pandora-kernel.git
9 years agocan: c_can: convert to use devm * api
Lad, Prabhakar [Fri, 20 Jun 2014 11:59:34 +0000 (12:59 +0100)]
can: c_can: convert to use devm * api

This patch uses devm_* APIs as they are device managed and make code simpler.

Signed-off-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
9 years agoMerge branch 'amd811e-cleanups'
David S. Miller [Tue, 15 Jul 2014 02:14:44 +0000 (19:14 -0700)]
Merge branch 'amd811e-cleanups'

Varka Bhadram says:

====================
This series cleanup for AMD8111E ethernet driver

v1: fix checkpatch warnings.
v2: added new line in debug messages
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: fix 'foo* bar'
Varka Bhadram [Mon, 14 Jul 2014 08:39:09 +0000 (14:09 +0530)]
ethernet: amd: fix 'foo* bar'

This patch fix the 'foo*' bar with 'foo *bar' and (foo*) with (foo *).

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: fix pci device ids
Varka Bhadram [Mon, 14 Jul 2014 08:39:08 +0000 (14:09 +0530)]
ethernet: amd: fix pci device ids

Normally any device ids will be above the corresponding device driver
structure. This patch moves the pci device ids and MODULE_DEVICE_TABLE()
above the pci driver structure.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: fix comment styles
Varka Bhadram [Mon, 14 Jul 2014 08:39:07 +0000 (14:09 +0530)]
ethernet: amd: fix comment styles

This patch fixes the comment style issues

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: dynamic debug fixes
Varka Bhadram [Mon, 14 Jul 2014 08:39:06 +0000 (14:09 +0530)]
ethernet: amd: dynamic debug fixes

This patch convert printk() to netdev_dbg/info/err or dev_info/err/dbg

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: use devm_ioremap()
Varka Bhadram [Mon, 14 Jul 2014 08:39:05 +0000 (14:09 +0530)]
ethernet: amd: use devm_ioremap()

This patch replace ioremap() with the devm_ioremap() so that
the resource will be freed automatically with the probe failed.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: amd: move amd111e_remove_one after probe
Varka Bhadram [Mon, 14 Jul 2014 08:39:04 +0000 (14:09 +0530)]
ethernet: amd: move amd111e_remove_one after probe

This patch moves the remove functionalities after the probe
so that we can see the registered and released resources properly.
Every driver follows the same concept.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'sfc-next'
David S. Miller [Tue, 15 Jul 2014 02:11:59 +0000 (19:11 -0700)]
Merge branch 'sfc-next'

Shradha Shah says:

====================
sfc: Add 40G support

This patch series adds support for Solarflare 7000 series
40G Solarflare network adapters starting with the SFN7X42Q.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosfc: Add 40G link capability decoding
Edward Cree [Mon, 14 Jul 2014 07:39:07 +0000 (08:39 +0100)]
sfc: Add 40G link capability decoding

Needed to select 40G mode on a 10G/40G capable card.

Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosfc: Adding PCI ID for Solarflare 7000 series 40G network adapter.
Mateusz Wrzesinski [Mon, 14 Jul 2014 07:38:49 +0000 (08:38 +0100)]
sfc: Adding PCI ID for Solarflare 7000 series 40G network adapter.

Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'udp_tunnel'
David S. Miller [Mon, 14 Jul 2014 23:12:24 +0000 (16:12 -0700)]
Merge branch 'udp_tunnel'

Tom Herbert says:

====================
udp: UDP tunnel enhancements

- Add udp_sock_create in new helper module udp_tunnel. Tunnel
  implementations call this function to create listener UDP ports.
- Make vxlan and l2tp call udp_sock_create.
- Move udp_tunnel_segment into udp_offload.c.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoudp: Move udp_tunnel_segment into udp_offload.c
Tom Herbert [Mon, 14 Jul 2014 02:49:56 +0000 (19:49 -0700)]
udp: Move udp_tunnel_segment into udp_offload.c

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agol2tp: Call udp_sock_create
Tom Herbert [Mon, 14 Jul 2014 02:49:48 +0000 (19:49 -0700)]
l2tp: Call udp_sock_create

In l2tp driver call common function udp_sock_create to create the
listener UDP port.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovxlan: Call udp_sock_create
Tom Herbert [Mon, 14 Jul 2014 02:49:42 +0000 (19:49 -0700)]
vxlan: Call udp_sock_create

In vxlan driver call common function udp_sock_create to create the
listener UDP port.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoudp: Add udp_sock_create for UDP tunnels to open listener socket
Tom Herbert [Mon, 14 Jul 2014 02:49:37 +0000 (19:49 -0700)]
udp: Add udp_sock_create for UDP tunnels to open listener socket

Added udp_tunnel.c which can contain some common functions for UDP
tunnels. The first function in this is udp_sock_create which is used
to open the listener port for a UDP tunnel.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'unnecessary_break'
David S. Miller [Mon, 14 Jul 2014 06:45:55 +0000 (23:45 -0700)]
Merge branch 'unnecessary_break'

Fabian Frederick says:

====================
drivers/net: remove unnecessary break after goto

Small patchset addressing break redundancy on drivers/net branch
(suggested by Joe Perches).

