GitHub/LineageOS/android_kernel_motorola_exynos9610.git
10 years agonet: vmxnet3 calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:32:08 +0000 (23:32 -0800)]
net: vmxnet3 calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sun-niu calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:32:00 +0000 (23:32 -0800)]
net: sun-niu calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sfc calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:31:50 +0000 (23:31 -0800)]
net: sfc calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: vxge calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:31:36 +0000 (23:31 -0800)]
net: vxge calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: mlx4 calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:31:23 +0000 (23:31 -0800)]
net: mlx4 calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sky2 calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:28:13 +0000 (23:28 -0800)]
net: sky2 calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: emulex-benet calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:23:51 +0000 (23:23 -0800)]
net: emulex-benet calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: cisco-enic calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:23:42 +0000 (23:23 -0800)]
net: cisco-enic calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: cxgb4 calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:23:29 +0000 (23:23 -0800)]
net: cxgb4 calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: bnx2x calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:23:11 +0000 (23:23 -0800)]
net: bnx2x calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: bnx2 calls skb_set_hash
Tom Herbert [Wed, 18 Dec 2013 07:22:57 +0000 (23:22 -0800)]
net: bnx2 calls skb_set_hash

Drivers should call skb_set_hash to set the hash and its type
in an skbuff.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Wed, 18 Dec 2013 19:58:41 +0000 (14:58 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to i40e, ixgbevf, ixgbe and igb.

Don provides an ixgbevf patch to add DCB configuration into the queue
setup so that we won't have to allocate queues in a separate place when
enabling DCB.

Guenter Roeck provides 2 patches for ixgbe to simplify the code by
attaching hwmon sysfs attributes to hwmon device instead of PCI device.
Also fix an issues where the temperature sensor attribute index was
being started with the value 0 and not 1 as per the hwmon API.

Carolyn provides igb patches to fix queue allocation method to
accommodate changes during runtime.  This includes changing how the
driver initializes MSIx and checks for MSIx configuration to make it
easier to reconfigure the device when queue changes happen at runtime.

Neerav and Shannon fixes i40e debugfs commands that dump hex information
by using print_hex_dump().

Shannon provides several i40e fixes which include the prevention of
null pointer exception in the dump descriptor by checking that rings
were allocated before trying to reference them.  Fixed up a couple of
scanfs to accept various base numbers instead of silently requiring hex.

Anjali fixes up i40e where the incorrect defines were being used for
misc interrupts.

Alan Cox provides a fix for i40e where we assume that the resulting
buffer is zero terminated when we then re-use it.  The sscanf is limited
to 512 bytes but needs to be 511 to allow for a terminator.

Stephen Hemminger fixes i40e by making local functions static and removes
unused code (i40e_aq_add/remove_vlan() functions).
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: convert tcf_proto_ops to use struct list_head
WANG Cong [Mon, 16 Dec 2013 04:15:11 +0000 (20:15 -0800)]
net_sched: convert tcf_proto_ops to use struct list_head

We don't need to maintain our own singly linked list code.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: convert tc_action_ops to use struct list_head
WANG Cong [Mon, 16 Dec 2013 04:15:10 +0000 (20:15 -0800)]
net_sched: convert tc_action_ops to use struct list_head

We don't need to maintain our own singly linked list code.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: convert tcf_hashinfo to hlist and use spinlock
WANG Cong [Mon, 16 Dec 2013 04:15:09 +0000 (20:15 -0800)]
net_sched: convert tcf_hashinfo to hlist and use spinlock

So that we don't need to play with singly linked list,
and since the code is not on hot path, we can use spinlock
instead of rwlock.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: init struct tcf_hashinfo at register time
WANG Cong [Mon, 16 Dec 2013 04:15:08 +0000 (20:15 -0800)]
net_sched: init struct tcf_hashinfo at register time

It looks weird to store the lock out of the struct but
still points to a static variable. Just move them into the struct.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: cls: refactor out struct tcf_ext_map
WANG Cong [Mon, 16 Dec 2013 04:15:07 +0000 (20:15 -0800)]
net_sched: cls: refactor out struct tcf_ext_map

These information can be saved in tcf_exts, and this will
simplify the code.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: act: use standard struct list_head
WANG Cong [Mon, 16 Dec 2013 04:15:05 +0000 (20:15 -0800)]
net_sched: act: use standard struct list_head

Currently actions are chained by a singly linked list,
therefore it is a bit hard to add and remove a specific
entry. Convert it to struct list_head so that in the
latter patch we can remove an action without finding
its head.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: remove get_stats from tc_action_ops
WANG Cong [Mon, 16 Dec 2013 04:15:04 +0000 (20:15 -0800)]
net_sched: remove get_stats from tc_action_ops

It is not used.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoigb: Change to use statically allocated array for MSIx entries
Carolyn Wyborny [Tue, 10 Dec 2013 07:58:34 +0000 (07:58 +0000)]
igb: Change to use statically allocated array for MSIx entries

This patch changes how the driver initializes MSIx and checks
for MSIx configuration.  This change makes it easier to reconfigure the
device when queue changes happen at runtime using ethtool's set_channels
feature.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Fix queue allocation method to accommodate changing during runtime
Carolyn Wyborny [Tue, 10 Dec 2013 07:58:29 +0000 (07:58 +0000)]
igb: Fix queue allocation method to accommodate changing during runtime

