pandora-kernel.git
10 years agobnx2x: Add Private Flags Support
Yuval Mintz [Wed, 22 May 2013 21:21:49 +0000 (21:21 +0000)]
bnx2x: Add Private Flags Support

Utilize ethtool's callback `get_priv_flags' - shed more light on the
feasibility of devices as storage interfaces.

CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Yuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'xen_split_event_channels'
David S. Miller [Fri, 24 May 2013 01:40:46 +0000 (18:40 -0700)]
Merge branch 'xen_split_event_channels'

Wei Liu says:

====================
This series adds a new feature called split event channels.

In the original implementation, only one event channel is setup between
frontend and backend. This is not ideal as TX notification interferes with RX
notification. Using dedicated event channels for TX and RX solves this issue.

Changes since V2:
* feature_split_event_channels -> separate_tx_rx_irq
* make separate_tx_rx_irq bool
* document this feature in header file
* don't report fatal if writing feature to xenstore fails
* frontend will fall back to single event channel if new feature setup fails

Changes since V1:
* change subject lines of commits to be more specific
* add parameter feature_split_event_channels for xen-netback
* remove two dev_info
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen: netif.h: document feature-split-event-channels
Wei Liu [Wed, 22 May 2013 06:34:47 +0000 (06:34 +0000)]
xen: netif.h: document feature-split-event-channels

This patch synchronises documentation for feature-split-event-channels from
Xen canonical header file.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netfront: split event channels support for Xen frontend driver
Wei Liu [Wed, 22 May 2013 06:34:46 +0000 (06:34 +0000)]
xen-netfront: split event channels support for Xen frontend driver

This patch adds a new feature called feature-split-event-channels for
netfront, enabling it to handle TX and RX events separately.

If netback does not support this feature, it falls back to use single event
channel.

If netfront fails to setup split event channels, it will try falling back to
single event channel.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: split event channels support for Xen backend driver
Wei Liu [Wed, 22 May 2013 06:34:45 +0000 (06:34 +0000)]
xen-netback: split event channels support for Xen backend driver

Netback and netfront only use one event channel to do TX / RX notification,
which may cause unnecessary wake-up of processing routines. This patch adds a
new feature called feature-split-event-channels to netback, enabling it to
handle TX and RX events separately.

Netback will use tx_irq to notify guest for TX completion, rx_irq for RX
notification.

If frontend doesn't support this feature, tx_irq equals to rx_irq.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: use ipv6_addr_scope() helper
Cong Wang [Wed, 22 May 2013 05:52:22 +0000 (05:52 +0000)]
ipv6: use ipv6_addr_scope() helper

ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK could be replaced
by ipv6_addr_scope(), which is slightly faster.

Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: use ipv6_addr_any() helper
Cong Wang [Wed, 22 May 2013 05:41:06 +0000 (05:41 +0000)]
ipv6: use ipv6_addr_any() helper

ipv6_addr_any() is a faster way to determine if an addr
is ipv6 any addr, no need to compute the addr type.

Cc: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosch_tbf: segment too big GSO packets
Eric Dumazet [Tue, 21 May 2013 08:16:46 +0000 (08:16 +0000)]
sch_tbf: segment too big GSO packets

If a GSO packet has a length above tbf burst limit, the packet
is currently silently dropped.

Current way to handle this is to set the device in non GSO/TSO mode, or
setting high bursts, and its sub optimal.

We can actually segment too big GSO packets, and send individual
segments as tbf parameters allow, allowing for better interoperability.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofec: Use DIV_ROUND_UP macro
Fabio Estevam [Tue, 21 May 2013 05:44:26 +0000 (05:44 +0000)]
fec: Use DIV_ROUND_UP macro

Use the standard DIV_ROUND_UP macro in order to provide better readability.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb3: Check and handle the dma mapping errors
Santosh Rastapur [Tue, 21 May 2013 04:21:29 +0000 (04:21 +0000)]
cxgb3: Check and handle the dma mapping errors

This patch adds checks at approprate places whether *dma_map*() call has
succeeded or not.

Signed-off-by: Santosh Rastapur <santosh@chelsio.com>
Reviewed-by: Jay Fenlason <fenlason@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb3: Fix warning about using rcu_dereference when not in a rcu-locked section
Jay Fenlason [Tue, 21 May 2013 04:21:28 +0000 (04:21 +0000)]
cxgb3: Fix warning about using rcu_dereference when not in a rcu-locked section

It is about using rcu_dereference() when not in a rcu-locked section. It only
happens on initialization hence fix the initialization to not rcu_dereference()

Signed-off-by: Jay Fenlason <fenlason@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: smsc911x: don't artificially limit build
Mark Rutland [Tue, 21 May 2013 00:22:37 +0000 (00:22 +0000)]
net: smsc911x: don't artificially limit build

Currently the SMSC911X driver may only be built for a specific set of
architectures, being limited to do so by a Kconfig depends line. This
means that if a platform wishes to use the driver, its architecture must
be added to the list explicitly, introducing pointless churn.

This may have been due to the driver's use of the {read,write}s{b,w,l}
functions, which have since been replaced with the more standard
io{read,write}{8,16,32}_rep. We can instead depend on HAS_IOMEM, which
should prevent build issues while allowing the driver to be built for
currently unlisted architectures, including x86 and arm64.

This patch removes the explicit list of architectures from the driver's
depend line, and replaces it with a dependency on HAS_IOMEM.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Loosen constraints for recalculating checksum in skb_segment()
Simon Horman [Sun, 19 May 2013 15:46:49 +0000 (15:46 +0000)]
net: Loosen constraints for recalculating checksum in skb_segment()

This is a generic solution to resolve a specific problem that I have observed.

