[Intel-wired-lan] [PATCH 0/2 net-next v2] Bring sizeof(net_device) down to < 2K bytes

[Intel-wired-lan] [PATCH 0/2 net-next v2] Bring sizeof(net_device) down to < 2K bytes

[Thomas Graf]

The size of struct net_device crossed the 2K boundary a while ago which
is a waste in combination with many net namespaces. This series brings
the size of struct net_device down to well below 2K in total size with
a typical configuration. Some reserves a several holes leave room for
further expansion.

Before:
/* size: 2176, cachelines: 34, members: 121 */

After:
/* size: 1984, cachelines: 31, members: 120 */


Thomas Graf (2):
  e1000e: Move pm_qos_req to e1000e adapter
  net_device: Reorder members to fill holes

 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 +
 drivers/net/ethernet/intel/e1000e/netdev.c |  8 +++---
 include/linux/netdevice.h                  | 46 ++++++++++++++----------------
 3 files changed, 26 insertions(+), 29 deletions(-)

---
v2:
 - Moved pm_qos_req to e1000e adapter entirely as suggested by Daniel
 - Dropped patch 2 based on comment from Eric. __alignof__ is not
   reliable to derive alignment of kmalloc returned pointer.

[PATCH 0/2 net-next v2] Bring sizeof(net_device) down to < 2K bytes

[Thomas Graf]

The size of struct net_device crossed the 2K boundary a while ago which
is a waste in combination with many net namespaces. This series brings
the size of struct net_device down to well below 2K in total size with
a typical configuration. Some reserves a several holes leave room for
further expansion.

Before:
/* size: 2176, cachelines: 34, members: 121 */

After:
/* size: 1984, cachelines: 31, members: 120 */


Thomas Graf (2):
  e1000e: Move pm_qos_req to e1000e adapter
  net_device: Reorder members to fill holes

 drivers/net/ethernet/intel/e1000e/e1000.h  |  1 +
 drivers/net/ethernet/intel/e1000e/netdev.c |  8 +++---
 include/linux/netdevice.h                  | 46 ++++++++++++++----------------
 3 files changed, 26 insertions(+), 29 deletions(-)

---
v2:
 - Moved pm_qos_req to e1000e adapter entirely as suggested by Daniel
 - Dropped patch 2 based on comment from Eric. __alignof__ is not
   reliable to derive alignment of kmalloc returned pointer.

[Intel-wired-lan] [PATCH 1/2 net-next] e1000e: Move pm_qos_req to e1000e adapter

[Thomas Graf]

e1000e is the only driver requiring pm_qos_req, instead of causing
every device to waste up to 240 bytes. Allocate it for the specific
driver.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
 drivers/net/ethernet/intel/e1000e/e1000.h  | 1 +
 drivers/net/ethernet/intel/e1000e/netdev.c | 8 ++++----
 include/linux/netdevice.h                  | 1 -
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/intel/e1000e/e1000.h b/drivers/net/ethernet/intel/e1000e/e1000.h
index a69f09e..5d9ceb1 100644
--- a/drivers/net/ethernet/intel/e1000e/e1000.h
+++ b/drivers/net/ethernet/intel/e1000e/e1000.h
@@ -343,6 +343,7 @@ struct e1000_adapter {
 	struct timecounter tc;
 	struct ptp_clock *ptp_clock;
 	struct ptp_clock_info ptp_clock_info;
+	struct pm_qos_request pm_qos_req;
 
 	u16 eee_advert;
 };
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c
index 74ec185..c509a5c 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -3297,9 +3297,9 @@ static void e1000_configure_rx(struct e1000_adapter *adapter)
 			ew32(RXDCTL(0), rxdctl | 0x3);
 		}
 
-		pm_qos_update_request(&adapter->netdev->pm_qos_req, lat);
+		pm_qos_update_request(&adapter->pm_qos_req, lat);
 	} else {
-		pm_qos_update_request(&adapter->netdev->pm_qos_req,
+		pm_qos_update_request(&adapter->pm_qos_req,
 				      PM_QOS_DEFAULT_VALUE);
 	}
 
