapi/ instead as then we could also drop the
fallback definition of ALTIFNAMSIZ and perhaps more similar hacks.

Michal

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH ethtool 2/3] netlink: Fix maybe uninitialized 'meters' variable

[Michal Kubecek]

[-- Attachment #1: Type: text/plain, Size: 1191 bytes --]

On Fri, Jan 13, 2023 at 03:31:47PM -0800, Florian Fainelli wrote:
> GCC12 warns that 'meters' may be uninitialized, initialize it
> accordingly.
> 
> Fixes: 9561db9b76f4 ("Add cable test TDR support")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  netlink/parser.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/netlink/parser.c b/netlink/parser.c
> index f982f229a040..6f863610a490 100644
> --- a/netlink/parser.c
> +++ b/netlink/parser.c
> @@ -237,7 +237,7 @@ int nl_parse_direct_m2cm(struct nl_context *nlctx, uint16_t type,
>  			 struct nl_msg_buff *msgbuff, void *dest)
>  {
>  	const char *arg = *nlctx->argp;
> -	float meters;
> +	float meters = 0.0;
>  	uint32_t cm;
>  	int ret;
>  

No problem here either but it's quite surprising as I check build with
gcc versions 7 and 11-13 (10-12 until recently) for each new commit and
I never saw this warning. As the warning is actually incorrect (either
parse_float() returns an error and we bail out or it assigns a value to
meters), it may be a gcc issue that was fixed in a later version. But
initializing the variable does no harm so let's do it.

Michal

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH ethtool 1/3] misc: Fix build with kernel headers < v4.11

[Florian Fainelli]

On 1/13/2023 3:57 PM, Michal Kubecek wrote:
> On Fri, Jan 13, 2023 at 03:31:46PM -0800, Florian Fainelli wrote:
>> Not all toolchain kernel headers may contain upstream commit
>> 2618be7dccf8739b89e1906b64bd8d551af351e6 ("uapi: fix linux/if.h
>> userspace compilation errors") which is included in v4.11 and onwards.
>> Err on the side of caution by including sys/socket.h ahead of including
>> linux/if.h.
>>
>> Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
>> Reported-by: Markus Mayer <mmayer@broadcom.com>
>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>> ---
>>   internal.h | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/internal.h b/internal.h
>> index b80f77afa4c0..f7aaaf5229f4 100644
>> --- a/internal.h
>> +++ b/internal.h
>> @@ -21,6 +21,7 @@
>>   #include <unistd.h>
>>   #include <endian.h>
>>   #include <sys/ioctl.h>
>> +#include <sys/socket.h>
>>   #include <linux/if.h>
>>   
>>   #include "json_writer.h"
> 
> No objection but I wonder if it wouldn't make sense to add linux/if.h to
> the header copies in uapi/ instead as then we could also drop the
> fallback definition of ALTIFNAMSIZ and perhaps more similar hacks.

Humm, I wondered about that but it seems like opening a possible can of 
worms as the history of include/uapi/linux/if.h changed across multiple 
versions. As long as it does compile, why not :)
-- 
Florian

Re: [PATCH ethtool 2/3] netlink: Fix maybe uninitialized 'meters' variable

[Florian Fainelli]

On 1/13/2023 4:09 PM, Michal Kubecek wrote:
> On Fri, Jan 13, 2023 at 03:31:47PM -0800, Florian Fainelli wrote:
>> GCC12 warns that 'meters' may be uninitialized, initialize it
>> accordingly.
>>
>> Fixes: 9561db9b76f4 ("Add cable test TDR support")
>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>> ---
>>   netlink/parser.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/netlink/parser.c b/netlink/parser.c
>> index f982f229a040..6f863610a490 100644
>> --- a/netlink/parser.c
>> +++ b/netlink/parser.c
>> @@ -237,7 +237,7 @@ int nl_parse_direct_m2cm(struct nl_context *nlctx, uint16_t type,
>>   			 struct nl_msg_buff *msgbuff, void *dest)
>>   {
>>   	const char *arg = *nlctx->argp;
>> -	float meters;
>> +	float meters = 0.0;
>>   	uint32_t cm;
>>   	int ret;
>>   
> 
> No problem here either but it's quite surprising as I check build with
> gcc versions 7 and 11-13 (10-12 until recently) for each new commit and
> I never saw this warning. As the warning is actually incorrect (either
> parse_float() returns an error and we bail out or it assigns a value to
> meters), it may be a gcc issue that was fixed in a later version. But
> initializing the variable does no harm so let's do it.