If the encapsulation of an skb changes then ability to offload checksums
may also change. In particular it may be necessary to perform checksumming
in software.

An example of such a case is where a non-GRE packet is received but
is to be encapsulated and transmitted as GRE.

Another example relates to my proposed support for for packets
that are non-MPLS when received but MPLS when transmitted.

The cost of this change is that the value of the csum variable may be
checked when it previously was not. In the case where the csum variable is
true this is pure overhead. In the case where the csum variable is false it
leads to software checksumming, which I believe also leads to correct
checksums in transmitted packets for the cases described above.

Further analysis:

This patch relies on the return value of can_checksum_protocol()
being correct and in turn the return value of skb_network_protocol(),
used to provide the protocol parameter of can_checksum_protocol(),
being correct. It also relies on the features passed to skb_segment()
and in turn to can_checksum_protocol() being correct.

I believe that this problem has not been observed for VLANs because it
appears that almost all drivers, the exception being xgbe, set
vlan_features such that that the checksum offload support for VLAN packets
is greater than or equal to that of non-VLAN packets.

I wonder if the code in xgbe may be an oversight and the hardware does
support checksumming of VLAN packets.  If so it may be worth updating the
vlan_features of the driver as this patch will force such checksums to be
performed in software rather than hardware.

Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: send query as soon as leave is received
Cong Wang [Tue, 21 May 2013 21:52:56 +0000 (21:52 +0000)]
bridge: send query as soon as leave is received

Continue sending queries when leave is received if the user marks
it as a querier.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Cong Wang <amwang@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: only expire the mdb entry when query is received
Cong Wang [Tue, 21 May 2013 21:52:55 +0000 (21:52 +0000)]
bridge: only expire the mdb entry when query is received

Currently we arm the expire timer when the mdb entry is added,
however, this causes problem when there is no querier sent
out after that.

So we should only arm the timer when a corresponding query is
received, as suggested by Herbert.

And he also mentioned "if there is no querier then group
subscriptions shouldn't expire. There has to be at least one querier
in the network for this thing to work.  Otherwise it just degenerates
into a non-snooping switch, which is OK."

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Cong Wang <amwang@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: use the bridge IP addr as source addr for querier
Cong Wang [Tue, 21 May 2013 21:52:54 +0000 (21:52 +0000)]
bridge: use the bridge IP addr as source addr for querier

Quote from Adam:
"If it is believed that the use of 0.0.0.0
as the IP address is what is causing strange behaviour on other devices
then is there a good reason that a bridge rather than a router shouldn't
be the active querier? If not then using the bridge IP address and
having the querier enabled by default may be a reasonable solution
(provided that our querier obeys the election rules and shuts up if it
sees a query from a lower IP address that isn't 0.0.0.0). Just because a
device is the elected querier for IGMP doesn't appear to mean it is
required to perform any other routing functions."

And introduce a new troggle for it, as suggested by Herbert.

Suggested-by: Adam Baker <linux@baker-net.org.uk>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Adam Baker <linux@baker-net.org.uk>
Signed-off-by: Cong Wang <amwang@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/nvidia/forcedeth: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 13:42:56 +0000 (13:42 +0000)]
net/ethernet/nvidia/forcedeth: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

The name of the pci_driver struct had to be changed in order to prevent
a build failure.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/chelsio/cxgb/cxgb2: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 13:42:55 +0000 (13:42 +0000)]
net/ethernet/chelsio/cxgb/cxgb2: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

The name of the pci_driver struct had to be changed in order to prevent
a build failure.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/fddi/skfp/skfddi: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 13:42:54 +0000 (13:42 +0000)]
net/fddi/skfp/skfddi: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/hippi/rrunner: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 13:42:53 +0000 (13:42 +0000)]
net/hippi/rrunner: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/amd/amd8111e: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:10 +0000 (12:58 +0000)]
net/ethernet/amd/amd8111e: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/sun/sungem: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:09 +0000 (12:58 +0000)]
net/ethernet/sun/sungem: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/qlogic/qlge/qlge_main: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:08 +0000 (12:58 +0000)]
net/ethernet/qlogic/qlge/qlge_main: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/sgi/ioc3-eth: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:07 +0000 (12:58 +0000)]
net/ethernet/sgi/ioc3-eth: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/broadcom/tg3: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:06 +0000 (12:58 +0000)]
net/ethernet/broadcom/tg3: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/broadcom/bnx2: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:58:05 +0000 (12:58 +0000)]
net/ethernet/broadcom/bnx2: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/alteon/acenic: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:14 +0000 (12:42 +0000)]
net/ethernet/alteon/acenic: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/icplus/ipg: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:13 +0000 (12:42 +0000)]
net/ethernet/icplus/ipg: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/toshiba/tc35815: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:12 +0000 (12:42 +0000)]
net/ethernet/toshiba/tc35815: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/dec/tulip/xircom_cb: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:11 +0000 (12:42 +0000)]
net/ethernet/dec/tulip/xircom_cb: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/sis/sis190: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:10 +0000 (12:42 +0000)]
net/ethernet/sis/sis190: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/atheros/atlx/atl1: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:09 +0000 (12:42 +0000)]
net/ethernet/atheros/atlx/atl1: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/atheros/atl1e/atl1e_main: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:08 +0000 (12:42 +0000)]
net/ethernet/atheros/atl1e/atl1e_main: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/atheros/atl1c/atl1c_main: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:07 +0000 (12:42 +0000)]
net/ethernet/atheros/atl1c/atl1c_main: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ethernet/silan/sc92031: Use module_pci_driver to register driver
Peter Hüwe [Tue, 21 May 2013 12:42:06 +0000 (12:42 +0000)]
net/ethernet/silan/sc92031: Use module_pci_driver to register driver

Removing some boilerplate by using module_pci_driver instead of calling
register and unregister in the otherwise empty init/exit functions.