V2: cc to maintainers of each section.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoslip: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:48 +0000 (20:09 +0200)]
slip: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agortlwifi: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:47 +0000 (20:09 +0200)]
rtlwifi: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agowcn36xx: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:46 +0000 (20:09 +0200)]
wcn36xx: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoorinoco_usb: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:45 +0000 (20:09 +0200)]
orinoco_usb: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agops3_gelic: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:44 +0000 (20:09 +0200)]
ps3_gelic: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40evf: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:43 +0000 (20:09 +0200)]
i40evf: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40e: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:42 +0000 (20:09 +0200)]
i40e: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoixgbe: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:41 +0000 (20:09 +0200)]
ixgbe: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoigb: remove unnecessary break after goto
Fabian Frederick [Sat, 12 Jul 2014 18:09:40 +0000 (20:09 +0200)]
igb: remove unnecessary break after goto

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: filter: sk_chk_filter() no longer mangles filter
Eric Dumazet [Sat, 12 Jul 2014 13:49:16 +0000 (15:49 +0200)]
net: filter: sk_chk_filter() no longer mangles filter

Add const attribute to filter argument to make clear it is no
longer modified.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Mon, 14 Jul 2014 05:42:17 +0000 (22:42 -0700)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
Please pull this batch of updates intended for the 3.17 stream...

This is primarily a Bluetooth pull.  Gustavo says:

"A lot of patches to 3.17. The bulk of changes here are for LE support.
The 6loWPAN over Bluetooth now has it own module, we also have support for
background auto-connection and passive scanning, Bluetooth device address
provisioning, support for reading Bluetooth clock values and LE connection
parameters plus many many fixes."

The balance is just a pull of the wireless.git tree, to avoid some
pending merge problems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: Use BUG_ON
Himangi Saraogi [Fri, 11 Jul 2014 20:27:17 +0000 (01:57 +0530)]
ipv6: Use BUG_ON

The semantic patch that makes this transformation is as follows:

// <smpl>
@@ expression e; @@
-if (e) BUG();
+BUG_ON(e);
// </smpl>

Signed-off-by: Himangi Saraogi <himangi774@gmail.com>
Acked-by: Julia Lawall <julia.lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: ipv6: Use BUG_ON
Himangi Saraogi [Fri, 11 Jul 2014 20:25:38 +0000 (01:55 +0530)]
net: ipv6: Use BUG_ON

The semantic patch that makes the transformation is as follows:

// <smpl>
@@ expression e; @@
-if (e) BUG();
+BUG_ON(e);
// </smpl>

Signed-off-by: Himangi Saraogi <himangi774@gmail.com>
Acked-by: Julia Lawall <julia.lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: addrconf: implement address generation modes
Jiri Pirko [Fri, 11 Jul 2014 19:10:18 +0000 (21:10 +0200)]
ipv6: addrconf: implement address generation modes

This patch introduces a possibility for userspace to set various (so far
two) modes of generating addresses. This is useful for example for
NetworkManager because it can set the mode to NONE and take care of link
local addresses itself. That allow it to have the interface up,
monitoring carrier but still don't have any addresses on it.

One more use-case by Dan Williams:
<quote>
WWAN devices often have their LL address provided by the firmware of the
device, which sometimes refuses to respond to incorrect LL addresses
when doing DHCPv6 or IPv6 ND.  The kernel cannot generate the correct LL
address for two reasons:

1) WWAN pseudo-ethernet interfaces often construct a fake MAC address,
or read a meaningless MAC address from the firmware.  Thus the EUI64 and
the IPv6LL address the kernel assigns will be wrong.  The real LL
address is often retrieved from the firmware with AT or proprietary
commands.

2) WWAN PPP interfaces receive their LL address from IPV6CP, not from
kernel assignments.  Only after IPV6CP has completed do we know the LL
address of the PPP interface and its peer.  But the kernel has already
assigned an incorrect LL address to the interface.

So being able to suppress the kernel LL address generation and assign
the one retrieved from the firmware is less complicated and more robust.
</quote>

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/hsr: Remove left-over never-true conditional code.
Arvid Brodin [Fri, 11 Jul 2014 16:21:12 +0000 (18:21 +0200)]
net/hsr: Remove left-over never-true conditional code.

MacAddressB is an array (unsigned char MacAddressB[ETH_ALEN]) and is allocated
as a part of *node_dst (which is a struct hsr_node). So the condition is always
false.

Detected by Dan Carpenter.

