[PATCH 1/2] ethtool: allow const ethtool_ops

[PATCH 1/2] ethtool: allow const ethtool_ops

[Stephen Hemminger]

The ethtool_ops structure is immutable, it expected to be setup
by the driver and is never changed. This patch allows drivers to
declare there ethtool_ops structure read-only.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>


--- linux-2.6.orig/include/linux/netdevice.h
+++ linux-2.6/include/linux/netdevice.h
@@ -342,7 +342,7 @@ struct net_device
 	/* Instance data managed by the core of Wireless Extensions. */
 	struct iw_public_data *	wireless_data;
 
-	struct ethtool_ops *ethtool_ops;
+	const struct ethtool_ops *ethtool_ops;
 
 	/*
 	 * This marks the end of the "visible" part of the structure. All
--- linux-2.6.orig/net/core/ethtool.c
+++ linux-2.6/net/core/ethtool.c
@@ -143,7 +143,7 @@ static int ethtool_set_settings(struct n
 static int ethtool_get_drvinfo(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_drvinfo info;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 
 	if (!ops->get_drvinfo)
 		return -EOPNOTSUPP;
@@ -169,7 +169,7 @@ static int ethtool_get_drvinfo(struct ne
 static int ethtool_get_regs(struct net_device *dev, char __user *useraddr)
 {
 	struct ethtool_regs regs;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	void *regbuf;
 	int reglen, ret;
 
@@ -282,7 +282,7 @@ static int ethtool_get_link(struct net_d
 static int ethtool_get_eeprom(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_eeprom eeprom;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	u8 *data;
 	int ret;
 
@@ -327,7 +327,7 @@ static int ethtool_get_eeprom(struct net
 static int ethtool_set_eeprom(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_eeprom eeprom;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	u8 *data;
 	int ret;
 
@@ -640,7 +640,7 @@ static int ethtool_set_gso(struct net_de
 static int ethtool_self_test(struct net_device *dev, char __user *useraddr)
 {
 	struct ethtool_test test;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	u64 *data;
 	int ret;
 
@@ -673,7 +673,7 @@ static int ethtool_self_test(struct net_
 static int ethtool_get_strings(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_gstrings gstrings;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	u8 *data;
 	int ret;
 
@@ -733,7 +733,7 @@ static int ethtool_phys_id(struct net_de
 static int ethtool_get_stats(struct net_device *dev, void __user *useraddr)
 {
 	struct ethtool_stats stats;
-	struct ethtool_ops *ops = dev->ethtool_ops;
+	const struct ethtool_ops *ops = dev->ethtool_ops;
 	u64 *data;
 	int ret;
 

[PATCH 2/2] loopback: minor statistics optimization

[Stephen Hemminger]

Minor loopback enhancements for 2.6.19

The loopback device status structure is a singleton and doesn't
need to be allocated. Add ethtool_ops hooks to show checksum always on,
and make ethtool_ops const.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>


--- linux-2.6.orig/drivers/net/loopback.c
+++ linux-2.6/drivers/net/loopback.c
@@ -161,15 +161,13 @@ static int loopback_xmit(struct sk_buff 
 	return(0);
 }
 
+static struct net_device_stats loopback_stats;
+
 static struct net_device_stats *get_stats(struct net_device *dev)
 {
-	struct net_device_stats *stats = dev->priv;
+	struct net_device_stats *stats = &loopback_stats;
 	int i;
 
-	if (!stats) {
-		return NULL;
-	}
-
 	memset(stats, 0, sizeof(struct net_device_stats));
 
 	for_each_possible_cpu(i) {
@@ -185,19 +183,28 @@ static struct net_device_stats *get_stat
 	return stats;
 }
 
-static u32 loopback_get_link(struct net_device *dev)
+static u32 always_on(struct net_device *dev)
 {
 	return 1;
 }
 
-static struct ethtool_ops loopback_ethtool_ops = {
-	.get_link		= loopback_get_link,
+static const struct ethtool_ops loopback_ethtool_ops = {
+	.get_link		= always_on,
 	.get_tso		= ethtool_op_get_tso,
 	.set_tso		= ethtool_op_set_tso,
+	.get_tx_csum		= always_on,
+	.get_sg			= always_on,
+	.get_rx_csum		= always_on,
 };
 
+/*
+ * The loopback device is special. There is only one instance and
+ * it is statically allocated. Don't do this for other devices.
+ */
 struct net_device loopback_dev = {
 	.name	 		= "lo",
+	.get_stats		= &get_stats,
+	.priv			= &loopback_stats,
 	.mtu			= (16 * 1024) + 20 + 20 + 12,
 	.hard_start_xmit	= loopback_xmit,
 	.hard_header		= eth_header,
@@ -221,16 +228,6 @@ struct net_device loopback_dev = {
 /* Setup and register the loopback device. */
 int __init loopback_init(void)
 {
-	struct net_device_stats *stats;
-
-	/* Can survive without statistics */
-	stats = kmalloc(sizeof(struct net_device_stats), GFP_KERNEL);
-	if (stats) {
-		memset(stats, 0, sizeof(struct net_device_stats));
-		loopback_dev.priv = stats;
-		loopback_dev.get_stats = &get_stats;
-	}
-	
 	return register_netdev(&loopback_dev);
 };
 

Re: [PATCH 2/2] loopback: minor statistics optimization

[David Miller]

From: Stephen Hemminger <shemminger@osdl.org>
Date: Fri, 8 Sep 2006 11:20:56 -0700

> Minor loopback enhancements for 2.6.19
> 
> The loopback device status structure is a singleton and doesn't
> need to be allocated. Add ethtool_ops hooks to show checksum always on,
> and make ethtool_ops const.
> 
> Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

The ethtool bits are OK, but putting the loopback stats into
the image adds nearly 200 bytes of .data section space on
64-bit.