Signed-off-by: Peter Huewe <peterhuewe@gmx.de>
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, 22 May 2013 20:56:56 +0000 (13:56 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
This series contains updates to e1000e, igb and ixgbe.

Bruce Allan provide 2 minor cleanups for e1000e to resolve whitespace
issues and build warnings about unused parameters.

Carolyn provides a couple of fixes for igb, one being a fix for a
possible panic when the interface is down and receive traffic
arrives.  The second fix resolves an issue on newer parts which have
multiple checksum fields and set_ethtool was only checking to update
the first checksum of the NVM image.

Akeem provides majority of the changes in this patch set.  Akeem
provides a fix for e1000e on an issue reported from the community to
resolve the issue of unlocking swflag_mutex for 82574 and 82583
devices even if the hardware semaphore was successfully acquired.
The other patches from Akeem are against igb, where he adds support
SFP module discovery, LED blink mechanism for devices using cathodes,
LED support for i210/i211 parts and cleanup of a i2c function which
was not being used.

Matthew provides an update for igb to support a more accurate check
for a PTP RX hang.

Amir provides a patch for ixgbe to set the software prio_tc values at
initialization to the hardware setting to remove the need to reset the
device at the first time we call ixgbe_dcbnl_ieee_setets.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoIXGBE: Set the SW prio_tc values at initialization to the HW setting.
Amir Hanania [Thu, 18 Apr 2013 04:23:52 +0000 (04:23 +0000)]
IXGBE: Set the SW prio_tc values at initialization to the HW setting.

Set the SW prio_tc values at initialization to the HW setting.
Setting the SW prio_tc default values to be the HW setting by reading the
rtrup2tc register. For any TC change we need to reset the device.
This will remove the need to reset the device at the first
time we call ixgbe_dcbnl_ieee_setets.

Signed-off-by: Amir Hanania <amir.hanania@intel.com>
Tested-by: Jack Morgan<jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Removed unused i2c function
Akeem G. Abodunrin [Wed, 15 May 2013 07:41:30 +0000 (07:41 +0000)]
igb: Removed unused i2c function

This patch removes unused i2c function definition.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Implementation of i210/i211 LED support
Akeem G. Abodunrin [Wed, 1 May 2013 05:44:45 +0000 (05:44 +0000)]
igb: Implementation of i210/i211 LED support

This patch fixes LED issues with i210 and i211 devices, due to changes in the
device registers.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Fix possible panic caused by Rx traffic arrival while interface is down
Carolyn Wyborny [Tue, 30 Apr 2013 00:21:32 +0000 (00:21 +0000)]
igb: Fix possible panic caused by Rx traffic arrival while interface is down

This patch reorders disabling napi and irqs during igb_down.
This is done to avoid possible panic's found in other Intel drivers
when Rx traffic arrives while interface is going down.

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 set_ethtool function to call update nvm for entire image
Carolyn Wyborny [Thu, 25 Apr 2013 17:22:34 +0000 (17:22 +0000)]
igb: Fix set_ethtool function to call update nvm for entire image

This patch fixes a problem where we were only checking to update checksum
on first part of nvm image.  Newer parts have multiple checksum fields and
checksum function will accommodate that as long as we call it in the first
place for any changes made.

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: SerDes flow control setting
Akeem G. Abodunrin [Fri, 29 Mar 2013 15:22:17 +0000 (15:22 +0000)]
igb: SerDes flow control setting

This path allows users to get appropriate flow control setting on SerDes
devices, based on original implementation for Copper devices.
Also, since 100baseFX does not support setting flow control, so exclude
it from the setting mechanism.

Signed-off-by: Akeem G. Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Support for SFP modules discovery
Akeem G. Abodunrin [Wed, 24 Apr 2013 16:54:50 +0000 (16:54 +0000)]
igb: Support for SFP modules discovery

This patch adds support for SFP modules media type discovery for
SGMII, which will enable driver to detect supported external PHYs,
including 100baseFXSFP module.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Add update to last_rx_timestamp in Rx rings
Matthew Vick [Wed, 24 Apr 2013 07:42:06 +0000 (07:42 +0000)]
igb: Add update to last_rx_timestamp in Rx rings

In order to support a more accurate check for a PTP Rx hang where the
device can no longer timestamp received packets, we need to update, per
ring, when the last Rx timestamp was. Because of how the PTP Rx hang logic
works, the current logic is valid, but properly updating the ring variable
increases the accuracy of the check.

Signed-off-by: Matthew Vick <matthew.vick@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoigb: Changed LEDs blink mechanism to include designs using cathode
Akeem G. Abodunrin [Fri, 29 Mar 2013 08:22:25 +0000 (08:22 +0000)]
igb: Changed LEDs blink mechanism to include designs using cathode

This patch addresses the changes needed to make LEDs work properly with
negative logic. This implementation uses LED Invert bit to reverse the
logic issue that occurred when LEDs are driven by cathode. Keep LEDs
blinking for SerDes devices. Also made changes to magic number and the
for loop to reduce number of shifts.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: Release mutex lock only if it has been initially acquired
Akeem G. Abodunrin [Thu, 2 May 2013 02:57:44 +0000 (02:57 +0000)]
e1000e: Release mutex lock only if it has been initially acquired

This patch fixes the issue of unlocking swflag_mutex for 82574 and 82583
devices regardless of if the hw semaphore has been successfully acquired via
e1000_get_hw_semaphore_82574(). With this patch, unlocking mutex now depends
on if the hw semaphore was successfully acquired before. And 82574/82583
devices are reset regardless of whether e1000_get_hw_semaphore_82574()
returns success or failure.