Signed-off-by: Arvid Brodin <arvid.brodin@alten.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'r8169_tso_ipv6'
David S. Miller [Fri, 11 Jul 2014 21:33:31 +0000 (14:33 -0700)]
Merge branch 'r8169_tso_ipv6'

Hayes Wang says:

====================
r8169: support IPv6

The RTL8168C and the later chips support the hardware checksum
for IPv6. Adjust some code and add the relative code.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agor8169: support IPv6
hayeswang [Fri, 11 Jul 2014 08:25:58 +0000 (16:25 +0800)]
r8169: support IPv6

Support the IPv6 hw checksum for RTL8111C and later chips. Note
that the hw has the limitation for the transport offset. The
checksum must be calculated by sw, when the transport offset is
out of the range which the hw accepts.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agor8169: use Giant Send
hayeswang [Fri, 11 Jul 2014 08:25:57 +0000 (16:25 +0800)]
r8169: use Giant Send

Replace large send with giant send for TSO for RTL8111C and later ICs.
The large send setting of the RTL8111DP is different from the other
chips. However, the giant send setting is the same for all the chips
which support it. Use the giant send to synchronize the settings.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agor8169: split rtl8169_tso_csum
hayeswang [Fri, 11 Jul 2014 08:25:56 +0000 (16:25 +0800)]
r8169: split rtl8169_tso_csum

According to the txd_version, split rtl8169_tso_csum() into
rtl8169_tso_csum_v1() and rtl8169_tso_csum_v2().

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv4: remove the unnecessary variable in udp_mcast_next
Li RongQing [Fri, 11 Jul 2014 06:32:17 +0000 (14:32 +0800)]
ipv4: remove the unnecessary variable in udp_mcast_next

Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodp83640: Adjust ptp event timestamps
Stefan Sørensen [Fri, 11 Jul 2014 06:18:26 +0000 (08:18 +0200)]
dp83640: Adjust ptp event timestamps

  Event timestamp values should be adjusted by 3*reference clock period +
  11 ns = 35 ns to compensate for input path and synchronization delays.

So subtract 35ns from event timestamps.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Fri, 11 Jul 2014 19:52:57 +0000 (15:52 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

9 years agoMerge branch 'mvebu'
David S. Miller [Fri, 11 Jul 2014 00:19:38 +0000 (17:19 -0700)]
Merge branch 'mvebu'

Ezequiel Garcia says:

====================
Network driver for Armada 375 SoC

This is the fourth round of the Armada 375 network support patchset. I've
tried to address all the feedback provided for last version and I hope the
driver looks better now.

If there's nothing else to fix, we'd like to merge this for v3.17. The first
patch should go through the network tree, and the other patches through
the mvebu tree.

Thanks a lot for all the great review, and feel free to comment some more!

Changes from v3:

  * Further optimization of the MTU, MAC and ring parameter change to make
    it smoothier.

  * Lots of cleanups in the parser configuration code, most of them addressing
    the feedback from Francois. This include fixing: missing curly braces,
    excessive parenthesis, excessive scope, and making several functions
    more readable.

  * Removed the Rx/Tx queue number module parameter. There's no reason to
    use any other than the default hardware-defined value.

Changes from v2:

  * Reworked mvpp2_prs_tcam_first_free() as suggested by Joe and Francois,
    to have a single loop instead of two.

  * Replaced mvpp2_cpu_interrupts_enable/disable(pp, cpu) with one function
    that enables/disable interrupts on all the CPUs at once.

  * Factor out Tx descriptor DMA unmap + descriptor put sequence to have
    more readable code, as suggested by Francois.

  * Remove redundant netif_running() checks in the ingress and egress path,
    as suggested by Francois.

  * Reworked ring parameter, MTU and MAC address setting to produce a
    more gentle modification of the parameter, and have a fallback in the
    event of a failure.

  * Fixed a percpu memory leak on error path, also noted by Francois.

  * Removed the usage of the legacy net_device irq field, requested by
    Francois.

  * Removed the unneeded multiple Tx port support. It was hardcoded to a single
    Tx port in the previous version so we decided to drop it and simplify the
    code.

  * Optimize the on_each_cpu() calls to clear the sent counters and the
    TX_DONE pkts coalescing setting. on_each_cpu is expensive so it's better
    to minize the calls to it.

Changes from v1:

  * Marcin Wojtas is the author of the driver, so I fixed authorship
    for patch 1/3:
    "ethernet: Add new driver for Marvell Armada 375 network unit"

This patchset adds a new network driver to support the network controller
in Armada 375 SoC.

The network interfaces share a common hardware unit called Packet Processor,
which contains a common register space and per-port register spaces.

The new network unit has different RXQ and TXQ management. The ports
associate so-called per-port "logical queues" which are mapped to "physical
queues". The latter are shared among the ports.

Fo the egress part, the mapping for each port is predefined by hardware.
The egress path incorporates so-called aggregation queues (one per CPU),
from where the data is passed to the physical queues and then via prefetch
buffer to the TxDMA.

The ingress path has a Parser and Classifier (PnC) and a Buffer Manager (BM)
whose usage is obligatory. We are only implementing a simple configuration
for the Parser and Classifier, yet the code is considerably large.

This network unit has other optional features like xPON, WoL, Hardware
Forwarding, and more. This initial commit doesn't provide support for these.

The mvpp2 network driver has been written by Marcin Wojtas and then reviewed
and cleaned up by Ezequiel Garcia.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoethernet: Add new driver for Marvell Armada 375 network unit
Marcin Wojtas [Thu, 10 Jul 2014 19:52:13 +0000 (16:52 -0300)]
ethernet: Add new driver for Marvell Armada 375 network unit

This commit adds a new network driver for the network controller in Marvell
Armada 375 SoC.