When changing number of queues using ethtool's set_channels during runtime,
a queue allocation could fail, which can leave the device in a down state.
In order to preserve the usability of the device in this scenario, this patch
changes the driver to allocate the  number of queues only if they have not
been allocated already. The first allocation is then done for the max number
of queues, which is the default queues for this driver.   With this change,
queue quantity changes are not subject to queue allocation failures.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoixgbe: Start temperature sensor attribute index with 1
Guenter Roeck [Tue, 26 Nov 2013 07:15:39 +0000 (07:15 +0000)]
ixgbe: Start temperature sensor attribute index with 1

Per hwmon ABI, temperature sensor attribute index starts with 1, not 0.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoixgbe: Convert to use devm_hwmon_device_register_with_groups
Guenter Roeck [Tue, 26 Nov 2013 07:15:28 +0000 (07:15 +0000)]
ixgbe: Convert to use devm_hwmon_device_register_with_groups

Simplify the code. Attach hwmon sysfs attributes to hwmon device
instead of pci device. Avoid race conditions caused by attributes
being created after hwmon device registration. Implicitly
(through hwmon API) add mandatory 'name' sysfs attribute.

Other cleanup:

Instead of allocating memory for hwmon attributes, move attributes
and all other hwmon related data into struct hwmon_buff and allocate
the entire structure using devm_kzalloc.

Check return value from calls to igb_add_hwmon_attr() one by one instead
of logically combining them all together.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoixgbevf: add DCB configuration into queue setup
Don Skidmore [Sat, 21 Sep 2013 01:40:49 +0000 (01:40 +0000)]
ixgbevf: add DCB configuration into queue setup

This patch takes the DCB config checks and adds them to the normal setting
up of the queues. This way we won't have to allocation queues in a separate
place for enabling DCB.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Tested-By: Jack Morgan<jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: make functions static and remove dead code
Stephen Hemminger [Fri, 13 Dec 2013 04:37:50 +0000 (04:37 +0000)]
i40e: make functions static and remove dead code

Make local functions static in the file they are used.

Remove functions i40e_aq_add_vlan and i40e_aq_remove_vlan since
they are not used anywhere by current code.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Fix off by one in i40e_dbg_command_write
Alan Cox [Thu, 12 Dec 2013 02:44:24 +0000 (02:44 +0000)]
i40e: Fix off by one in i40e_dbg_command_write

We assume that the resulting buffer is zero terminated when we then
re-use it. The sscanf is limited to 512 bytes but needs to be 511
to allow for a terminator.

One of a set of problems noted by Jackie Chang

Signed-off-by: Alan Cox <alan@linux.intel.com>
Acked-by: Shannon Nelson <Shannon.nelson@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Bump version number
Catherine Sullivan [Tue, 26 Nov 2013 10:49:33 +0000 (10:49 +0000)]
i40e: Bump version number

Version updated to 0.3.14-k

Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: Fix wrong mask bits being used in misc interrupt
Anjali Singhai Jain [Tue, 26 Nov 2013 10:49:32 +0000 (10:49 +0000)]
i40e: Fix wrong mask bits being used in misc interrupt

No functional change, but the wrong defines were being used.

Change-Id: Ica2afd2dfe18154ca0f1260a508f31e372319ba7
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: more print_hex_dump use
Shannon Nelson [Tue, 26 Nov 2013 10:49:31 +0000 (10:49 +0000)]
i40e: more print_hex_dump use

Take advantage of print_hex_dump() in another couple places to
clean up the code.

Change-Id: Ib618e75f928308c0afd0d8d74105da0c6577a024
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: fix up scanf decoders
Shannon Nelson [Tue, 26 Nov 2013 10:49:30 +0000 (10:49 +0000)]
i40e: fix up scanf decoders

Fix up a couple of scanfs to accept various base numbers instead of silently
requiring hex.

Change-Id: I1cc4dffbb1d011bf603cbf34a8db093da57fad7a
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: simplify error messages for dump descriptor
Shannon Nelson [Tue, 26 Nov 2013 10:49:29 +0000 (10:49 +0000)]
i40e: simplify error messages for dump descriptor

debugfs fixes: We don't really need to give usage messages for data errors,
only for invalid command errors.

Change-Id: If3f74ac49e43c3ced7fd388323fa738ac145e055
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoi40e: prevent null pointer exception in dump descriptor
Shannon Nelson [Tue, 26 Nov 2013 10:49:26 +0000 (10:49 +0000)]
i40e: prevent null pointer exception in dump descriptor

Check that rings were allocated before trying to reference them.

Change-Id: I33151e55ab7a7a305fecdb88ccb2709ac246b7c7
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoMerge branch 'phy_checkpatch'
David S. Miller [Wed, 18 Dec 2013 05:49:35 +0000 (00:49 -0500)]
Merge branch 'phy_checkpatch'

Florian Fainelli says:

====================
net: phy: fix checkpatch errors

This patchset fixes trivial checkpatch errors, no functional change
introduced.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: spi_ks8995: fix checkpatch errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:13 +0000 (21:38 -0800)]
net: phy: spi_ks8995: fix checkpatch errors

checkpatch spotted two errors, fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: fix checkpatch errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:12 +0000 (21:38 -0800)]
net: phy: fix checkpatch errors

checkpatch spotted a few checkpatch errors such as whitespace damages
and switch/case labels not being on the same column, fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: micrel: fix checkpath errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:11 +0000 (21:38 -0800)]
net: phy: micrel: fix checkpath errors

checkpatch spotted a few spaces vs tabs errors, fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: mdio_bus: fix checkpath error
Florian Fainelli [Wed, 18 Dec 2013 05:38:10 +0000 (21:38 -0800)]
net: phy: mdio_bus: fix checkpath error