@@ -4403,7 +4403,7 @@ static int e1000_open(struct net_device *netdev)
 		e1000_update_mng_vlan(adapter);
 
 	/* DMA latency requirement to workaround jumbo issue */
-	pm_qos_add_request(&adapter->netdev->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
+	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
 			   PM_QOS_DEFAULT_VALUE);
 
 	/* before we allocate an interrupt, we must be ready to handle it.
@@ -4514,7 +4514,7 @@ static int e1000_close(struct net_device *netdev)
 	    !test_bit(__E1000_TESTING, &adapter->state))
 		e1000e_release_hw_control(adapter);
 
-	pm_qos_remove_request(&adapter->netdev->pm_qos_req);
+	pm_qos_remove_request(&adapter->pm_qos_req);
 
 	pm_runtime_put_sync(&pdev->dev);
 
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index bf6d9df..d66c0f9 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1743,7 +1743,6 @@ struct net_device {
 #endif
 	struct phy_device *phydev;
 	struct lock_class_key *qdisc_tx_busylock;
-	struct pm_qos_request	pm_qos_req;
 };
 #define to_net_dev(d) container_of(d, struct net_device, dev)
 
-- 
1.9.3

[PATCH 1/2 net-next] e1000e: Move pm_qos_req to e1000e adapter

[Thomas Graf]

e1000e is the only driver requiring pm_qos_req, instead of causing
every device to waste up to 240 bytes. Allocate it for the specific
driver.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
 drivers/net/ethernet/intel/e1000e/e1000.h  | 1 +
 drivers/net/ethernet/intel/e1000e/netdev.c | 8 ++++----
 include/linux/netdevice.h                  | 1 -
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/intel/e1000e/e1000.h b/drivers/net/ethernet/intel/e1000e/e1000.h
index a69f09e..5d9ceb1 100644
--- a/drivers/net/ethernet/intel/e1000e/e1000.h
+++ b/drivers/net/ethernet/intel/e1000e/e1000.h
@@ -343,6 +343,7 @@ struct e1000_adapter {
 	struct timecounter tc;
 	struct ptp_clock *ptp_clock;
 	struct ptp_clock_info ptp_clock_info;
+	struct pm_qos_request pm_qos_req;
 
 	u16 eee_advert;
 };
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c
index 74ec185..c509a5c 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -3297,9 +3297,9 @@ static void e1000_configure_rx(struct e1000_adapter *adapter)
 			ew32(RXDCTL(0), rxdctl | 0x3);
 		}
 
-		pm_qos_update_request(&adapter->netdev->pm_qos_req, lat);
+		pm_qos_update_request(&adapter->pm_qos_req, lat);
 	} else {
-		pm_qos_update_request(&adapter->netdev->pm_qos_req,
+		pm_qos_update_request(&adapter->pm_qos_req,
 				      PM_QOS_DEFAULT_VALUE);
 	}
 
@@ -4403,7 +4403,7 @@ static int e1000_open(struct net_device *netdev)
 		e1000_update_mng_vlan(adapter);
 
 	/* DMA latency requirement to workaround jumbo issue */
-	pm_qos_add_request(&adapter->netdev->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
+	pm_qos_add_request(&adapter->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
 			   PM_QOS_DEFAULT_VALUE);
 
 	/* before we allocate an interrupt, we must be ready to handle it.
@@ -4514,7 +4514,7 @@ static int e1000_close(struct net_device *netdev)
 	    !test_bit(__E1000_TESTING, &adapter->state))
 		e1000e_release_hw_control(adapter);
 
-	pm_qos_remove_request(&adapter->netdev->pm_qos_req);
+	pm_qos_remove_request(&adapter->pm_qos_req);
 
 	pm_runtime_put_sync(&pdev->dev);
 
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index bf6d9df..d66c0f9 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1743,7 +1743,6 @@ struct net_device {
 #endif
 	struct phy_device *phydev;
 	struct lock_class_key *qdisc_tx_busylock;
-	struct pm_qos_request	pm_qos_req;
 };
 #define to_net_dev(d) container_of(d, struct net_device, dev)
 