Given the controller is very different from the ones in the other Marvell
SoCs that use the mv643xx_eth (Kirkwood, Orion, Discovery) and mvneta
(Armada 370/38x/XP) drivers, a new driver is needed.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>
[Ezequiel: coding style cleanup]
Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: fix the check when handle RA
Li RongQing [Thu, 10 Jul 2014 10:02:46 +0000 (18:02 +0800)]
ipv6: fix the check when handle RA

d9333196572(ipv6:  Allow accepting RA from local IP addresses.) made the wrong
check, whether or not to accept RA with source-addr found on local machine, when
accept_ra_from_local is 0.

Fixes: d9333196572(ipv6:  Allow accepting RA from local IP addresses.)
Cc: Ben Greear <greearb@candelatech.com>
Cc: Hannes Frederic Sowa <hannes@redhat.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix in debug messages
Varka Bhadram [Thu, 10 Jul 2014 09:59:38 +0000 (15:29 +0530)]
net: cpmac: fix in debug messages

This patch fix the debug message format. This patch changes to the
commit f160a2d0b524eeebd97a68e2fbb59fad4cdd3fee: net: cpmac: dynamic debug fixes

When we use pr_debug()/netdev_dbg() new lines are inserting in b/w
the values. The format when i use the printk()

These formats used in skb dump and reg dump. This functions
called from the entire code. So this will be enabled all the lines.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
John W. Linville [Thu, 10 Jul 2014 21:00:34 +0000 (17:00 -0400)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
John W. Linville [Thu, 10 Jul 2014 21:00:24 +0000 (17:00 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless

9 years agoMerge branch 'bridge-fdb-dump-filter'
David S. Miller [Thu, 10 Jul 2014 19:38:07 +0000 (12:38 -0700)]
Merge branch 'bridge-fdb-dump-filter'

Jamal Hadi Salim says:

====================
bridge: fdb dumping takes a filter device

v7:
Vxlan driver was not updated with new API. Found by DaveM

v6:
Missed checkpatch > 80 chars lines found by Varka Bhadram

v5:
 Embarassing qlnic compile failure found by DaveM

v4:
 Request from DaveM to use proper comment tagging and remove if-stmnt braces

V3:
 Suggestion from Eric D. to use for_each_netdev
 Suggestion from Stephen H. to reduce level of indentation

V2:
Suggestions from Vlad
 Get rid of rcu read lock since rtnl_lock is being held
 simplify for readability
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: netlink dump interface at par with brctl
Jamal Hadi Salim [Thu, 10 Jul 2014 11:01:59 +0000 (07:01 -0400)]
bridge: netlink dump interface at par with brctl

Actually better than brctl showmacs because we can filter by bridge
port in the kernel.
The current bridge netlink interface doesnt scale when you have many
bridges each with large fdbs or even bridges with many bridge ports

And now for the science non-fiction novel you have all been
waiting for..

//lets see what bridge ports we have
root@moja-1:/configs/may30-iprt/bridge# ./bridge link show
8: eth1 state DOWN : <BROADCAST,MULTICAST> mtu 1500 master br0 state
disabled priority 32 cost 19
17: sw1-p1 state DOWN : <BROADCAST,NOARP> mtu 1500 master br0 state
disabled priority 32 cost 100

// show all..
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show
33:33:00:00:00:01 dev bond0 self permanent
33:33:00:00:00:01 dev dummy0 self permanent
33:33:00:00:00:01 dev ifb0 self permanent
33:33:00:00:00:01 dev ifb1 self permanent
33:33:00:00:00:01 dev eth0 self permanent
01:00:5e:00:00:01 dev eth0 self permanent
33:33:ff:22:01:01 dev eth0 self permanent
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
33:33:00:00:00:01 dev gretap0 self permanent
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

//filter by bridge
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

// bridge sw1 has no ports attached..
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br sw1

//filter by port
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show brport eth1
02:00:00:12:01:02 vlan 0 master br0 permanent
00:17:42:8a:b4:05 vlan 0 master br0 permanent
00:17:42:8a:b4:07 self permanent
33:33:00:00:00:01 self permanent

// filter by port + bridge
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0 brport
sw1-p1
da:ac:46:27:d9:53 vlan 0 master br0 permanent
33:33:00:00:00:01 self permanent

// for shits and giggles (as they say in New Brunswick), lets
// change the mac that br0 uses
// Note: a magical fdb entry with no brport is added ...
root@moja-1:/configs/may30-iprt/bridge# ip link set dev br0 address
02:00:00:12:01:04

// lets see if we can see the unicorn ..
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show
33:33:00:00:00:01 dev bond0 self permanent
33:33:00:00:00:01 dev dummy0 self permanent
33:33:00:00:00:01 dev ifb0 self permanent
33:33:00:00:00:01 dev ifb1 self permanent
33:33:00:00:00:01 dev eth0 self permanent
01:00:5e:00:00:01 dev eth0 self permanent
33:33:ff:22:01:01 dev eth0 self permanent
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
33:33:00:00:00:01 dev gretap0 self permanent
02:00:00:12:01:04 dev br0 vlan 0 master br0 permanent <=== there it is
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

//can we see it if we filter by bridge?
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
02:00:00:12:01:04 dev br0 vlan 0 master br0 permanent <=== there it is
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: fdb dumping takes a filter device
Jamal Hadi Salim [Thu, 10 Jul 2014 11:01:58 +0000 (07:01 -0400)]
bridge: fdb dumping takes a filter device

Dumping a bridge fdb dumps every fdb entry
held. With this change we are going to filter
on selected bridge port.