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: lxt: fix checkpath errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:09 +0000 (21:38 -0800)]
net: phy: lxt: fix checkpath errors

checkpath spotted a few errors in this file, fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: icplus: fix checkpath error
Florian Fainelli [Wed, 18 Dec 2013 05:38:08 +0000 (21:38 -0800)]
net: phy: icplus: fix checkpath error

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: dp83640: fix checkpath error
Florian Fainelli [Wed, 18 Dec 2013 05:38:07 +0000 (21:38 -0800)]
net: phy: dp83640: fix checkpath error

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: davicom: fix checkpath errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:06 +0000 (21:38 -0800)]
net: phy: davicom: fix checkpath errors

checkpath spotted a few stylistic errors, fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: cicada: fix checkpath errors
Florian Fainelli [Wed, 18 Dec 2013 05:38:05 +0000 (21:38 -0800)]
net: phy: cicada: fix checkpath errors

checkpath spotted a few stylistic errors fix them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'vlan_tpid'
David S. Miller [Wed, 18 Dec 2013 05:36:51 +0000 (00:36 -0500)]
Merge branch 'vlan_tpid'

Atzm Watanabe says:

====================
packet: deliver VLAN TPID to userspace

This patchset enables userspace to get VLAN TPID as well as the VLAN TCI.

After the 802.1AD support, userspace packet receivers (packet dumper,
software switch, and the like) need how to know VLAN TPID in order to
reconstruct original tagged frame.

v4: Simply use sizeof(tp_padding) for zeroing the padding bytes,
    commented by David Laight.
    Use __u16 for tp_vlan_tpid in tpacket_hdr_variant1,
    commented by Daniel Borkmann.

v3: Add a definition which indicates whether tp_vlan_tpid is valid.
    Explicitly define pad bytes for tpacket{2,3}_hdr and pick the area
    for tp_vlan_tpid from the definition.  Commented by David Laight.

v2: Add BUILD_BUG_ON() to make current aligned size of
    struct tpacket{2,3}_hdr clear.  Commented by Ben Hutchings.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopacket: deliver VLAN TPID to userspace
Atzm Watanabe [Tue, 17 Dec 2013 13:53:40 +0000 (22:53 +0900)]
packet: deliver VLAN TPID to userspace

This enables userspace to get VLAN TPID as well as the VLAN TCI.

Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopacket: fill the gap of TPACKET_ALIGNMENT with zeros
Atzm Watanabe [Tue, 17 Dec 2013 13:53:36 +0000 (22:53 +0900)]
packet: fill the gap of TPACKET_ALIGNMENT with zeros

struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT.
Explicitly defining and zeroing the gap of this makes additional changes
easier.

Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopacket: make aligned size of struct tpacket{2,3}_hdr clear
Atzm Watanabe [Tue, 17 Dec 2013 13:53:32 +0000 (22:53 +0900)]
packet: make aligned size of struct tpacket{2,3}_hdr clear

struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT.
We may add members to them until current aligned size without forcing
userspace to call getsockopt(..., PACKET_HDRLEN, ...).

Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'bna'
David S. Miller [Wed, 18 Dec 2013 05:31:01 +0000 (00:31 -0500)]
Merge branch 'bna'

Rasesh Mody says:

====================
bna: Update the Driver to v3.2.23.0

This patch set consists of feature additions like s/w timestamping support,
multi-buffer RX, firmware patch simplification, enhancements for RX filters,
RX processing changes, bug fixes and updates the firmware version to v3.2.3.0.
The patch set addressed the review commnets recieved.

This patch set updates the BNA driver to v3.2.23.0 and was tested against
net-next 3.12.0-rc6 kernel.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Update the Driver Version to 3.2.23.0
Rasesh Mody [Wed, 18 Dec 2013 01:07:42 +0000 (17:07 -0800)]
bna: Update the Driver Version to 3.2.23.0

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Firmware Patch Simplification
Rasesh Mody [Wed, 18 Dec 2013 01:07:41 +0000 (17:07 -0800)]
bna: Firmware Patch Simplification

This patch includes change to enable firmware patch simplication feature.
This feature is targeted to address the requirement to have independent patch
release for firmware. Prior to the 3.2.3.0 firmware, releasing a patch fix for
firmware would require changes to bna driver, to use new firmware images.
However with these changes, if the new firmware is flashed on to the Adapter,
the driver will use the new firmware after checking the patch release byte in
the firmware version.

Update the f/w version to 3.2.3.0

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Embed SKB Length in TX Vector
Rasesh Mody [Wed, 18 Dec 2013 01:07:40 +0000 (17:07 -0800)]
bna: Embed SKB Length in TX Vector

- Store the length of the skb buffer mapped along with the handle and use it
   while unmapping the buffer.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Handle the TX Setup Failures
Rasesh Mody [Wed, 18 Dec 2013 01:07:39 +0000 (17:07 -0800)]
bna: Handle the TX Setup Failures

Change details:
 - When bnad_setup_tx() returns NULL, the error is NOT returned to the caller.
   The caller will incorrectly assume success. So Return ENOMEM when bna_tx_create()
   fails.
 - If bnad_tx_msix_register() fails, call bna_tx_destroy() to free tx & to NULL
   the bnad reference to tcb.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Add NULL Check Before Dereferencing TCB
Rasesh Mody [Wed, 18 Dec 2013 01:07:38 +0000 (17:07 -0800)]
bna: Add NULL Check Before Dereferencing TCB

Currently we already check to see whether the BNAD_TXQ_TX_STARTED cleared.
But if the tcb structure which contains this flag is also already freed by that
time, we would dereference the NULL pointer. This patch is to check tcb for NULL
pointer, before dereferencing it.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: CQ Read Fix
Rasesh Mody [Wed, 18 Dec 2013 01:07:37 +0000 (17:07 -0800)]
bna: CQ Read Fix