-- 
1.9.3

[Intel-wired-lan] [PATCH 1/2 net-next] e1000e: Move pm_qos_req to e1000e adapter

[Jeff Kirsher]

On Fri, 2015-04-10 at 15:52 +0200, Thomas Graf wrote:
> e1000e is the only driver requiring pm_qos_req, instead of causing
> every device to waste up to 240 bytes. Allocate it for the specific
> driver.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>

Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

> ---
>  drivers/net/ethernet/intel/e1000e/e1000.h  | 1 +
>  drivers/net/ethernet/intel/e1000e/netdev.c | 8 ++++----
>  include/linux/netdevice.h                  | 1 -
>  3 files changed, 5 insertions(+), 5 deletions(-)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20150410/bd78a132/attachment.asc>

Re: [PATCH 1/2 net-next] e1000e: Move pm_qos_req to e1000e adapter

[Jeff Kirsher]

[-- Attachment #1: Type: text/plain, Size: 545 bytes --]

On Fri, 2015-04-10 at 15:52 +0200, Thomas Graf wrote:
> e1000e is the only driver requiring pm_qos_req, instead of causing
> every device to waste up to 240 bytes. Allocate it for the specific
> driver.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>

Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

> ---
>  drivers/net/ethernet/intel/e1000e/e1000.h  | 1 +
>  drivers/net/ethernet/intel/e1000e/netdev.c | 8 ++++----
>  include/linux/netdevice.h                  | 1 -
>  3 files changed, 5 insertions(+), 5 deletions(-)



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[Intel-wired-lan] [PATCH 2/2 net-next] net_device: Reorder members to fill holes

[Thomas Graf]

Some trivial reorders while preserving the RX/TX cache lines
split to fill a couple of holes.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
 include/linux/netdevice.h | 45 +++++++++++++++++++++------------------------
 1 file changed, 21 insertions(+), 24 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d66c0f9..908f0e7 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1305,6 +1305,8 @@ enum netdev_priv_flags {
  *	@base_addr:	Device I/O address
  *	@irq:		Device IRQ number
  *
+ *	@carrier_changes:	Stats to monitor carrier on<->off transitions
+ *
  *	@state:		Generic network queuing layer state, see netdev_state_t
  *	@dev_list:	The global list of network devices
  *	@napi_list:	List entry, that is used for polling napi devices
@@ -1338,8 +1340,6 @@ enum netdev_priv_flags {
  *	@tx_dropped:	Dropped packets by core network,
  *			do not use this in drivers
  *
- *	@carrier_changes:	Stats to monitor carrier on<->off transitions
- *
  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
  *				instead of ioctl,
  *				see <net/iw_handler.h> for details.
@@ -1382,14 +1382,14 @@ enum netdev_priv_flags {
  * 	@dev_port:		Used to differentiate devices that share
  * 				the same function
  *	@addr_list_lock:	XXX: need comments on this one
- *	@uc:			unicast mac addresses
- *	@mc:			multicast mac addresses
- *	@dev_addrs:		list of device hw addresses
- *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
  *	@uc_promisc:		Counter, that indicates, that promiscuous mode
  *				has been enabled due to the need to listen to
  *				additional unicast addresses in a device that
  *				does not implement ndo_set_rx_mode()
+ *	@uc:			unicast mac addresses
+ *	@mc:			multicast mac addresses
+ *	@dev_addrs:		list of device hw addresses
+ *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
  *	@promiscuity:		Number of times, the NIC is told to work in
  *				Promiscuous mode, if it becomes 0 the NIC will
  *				exit from working in Promiscuous mode
@@ -1419,6 +1419,12 @@ enum netdev_priv_flags {
  *	@ingress_queue:		XXX: need comments on this one
  *	@broadcast:		hw bcast address
  *
+ *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
+ *			indexed by RX queue number. Assigned by driver.
+ *			This must only be set if the ndo_rx_flow_steer
+ *			operation is defined
+ *	@index_hlist:		Device index hash chain
+ *
  *	@_tx:			Array of TX queues
  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
  *	@real_num_tx_queues: 	Number of TX queues currently active in device
@@ -1428,11 +1434,6 @@ enum netdev_priv_flags {
  *
  *	@xps_maps:	XXX: need comments on this one
  *
- *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
- *			indexed by RX queue number. Assigned by driver.
- *			This must only be set if the ndo_rx_flow_steer
- *			operation is defined
- *
  *	@trans_start:		Time (in jiffies) of last Tx
  *	@watchdog_timeo:	Represents the timeout that is used by
  *				the watchdog ( see dev_watchdog() )
@@ -1440,7 +1441,6 @@ enum netdev_priv_flags {
  *
  *	@pcpu_refcnt:		Number of references to this device
  *	@todo_list:		Delayed register/unregister
- *	@index_hlist:		Device index hash chain
  *	@link_watch_list:	XXX: need comments on this one
  *
  *	@reg_state:		Register/unregister state machine
@@ -1507,6 +1507,8 @@ struct net_device {
 	unsigned long		base_addr;
 	int			irq;
 