I don't really think it's that big of a deal to avoid the
kmalloc() call.

Re: [PATCH 2/2] loopback: minor statistics optimization

[Stephen Hemminger]

David Miller wrote:
> From: Stephen Hemminger <shemminger@osdl.org>
> Date: Fri, 8 Sep 2006 11:20:56 -0700
>
>   
>> Minor loopback enhancements for 2.6.19
>>
>> The loopback device status structure is a singleton and doesn't
>> need to be allocated. Add ethtool_ops hooks to show checksum always on,
>> and make ethtool_ops const.
>>
>> Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
>>     
>
> The ethtool bits are OK, but putting the loopback stats into
> the image adds nearly 200 bytes of .data section space on
> 64-bit.
>
> I don't really think it's that big of a deal to avoid the
> kmalloc() call.
>   
Yeah but doing the kmalloc is going to get a bigger chunk (rounded up),
anyway. The statistics should show up in BSS so it won't add to kernel size.

Re: [PATCH 2/2] loopback: minor statistics optimization

[David Miller]

From: Stephen Hemminger <shemminger@osdl.org>
Date: Wed, 27 Sep 2006 20:13:14 -0700

> David Miller wrote:
> > From: Stephen Hemminger <shemminger@osdl.org>
> > Date: Fri, 8 Sep 2006 11:20:56 -0700
> >
> >   
> >> Minor loopback enhancements for 2.6.19
> >>
> >> The loopback device status structure is a singleton and doesn't
> >> need to be allocated. Add ethtool_ops hooks to show checksum always on,
> >> and make ethtool_ops const.
> >>
> >> Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
> >>     
> >
> > The ethtool bits are OK, but putting the loopback stats into
> > the image adds nearly 200 bytes of .data section space on
> > 64-bit.
> >
> > I don't really think it's that big of a deal to avoid the
> > kmalloc() call.
> >   
> Yeah but doing the kmalloc is going to get a bigger chunk (rounded up),
> anyway. The statistics should show up in BSS so it won't add to kernel size.

Good point, I'll apply your patch.

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Jeff Garzik]

Stephen Hemminger wrote:
> The ethtool_ops structure is immutable, it expected to be setup
> by the driver and is never changed. This patch allows drivers to
> declare there ethtool_ops structure read-only.
> 
> Signed-off-by: Stephen Hemminger <shemminger@osdl.org>

ACK, but I need the associated change-all-drivers patch, in order to 
apply this.

	Jeff

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Stephen Hemminger]

On Tue, 12 Sep 2006 12:08:03 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> Stephen Hemminger wrote:
> > The ethtool_ops structure is immutable, it expected to be setup
> > by the driver and is never changed. This patch allows drivers to
> > declare there ethtool_ops structure read-only.
> > 
> > Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
> 
> ACK, but I need the associated change-all-drivers patch, in order to 
> apply this.
> 
> 	Jeff
> 
> 

No, they can pass non-const and everthing works.
Changing the ethtool_ops to have const args would require a global change.

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Jeff Garzik]

Stephen Hemminger wrote:
> No, they can pass non-const and everthing works.
> Changing the ethtool_ops to have const args would require a global change.

Re-stating previous email:  I need a patch which updates all drivers to 
declare their ethtool_ops structs const.

	Jeff

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Stephen Hemminger]

On Tue, 12 Sep 2006 22:12:30 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> Stephen Hemminger wrote:
> > No, they can pass non-const and everthing works.
> > Changing the ethtool_ops to have const args would require a global change.
> 
> Re-stating previous email:  I need a patch which updates all drivers to 
> declare their ethtool_ops structs const.
> 
> 	Jeff
> 
> 

I'll do one mondo patch for all drivers.

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Stephen Hemminger]

On Tue, 12 Sep 2006 12:08:03 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> Stephen Hemminger wrote:
> > The ethtool_ops structure is immutable, it expected to be setup
> > by the driver and is never changed. This patch allows drivers to
> > declare there ethtool_ops structure read-only.
> > 
> > Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
> 
> ACK, but I need the associated change-all-drivers patch, in order to 
> apply this.
> 
> 	Jeff
> 
> 

Actually, no driver is immediately required since this is valid:

exiting_driver.c:

static struct ethtool_ops myops= { ... };

static int mydriver_probe(...) {
...
	dev = alloc_etherdev(sizeof(struct mypriv));
...
	dev->ethtool_ops = &myops;
...
	return register_netdev(dev);
}

Remember difference between:
struct net_device {
 
	struct ethtool_ops *ethtool_ops;

Existing definition, allow network core to change contents of
dev->ethtool_ops.


	const struct ethtool_ops *ethtool_ops;
Proposed patch, network code treats dev->ethtool_ops contents
as read-only. Devices MAY make ethtool_ops const.

Radical change would be:
	struct ethtool_ops *const ethtool_ops;

This would mean every device would have to define ethtool_ops
as const. Devices MUST make ethtool_ops const.

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Jeff Garzik]

How hard can it be to grep?  Can you please just provide the patch??

	Jeff

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Jeff Garzik]

applied, then I fixed up all the drivers

Otherwise, it was a half-complete job that would take years to get 
cleaned up.

Re: [PATCH 1/2] ethtool: allow const ethtool_ops

[Stephen Hemminger]

On Wed, 13 Sep 2006 13:45:42 -0400
Jeff Garzik <jgarzik@pobox.com> wrote:

> applied, then I fixed up all the drivers
> 
> Otherwise, it was a half-complete job that would take years to get 
> cleaned up.
> 
> 

Thanks, I did the same effective patch last night but doing a "make allmodconfig"
build took overnight on a laptop.