Valid bit check for completion needs read fence, so that it does not get
reordered with other loads.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: RX Processing and Config Changes
Rasesh Mody [Wed, 18 Dec 2013 01:07:36 +0000 (17:07 -0800)]
bna: RX Processing and Config Changes

Change Details:
 - Prefetch header in GRO path. This reduces napi_frags_skb time from 9% to 5%.
 - Changed the configurable limit of RxQ depth to 16384 (was 2048).
 - bnad_rx_unmap_q elements are cachealigned.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Enable Multi Buffer RX
Rasesh Mody [Wed, 18 Dec 2013 01:07:35 +0000 (17:07 -0800)]
bna: Enable Multi Buffer RX

The CT2 HW supports multi-buffer Rx. This patch provides the necessary changes
for bnad to use multi-buffer Rx feature. For BNAD, multi-buffer Rx is by
default enabled when MTU is > 4096. For >4096 MTU, q0 data/large buffers are of
2048 size. As the resource requirements of multi-buffer Rx are different new Rx
needs to be created to use this feature. ASIC posts multiple completions if
frame exceeds buffer size. The last completion is marked with EOP flag.
 - Separate HQ and DQ enums for resource allocations and configurations.
 - rx_config and rxq structure changes to pass the correct info from bnad.
 - DQ depth need not be same as HQ depth. So CQ depth is adjusted accordingly.
 - Rx CFG frame size is taken from configured MTU.
 - Rx q0 buffer size is configured from bnad s rx_config when multi-buffer is
   enabled.
 - Poll for entire frame completion.
 - Once EOP completion is received gather the number of vectors used by the
   frame to submit it to the stack.
 - Changed MTU to frame size wherever necessary.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: RX Filter Enhancements
Rasesh Mody [Wed, 18 Dec 2013 01:07:34 +0000 (17:07 -0800)]
bna: RX Filter Enhancements

Change Details:
 - Added bna_rx_ucast_listset() for synchronous ucast listadd operation.
 - Clear mac->handle before adding it to free_q.
 - bnad_set_rx_mode() rewritten. bnad_set_rx_mode() adds the MACs in uc_list
   to UCAM. If it exceeds the max supported, DEFAULT mode is turned on. If
   MCAM limit is exceeded, ALLMULTI mode is turned on.
 - Clear CF flags, check for the new mode and reprogram the Rx approach.
 - Added bnad_set_rx_ucast_fltr() and bnad_set_rx_mcast_fltr().
 - Check for IFF_PROMISC to set the correct mode.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Fix Filter Add Del
Rasesh Mody [Wed, 18 Dec 2013 01:07:33 +0000 (17:07 -0800)]
bna: Fix Filter Add Del

Change Details:
 - bna_rx_mcast_listset() API first looks at free_q only and not at other
   pending Qs rendering it non-deterministic of giving an upper limit.
   Modify bna_rx_mcast_listset() implementation to not use only half of the
   limit.
 - Allocate and initialize queue for deleting
 - Segregate the adding and deleting process by using separate queues.
 - The filter framework in bna does not let adding addresses to its max capacity
   due to asynchronous operations involved.
   Provide a synchronous option to set a given list.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Set Get IOC fw State
Rasesh Mody [Wed, 18 Dec 2013 01:07:32 +0000 (17:07 -0800)]
bna: Set Get IOC fw State

Add APIs to set and get IOC currnet fw state and alt IOC fw state
 - bfa_ioc_ct_set_cur_ioc_fwstate()
 - bfa_ioc_ct_get_cur_ioc_fwstate()
 - bfa_ioc_ct_set_alt_ioc_fwstate()
 - bfa_ioc_ct_get_alt_ioc_fwstate()

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Add software timestamping support
Rasesh Mody [Wed, 18 Dec 2013 01:07:31 +0000 (17:07 -0800)]
bna: Add software timestamping support

- Invoke skb_tx_timestamp() API just before invoking txq_doorbell()
 - Add ethtool (-T) support

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agolib: Add missing arch generic-y entries for asm-generic/hash.h
David S. Miller [Wed, 18 Dec 2013 02:26:19 +0000 (21:26 -0500)]
lib: Add missing arch generic-y entries for asm-generic/hash.h

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoi40e: Fix dump output from debugfs calls
Neerav Parikh [Tue, 26 Nov 2013 10:49:25 +0000 (10:49 +0000)]
i40e: Fix dump output from debugfs calls

The debugfs commands that dump hex information are not doing these as
expected viz. "lldp get local", "nvm read", "dump debug fwdata", etc.

Use print_hex_dump() instead to hex dump and remove the print buffer
stuff from the code.

Change-Id: I507bd8b2187aae8bad5055b7872978c309cf143e
Signed-off-by: Neerav Parikh <Neerav.Parikh@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoqeth: Accurate ethtool output
Eugene Crosser [Mon, 16 Dec 2013 08:44:52 +0000 (09:44 +0100)]
qeth: Accurate ethtool output

For OSA devices that support the QUERY_CARD_INFO command, supply
accurate data based on the card type, port mode and link speed
via the 'ethtool' interface.

