diff mbox

[2/4] batman-adv: Import sparse fixes from mainline.

Message ID 1272719410-406-3-git-send-email-andrew@lunn.ch (mailing list archive)
State Accepted, archived
Commit 3341201c1fdb41f437cab56543b53a0303724860
Headers

Commit Message

Andrew Lunn May 1, 2010, 1:10 p.m. UTC 
  I fixed sparce warnings as i moved patches from maint to mainline.
This patch move the fixes back to maint.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
---
 bat_sysfs.c |    9 ++++++---
 vis.c       |    8 ++++----
 2 files changed, 10 insertions(+), 7 deletions(-)

Comments

Sven Eckelmann May 1, 2010, 4:34 p.m. UTC | #1 
Andrew Lunn wrote:
> diff --git a/bat_sysfs.c b/bat_sysfs.c
> index 7acf8d8..e2c000b 100644
> --- a/bat_sysfs.c
> +++ b/bat_sysfs.c
> @@ -357,7 +357,8 @@ void sysfs_del_meshif(struct net_device *dev)
>  static ssize_t show_mesh_iface(struct kobject *kobj, struct attribute
> *attr, char *buff)
>  {
> -	struct net_device *net_dev = to_net_dev(to_dev(kobj->parent));
> +	struct device *dev = to_dev(kobj->parent);
> +	struct net_device *net_dev = to_net_dev(dev);
>  	struct batman_if *batman_if = get_batman_if_by_netdev(net_dev);

This has not real something to do with this patch (which is good), but 
to_net_dev was added in v2.6.20-236-g43cb76d and the README still states that 
we are compatible with linux 2.6.20. I think we must change that to 2.6.25 
(explanation below).

I did a build test with 2.6.19-2.6.28 (have to update my buildenv...):

 * 2.6.19: fails in many ways
 * 2.6.20: fails in bat_sys.c
 * 2.6.21: warnings in bat_sys.c (better call it fail)
 * 2.6.22: warnings in bat_sys.c (better call it fail)
 * 2.6.23: fails
 * 2.6.24: fails
 * 2.6.25: ok
 * 2.6.26: ok
 * 2.6.27: ok
 * 2.6.28: ok

But to your patch:

Acked-by: Sven Eckelmann <sven.eckelmann@gmx.de>

Best regards,
	Sven
Andrew Lunn May 1, 2010, 4:48 p.m. UTC | #2 
On Sat, May 01, 2010 at 06:34:56PM +0200, Sven Eckelmann wrote:
> Andrew Lunn wrote:
> > diff --git a/bat_sysfs.c b/bat_sysfs.c
> > index 7acf8d8..e2c000b 100644
> > --- a/bat_sysfs.c
> > +++ b/bat_sysfs.c
> > @@ -357,7 +357,8 @@ void sysfs_del_meshif(struct net_device *dev)
> >  static ssize_t show_mesh_iface(struct kobject *kobj, struct attribute
> > *attr, char *buff)
> >  {
> > -	struct net_device *net_dev = to_net_dev(to_dev(kobj->parent));
> > +	struct device *dev = to_dev(kobj->parent);
> > +	struct net_device *net_dev = to_net_dev(dev);
> >  	struct batman_if *batman_if = get_batman_if_by_netdev(net_dev);
> 
> This has not real something to do with this patch (which is good), but 
> to_net_dev was added in v2.6.20-236-g43cb76d and the README still states that 
> we are compatible with linux 2.6.20. I think we must change that to 2.6.25 
> (explanation below).
> 
> I did a build test with 2.6.19-2.6.28 (have to update my buildenv...):
> 
>  * 2.6.19: fails in many ways
>  * 2.6.20: fails in bat_sys.c
>  * 2.6.21: warnings in bat_sys.c (better call it fail)
>  * 2.6.22: warnings in bat_sys.c (better call it fail)
>  * 2.6.23: fails
>  * 2.6.24: fails
>  * 2.6.25: ok
>  * 2.6.26: ok
>  * 2.6.27: ok
>  * 2.6.28: ok

Simon is good at testing with older kernels and fixing problems. I
guess we can expect a patch to compat.h sometime soon to fix this....

> But to your patch:
> 
> Acked-by: Sven Eckelmann <sven.eckelmann@gmx.de>

Thanks
	Andrew
Marek Lindner May 1, 2010, 5:13 p.m. UTC | #3 
On Sunday 02 May 2010 00:34:56 Sven Eckelmann wrote:
> This has not real something to do with this patch (which is good), but 
> to_net_dev was added in v2.6.20-236-g43cb76d and the README still states
>  that  we are compatible with linux 2.6.20. I think we must change that to
>  2.6.25 (explanation below).