Reported-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: prevent warning from -Wunused-parameter
Bruce Allan [Wed, 1 May 2013 03:48:11 +0000 (03:48 +0000)]
e1000e: prevent warning from -Wunused-parameter

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agoe1000e: cleanup whitespace
Bruce Allan [Wed, 1 May 2013 01:19:46 +0000 (01:19 +0000)]
e1000e: cleanup whitespace

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
10 years agophy: add phy_mac_interrupt() to use with PHY_IGNORE_INTERRUPT
Florian Fainelli [Sun, 19 May 2013 22:53:43 +0000 (22:53 +0000)]
phy: add phy_mac_interrupt() to use with PHY_IGNORE_INTERRUPT

There is currently no way for an Ethernet MAC driver servicing PHY link
interrupts to notify this to the PHY state machine without defining its
own state machine. Since most drivers are not so special, introduce a
helper: phy_mac_interrupt() which can be called from a link up/down
interrupt routine to update the PHY state machine. To avoid code
duplication some refactoring has been done to expose the workqueue and
its corresponding callback internally.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agophy: fix the use of PHY_IGNORE_INTERRUPT
Florian Fainelli [Sun, 19 May 2013 22:53:42 +0000 (22:53 +0000)]
phy: fix the use of PHY_IGNORE_INTERRUPT

When a PHY device is registered with the special IRQ value
PHY_IGNORE_INTERRUPT (-2) it will not properly be handled by the PHY
library:

- it continues to poll its register, while we do not want this
  because such PHY link events or register changes are serviced by an
  Ethernet MAC
- it will still try to configure PHY interrupts at the PHY level, such
  interrupts do not exist at the PHY but at the MAC level
- the state machine only handles PHY_POLL, but should also handle
  PHY_IGNORE_INTERRUPT similarly

This patch updates the PHY state machine and initialization paths to
account for the specific PHY_IGNORE_INTERRUPT. Based on an earlier patch
by Thomas Petazzoni, and reworked to add the missing bits. Add a helper
phy_interrupt_is_valid() which specifically tests for a PHY interrupt
not to be PHY_POLL or PHY_IGNORE_INTERRUPT and use it throughout the
code.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Driver and Firmware Updated
Rasesh Mody [Mon, 20 May 2013 10:08:04 +0000 (10:08 +0000)]
bna: Driver and Firmware Updated

Driver and Firmware versions updated to 3.2.21.1.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Enahncement to Identify Default IOC Function
Rasesh Mody [Mon, 20 May 2013 10:08:03 +0000 (10:08 +0000)]
bna: Enahncement to Identify Default IOC Function

User should not be allowed to delete base function of eth port. Add a new field
to the bfa ioc attributes structure to indicate if the given ioc is default
function on the port or not.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Fix Ucast Failure Handling
Rasesh Mody [Mon, 20 May 2013 10:08:02 +0000 (10:08 +0000)]
bna: Fix Ucast Failure Handling

Failure of the UCAST set for base mac address fails when user configures a
duplicate mac address that matches that of another vNIC on the same port.
The bna does not handle the ucast failure and keeps this address in cache.
On disable of the vNIC, bna tries to delete the failed base mac address and the
fw asserts.

On failure of ucast address, mark ucast address set to false.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobna: Clear Driver Config Flags When HW Resets
Rasesh Mody [Mon, 20 May 2013 10:08:01 +0000 (10:08 +0000)]
bna: Clear Driver Config Flags When HW Resets

Driver configuration flags are retained across open/stop operations preventing
configurations to be set in next open/stop. Setting MTU on a 1020 causes
network to fail until a reboot is performed on the host.

Clear the flags when configuration resets in hardware.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: dm9000: Allow instantiation using device tree
Tomasz Figa [Mon, 20 May 2013 09:16:58 +0000 (09:16 +0000)]
net: dm9000: Allow instantiation using device tree

This patch adds Device Tree support to dm9000 driver.

Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
Reviewed-by: Sylwester Nawrocki <sylvester.nawrocki@gmail.com>
Reviewed-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoarm: bpf_jit: can call module_free() from any context
Daniel Borkmann [Mon, 20 May 2013 08:05:51 +0000 (08:05 +0000)]
arm: bpf_jit: can call module_free() from any context

Follow-up on module_free()/vfree() that takes care of the rest, so no
longer this workaround with work_struct needed.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Mircea Gherzan <mgherzan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoppc: bpf_jit: can call module_free() from any context
Daniel Borkmann [Mon, 20 May 2013 08:05:50 +0000 (08:05 +0000)]
ppc: bpf_jit: can call module_free() from any context

Followup patch on module_free()/vfree() that takes care of the rest, so
no longer this workaround with work_struct is needed.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Matt Evans <matt@ozlabs.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: md5: remove spinlock usage in fast path
Eric Dumazet [Mon, 20 May 2013 06:52:26 +0000 (06:52 +0000)]
tcp: md5: remove spinlock usage in fast path

TCP md5 code uses per cpu variables but protects access to them with
a shared spinlock, which is a contention point.

[ tcp_md5sig_pool_lock is locked twice per incoming packet ]

Makes things much simpler, by allocating crypto structures once, first
time a socket needs md5 keys, and not deallocating them as they are
really small.