+	atomic_t		carrier_changes;
+
 	/*
 	 *	Some hardware also needs these fields (state,dev_list,
 	 *	napi_list,unreg_list,close_list) but they are not
@@ -1547,8 +1549,6 @@ struct net_device {
 	atomic_long_t		rx_dropped;
 	atomic_long_t		tx_dropped;
 
-	atomic_t		carrier_changes;
-
 #ifdef CONFIG_WIRELESS_EXT
 	const struct iw_handler_def *	wireless_handlers;
 	struct iw_public_data *	wireless_data;
@@ -1588,6 +1588,8 @@ struct net_device {
 	unsigned short          dev_id;
 	unsigned short          dev_port;
 	spinlock_t		addr_list_lock;
+	unsigned char		name_assign_type;
+	bool			uc_promisc;
 	struct netdev_hw_addr_list	uc;
 	struct netdev_hw_addr_list	mc;
 	struct netdev_hw_addr_list	dev_addrs;
@@ -1595,10 +1597,6 @@ struct net_device {
 #ifdef CONFIG_SYSFS
 	struct kset		*queues_kset;
 #endif
-
-	unsigned char		name_assign_type;
-
-	bool			uc_promisc;
 	unsigned int		promiscuity;
 	unsigned int		allmulti;
 
@@ -1645,7 +1643,10 @@ struct net_device {
 
 	struct netdev_queue __rcu *ingress_queue;
 	unsigned char		broadcast[MAX_ADDR_LEN];
-
+#ifdef CONFIG_RFS_ACCEL
+	struct cpu_rmap		*rx_cpu_rmap;
+#endif
+	struct hlist_node	index_hlist;
 
 /*
  * Cache lines mostly used on transmit path
@@ -1656,13 +1657,11 @@ struct net_device {
 	struct Qdisc		*qdisc;
 	unsigned long		tx_queue_len;
 	spinlock_t		tx_global_lock;
+	int			watchdog_timeo;
 
 #ifdef CONFIG_XPS
 	struct xps_dev_maps __rcu *xps_maps;
 #endif
-#ifdef CONFIG_RFS_ACCEL
-	struct cpu_rmap		*rx_cpu_rmap;
-#endif
 
 	/* These may be needed for future network-power-down code. */
 
@@ -1672,13 +1671,11 @@ struct net_device {
 	 */
 	unsigned long		trans_start;
 
-	int			watchdog_timeo;
 	struct timer_list	watchdog_timer;
 
 	int __percpu		*pcpu_refcnt;
 	struct list_head	todo_list;
 
-	struct hlist_node	index_hlist;
 	struct list_head	link_watch_list;
 
 	enum { NETREG_UNINITIALIZED=0,
-- 
1.9.3

[PATCH 2/2 net-next] net_device: Reorder members to fill holes

[Thomas Graf]