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agor8152: support jumbo frame for RTL8153
hayeswang [Thu, 10 Jul 2014 02:58:54 +0000 (10:58 +0800)]
r8152: support jumbo frame for RTL8153

The maximum jumbo frame size for RTL8153 is 9K bytes.
Change the max rx packet size to 9K.
Change the use of the shared fifo from 6K (default) to 12K for tx.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'cpmac-next'
David S. Miller [Thu, 10 Jul 2014 07:41:06 +0000 (00:41 -0700)]
Merge branch 'cpmac-next'

Varka Bhadram says:

====================
This patch series cleanup for AR7 CPMAC Ethernet controller driver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix in releasing resources
Varka Bhadram [Thu, 10 Jul 2014 05:35:45 +0000 (11:05 +0530)]
net: cpmac: fix in releasing resources

before registering the the net device this code freeing net device
by using the label 'fail'

fixed by introducing an another label 'out'

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix proper spacing before return statement
Varka Bhadram [Thu, 10 Jul 2014 05:35:44 +0000 (11:05 +0530)]
net: cpmac: fix proper spacing before return statement

This patch insert proper spaces before return statement.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix missing a blank line after declarations
Varka Bhadram [Thu, 10 Jul 2014 05:35:43 +0000 (11:05 +0530)]
net: cpmac: fix missing a blank line after declarations

This patch insert a blank line after declaration

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix cpmac driver structure
Varka Bhadram [Thu, 10 Jul 2014 05:35:42 +0000 (11:05 +0530)]
net: cpmac: fix cpmac driver structure

This patch changes to style of declarattion which follows every driver

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: dynamic debug fixes
Varka Bhadram [Thu, 10 Jul 2014 05:35:41 +0000 (11:05 +0530)]
net: cpmac: dynamic debug fixes

This patch does the following changes
1. convert printk(KERN_DEBUG.. to netdev_dbg() if we have net_device object
   or convert to dev_dbg() if we have device object.
2. convert printk(KERN_WARNING.. to netdev_warn() if we have net_device object
   or convert to dev_warn() if we have device object
3. convert printk() to pr_*

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: fix comments
Varka Bhadram [Thu, 10 Jul 2014 05:35:40 +0000 (11:05 +0530)]
net: cpmac: fix comments

This patch convert the normal comments to networking subsystem
style comments.

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: cpmac: remove space in macro defination
Varka Bhadram [Thu, 10 Jul 2014 05:35:39 +0000 (11:05 +0530)]
net: cpmac: remove space in macro defination

This patch fix the space after '#' in macro defination

Signed-off-by: Varka Bhadram <varkab@cdac.in>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipconfig: Only bootp paths should reference ic_dev_xid.
David S. Miller [Thu, 10 Jul 2014 05:25:18 +0000 (22:25 -0700)]
ipconfig: Only bootp paths should reference ic_dev_xid.

It is only tested, and declared, in the bootp code.

So, in ic_dynamic() guard it's setting with IPCONFIG_BOOTP.

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'systemport-next'
David S. Miller [Thu, 10 Jul 2014 01:20:23 +0000 (18:20 -0700)]
Merge branch 'systemport-next'

Florian Fainelli says:

====================
net: systemport: checkpatch fixes

These two patches fixes the multiple line alignment issues and
kzalloc on array reported by checkpatch.pl
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: systemport: use kcalloc instead of kzalloc
Florian Fainelli [Thu, 10 Jul 2014 00:36:47 +0000 (17:36 -0700)]
net: systemport: use kcalloc instead of kzalloc

checkpatch.pl flagged two uses of kzalloc() for allocating and zeroing
arrays, use kcalloc() instead as recommended.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: systemport: align multiple lines correctly
Florian Fainelli [Thu, 10 Jul 2014 00:36:46 +0000 (17:36 -0700)]
net: systemport: align multiple lines correctly

checkpatch.pl flagged a bunch of: "CHECK: Alignment should match open
parenthesis" problems, fix all of them.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agohyperv: Add netpoll support
Richard Weinberger [Wed, 9 Jul 2014 14:23:59 +0000 (16:23 +0200)]
hyperv: Add netpoll support

In order to have at least a netconsole to debug kernel issues on
Windows Azure this patch implements netpoll support.
Sending packets is easy, netvsc_start_xmit() does already everything
needed.