Next step would be to allow crypto allocations being done in a NUMA
aware way.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ipv6: remove 'next' member from inet6_dev
Daniel Borkmann [Mon, 20 May 2013 04:53:38 +0000 (04:53 +0000)]
net: ipv6: remove 'next' member from inet6_dev

The next pointer within the inet6_dev structure seems not to be used
anywhere. So just remove it. Tested with allmodconfig on x86_64.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agorps: selective flow shedding during softnet overflow
Willem de Bruijn [Mon, 20 May 2013 04:02:32 +0000 (04:02 +0000)]
rps: selective flow shedding during softnet overflow

A cpu executing the network receive path sheds packets when its input
queue grows to netdev_max_backlog. A single high rate flow (such as a
spoofed source DoS) can exceed a single cpu processing rate and will
degrade throughput of other flows hashed onto the same cpu.

This patch adds a more fine grained hashtable. If the netdev backlog
is above a threshold, IRQ cpus track the ratio of total traffic of
each flow (using 4096 buckets, configurable). The ratio is measured
by counting the number of packets per flow over the last 256 packets
from the source cpu. Any flow that occupies a large fraction of this
(set at 50%) will see packet drop while above the threshold.

Tested:
Setup is a muli-threaded UDP echo server with network rx IRQ on cpu0,
kernel receive (RPS) on cpu0 and application threads on cpus 2--7
each handling 20k req/s. Throughput halves when hit with a 400 kpps
antagonist storm. With this patch applied, antagonist overload is
dropped and the server processes its complete load.

The patch is effective when kernel receive processing is the
bottleneck. The above RPS scenario is a extreme, but the same is
reached with RFS and sufficient kernel processing (iptables, packet
socket tap, ..).

Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofec: Let device core handle pinctrl
Fabio Estevam [Mon, 20 May 2013 03:06:17 +0000 (03:06 +0000)]
fec: Let device core handle pinctrl

Since commit ab78029 (drivers/pinctrl: grab default handles from device core)
we can rely on device core for handling pinctrl, so remove
devm_pinctrl_get_select_default() from the driver.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netfront: avoid leaking resources when setup_netfront fails
Wei Liu [Mon, 20 May 2013 01:05:12 +0000 (01:05 +0000)]
xen-netfront: avoid leaking resources when setup_netfront fails

We should correctly free related resources (grant ref, memory page, evtchn)
when setup_netfront fails.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: velocity: Add platform device support to VIA velocity driver
Tony Prisk [Sat, 18 May 2013 09:39:07 +0000 (09:39 +0000)]
net: velocity: Add platform device support to VIA velocity driver

Add support for the VIA Velocity network driver to be bound to a
OF created platform device.

Signed-off-by: Tony Prisk <linux@prisktech.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: velocity: Convert to generic dma functions
Tony Prisk [Sat, 18 May 2013 09:39:06 +0000 (09:39 +0000)]
net: velocity: Convert to generic dma functions

Remove the pci_* dma functions and replace with the more generic
versions.

In preparation of adding platform support, a new struct device *dev
is added to struct velocity_info which can be used by both the pci
and platform code.

Signed-off-by: Tony Prisk <linux@prisktech.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: velocity: Rename vptr->dev to vptr->netdev
Tony Prisk [Sat, 18 May 2013 09:39:05 +0000 (09:39 +0000)]
net: velocity: Rename vptr->dev to vptr->netdev

Improve the clarity of the code in preparation for converting the
dma functions to generic versions, which require a struct device *.

This makes it possible to store a 'struct device *dev' in the
velocity_info structure.

Signed-off-by: Tony Prisk <linux@prisktech.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years ago3c59x: remove useless VORTEX_PCI() invocations
Sergei Shtylyov [Sun, 19 May 2013 10:17:13 +0000 (10:17 +0000)]
3c59x: remove useless VORTEX_PCI() invocations

It's suboptimal to invoke quite complex VORTEX_PCI() macro every time we want
to get a 'struct pci_dev *' when we already have it in a variable...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoThunderLAN: remove is_eisa flag
Rolf Eike Beer [Sat, 18 May 2013 11:50:17 +0000 (11:50 +0000)]
ThunderLAN: remove is_eisa flag

These 2 places are the only matches for is_eisa in the whole tree.

Signed-off-by: Rolf Eike Beer <eike-kernel@sf-tec.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet-bnx2x: dont reload on GRO change
Eric Dumazet [Sat, 18 May 2013 07:14:53 +0000 (07:14 +0000)]
net-bnx2x: dont reload on GRO change

bnx2x_set_features() forces a driver reload if GRO setting is changed.

A reload makes the ethernet port unresponsive for about 5 seconds.

This is not needed in the common case LRO is enabled, as LRO
(TPA_ENABLE_FLAG) has precedence over GRO (GRO_ENABLE_FLAG)

Tested:
 Verified that "ethtool -K eth0 gro {on|off}" doesn't blackout
 the NIC anymore

Google-Bug-Id: 8440442
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Dmitry Kravkov <dmitry@broadcom.com>
Acked-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'tg3_eee'
David S. Miller [Mon, 20 May 2013 07:13:54 +0000 (00:13 -0700)]
Merge branch 'tg3_eee'

Nithin Nayak Sujir says:

====================
This series adds support for modifying EEE settings via ethtool. Since this can
impact Link Flap Avoidance, the driver pulls the current hardware settings if
LFA is enabled. This is similar to how we do the link settings to avoid a flap.

v2: Fixes pointed out by Ben Hutchings.
 - Use MDIO_AN_EEE_LPABLE to set the lp_advertised field.
 - Check that tx_lpi_timer is within valid range.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Implement set/get_eee handlers
Nithin Sujir [Sat, 18 May 2013 06:26:55 +0000 (06:26 +0000)]
tg3: Implement set/get_eee handlers

Reviewed-by: Ben Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Simplify tg3_phy_eee_config_ok() by reusing tg3_eee_pull_config()
Nithin Sujir [Sat, 18 May 2013 06:26:54 +0000 (06:26 +0000)]
tg3: Simplify tg3_phy_eee_config_ok() by reusing tg3_eee_pull_config()

eee_config_ok() was checking only for mismatch in advertised settings.
This patch expands the scope of eee_config_ok() to check for mismatch in
the other eee settings. On mismatch we will require a call to
tg3_setup_eee() to push the configured settings to the hardware.