Signed-off-by: Eugene Crosser <Eugene.Crosser@ru.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetiucv: improve state checking in conn_action_txdone
Ursula Braun [Mon, 16 Dec 2013 08:44:51 +0000 (09:44 +0100)]
netiucv: improve state checking in conn_action_txdone

state checking in conn_action_txdone() is inconsistent.
This patch makes it consistent and issues a trace message
if an unexpected state is detected for the netiucv device.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobpf_exp: free duplicated labels at exit time
Daniel Borkmann [Mon, 16 Dec 2013 10:45:01 +0000 (11:45 +0100)]
bpf_exp: free duplicated labels at exit time

Valgrind found that extracted labels that are passed from the lexer
weren't freed upon exit. Therefore, add a small helper function that
walks label tables and frees them. Since also NULL can be passed to
free(3), we do not need to take care of that here. While at it, fix
up a spacing error in bpf_set_curr_label().

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobpf_dbg: always close socket in bpf_runnable
Daniel Borkmann [Mon, 16 Dec 2013 10:45:00 +0000 (11:45 +0100)]
bpf_dbg: always close socket in bpf_runnable

We must not leave the socket intact in bpf_runnable(). The socket
is used to test if the filter code is being accepted by the kernel
or not. So right after we do the setsockopt(2), we need to close
it again.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'qlcnic'
David S. Miller [Tue, 17 Dec 2013 22:09:49 +0000 (17:09 -0500)]
Merge branch 'qlcnic'

Manish Chopra says:

====================
qlcnic: Refactoring and enhancements

This patch series includes follwing changes

* Refactor DCBX code. Do not allow DCBX operations for VFs
* Issue INIT_NIC mailbox command only once
* Refactor initialize nic code path
* Allow configuration for single TX/RX queue
* VLAN enhancement for 84xx adapters
* Support for 16 virtual NIC functions for 84XX series adapters

Please apply to net-next
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: update version to 5.3.53
Manish Chopra [Tue, 17 Dec 2013 14:01:55 +0000 (09:01 -0500)]
qlcnic: update version to 5.3.53

Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Support for 16 virtual NIC functions.
Jitendra Kalsaria [Tue, 17 Dec 2013 14:01:54 +0000 (09:01 -0500)]
qlcnic: Support for 16 virtual NIC functions.

Extend virtual NIC functions from 8 to 16 for 84xx adapter.

Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: VLAN enhancement for 84XX adapters
Manish Chopra [Tue, 17 Dec 2013 14:01:53 +0000 (09:01 -0500)]
qlcnic: VLAN enhancement for 84XX adapters

o Support multiple VLANs on 84xx VF devices

Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Allow single Tx/Rx queue for all adapters.
Himanshu Madhani [Tue, 17 Dec 2013 14:01:52 +0000 (09:01 -0500)]
qlcnic: Allow single Tx/Rx queue for all adapters.

o Allow user to set sigle Tx/Rx queue in MSI-x mode,
  for ALL supported adapters.

Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Refactor initialize nic code path.
Sucheta Chakraborty [Tue, 17 Dec 2013 14:01:51 +0000 (09:01 -0500)]
qlcnic: Refactor initialize nic code path.

o Change function name from qlcnic_83xx_register_nic_idc_func to
  qlcnic_83xx_initialize_nic

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Issue INIT_NIC command only once.
Sucheta Chakraborty [Tue, 17 Dec 2013 14:01:50 +0000 (09:01 -0500)]
qlcnic: Issue INIT_NIC command only once.

o DCB AEN registration was reissuing INIT_NIC command. Instead, club
  all options of INIT NIC command and issue this command only once.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Disable DCB operations from SR-IOV VFs.
Sucheta Chakraborty [Tue, 17 Dec 2013 14:01:49 +0000 (09:01 -0500)]
qlcnic: Disable DCB operations from SR-IOV VFs.

o These operations will be supported only through PFs (SR-IOV and non-SR-IOV).

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/bwh/sfc...
David S. Miller [Tue, 17 Dec 2013 22:08:21 +0000 (17:08 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/bwh/sfc-next

Ben Hutchings says:

====================
Miscellaneous changes for 3.14:

1. Add more information to some WARN messages.
2. Refactor pushing of RSS configuration, from Andrew Rybchenko.
3. Refactor handling of automatic (device address list) vs manual (RX
NFC) MAC filters.
4. Implement clearing of manual RX filters on EF10 when ntuple offload
is disabled.
5. Remove definitions that are unused since the RX buffer allocation
changes, from Andrew Rybchenko.
6. Improve naming of some statistics, from Shradha Shah.
7. Add statistics for PTP support code.
8. Fix insertion of RX drop filters on EF10.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'skb_hash'
David S. Miller [Tue, 17 Dec 2013 21:36:33 +0000 (16:36 -0500)]
Merge branch 'skb_hash'

Tom Herbert says:

====================
net: Add rxhash utility hash functions

v3:

There's really nothing specific about rxhash that constrains
it to be a value just for these receive path. Drop the 'rx'
part in utility functions, including skb_get_rxhash. In subsequent
patches, we can change the rxhash and l4_rxhash names also, as
well as abstracting out the interface to the hash.

Added comments about hash types per feedback.

In this version I'm omitting the changes to drivers to make the
patch set manageable. Will add those changes in followup pathes.

-----
This patch series introduce skb_set_rxhash and skb_clear_rxhash
which are called to set the rxhash (from network drivers) and
to clear the rxhash. This API should be used instead of updating
fields in the skbuff directly.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Add utility function to copy skb hash
Tom Herbert [Mon, 16 Dec 2013 06:16:29 +0000 (22:16 -0800)]
net: Add utility function to copy skb hash

Adds skb_copy_hash to copy rxhash and l4_rxhash from one skb to another.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Add function to set the rxhash
Tom Herbert [Mon, 16 Dec 2013 06:16:19 +0000 (22:16 -0800)]
net: Add function to set the rxhash

The function skb_set_rxash was added for drivers to call to set
the rxhash in an skb. The type of hash is also specified as
a parameter (L2, L3, L4, or unknown type).