Signed-off-by: Richard Weinberger <richard@nod.at>
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoarc_emac: Remove unused pointer to net_device from arc_emac_priv
Tobias Klauser [Wed, 9 Jul 2014 09:07:56 +0000 (11:07 +0200)]
arc_emac: Remove unused pointer to net_device from arc_emac_priv

The pointer to the struct net_device in the private data is only
assigned but never used, so delete it.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoarc_emac: Use net_device_stats from struct net_device
Tobias Klauser [Wed, 9 Jul 2014 09:07:37 +0000 (11:07 +0200)]
arc_emac: Use net_device_stats from struct net_device

Instead of using a private copy of struct net_device_stats in struct
arc_emac_priv, use stats from struct net_device.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipconfig: move ic_dev_xid under IPCONFIG_BOOTP
Fabian Frederick [Wed, 9 Jul 2014 18:35:21 +0000 (20:35 +0200)]
ipconfig: move ic_dev_xid under IPCONFIG_BOOTP

ic_dev_xid is only used in __init ic_bootp_recv under IPCONFIG_BOOTP
and __init ic_dynamic under IPCONFIG_DYNAMIC(which is itself defined
with the same IPCONFIG_BOOTP)

This patch fixes the following warning when IPCONFIG_BOOTP is not set:
>> net/ipv4/ipconfig.c:146:15: warning: 'ic_dev_xid' defined but not used [-Wunused-variable]
    static __be32 ic_dev_xid;  /* Device under configuration */

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: netdev@vger.kernel.org
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoBluetooth: Don't send ERTM configuration option when disabled
Marcel Holtmann [Wed, 9 Jul 2014 09:53:35 +0000 (11:53 +0200)]
Bluetooth: Don't send ERTM configuration option when disabled

When ERTM support is disabled, then do not even send ERTM configuration
option even if the remote side supports it.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
9 years agoBluetooth: Fix setting HCI_CONNECTABLE from ioctl code
Johan Hedberg [Wed, 9 Jul 2014 10:28:26 +0000 (13:28 +0300)]
Bluetooth: Fix setting HCI_CONNECTABLE from ioctl code

When the white list is in use the code would not update the
HCI_CONNECTABLE flag if it gets changed through the ioctl code (e.g.
hciconfig hci0 pscan). Since the flag is important for properly
accepting incoming connections add code to fix it up if necessary and
emit a New Settings mgmt event.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Add white list lookup for incoming connection requests
Johan Hedberg [Wed, 9 Jul 2014 09:59:19 +0000 (12:59 +0300)]
Bluetooth: Add white list lookup for incoming connection requests

This patch adds support for looking up entries in the white list when
HCI_CONNECTABLE is not set. The logic is fairly simple: if we're
connectable check the black list, if we're not connectable check the
white list.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Enable page scan also if there are white list entries
Johan Hedberg [Wed, 9 Jul 2014 09:59:18 +0000 (12:59 +0300)]
Bluetooth: Enable page scan also if there are white list entries

Page scan should be enabled either if the connectable setting is set or
if there are any entries in the BR/EDR white list. This patch implements
such behavior by updating the two places that were making decisions on
whether to enable page scan or not.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Refactor connection request handling
Johan Hedberg [Wed, 9 Jul 2014 09:59:17 +0000 (12:59 +0300)]
Bluetooth: Refactor connection request handling

The conditions for accepting an incoming connections are already
non-trivial and will become more so once a white list is added. This
patch breaks up the checks for when to reject the request by creating a
helper function for it.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Fix incorrectly setting HCI_CONNECTABLE
Johan Hedberg [Wed, 9 Jul 2014 09:59:16 +0000 (12:59 +0300)]
Bluetooth: Fix incorrectly setting HCI_CONNECTABLE

Since page scan might be enabled by Add Device we should not implicitly
set connectable whenever something else than Set Connectable changes it.
This patch makes sure that we don't set HCI_CONNECTABLE for these cases
if there are any entries in the white list.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Update page scan when necessary for Add/Remove Device
Johan Hedberg [Wed, 9 Jul 2014 09:59:15 +0000 (12:59 +0300)]
Bluetooth: Update page scan when necessary for Add/Remove Device

When we're removing the last item in the white list or adding the first
one to it and HCI_CONNECTABLE is not set we need to update the current
page scan. This patch adds a simple helper function for the purpose and
calls it from the respective mgmt command handlers.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Introduce a whitelist for BR/EDR devices
Johan Hedberg [Wed, 9 Jul 2014 09:59:14 +0000 (12:59 +0300)]
Bluetooth: Introduce a whitelist for BR/EDR devices

This patch extends the Add/Remove device commands by letting user space
pass BR/EDR addresses to them. The resulting entries get stored in a new
hdev->whitelist list. The idea is that we can now selectively accept
connections from devices in the list even though HCI_CONNECTABLE is not
set (the actual implementation of this is coming in a subsequent patch).

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Unify helpers for bdaddr_list manipulations
Johan Hedberg [Wed, 9 Jul 2014 09:59:13 +0000 (12:59 +0300)]
Bluetooth: Unify helpers for bdaddr_list manipulations

We already have several lists with struct bdaddr_list entries, and there
will be more in the future. Since the operations for adding, removing,
looking up and clearing entries in these lists are exactly the same it
doesn't make sense to define new functions for every single list. This
patch unifies the functions by passing the list_head to them instead of
a hci_dev pointer.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
9 years agoBluetooth: Fix enabling Authenticated Payload Timeout Expired event
Marcel Holtmann [Wed, 9 Jul 2014 07:49:05 +0000 (09:49 +0200)]
Bluetooth: Fix enabling Authenticated Payload Timeout Expired event

