[PATCH iproute2] Fix compilation with kernel headers < 3.4

Since commit 596b1c94aa38e21b7a8c8562e8b61ccb744255d2, iproute2 uses types
__kernel_long_t and __kernel_ulong_t but does not provide internal
definitions for it.

This means that compilation using kernel headers that are older than 3.4
(where these types were added) will fail. This situation may be uncommon for
native compilation, but not uncommon for cross compilation where the
toolchains may be a bit older.

Provide the necessary types internally if not provided by the kernel
headers to fix compilation in such cases.

Co-Developed-by: Serhii Popovych <redacted>
Signed-off-by: Thomas De Schampheleire <redacted>
Signed-off-by: Serhey Popovych <redacted>
---
 include/linux/sysinfo.h |   14 ++++++++++++++
 misc/ss.c               |   10 ++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 include/linux/sysinfo.h

diff --git a/include/linux/sysinfo.h b/include/linux/sysinfo.h
new file mode 100644
index 0000000..766de8d
--- /dev/null
+++ b/include/linux/sysinfo.h

@@ -0,0 +1,14 @@
+#ifndef _SYSINFO_COMPAT_H
+#define _SYSINFO_COMPAT_H
+
+/* In case the kernel header asm/posix_types.h is too old (< 3.4) to provide
+ * __kernel_long_t, provide it here
+ */
+#ifndef __kernel_long_t
+typedef long		__kernel_long_t;
+typedef unsigned long	__kernel_ulong_t;
+#endif
+
+#include_next <linux/sysinfo.h>
+
+#endif /* _SYSINFO_COMPAT_H */

diff --git a/misc/ss.c b/misc/ss.c
index e047f9c..de5a673 100644
--- a/misc/ss.c
+++ b/misc/ss.c

@@ -46,6 +46,16 @@
 #include <linux/sctp.h>
 #include <linux/vm_sockets_diag.h>
 
+#ifndef AF_VSOCK
+#define AF_VSOCK	40
+#if defined(AF_MAX) && AF_MAX < 41
+#undef AF_MAX
+#endif
+#ifndef AF_MAX
+#define AF_MAX		41
+#endif /* AF_MAX */
+#endif /* AF_VSOCK */
+
 #define MAGIC_SEQ 123456
 #define BUF_CHUNK (1024 * 1024)
 #define LEN_ALIGN(x) (((x) + 1) & ~1)

-- 
1.7.10.4

Serhey Popovych wrote:

Since commit 596b1c94aa38e21b7a8c8562e8b61ccb744255d2, iproute2 uses types
__kernel_long_t and __kernel_ulong_t but does not provide internal
definitions for it.

This means that compilation using kernel headers that are older than 3.4
(where these types were added) will fail. This situation may be uncommon for
native compilation, but not uncommon for cross compilation where the
toolchains may be a bit older.

Provide the necessary types internally if not provided by the kernel
headers to fix compilation in such cases.

Co-Developed-by: Serhii Popovych <redacted>
Signed-off-by: Thomas De Schampheleire <redacted>
Signed-off-by: Serhey Popovych <redacted>
---
 include/linux/sysinfo.h |   14 ++++++++++++++
 misc/ss.c               |   10 ++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 include/linux/sysinfo.h

Oops, forgot to mention in subject: this is RFC one.

On Mon, Feb 26, 2018 at 09:46:41PM +0200, Serhey Popovych wrote:

Thanks, this patch works fine in my use case too.

I can't judge the changes in ss.c. I did not see such problems, but my toolchain
is based on 3.3 kernel headers.

Best regards,
Thomas

Thomas De Schampheleire wrote:

On Mon, Feb 26, 2018 at 09:46:41PM +0200, Serhey Popovych wrote:

quoted

Since commit 596b1c94aa38e21b7a8c8562e8b61ccb744255d2, iproute2 uses types
__kernel_long_t and __kernel_ulong_t but does not provide internal
definitions for it.

This means that compilation using kernel headers that are older than 3.4
(where these types were added) will fail. This situation may be uncommon for
native compilation, but not uncommon for cross compilation where the
toolchains may be a bit older.

Provide the necessary types internally if not provided by the kernel
headers to fix compilation in such cases.

