On Wed, Jul 21, 2021 at 9:32 AM Christoph Hellwig [off-list ref] wrote:

quoted

The implementation can be simplified further, based on the
knowledge that the dynamic registration is only ever used

quoted

for IPv4.

I think dropping register_gifconf (which seems like a nice cleanup!)
needs to be a separate prep patch to not make this too confusing.

Right, good idea.

quoted

index e6231837aff5..4727c7a3a988 100644

--- a/include/linux/compat.h
+++ b/include/linux/compat.h

@@ -104,6 +104,11 @@ struct compat_ifmap {
      unsigned char port;
 };

+struct compat_ifconf {
+     compat_int_t    ifc_len;                /* size of buffer */
+     compat_uptr_t   ifcbuf;
+};
+
 #ifdef CONFIG_COMPAT

 #ifndef compat_user_stack_pointer

@@ -326,12 +331,6 @@ typedef struct compat_sigevent {
      } _sigev_un;
 } compat_sigevent_t;

-struct compat_if_settings {
-     unsigned int type;      /* Type of physical device or protocol */
-     unsigned int size;      /* Size of the data allocated by the caller */
-     compat_uptr_t ifs_ifsu; /* union of pointers */
-};

Does this actually compile as-is?  It adds a second definition of
compat_ifconf but removes the still used compat_if_settings?

Indeed it does not. I must have applied a hunk to the wrong patch
in an earlier rebase, and the build bots never picked up on it because
the series was fine in the end.

Maybe it would be a better idea to add a prep patch that makes as much
as possible of compat.h available unconditionally instead of all these
little moves.

Ok, I'll figure something out.

        Arnd