Re: [PATCH] net: Fix struct sock bitfield annotation

[Eric Dumazet <eric.dumazet@gmail.com>]

David Miller a écrit :

> 
> I think from a practical standpoint, you are right.
> 
> But Vegard is right too, as we should be able to put the annotation
> right next to the ":" statements.
> 
> So if you really want why don't you put the sk_protocol and
> sk_type into the ":" block as you mentioned.
> 
> And then you can use Arnaldo's 'pahole' instead of the kludgy
> offsetof() which doesn't work with bitfields :-)
> 
> I want the 8 bytes back just like you, but seperating the annotation
> from the real C bitfields looks definitely wrong to me.


Let's hope nobody wants to use &sk->sk_protocol, &sk->sk_type,
(or offsetof(..., sk_somefield) if that matters)

Only compile tested on 'allyesconfig' build on x86_64

[PATCH] net: Fix struct sock bitfield annotation

Since commit a98b65a3 (net: annotate struct sock bitfield), we lost
8 bytes in struct sock on 64bit arches because of 
kmemcheck_bitfield_end(flags) misplacement.

Fix this by putting together sk_shutdown, sk_no_check, sk_userlocks,
sk_protocol and sk_type in the 'flags' 32bits bitfield

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 include/net/sock.h |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 1621935..9f96394 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -226,12 +226,12 @@ struct sock {
 #define sk_prot			__sk_common.skc_prot
 #define sk_net			__sk_common.skc_net
 	kmemcheck_bitfield_begin(flags);
-	unsigned char		sk_shutdown : 2,
-				sk_no_check : 2,
-				sk_userlocks : 4;
+	unsigned int		sk_shutdown  : 2,
+				sk_no_check  : 2,
+				sk_userlocks : 4,
+				sk_protocol  : 8,
+				sk_type      : 16;
 	kmemcheck_bitfield_end(flags);
-	unsigned char		sk_protocol;
-	unsigned short		sk_type;
 	int			sk_rcvbuf;
 	socket_lock_t		sk_lock;
 	/*