Reviewed-by: Ben Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Add tg3_eee_pull_config() function
Nithin Sujir [Sat, 18 May 2013 06:26:53 +0000 (06:26 +0000)]
tg3: Add tg3_eee_pull_config() function

Add tg3_eee_pull_config() to pull the settings from the hardware and
populate the eee structure.

If Link Flap Avoidance is enabled, we pull the eee settings from the hw
so as not to cause a phy reset on eee config mismatch later. This
requires moving down tg3_setup_eee() below the tg3_pull_config() to not
trample existing settings.

Reviewed-by: Ben Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Add ethtool_eee struct and tg3_setup_eee()
Nithin Sujir [Sat, 18 May 2013 06:26:52 +0000 (06:26 +0000)]
tg3: Add ethtool_eee struct and tg3_setup_eee()

Add an eee structure and update it with eee settings. This will be used
for set/get_eee operations. Add common function tg3_setup_eee() that
will be used in the subsequent patches.

Reviewed-by: Ben Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofilter: do not output bpf image address for security reason
Eric Dumazet [Fri, 17 May 2013 16:57:37 +0000 (16:57 +0000)]
filter: do not output bpf image address for security reason

Do not leak starting address of BPF JIT code for non root users,
as it might help intruders to perform an attack.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agox86: bpf_jit_comp: secure bpf jit against spraying attacks
Eric Dumazet [Fri, 17 May 2013 16:37:03 +0000 (16:37 +0000)]
x86: bpf_jit_comp: secure bpf jit against spraying attacks

hpa bringed into my attention some security related issues
with BPF JIT on x86.

This patch makes sure the bpf generated code is marked read only,
as other kernel text sections.

It also splits the unused space (we vmalloc() and only use a fraction of
the page) in two parts, so that the generated bpf code not starts at a
known offset in the page, but a pseudo random one.

Refs:
http://mainisusuallyafunction.blogspot.com/2012/11/attacking-hardened-linux-systems-with.html

Reported-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: remove bad timeout logic in fast recovery
Yuchung Cheng [Fri, 17 May 2013 13:45:05 +0000 (13:45 +0000)]
tcp: remove bad timeout logic in fast recovery

tcp_timeout_skb() was intended to trigger fast recovery on timeout,
unfortunately in reality it often causes spurious retransmission
storms during fast recovery. The particular sign is a fast retransmit
over the highest sacked sequence (SND.FACK).

Currently the RTO timer re-arming (as in RFC6298) offers a nice cushion
to avoid spurious timeout: when SND.UNA advances the sender re-arms
RTO and extends the timeout by icsk_rto. The sender does not offset
the time elapsed since the packet at SND.UNA was sent.

But if the next (DUP)ACK arrives later than ~RTTVAR and triggers
tcp_fastretrans_alert(), then tcp_timeout_skb() will mark any packet
sent before the icsk_rto interval lost, including one that's above the
highest sacked sequence. Most likely a large part of scorebard will be
marked.

If most packets are not lost then the subsequent DUPACKs with new SACK
blocks will cause the sender to continue to retransmit packets beyond
SND.FACK spuriously. Even if only one packet is lost the sender may
falsely retransmit almost the entire window.

The situation becomes common in the world of bufferbloat: the RTT
continues to grow as the queue builds up but RTTVAR remains small and
close to the minimum 200ms. If a data packet is lost and the DUPACK
triggered by the next data packet is slightly delayed, then a spurious
retransmission storm forms.

As the original comment on tcp_timeout_skb() suggests: the usefulness
of this feature is questionable. It also wastes cycles walking the
sack scoreboard and is actually harmful because of false recovery.

It's time to remove this.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoDocumentation/sysctl/net.txt: fix (attribute removal).
Rami Rosen [Fri, 17 May 2013 09:10:34 +0000 (09:10 +0000)]
Documentation/sysctl/net.txt: fix (attribute removal).

This patch removes mentioning the sysfsf net_device weight attribute
(class/net/<device>/weight)
in Documentation/sysctl/net.txt, since the net sysfs weight attribute
was removed by the following patch:

[NET]: Make NAPI polling independent of struct net_device objects
 bea3348eef27e6044b6161fd04c3152215f96411

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: add support of peer address
Nicolas Dichtel [Thu, 16 May 2013 22:32:00 +0000 (22:32 +0000)]
ipv6: add support of peer address

This patch adds the support of peer address for IPv6. For example, it is
possible to specify the remote end of a 6inY tunnel.
This was already possible in IPv4:
 ip addr add ip1 peer ip2 dev dev1

The peer address is specified with IFA_ADDRESS and the local address with
IFA_LOCAL (like explained in include/uapi/linux/if_addr.h).
Note that the API is not changed, because before this patch, it was not
possible to specify two different addresses in IFA_LOCAL and IFA_REMOTE.
There is a small change for the dump: if the peer is different from ::,
IFA_ADDRESS will contain the peer address instead of the local address.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosparc: bpf_jit_comp: can call module_free() from any context
Eric Dumazet [Fri, 17 May 2013 12:12:34 +0000 (12:12 +0000)]
sparc: bpf_jit_comp: can call module_free() from any context

module_free()/vfree() takes care of details, we no longer need a wrapper
and a work_struct.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodev: remove duplicate 'skb->dev = dev' in dev_forward_skb()
Nicolas Dichtel [Thu, 16 May 2013 23:36:32 +0000 (23:36 +0000)]
dev: remove duplicate 'skb->dev = dev' in dev_forward_skb()

This was added by commit 59b9997baba5 (Revert "net: maintain namespace
isolation between vlan and real device").
In fact, before the initial commit - the one that is reverted -, this
statement was not present.
'skb->dev = dev' is already done in eth_type_trans(), which is call just
after.