Some trivial reorders while preserving the RX/TX cache lines
split to fill a couple of holes.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
 include/linux/netdevice.h | 45 +++++++++++++++++++++------------------------
 1 file changed, 21 insertions(+), 24 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d66c0f9..908f0e7 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1305,6 +1305,8 @@ enum netdev_priv_flags {
  *	@base_addr:	Device I/O address
  *	@irq:		Device IRQ number
  *
+ *	@carrier_changes:	Stats to monitor carrier on<->off transitions
+ *
  *	@state:		Generic network queuing layer state, see netdev_state_t
  *	@dev_list:	The global list of network devices
  *	@napi_list:	List entry, that is used for polling napi devices
@@ -1338,8 +1340,6 @@ enum netdev_priv_flags {
  *	@tx_dropped:	Dropped packets by core network,
  *			do not use this in drivers
  *
- *	@carrier_changes:	Stats to monitor carrier on<->off transitions
- *
  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
  *				instead of ioctl,
  *				see <net/iw_handler.h> for details.
@@ -1382,14 +1382,14 @@ enum netdev_priv_flags {
  * 	@dev_port:		Used to differentiate devices that share
  * 				the same function
  *	@addr_list_lock:	XXX: need comments on this one
- *	@uc:			unicast mac addresses
- *	@mc:			multicast mac addresses
- *	@dev_addrs:		list of device hw addresses
- *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
  *	@uc_promisc:		Counter, that indicates, that promiscuous mode
  *				has been enabled due to the need to listen to
  *				additional unicast addresses in a device that
  *				does not implement ndo_set_rx_mode()
+ *	@uc:			unicast mac addresses
+ *	@mc:			multicast mac addresses
+ *	@dev_addrs:		list of device hw addresses
+ *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
  *	@promiscuity:		Number of times, the NIC is told to work in
  *				Promiscuous mode, if it becomes 0 the NIC will
  *				exit from working in Promiscuous mode
@@ -1419,6 +1419,12 @@ enum netdev_priv_flags {
  *	@ingress_queue:		XXX: need comments on this one
  *	@broadcast:		hw bcast address
  *
+ *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
+ *			indexed by RX queue number. Assigned by driver.
+ *			This must only be set if the ndo_rx_flow_steer
+ *			operation is defined
+ *	@index_hlist:		Device index hash chain
+ *
  *	@_tx:			Array of TX queues
  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
  *	@real_num_tx_queues: 	Number of TX queues currently active in device
@@ -1428,11 +1434,6 @@ enum netdev_priv_flags {
  *
  *	@xps_maps:	XXX: need comments on this one
  *
- *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
- *			indexed by RX queue number. Assigned by driver.
- *			This must only be set if the ndo_rx_flow_steer
- *			operation is defined
- *
  *	@trans_start:		Time (in jiffies) of last Tx
  *	@watchdog_timeo:	Represents the timeout that is used by
  *				the watchdog ( see dev_watchdog() )
@@ -1440,7 +1441,6 @@ enum netdev_priv_flags {
  *
  *	@pcpu_refcnt:		Number of references to this device
  *	@todo_list:		Delayed register/unregister
- *	@index_hlist:		Device index hash chain
  *	@link_watch_list:	XXX: need comments on this one
  *
  *	@reg_state:		Register/unregister state machine
@@ -1507,6 +1507,8 @@ struct net_device {
 	unsigned long		base_addr;
 	int			irq;
 
+	atomic_t		carrier_changes;
+
 	/*
 	 *	Some hardware also needs these fields (state,dev_list,
 	 *	napi_list,unreg_list,close_list) but they are not
@@ -1547,8 +1549,6 @@ struct net_device {
 	atomic_long_t		rx_dropped;
 	atomic_long_t		tx_dropped;
 
-	atomic_t		carrier_changes;
-
 #ifdef CONFIG_WIRELESS_EXT
 	const struct iw_handler_def *	wireless_handlers;
 	struct iw_public_data *	wireless_data;
@@ -1588,6 +1588,8 @@ struct net_device {
 	unsigned short          dev_id;
 	unsigned short          dev_port;
 	spinlock_t		addr_list_lock;
+	unsigned char		name_assign_type;
+	bool			uc_promisc;
 	struct netdev_hw_addr_list	uc;
 	struct netdev_hw_addr_list	mc;
 	struct netdev_hw_addr_list	dev_addrs;
@@ -1595,10 +1597,6 @@ struct net_device {
 #ifdef CONFIG_SYSFS
 	struct kset		*queues_kset;
 #endif
-
-	unsigned char		name_assign_type;
-
-	bool			uc_promisc;
 	unsigned int		promiscuity;
 	unsigned int		allmulti;
 
