private netdev flags into UAPI?

private netdev flags into UAPI?

[Or Gerlitz]

not sure if this has been brought up before, but I realized that the 
private IFF_yyy netdevice flags which weren't exposed to user space so 
far have been moved to include/uapi/linux/if.h, isn't that wrong?

Or.

> /* Private (from user) interface flags (netdevice->priv_flags). */
> #define IFF_802_1Q_VLAN 0x1             /* 802.1Q VLAN device.          */
> #define IFF_EBRIDGE     0x2             /* Ethernet bridging device.    */
> [...]                                  * change when it's running */

Re: private netdev flags into UAPI?

[David Howells]

Or Gerlitz <ogerlitz@mellanox.com> wrote:

> not sure if this has been brought up before, but I realized that the private
> IFF_yyy netdevice flags which weren't exposed to user space so far have been
> moved to include/uapi/linux/if.h, isn't that wrong?

They were exposed to userspace already.  On my Fedora 16 test box, I see:

	[root@andromeda ~]# grep -r IFF_EBRIDGE /usr/include/linux
	/usr/include/linux/if.h:#define IFF_EBRIDGE     0x2             /* Ethernet bridging device.    */

which is from:

	kernel-headers-3.3.6-3.fc16.x86_64

which was built before UAPI proper started going in:

	[root@andromeda ~]# rpm -qfi /usr/include/linux/if.h | grep "Build Date"
	Build Date  : Wed 16 May 2012 23:05:59 BST

Looking at commit 607ca46e97a1b6594b29647d98a32d545c24bdff, I see:

	--------------------------- include/uapi/linux/if.h ---------------------------
	similarity index 100%
	rename from include/linux/if.h
	rename to include/uapi/linux/if.h

so my script just moved it without alteration.

David

Re: private netdev flags into UAPI?

[Or Gerlitz]

On Mon, Nov 26, 2012 at 11:22 AM, David Howells <dhowells@redhat.com> wrote:
> They were exposed to userspace already

So the script carries the bug into a new directory... why? AFAIK,
intentionally there's no way to read private flags from user space, so
what's the point in defining them there?

Or.

Re: private netdev flags into UAPI?

[David Howells]

Or Gerlitz <or.gerlitz@gmail.com> wrote:

> On Mon, Nov 26, 2012 at 11:22 AM, David Howells <dhowells@redhat.com> wrote:
> > They were exposed to userspace already
> 
> So the script carries the bug into a new directory... why? AFAIK,
> intentionally there's no way to read private flags from user space, so
> what's the point in defining them there?

How should the script know what's private and what's not?  By the
encapsulation of code inside __KERNEL__ blocks.  In their absence, everything
is assumed to be public - given it is already part of the UAPI.  I don't know
that the code is private rather than the comment is wrong.

David

Re: private netdev flags into UAPI?

[Or Gerlitz]

On 27/11/2012 04:03, David Howells wrote:
> Or Gerlitz <or.gerlitz@gmail.com> wrote:
>
>> On Mon, Nov 26, 2012 at 11:22 AM, David Howells <dhowells@redhat.com> wrote:
>>> They were exposed to userspace already
>> So the script carries the bug into a new directory... why? AFAIK,
>> intentionally there's no way to read private flags from user space, so
>> what's the point in defining them there?
> How should the script know what's private and what's not?  By the
> encapsulation of code inside __KERNEL__ blocks.  In their absence, everything
> is assumed to be public - given it is already part of the UAPI.  I don't know
> that the code is private rather than the comment is wrong.
>
>

makes sense, but I have pointed on a bug in the final result, so this 
way or another, the fact that the bug
existed before doesn't mean we should carry it over.

Or.