Co-Developed-by: Serhii Popovych <redacted>
Signed-off-by: Thomas De Schampheleire <redacted>
Signed-off-by: Serhey Popovych <redacted>
---
 include/linux/sysinfo.h |   14 ++++++++++++++
 misc/ss.c               |   10 ++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 include/linux/sysinfo.h

diff --git a/include/linux/sysinfo.h b/include/linux/sysinfo.h
new file mode 100644
index 0000000..766de8d
--- /dev/null
+++ b/include/linux/sysinfo.h

@@ -0,0 +1,14 @@
+#ifndef _SYSINFO_COMPAT_H
+#define _SYSINFO_COMPAT_H
+
+/* In case the kernel header asm/posix_types.h is too old (< 3.4) to provide
+ * __kernel_long_t, provide it here
+ */
+#ifndef __kernel_long_t
+typedef long		__kernel_long_t;
+typedef unsigned long	__kernel_ulong_t;
+#endif
+
+#include_next <linux/sysinfo.h>
+
+#endif /* _SYSINFO_COMPAT_H */

diff --git a/misc/ss.c b/misc/ss.c
index e047f9c..de5a673 100644
--- a/misc/ss.c
+++ b/misc/ss.c

@@ -46,6 +46,16 @@
 #include <linux/sctp.h>
 #include <linux/vm_sockets_diag.h>
 
+#ifndef AF_VSOCK
+#define AF_VSOCK	40
+#if defined(AF_MAX) && AF_MAX < 41
+#undef AF_MAX
+#endif
+#ifndef AF_MAX
+#define AF_MAX		41
+#endif /* AF_MAX */
+#endif /* AF_VSOCK */
+
 #define MAGIC_SEQ 123456
 #define BUF_CHUNK (1024 * 1024)
 #define LEN_ALIGN(x) (((x) + 1) & ~1)

Thanks, this patch works fine in my use case too.

I can't judge the changes in ss.c. I did not see such problems, but my toolchain
is based on 3.3 kernel headers.

Well, I tested this with 3.2 kernels (oldest supported) in Debian 7
(Wheezy) environment:

ss.c:294:15: error: 'AF_VSOCK' undeclared here (not in a function)
ss.c:323:2: error: array index in initializer not of integer type
ss.c:323:2: error: (near initialization for 'default_afs')
make[1]: *** [ss.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [all] Error 2

Look at your 3.3 headers, probably they provide some redefine for
AF_VSOCK. In stock linux-stable/linux-3.3.y I see no AF_VSOCK, so
probably I will get same error as above (however not checked yet).

Best regards,
Thomas

On Mon, Feb 26, 2018 at 09:46:41PM +0200, Serhey Popovych wrote:

Actually, I now wonder: instead of applying this trick with #include_next on
sysinfo.h, why not do it on linux/types.h ? That would be more correct and more
robust for the future, no?

/Thomas

Thomas De Schampheleire wrote:

On Mon, Feb 26, 2018 at 09:46:41PM +0200, Serhey Popovych wrote:

quoted

Since commit 596b1c94aa38e21b7a8c8562e8b61ccb744255d2, iproute2 uses types
__kernel_long_t and __kernel_ulong_t but does not provide internal
definitions for it.

This means that compilation using kernel headers that are older than 3.4
(where these types were added) will fail. This situation may be uncommon for
native compilation, but not uncommon for cross compilation where the
toolchains may be a bit older.

Provide the necessary types internally if not provided by the kernel
headers to fix compilation in such cases.

Co-Developed-by: Serhii Popovych <redacted>
Signed-off-by: Thomas De Schampheleire <redacted>
Signed-off-by: Serhey Popovych <redacted>
---
 include/linux/sysinfo.h |   14 ++++++++++++++
 misc/ss.c               |   10 ++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 include/linux/sysinfo.h

diff --git a/include/linux/sysinfo.h b/include/linux/sysinfo.h
new file mode 100644
index 0000000..766de8d
--- /dev/null
+++ b/include/linux/sysinfo.h

@@ -0,0 +1,14 @@
+#ifndef _SYSINFO_COMPAT_H
+#define _SYSINFO_COMPAT_H
+
+/* In case the kernel header asm/posix_types.h is too old (< 3.4) to provide
+ * __kernel_long_t, provide it here
+ */
+#ifndef __kernel_long_t
+typedef long		__kernel_long_t;
+typedef unsigned long	__kernel_ulong_t;
+#endif
+
+#include_next <linux/sysinfo.h>
+
+#endif /* _SYSINFO_COMPAT_H */

Actually, I now wonder: instead of applying this trick with #include_next on
sysinfo.h, why not do it on linux/types.h ? That would be more correct and more
robust for the future, no?

No. Headers in include/uapi updated automatically from newer kernel
versions. If we do any hacks in iproute2 uapi directory updating headers
automatically would be problematic.

As for me using #include_next presents better approach.

Better to provide means for compatibility and that I propose to do with
series "ip: Provide compatibility bits to build with old glibc/kernel
headers".

/Thomas