@@ -1645,7 +1643,10 @@ struct net_device {
 
 	struct netdev_queue __rcu *ingress_queue;
 	unsigned char		broadcast[MAX_ADDR_LEN];
-
+#ifdef CONFIG_RFS_ACCEL
+	struct cpu_rmap		*rx_cpu_rmap;
+#endif
+	struct hlist_node	index_hlist;
 
 /*
  * Cache lines mostly used on transmit path
@@ -1656,13 +1657,11 @@ struct net_device {
 	struct Qdisc		*qdisc;
 	unsigned long		tx_queue_len;
 	spinlock_t		tx_global_lock;
+	int			watchdog_timeo;
 
 #ifdef CONFIG_XPS
 	struct xps_dev_maps __rcu *xps_maps;
 #endif
-#ifdef CONFIG_RFS_ACCEL
-	struct cpu_rmap		*rx_cpu_rmap;
-#endif
 
 	/* These may be needed for future network-power-down code. */
 
@@ -1672,13 +1671,11 @@ struct net_device {
 	 */
 	unsigned long		trans_start;
 
-	int			watchdog_timeo;
 	struct timer_list	watchdog_timer;
 
 	int __percpu		*pcpu_refcnt;
 	struct list_head	todo_list;
 
-	struct hlist_node	index_hlist;
 	struct list_head	link_watch_list;
 
 	enum { NETREG_UNINITIALIZED=0,
-- 
1.9.3

[Intel-wired-lan] [PATCH 2/2 net-next] net_device: Reorder members to fill holes

[Eric Dumazet]

On Fri, 2015-04-10 at 15:52 +0200, Thomas Graf wrote:
> Some trivial reorders while preserving the RX/TX cache lines
> split to fill a couple of holes.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>
> ---
>  include/linux/netdevice.h | 45 +++++++++++++++++++++------------------------
>  1 file changed, 21 insertions(+), 24 deletions(-)

Acked-by: Eric Dumazet <edumazet@google.com>

Thanks Thomas !

Re: [PATCH 2/2 net-next] net_device: Reorder members to fill holes

[Eric Dumazet]

On Fri, 2015-04-10 at 15:52 +0200, Thomas Graf wrote:
> Some trivial reorders while preserving the RX/TX cache lines
> split to fill a couple of holes.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>
> ---
>  include/linux/netdevice.h | 45 +++++++++++++++++++++------------------------
>  1 file changed, 21 insertions(+), 24 deletions(-)

Acked-by: Eric Dumazet <edumazet@google.com>

Thanks Thomas !

[Intel-wired-lan] [PATCH 0/2 net-next v2] Bring sizeof(net_device) down to < 2K bytes

[David Miller]

From: Thomas Graf <tgraf@suug.ch>
Date: Fri, 10 Apr 2015 15:52:36 +0200

> The size of struct net_device crossed the 2K boundary a while ago which
> is a waste in combination with many net namespaces. This series brings
> the size of struct net_device down to well below 2K in total size with
> a typical configuration. Some reserves a several holes leave room for
> further expansion.
> 
> Before:
> /* size: 2176, cachelines: 34, members: 121 */
> 
> After:
> /* size: 1984, cachelines: 31, members: 120 */

Series applied, thanks Thomas

Re: [PATCH 0/2 net-next v2] Bring sizeof(net_device) down to < 2K bytes

[David Miller]

From: Thomas Graf <tgraf@suug.ch>
Date: Fri, 10 Apr 2015 15:52:36 +0200

> The size of struct net_device crossed the 2K boundary a while ago which
> is a waste in combination with many net namespaces. This series brings
> the size of struct net_device down to well below 2K in total size with
> a typical configuration. Some reserves a several holes leave room for
> further expansion.
> 
> Before:
> /* size: 2176, cachelines: 34, members: 121 */
> 
> After:
> /* size: 1984, cachelines: 31, members: 120 */

Series applied, thanks Thomas