Compilation issues using netlink.h

[Matthew Burgess]

Hi,

The test case below simplifies a failure I see when trying to compile
strace using kernel headers from Linux-2.6.32.2:

#include <linux/netlink.h>
struct sockaddr_nl nl;
int main() {
return 0;
}

$ gcc -o test test.c
In file included from test.c:1:
/usr/include/linux/netlink.h:34: error: expected
specifier-qualifier-list before 'sa_family_t'

The comment in netlink.h suggests that 'sa_family_t' is expected to be
found in linux/socket.h, but since commit 9c501935a3 ("net: Support
inclusion of <linux/socket.h> before <sys/socket.h>") that appears to
not be true anymore, it's now in sys/socket.h. Sure enough, if I change
the include in netlink.h to pull in sys/socket.h instead of
linux/socket.h, that enables the test case (and strace) to compile
again, but I wasn't sure if it was really the right thing to do.

Thanks,

Matt.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majo...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

[Ben Hutchings]

On Mon, 2009-12-28 at 00:32 +0000, Matthew Burgess wrote:
> Hi,
>
> The test case below simplifies a failure I see when trying to compile
> strace using kernel headers from Linux-2.6.32.2:
>
> #include <linux/netlink.h>
> struct sockaddr_nl nl;
> int main() {
> return 0;
> }
>
> $ gcc -o test test.c
> In file included from test.c:1:
> /usr/include/linux/netlink.h:34: error: expected
> specifier-qualifier-list before 'sa_family_t'
>
> The comment in netlink.h suggests that 'sa_family_t' is expected to be
> found in linux/socket.h, but since commit 9c501935a3 ("net: Support
> inclusion of <linux/socket.h> before <sys/socket.h>") that appears to
> not be true anymore, it's now in sys/socket.h.

Right. I'm aware of this and suggested that it should be reverted, but
David Miller reckons I was right in the first place.

> Sure enough, if I change
> the include in netlink.h to pull in sys/socket.h instead of
> linux/socket.h, that enables the test case (and strace) to compile
> again, but I wasn't sure if it was really the right thing to do.

I don't think it is - that will bring in many more definitions.

There are actually a whole lot of kernel networking headers which
require sa_family_t when included by user-space code. Until
<linux/socket.h> and <sys/socket.h> cooperate to define sa_family_t
once, you will just have to include <sys/socket.h> before kernel
networking headers.

Ben.

--
Ben Hutchings
The obvious mathematical breakthrough [to break modern encryption] would be
development of an easy way to factor large prime numbers. - Bill Gates