The Authenticated Payload Timeout Expired event is valid for
controllers with BR/EDR Secure Connections support, but also for
LE only controllers supporting LE Ping feature. When either of them
is available enable this event. Previous it was not enabled when
the controller was only supporting LE operation.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
9 years agonet: phy: export phy_suspend and phy_resume
Florian Fainelli [Tue, 8 Jul 2014 17:38:36 +0000 (10:38 -0700)]
net: phy: export phy_suspend and phy_resume

phy_suspend and phy_resume are two commonly used helper functions that
need to be exported for Ethernet drivers to be built as modules

Fixes: 40755a0fce17 ("net: systemport: add suspend and resume support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'fec-next'
David S. Miller [Wed, 9 Jul 2014 03:57:51 +0000 (20:57 -0700)]
Merge branch 'fec-next'

Russell King says:

====================
Freescale ethernet driver updates (part 3)

Here's the third batch of patches for the Freescale FEC ethernet driver,
based upon the previous set of patches.  This concludes the changes I
currently have prepared and have been reviewed for the next merge window
at this time.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: consolidate hwtstamp implementation
Russell King [Tue, 8 Jul 2014 12:02:09 +0000 (13:02 +0100)]
net: fec: consolidate hwtstamp implementation

Both transmit and receive use the same infrastructure for calculating
the packet timestamp.  Rather than duplicating the code, provide a
function to do this common work.  Model this function in the Intel
e1000e version which avoids calling ns_to_ktime() within the spinlock;
the spinlock is critical for timecounter_cyc2time() but not
ns_to_ktime().

Acked-by: Richard Cochran <richardcochran@gmail.com>
Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: remove useless status check in tx reap path
Russell King [Tue, 8 Jul 2014 12:02:04 +0000 (13:02 +0100)]
net: fec: remove useless status check in tx reap path

Remove a useless status check in the transmit reap path - we have
already checked that the BD_ENET_TX_READY bit is clear, and as the
hardware only ever clears this bit, there is no way this test can ever
be true.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: add support for dumping transmit ring on timeout
Russell King [Tue, 8 Jul 2014 12:01:59 +0000 (13:01 +0100)]
net: fec: add support for dumping transmit ring on timeout

When we timeout on transmit, it would be useful to dump the transmit
ring, so we can see the ring state.  This can be helpful to diagnose
the cause of transmit timeouts.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: reorder ethtool ops to match order in struct declaration
Russell King [Tue, 8 Jul 2014 12:01:54 +0000 (13:01 +0100)]
net: fec: reorder ethtool ops to match order in struct declaration

This allows us to merge two separate preprocessor conditionals together.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: clear receive interrupts before processing a packet
Russell King [Tue, 8 Jul 2014 12:01:49 +0000 (13:01 +0100)]
net: fec: clear receive interrupts before processing a packet

Clear any pending receive interrupt before we process a pending packet.
This helps to avoid any spurious interrupts being raised after we have
fully cleaned the receive ring, while still allowing an interrupt to be
raised if we receive another packet.

The position of this is critical: we must do this prior to reading the
next packet status to avoid potentially dropping an interrupt when a
packet is still pending.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: replace delayed work with standard work
Russell King [Tue, 8 Jul 2014 12:01:44 +0000 (13:01 +0100)]
net: fec: replace delayed work with standard work

As of "better implementation of iMX6 ERR006358 quirk", we no longer have
a requirement for a delayed work.  Moreover, the work is now only used
for timeout purposes, so the timeout flag is also pointless - we set it
each time we queue the work, and the work clears it.

Replace the fec_enet_delayed_work struct with a standard work_struct,
resulting in simplified timeout handling code.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: better implementation of iMX6 ERR006358 quirk
Russell King [Tue, 8 Jul 2014 12:01:38 +0000 (13:01 +0100)]
net: fec: better implementation of iMX6 ERR006358 quirk

Using a (delayed) workqueue for ERR006358 is not correct - a work queue
is a single-trigger device.  Once the work queue has been scheduled, it
can't be re-scheduled until it has been run.  This can cause problems -
with an appropriate packet timing, we can end up with packets queued,
but not sent by the hardware, resulting in the transmit timeout firing.

Re-implement this as per the workaround detailed in the ERR006358
documentation - if there are packets waiting to be sent when we service
the transmit ring, and we see that the transmitter is not running,
kick the transmitter to run the pending entries in the ring.

Testing here with a 10Mbit half duplex link sees the resulting iperf
TCP bandwidth increase from between 1 to 2Mbps to between 8 to 9Mbps.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetpoll: fix use after free
david decotigny [Tue, 8 Jul 2014 22:14:41 +0000 (15:14 -0700)]
netpoll: fix use after free

After a bonding master reclaims the netpoll info struct, slaves could
still hold a pointer to the reclaimed data. This patch fixes it: as
soon as netpoll_async_cleanup is called for a slave (eg. when
un-enslaved), we make sure that this slave doesn't point to the data.