Yes that might be the case, this was seen with GCC 11.3 as well. I 
definitively did use a GCC12 pre-release at some point, too.
-- 
Florian

Re: [PATCH ethtool 3/3] marvell.c: Fix build with musl-libc

[Michal Kubecek]

[-- Attachment #1: Type: text/plain, Size: 1083 bytes --]

On Fri, Jan 13, 2023 at 03:31:48PM -0800, Florian Fainelli wrote:
> After commit 1fa60003a8b8 ("misc: header includes cleanup") we stopped
> including net/if.h which resolved the proper defines to pull in
> sys/types.h and provide a definition for u_int32_t. With musl-libc we
> need to define _GNU_SOURCE to ensure that sys/types.h does provide a
> definition for u_int32_t.
> 
> Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  marvell.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/marvell.c b/marvell.c
> index d3d570e4d4ad..be2fc36b8fc5 100644
> --- a/marvell.c
> +++ b/marvell.c
> @@ -6,7 +6,7 @@
>   */
>  
>  #include <stdio.h>
> -
> +#define _GNU_SOURCE
>  #include "internal.h"
>  
>  static void dump_addr(int n, const u8 *a)

I would prefer replacing u_intXX_t types with standard uintXX_t and
including <stdint.h>. That would be consistent with the rest of the
code which uses ISO uintXX_t types or (older code) kernel uXX types.

Michal

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH ethtool 3/3] marvell.c: Fix build with musl-libc

[Florian Fainelli]

On 1/13/2023 4:13 PM, Michal Kubecek wrote:
> On Fri, Jan 13, 2023 at 03:31:48PM -0800, Florian Fainelli wrote:
>> After commit 1fa60003a8b8 ("misc: header includes cleanup") we stopped
>> including net/if.h which resolved the proper defines to pull in
>> sys/types.h and provide a definition for u_int32_t. With musl-libc we
>> need to define _GNU_SOURCE to ensure that sys/types.h does provide a
>> definition for u_int32_t.
>>
>> Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>> ---
>>   marvell.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/marvell.c b/marvell.c
>> index d3d570e4d4ad..be2fc36b8fc5 100644
>> --- a/marvell.c
>> +++ b/marvell.c
>> @@ -6,7 +6,7 @@
>>    */
>>   
>>   #include <stdio.h>
>> -
>> +#define _GNU_SOURCE
>>   #include "internal.h"
>>   
>>   static void dump_addr(int n, const u8 *a)
> 
> I would prefer replacing u_intXX_t types with standard uintXX_t and
> including <stdint.h>. That would be consistent with the rest of the
> code which uses ISO uintXX_t types or (older code) kernel uXX types.

Sounds good, I will do that in v2.
-- 
Florian

Re: [PATCH ethtool 1/3] misc: Fix build with kernel headers < v4.11

[Florian Fainelli]

On 1/13/2023 4:11 PM, Florian Fainelli wrote:
> 
> 
> On 1/13/2023 3:57 PM, Michal Kubecek wrote:
>> On Fri, Jan 13, 2023 at 03:31:46PM -0800, Florian Fainelli wrote:
>>> Not all toolchain kernel headers may contain upstream commit
>>> 2618be7dccf8739b89e1906b64bd8d551af351e6 ("uapi: fix linux/if.h
>>> userspace compilation errors") which is included in v4.11 and onwards.
>>> Err on the side of caution by including sys/socket.h ahead of including
>>> linux/if.h.
>>>
>>> Fixes: 1fa60003a8b8 ("misc: header includes cleanup")
>>> Reported-by: Markus Mayer <mmayer@broadcom.com>
>>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>>> ---
>>>   internal.h | 1 +
>>>   1 file changed, 1 insertion(+)
>>>
>>> diff --git a/internal.h b/internal.h
>>> index b80f77afa4c0..f7aaaf5229f4 100644
>>> --- a/internal.h
>>> +++ b/internal.h
>>> @@ -21,6 +21,7 @@
>>>   #include <unistd.h>
>>>   #include <endian.h>
>>>   #include <sys/ioctl.h>
>>> +#include <sys/socket.h>
>>>   #include <linux/if.h>
>>>   #include "json_writer.h"
>>
>> No objection but I wonder if it wouldn't make sense to add linux/if.h to
>> the header copies in uapi/ instead as then we could also drop the
>> fallback definition of ALTIFNAMSIZ and perhaps more similar hacks.
> 
> Humm, I wondered about that but it seems like opening a possible can of 
> worms as the history of include/uapi/linux/if.h changed across multiple 
> versions. As long as it does compile, why not :)

Just to be clear, I will give that one a shot and let you know how it 
goes, if it works fine, then it will be in v2.
-- 
Florian