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Add utility functions to clear rxhash
Tom Herbert [Mon, 16 Dec 2013 06:12:18 +0000 (22:12 -0800)]
net: Add utility functions to clear rxhash

In several places 'skb->rxhash = 0' is being done to clear the
rxhash value in an skb.  This does not clear l4_rxhash which could
still be set so that the rxhash wouldn't be recalculated on subsequent
call to skb_get_rxhash.  This patch adds an explict function to clear
all the rxhash related information in the skb properly.

skb_clear_hash_if_not_l4 clears the rxhash only if it is not marked as
l4_rxhash.

Fixed up places where 'skb->rxhash = 0' was being called.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Change skb_get_rxhash to skb_get_hash
Tom Herbert [Mon, 16 Dec 2013 06:12:06 +0000 (22:12 -0800)]
net: Change skb_get_rxhash to skb_get_hash

Changing name of function as part of making the hash in skbuff to be
generic property, not just for receive path.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/hsr: using kfree_rcu() to simplify the code
Wei Yongjun [Mon, 16 Dec 2013 06:05:50 +0000 (14:05 +0800)]
net/hsr: using kfree_rcu() to simplify the code

The callback function of call_rcu() just calls a kfree(), so we
can use kfree_rcu() instead of call_rcu() + callback function.

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Arvid Brodin <arvid.brodin@alten.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'bonding_netlink'
David S. Miller [Tue, 17 Dec 2013 21:09:15 +0000 (16:09 -0500)]
Merge branch 'bonding_netlink'

Scott Feldman says:

====================
bonding: add some more netlink attributes

The following series implements five more bonding netlink attributes:

primary
primary_reselect
fail_over_mac
xmit_hash_policy
resend_igmp

Tested with modified iproute2 to verify attributes can be set at bond creation
time or set later.  Verified sysfs interface to attributes continues to work.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: add resend_igmp attribute netlink support
sfeldma@cumulusnetworks.com [Mon, 16 Dec 2013 00:42:19 +0000 (16:42 -0800)]
bonding: add resend_igmp attribute netlink support

Add IFLA_BOND_RESEND_IGMP to allow get/set of bonding parameter
resend_igmp via netlink.

Signed-off-by: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: add xmit_hash_policy attribute netlink support
sfeldma@cumulusnetworks.com [Mon, 16 Dec 2013 00:42:12 +0000 (16:42 -0800)]
bonding: add xmit_hash_policy attribute netlink support

Add IFLA_BOND_XMIT_HASH_POLICY to allow get/set of bonding parameter
xmit_hash_policy via netlink.

Signed-off-by: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: add fail_over_mac attribute netlink support
sfeldma@cumulusnetworks.com [Mon, 16 Dec 2013 00:42:05 +0000 (16:42 -0800)]
bonding: add fail_over_mac attribute netlink support

Add IFLA_BOND_FAIL_OVER_MAC to allow get/set of bonding parameter
fail_over_mac via netlink.

Signed-off-by: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: add primary_select attribute netlink support
sfeldma@cumulusnetworks.com [Mon, 16 Dec 2013 00:41:58 +0000 (16:41 -0800)]
bonding: add primary_select attribute netlink support

Add IFLA_BOND_PRIMARY_SELECT to allow get/set of bonding parameter
primary_select via netlink.

Signed-off-by: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: add primary attribute netlink support
sfeldma@cumulusnetworks.com [Mon, 16 Dec 2013 00:41:51 +0000 (16:41 -0800)]
bonding: add primary attribute netlink support

Add IFLA_BOND_PRIMARY to allow get/set of bonding parameter
primary via netlink.

Signed-off-by: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopkt_sched: fq: more robust memory allocation
Eric Dumazet [Sun, 15 Dec 2013 21:15:25 +0000 (13:15 -0800)]
pkt_sched: fq: more robust memory allocation

This patch brings NUMA support and automatic fallback to vmalloc()
in case kmalloc() failed to allocate FQ hash table.

NUMA support depends on XPS being setup for the device before
qdisc allocation. After a XPS change, it might be worth creating
qdisc hierarchy again.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobondnl: use be32 nla put/get for be32 values
Jiri Pirko [Sat, 14 Dec 2013 11:32:10 +0000 (12:32 +0100)]
bondnl: use be32 nla put/get for be32 values

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnad: make local variable static
stephen hemminger [Sat, 14 Dec 2013 00:40:10 +0000 (16:40 -0800)]
bnad: make local variable static

Compile tested only.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: refine TSO splits
Eric Dumazet [Fri, 13 Dec 2013 21:51:23 +0000 (13:51 -0800)]
tcp: refine TSO splits

While investigating performance problems on small RPC workloads,
I noticed linux TCP stack was always splitting the last TSO skb
into two parts (skbs). One being a multiple of MSS, and a small one
with the Push flag. This split is done even if TCP_NODELAY is set,
or if no small packet is in flight.

Example with request/response of 4K/4K

IP A > B: . ack 68432 win 2783 <nop,nop,timestamp 6524593 6525001>
IP A > B: . 65537:68433(2896) ack 69632 win 2783 <nop,nop,timestamp 6524593 6525001>
IP A > B: P 68433:69633(1200) ack 69632 win 2783 <nop,nop,timestamp 6524593 6525001>
IP B > A: . ack 68433 win 2768 <nop,nop,timestamp 6525001 6524593>
IP B > A: . 69632:72528(2896) ack 69633 win 2768 <nop,nop,timestamp 6525001 6524593>
IP B > A: P 72528:73728(1200) ack 69633 win 2768 <nop,nop,timestamp 6525001 6524593>
IP A > B: . ack 72528 win 2783 <nop,nop,timestamp 6524593 6525001>
IP A > B: . 69633:72529(2896) ack 73728 win 2783 <nop,nop,timestamp 6524593 6525001>
IP A > B: P 72529:73729(1200) ack 73728 win 2783 <nop,nop,timestamp 6524593 6525001>

We can avoid this split by including the Nagle tests at the right place.