I ran into a bat_sysfs issue today regarding  kobject_create_and_add() on 
2.6.21 and made a fix for it. With that patch it compiles fine but it needs some 
more testing though. I did not see any complaints about to_net_dev. It seems 
to be there: include/linux/netdevice.h line 538. 
We still might have some issues on 2.6.20 ..

Cheers,
Marek
Sven Eckelmann May 1, 2010, 6:10 p.m. UTC | #4 
Marek Lindner wrote:
> On Sunday 02 May 2010 00:34:56 Sven Eckelmann wrote:
> > This has not real something to do with this patch (which is good), but
> > to_net_dev was added in v2.6.20-236-g43cb76d and the README still states
> > 
> >  that  we are compatible with linux 2.6.20. I think we must change that
> >  to 2.6.25 (explanation below).
> 
> I ran into a bat_sysfs issue today regarding  kobject_create_and_add() on
> 2.6.21 and made a fix for it. With that patch it compiles fine but it needs
> some more testing though. I did not see any complaints about to_net_dev.
> It seems to be there: include/linux/netdevice.h line 538.
> We still might have some issues on 2.6.20 ..

Yes, v2.6.20-236-g43cb76d was merged into v2.6.21-rc1.

I started to create a compat lib with the new kobject stuff, but as you have 
done it already... nice reaction time ;)

The v2.6.20 stuff seems even more tricky as the data stored in the structs was 
changed too. The new stuff looks nice, but it gets really ugly if we want to 
support both.

Best regards.
	Sven
diff mbox

Patch

diff --git a/bat_sysfs.c b/bat_sysfs.c
index 7acf8d8..e2c000b 100644
--- a/bat_sysfs.c
+++ b/bat_sysfs.c
@@ -357,7 +357,8 @@  void sysfs_del_meshif(struct net_device *dev)
 static ssize_t show_mesh_iface(struct kobject *kobj, struct attribute *attr,
 			       char *buff)
 {
-	struct net_device *net_dev = to_net_dev(to_dev(kobj->parent));
+	struct device *dev = to_dev(kobj->parent);
+	struct net_device *net_dev = to_net_dev(dev);
 	struct batman_if *batman_if = get_batman_if_by_netdev(net_dev);
 
 	if (!batman_if)
@@ -371,7 +372,8 @@  static ssize_t show_mesh_iface(struct kobject *kobj, struct attribute *attr,
 static ssize_t store_mesh_iface(struct kobject *kobj, struct attribute *attr,
 				char *buff, size_t count)
 {
-	struct net_device *net_dev = to_net_dev(to_dev(kobj->parent));
+	struct device *dev = to_dev(kobj->parent);
+	struct net_device *net_dev = to_net_dev(dev);
 	struct batman_if *batman_if = get_batman_if_by_netdev(net_dev);
 	int status_tmp = -1;
 
@@ -409,7 +411,8 @@  static ssize_t store_mesh_iface(struct kobject *kobj, struct attribute *attr,
 static ssize_t show_iface_status(struct kobject *kobj, struct attribute *attr,
 				 char *buff)
 {
-	struct net_device *net_dev = to_net_dev(to_dev(kobj->parent));
+	struct device *dev = to_dev(kobj->parent);
+	struct net_device *net_dev = to_net_dev(dev);
 	struct batman_if *batman_if = get_batman_if_by_netdev(net_dev);
 
 	if (!batman_if)
diff --git a/vis.c b/vis.c
index 3144439..bac1401 100644
--- a/vis.c
+++ b/vis.c
@@ -103,7 +103,7 @@  static int vis_info_choose(void *data, int size)
 
 /* insert interface to the list of interfaces of one originator, if it
  * does not already exist in the list */
-void vis_data_insert_interface(const uint8_t *interface,
+static void vis_data_insert_interface(const uint8_t *interface,
 				      struct hlist_head *if_list,
 				      bool primary)
 {
@@ -124,7 +124,7 @@  void vis_data_insert_interface(const uint8_t *interface,
 	hlist_add_head(&entry->list, if_list);
 }
 
-ssize_t vis_data_read_prim_sec(char *buff, struct hlist_head *if_list)
+static ssize_t vis_data_read_prim_sec(char *buff, struct hlist_head *if_list)
 {
 	struct if_list_entry *entry;
 	struct hlist_node *pos;
@@ -144,8 +144,8 @@  ssize_t vis_data_read_prim_sec(char *buff, struct hlist_head *if_list)
 }
 
 /* read an entry  */
-ssize_t vis_data_read_entry(char *buff, struct vis_info_entry *entry,
-			    uint8_t *src, bool primary)
+static ssize_t vis_data_read_entry(char *buff, struct vis_info_entry *entry,
+				   uint8_t *src, bool primary)
 {
 	char to[40];