Spotted-by: Alain Ritoux <alain.ritoux@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: enable user to unload netback module
Wei Liu [Thu, 16 May 2013 23:26:11 +0000 (23:26 +0000)]
xen-netback: enable user to unload netback module

This patch enables user to unload netback module, which is useful when user
wants to upgrade to a newer netback module without rebooting the host.

Netfront cannot handle netback removal event. As we cannot fix all possible
frontends we add module get / put along with vif get / put to avoid
mis-unloading of netback. To unload netback module, user needs to shutdown all
VMs or migrate them to another host or unplug all vifs before hand.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>¬
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: remove dead code
Wei Liu [Thu, 16 May 2013 23:24:28 +0000 (23:24 +0000)]
xen-netback: remove dead code

The array mmap_pages is never touched in the initialization function. This is
remnant of mapping mechanism, which does not exist upstream. In current
upstream code this array only tracks usage of pages inside netback. Those
pages are allocated when contructing a SKB and passed directly to network
subsystem.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agox86: bpf_jit_comp: can call module_free() from any context
Eric Dumazet [Thu, 16 May 2013 19:45:30 +0000 (19:45 +0000)]
x86: bpf_jit_comp: can call module_free() from any context

It looks like we can call module_free()/vfree() from softirq context,
so no longer need a wrapper and a work_struct.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/usb: r8152: Use module_usb_driver()
Sachin Kamat [Thu, 16 May 2013 17:48:08 +0000 (17:48 +0000)]
net/usb: r8152: Use module_usb_driver()

module_usb_driver() eliminates boilerplate and simplifies the code.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/usb: r8152: Remove redundant version.h header inclusion
Sachin Kamat [Thu, 16 May 2013 17:48:07 +0000 (17:48 +0000)]
net/usb: r8152: Remove redundant version.h header inclusion

version.h header inclusion is not necessary as detected by
checkversion.pl.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovxlan: listen on multiple ports
stephen hemminger [Thu, 16 May 2013 11:35:20 +0000 (11:35 +0000)]
vxlan: listen on multiple ports

The commit 823aa873bc782f1c51b1ce8ec6da7cfcaf93836e
  Author: stephen hemminger <stephen@networkplumber.org>
  Date:   Sat Apr 27 11:31:57 2013 +0000

    vxlan: allow choosing destination port per vxlan

introduced per-vxlan UDP port configuration but only did half of the
necessary work.  It added per vxlan destination for sending, but
overlooked the handling of multiple ports for incoming traffic.

This patch changes the listening port management to handle multiple
incoming UDP ports. The earlier per-namespace structure is now a hash
list per namespace.

It is also now possible to define the same virtual network id
but with different UDP port values which can be useful for migration.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: korina: initialize variables directly
Emilio López [Fri, 17 May 2013 10:42:56 +0000 (10:42 +0000)]
net: ethernet: korina: initialize variables directly

Clean up the code a bit to initialize the variables directly when
defining them.

Signed-off-by: Emilio López <emilio@elopez.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: davicom: dm9000: initialize variables directly
Emilio López [Fri, 17 May 2013 10:42:55 +0000 (10:42 +0000)]
net: ethernet: davicom: dm9000: initialize variables directly

Clean up the code a bit to initialize the variables directly when
defining them.

Signed-off-by: Emilio López <emilio@elopez.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: apple: initialize variables directly
Emilio López [Fri, 17 May 2013 10:42:54 +0000 (10:42 +0000)]
net: ethernet: apple: initialize variables directly

Clean up the code a bit to initialize the variables directly when
defining them.

Signed-off-by: Emilio López <emilio@elopez.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: sun: initialize variables directly
Emilio López [Fri, 17 May 2013 10:42:53 +0000 (10:42 +0000)]
net: ethernet: sun: initialize variables directly

Clean up the code a bit to initialize the variables directly when
defining them.

Signed-off-by: Emilio López <emilio@elopez.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-davem' of git://gitorious.org/linux-can/linux-can-next
David S. Miller [Fri, 17 May 2013 00:07:46 +0000 (17:07 -0700)]
Merge branch 'for-davem' of git://gitorious.org/linux-can/linux-can-next

Marc Kleine-Budde says:

====================
this is a pull-request for net-next/master. It consists of 4 patches by
Jingoo Han, which remove the unnecessary platform_set_drvdata() and a
patch by Laurent Navet converting the grcan driver to use
devm_ioremap_resource().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: 3com: 3c509: remove unnecessary code
govindarajulu.v [Thu, 16 May 2013 06:24:41 +0000 (06:24 +0000)]
net: 3com: 3c509: remove unnecessary code

This patch removes unnecessary #if 0 code from 3c509.c

Signed-off-by: govindarajulu.v <govindarajulu90@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/ethernet/renesas: don't check resource with devm_ioremap_resource
Wolfram Sang [Thu, 16 May 2013 01:15:41 +0000 (01:15 +0000)]
drivers/net/ethernet/renesas: don't check resource with devm_ioremap_resource

devm_ioremap_resource does sanity checks on the given resource. No need to
duplicate this in the driver.

Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: speedup tcp_fixup_rcvbuf()
Eric Dumazet [Wed, 15 May 2013 19:25:55 +0000 (19:25 +0000)]
tcp: speedup tcp_fixup_rcvbuf()

tcp_fixup_rcvbuf() contains a loop to estimate initial socket
rcv space needed for a given mss. With large MTU (like 64K on lo),
we can loop ~500 times and consume a lot of cpu cycles.

perf top of 200 concurrent netperf -t TCP_CRR

5.62%  netperf  [kernel.kallsyms]  [k] tcp_init_buffer_space
1.71%  netperf  [kernel.kallsyms]  [k] _raw_spin_lock
1.55%  netperf  [kernel.kallsyms]  [k] kmem_cache_free
1.51%  netperf  [kernel.kallsyms]  [k] tcp_transmit_skb
1.50%  netperf  [kernel.kallsyms]  [k] tcp_ack

Lets use a 100% factor, and remove the loop.

100% is needed anyway for tcp_adv_win_scale=1
default value, and is also the maximum factor.

Refs: commit b49960a05e32
      ("tcp: change tcp_adv_win_scale and tcp_rmem[2]")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: can: ti_hecc: remove unnecessary platform_set_drvdata()
Jingoo Han [Tue, 7 May 2013 04:17:38 +0000 (13:17 +0900)]
net: can: ti_hecc: remove unnecessary platform_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure, since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound).
Thus, it is not needed to manually clear the device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agonet: can: flexcan: remove unnecessary platform_set_drvdata()
Jingoo Han [Tue, 7 May 2013 04:15:49 +0000 (13:15 +0900)]
net: can: flexcan: remove unnecessary platform_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure, since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound).
Thus, it is not needed to manually clear the device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agonet: can: c_can: remove unnecessary platform_set_drvdata()
Jingoo Han [Tue, 7 May 2013 04:14:13 +0000 (13:14 +0900)]
net: can: c_can: remove unnecessary platform_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure, since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound).
Thus, it is not needed to manually clear the device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agonet: can: at91_can: remove unnecessary platform_set_drvdata()
Jingoo Han [Tue, 7 May 2013 04:11:29 +0000 (13:11 +0900)]
net: can: at91_can: remove unnecessary platform_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure, since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound).
Thus, it is not needed to manually clear the device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agodrivers: net: can: grcan: use devm_ioremap_resource()
Laurent Navet [Mon, 13 May 2013 15:27:51 +0000 (17:27 +0200)]
drivers: net: can: grcan: use devm_ioremap_resource()

Replace a call to deprecated devm_request_and_ioremap by devm_ioremap_resource.
dev_err() message is no more needed since it's already displayed in
devm_ioremap_resource().

Signed-off-by: Laurent Navet <laurent.navet@gmail.com>
Acked-by: Andreas Larsson <andreas@gaisler.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 13 May 2013 20:25:36 +0000 (13:25 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "Several small bug fixes all over:

   1) be2net driver uses wrong payload length when submitting MAC list
      get requests to the chip.  From Sathya Perla.

   2) Fix mwifiex memory leak on driver unload, from Amitkumar Karwar.

   3) Prevent random memory access in batman-adv, from Marek Lindner.

   4) batman-adv doesn't check for pskb_trim_rcsum() errors, also from
      Marek Lindner.

   5) Fix fec crashes on rapid link up/down, from Frank Li.

   6) Fix inner protocol grovelling in GSO, from Pravin B Shelar.

   7) Link event validation fix in qlcnic from Rajesh Borundia.

   8) Not all FEC chips can support checksum offload, fix from Shawn
      Guo.

   9) EXPORT_SYMBOL + inline doesn't make any sense, from Denis Efremov.

  10) Fix race in passthru mode during device removal in macvlan, from
      Jiri Pirko.

  11) Fix RCU hash table lookup socket state race in ipv6, leading to
      NULL pointer derefs, from Eric Dumazet.

  12) Add several missing HAS_DMA kconfig dependencies, from Geert
      Uyttterhoeven.

  13) Fix bogus PCI resource management in 3c59x driver, from Sergei
      Shtylyov.

  14) Fix info leak in ipv6 GRE tunnel driver, from Amerigo Wang.

  15) Fix device leak in ipv6 IPSEC policy layer, from Cong Wang.

  16) DMA mapping leak fix in qlge from Thadeu Lima de Souza Cascardo.

  17) Missing iounmap on probe failure in bna driver, from Wei Yongjun."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (40 commits)
  bna: add missing iounmap() on error in bnad_init()
  qlge: fix dma map leak when the last chunk is not allocated
  xfrm6: release dev before returning error
  ipv6,gre: do not leak info to user-space
  virtio_net: use default napi weight by default
  emac: Fix EMAC soft reset on 460EX/GT
  3c59x: fix PCI resource management
  caif: CAIF_VIRTIO should depend on HAS_DMA
  net/ethernet: MACB should depend on HAS_DMA
  net/ethernet: ARM_AT91_ETHER should depend on HAS_DMA
  net/wireless: ATH9K should depend on HAS_DMA
  net/ethernet: STMMAC_ETH should depend on HAS_DMA
  net/ethernet: NET_CALXEDA_XGMAC should depend on HAS_DMA
  ipv6: do not clear pinet6 field
  macvlan: fix passthru mode race between dev removal and rx path
  ipv4: ip_output: remove inline marking of EXPORT_SYMBOL functions
  net/mlx4: Strengthen VLAN tags/priorities enforcement in VST mode
  net/mlx4_core: Add missing report on VST and spoof-checking dev caps
  net: fec: enable hardware checksum only on imx6q-fec
  qlcnic: Fix validation of link event command.
  ...