Note : If some NIC had trouble sending TSO packets with a partial
last segment, we would have hit the problem in GRO/forwarding workload already.

tcp_minshall_update() is moved to tcp_output.c and is updated as we might
feed a TSO packet with a partial last segment.

This patch tremendously improves performance, as the traffic now looks
like :

IP A > B: . ack 98304 win 2783 <nop,nop,timestamp 6834277 6834685>
IP A > B: P 94209:98305(4096) ack 98304 win 2783 <nop,nop,timestamp 6834277 6834685>
IP B > A: . ack 98305 win 2768 <nop,nop,timestamp 6834686 6834277>
IP B > A: P 98304:102400(4096) ack 98305 win 2768 <nop,nop,timestamp 6834686 6834277>
IP A > B: . ack 102400 win 2783 <nop,nop,timestamp 6834279 6834686>
IP A > B: P 98305:102401(4096) ack 102400 win 2783 <nop,nop,timestamp 6834279 6834686>
IP B > A: . ack 102401 win 2768 <nop,nop,timestamp 6834687 6834279>
IP B > A: P 102400:106496(4096) ack 102401 win 2768 <nop,nop,timestamp 6834687 6834279>
IP A > B: . ack 106496 win 2783 <nop,nop,timestamp 6834280 6834687>
IP A > B: P 102401:106497(4096) ack 106496 win 2783 <nop,nop,timestamp 6834280 6834687>
IP B > A: . ack 106497 win 2768 <nop,nop,timestamp 6834688 6834280>
IP B > A: P 106496:110592(4096) ack 106497 win 2768 <nop,nop,timestamp 6834688 6834280>

Before :

lpq83:~# nstat >/dev/null;perf stat ./super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K
280774

 Performance counter stats for './super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K':

     205719.049006 task-clock                #    9.278 CPUs utilized
         8,449,968 context-switches          #    0.041 M/sec
         1,935,997 CPU-migrations            #    0.009 M/sec
           160,541 page-faults               #    0.780 K/sec
   548,478,722,290 cycles                    #    2.666 GHz                     [83.20%]
   455,240,670,857 stalled-cycles-frontend   #   83.00% frontend cycles idle    [83.48%]
   272,881,454,275 stalled-cycles-backend    #   49.75% backend  cycles idle    [66.73%]
   166,091,460,030 instructions              #    0.30  insns per cycle
                                             #    2.74  stalled cycles per insn [83.39%]
    29,150,229,399 branches                  #  141.699 M/sec                   [83.30%]
     1,943,814,026 branch-misses             #    6.67% of all branches         [83.32%]

      22.173517844 seconds time elapsed

lpq83:~# nstat | egrep "IpOutRequests|IpExtOutOctets"
IpOutRequests                   16851063           0.0
IpExtOutOctets                  23878580777        0.0

After patch :

lpq83:~# nstat >/dev/null;perf stat ./super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K
280877

 Performance counter stats for './super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K':

     107496.071918 task-clock                #    4.847 CPUs utilized
         5,635,458 context-switches          #    0.052 M/sec
         1,374,707 CPU-migrations            #    0.013 M/sec
           160,920 page-faults               #    0.001 M/sec
   281,500,010,924 cycles                    #    2.619 GHz                     [83.28%]
   228,865,069,307 stalled-cycles-frontend   #   81.30% frontend cycles idle    [83.38%]
   142,462,742,658 stalled-cycles-backend    #   50.61% backend  cycles idle    [66.81%]
    95,227,712,566 instructions              #    0.34  insns per cycle
                                             #    2.40  stalled cycles per insn [83.43%]
    16,209,868,171 branches                  #  150.795 M/sec                   [83.20%]
       874,252,952 branch-misses             #    5.39% of all branches         [83.37%]

      22.175821286 seconds time elapsed

lpq83:~# nstat | egrep "IpOutRequests|IpExtOutOctets"
IpOutRequests                   11239428           0.0
IpExtOutOctets                  23595191035        0.0

Indeed, the occupancy of tx skbs (IpExtOutOctets/IpOutRequests) is higher :
2099 instead of 1417, thus helping GRO to be more efficient when using FQ packet
scheduler.

Many thanks to Neal for review and ideas.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Van Jacobson <vanj@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Tested-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: remove dead code for add/del multiple
stephen hemminger [Fri, 13 Dec 2013 20:35:56 +0000 (12:35 -0800)]
net: remove dead code for add/del multiple

These function to manipulate multiple addresses are not used anywhere
in current net-next tree. Some out of tree code maybe using these but
too bad; they should submit their code upstream..

Also, make __hw_addr_flush local since only used by dev_addr_lists.c

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Tue, 17 Dec 2013 20:08:17 +0000 (15:08 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
Please pull this batch of updates for the 3.14 stream...

For the Bluetooth bits, Gustavo says:

"This is the first batch of patches intended for 3.14. There is
nothing big here.  Most of the code are refactors, clean up, small
fixes, plus some new device id support."

And...

"More patches to 3.14. Here we have the support for Low Energy
Connection Oriented Channels (LE CoC). Basically, as the name says,
this adds supports for connection oriented channels in the same way
we already have them for BR/EDR connections so profiles/protocols
that work on top of BR/EDR can now work on LE plus a plenty of new
possibilities for LE."

For the ath10k bits, Kalle says:

"Janusz and Marek implemented DFS support to ath10k, but the code is
not enabled yet due to missing cfg80211/mac80211 patches (it will be
enabled in the next pull request). Michal did some device reset fixes
and made it possible for ath10k to share an interrupt with another
device. And lots of smaller fixes from different people."

For the iwlwifi bits, Emmanuel says:

"I have here a big rework of the rate control by Eyal. This is obviously
the biggest part of this batch.
I also have enhancement of protection flags by Avri and a few bits for
WoWLAN by Eliad and Luca. Johannes cleans up the debugfs plus a few
fixes. I provided a few things for Bluetooth coexistence.
Besides this we have an implementation for low priority scan."

Along with all that, there are big batches of updates to mwifiex and
ath9k, Jeff Kirsher's FSF address fix patches, and a handful of other
bits here and there.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'phy_power'
David S. Miller [Tue, 17 Dec 2013 19:42:57 +0000 (14:42 -0500)]
Merge branch 'phy_power'

Sebastian Hesselbarth says:

====================
net: phy: Ethernet PHY powerdown optimization

This is v2 of the ethernet PHY power optimization patches to reduce
power consumption of network PHYs with link that are either unused or
the corresponding netdev is down.

Compared to the last version, this patch set drops a patch to disable
unused PHYs after late initcall, as it is not compatible with a modular
mdio bus [1]. I'll investigate different ways to have a modular mdio bus
driver get notified when driver loading is done.

Again, a branch with v2 applied to v3.13-rc2 can also be found at
https://github.com/shesselba/linux-dove.git topic/ethphy-power-v2

[1] http://www.spinics.net/lists/arm-kernel/msg293028.html
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: suspend phydev when going to HALTED
Sebastian Hesselbarth [Fri, 13 Dec 2013 09:20:29 +0000 (10:20 +0100)]
net: phy: suspend phydev when going to HALTED

When phydev is going to HALTED state, we can try to suspend it to
safe more power. phy_suspend helper will check if PHY can be suspended,
so just call it when entering HALTED state.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: resume/suspend PHYs on attach/detach
Sebastian Hesselbarth [Fri, 13 Dec 2013 09:20:28 +0000 (10:20 +0100)]
net: phy: resume/suspend PHYs on attach/detach

This ensures PHYs are resumed on attach and suspended on detach.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: provide phy_resume/phy_suspend helpers
Sebastian Hesselbarth [Fri, 13 Dec 2013 09:20:27 +0000 (10:20 +0100)]
net: phy: provide phy_resume/phy_suspend helpers

This adds helper functions to resume and suspend a given phy_device
by calling the corresponding driver callbacks if available.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: marvell: provide genphy suspend/resume
Sebastian Hesselbarth [Fri, 13 Dec 2013 09:20:26 +0000 (10:20 +0100)]
net: phy: marvell: provide genphy suspend/resume

Marvell PHYs support generic PHY suspend/resume, so provide those
callbacks to all marvell specific drivers.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: mv643xx_eth: properly start/stop phy device
Sebastian Hesselbarth [Fri, 13 Dec 2013 09:20:25 +0000 (10:20 +0100)]
net: mv643xx_eth: properly start/stop phy device

When using phydev, it should be phy_start/phy_stop'ed properly. This
driver doesn't do that, so add the corresponding calls to port_start/
stop respectively.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosctp: Reorder 'struc association' members to reduce its size
wangweidong [Fri, 13 Dec 2013 05:51:04 +0000 (13:51 +0800)]
sctp: Reorder 'struc association' members to reduce its size

Members of 'struct association' are not in appropriate order to
reuse compiler added padding on 64bit architectures. In this patch
we reorder those struct members and help reduce the size of the
structure from 2776 bytes to 2720 bytes on 64 bit architectures.

Signed-off-by: Wang Weidong <wangweidong1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Tue, 17 Dec 2013 19:31:17 +0000 (14:31 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to i40e only (again).

Jesse provides a fix for when tx_rings structure is NULL and we do not want
to panic. Then refactors the flow control set up and disables L2 flow control
by default.  Provides some trivial fixes as well as prevent compiler warnings.
Then to align to similar behaviour in ixgbe, use the total number of CPUs in
the system to suggest the number of transmit and receive queue pairs.

Shannon provides a i40e ethtool fix to get some more reasonable information
reports back out to the ethtool.  In addition, fixes PF reset after offline
test, where it reorders the test to put the register test last as it is the
only one that needs a reset, and we wait to trigger the reset until after we
clear the testing bit.  Lastly provides basic support for handling suspend
and resume for now, later on Wake-On-LAN support will be added.

Anjali provides changes to tell the stack about our actual number of queues
in order for RFS/RPS/XFS to work correctly.  Then provides several patches to
implement dynamically changing the queue count for the main VSI.  Adds
basic support for get/set channels for RSS so that the number of receive and
transmit queue pair can be changed via ethtool.  Cleans up the use of
rtnl_lock in the reset patch since it runs from a work time.

Neerav Parikh cleans up the VF interface to remove FCoE code as this
feature will not be supported on VF interfaces.

v2:
  - submitted patch 1 to net (since it was a fix needed for net), so dropped
    from this series (this patch will get added to net-next when Dave syncs
    his trees)
  - Dropped patches 4 & 11 from previous submission because of feedback
    received from Ben Hutchings and Sergei Shtylyov.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>