Signed-off-by: David Decotigny <decot@googlers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen-netback: Adding debugfs "io_ring_qX" files
Zoltan Kiss [Tue, 8 Jul 2014 18:49:14 +0000 (19:49 +0100)]
xen-netback: Adding debugfs "io_ring_qX" files

This patch adds debugfs capabilities to netback. There used to be a similar
patch floating around for classic kernel, but it used procfs. It is based on a
very similar blkback patch.
It creates xen-netback/[vifname]/io_ring_q[queueno] files, reading them output
various ring variables etc. Writing "kick" into it imitates an interrupt
happened, it can be useful to check whether the ring is just stalled due to a
missed interrupt.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: xen-devel@lists.xenproject.org
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: provide stubs for ip6_set_txhash and ip6_make_flowlabel
Florian Fainelli [Tue, 8 Jul 2014 18:15:03 +0000 (11:15 -0700)]
net: provide stubs for ip6_set_txhash and ip6_make_flowlabel

Commit cb1ce2ef387b ("ipv6: Implement automatic flow label generation
on transmit") introduced ip6_make_flowlabel, while commit b73c3d0e4f0e
("net: Save TX flow hash in sock and set in skbuf on xmit") introduced
ip6_set_txhash.

ip6_set_tx_hash() uses sk_v6_daddr which references
__sk_common.skc_v6_daddr from struct sock_common, which is gated with
IS_ENABLED(CONFIG_IPV6).

ip6_make_flowlabel() uses the ipv6 member from struct net which is
also gated with IS_ENABLED(CONFIG_IPV6).

When CONFIG_IPV6 is disabled, we will hit a build failure that looks
like this when the compiler attempts inlining these functions:

  CC [M]  drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.o
In file included from include/net/inet_sock.h:27:0,
                 from include/net/ip.h:30,
                 from drivers/net/ethernet/broadcom/cnic.c:37:
include/net/ipv6.h: In function 'ip6_set_txhash':
include/net/sock.h:327:33: error: 'struct sock_common' has no member named 'skc_v6_daddr'
 #define sk_v6_daddr  __sk_common.skc_v6_daddr
                                 ^
include/net/ipv6.h:696:49: note: in expansion of macro 'sk_v6_daddr'
  keys.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr);
                                                 ^
In file included from include/net/inetpeer.h:15:0,
                 from include/net/route.h:28,
                 from include/net/ip.h:31,
                 from drivers/net/ethernet/broadcom/cnic.c:37:
include/net/ipv6.h: In function 'ip6_make_flowlabel':
include/net/ipv6.h:706:37: error: 'struct net' has no member named 'ipv6'
  if (!flowlabel && (autolabel || net->ipv6.sysctl.auto_flowlabels)) {
                                     ^

Fixes: cb1ce2ef387b ("ipv6: Implement automatic flow label generation on transmit")
Fixes: b73c3d0e4f0e ("net: Save TX flow hash in sock and set in skbuf on xmit")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'fec-next'
David S. Miller [Wed, 9 Jul 2014 03:03:05 +0000 (20:03 -0700)]
Merge branch 'fec-next'

Russell King says:

====================
Freescale ethernet driver updates (part 2)

Here's the second batch of patches for the Freescale FEC ethernet driver,
based upon the previous set of patches.  One further set of 7 patches
remains.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: clean up duplex mode handling
Russell King [Tue, 8 Jul 2014 11:40:43 +0000 (12:40 +0100)]
net: fec: clean up duplex mode handling

Many places call fec_restart() with the second parameter being some kind
of previously saved duplex value, but only two places call it with some
other setting.  This is at odds with how the other link settings are
handled, and used to be racy before the rtnl locks were added to
fec_restart()'s various call paths.

Clean this up so all link capabilities are handled in the same way -
saved into the fec_enet_private structure, and then fec_restart() acts
on those settings.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: quiesce packet processing when taking link down in fec_enet_adjust_link()
Russell King [Tue, 8 Jul 2014 11:40:38 +0000 (12:40 +0100)]
net: fec: quiesce packet processing when taking link down in fec_enet_adjust_link()

When the link goes down, the adjust_link method will be called, but
there is no synchronisation to ensure that we won't be processing some
last remaining packets via the NAPI handlers while performing a reset of
the device.

Add the necessary synchronisation to ensure that packet processing
is complete before we stop and reset the FEC.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: quiesce packet processing before changing features
Russell King [Tue, 8 Jul 2014 11:40:33 +0000 (12:40 +0100)]
net: fec: quiesce packet processing before changing features

Changing the features (receive checksumming) requires the hardware to be
reprogrammed, and also changes the checks in the receive packet
processing.

The current implementation has a race - fec_set_features() changes the
flags which alter the receive packet processing while the adapter is
active, and potentially receiving frames.  Only after we've modified
the software flag do we shutdown and reconfigure the hardware.

This can lead to packets being received and marked with a valid checksum
(via CHECKSUM_UNNECESSARY) when the hardware checksum validation has not
yet been enabled.

We must quiesce the device, then change the software configuration for
this feature, and then resume the device if it was previously running.

The resulting code structure also allows us to add other configuration
features in this path without having to quiesce and resume the network
interface and device.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: quiesce packet processing before stopping device in fec_set_features()
Russell King [Tue, 8 Jul 2014 11:40:28 +0000 (12:40 +0100)]
net: fec: quiesce packet processing before stopping device in fec_set_features()

fec_set_features() calls fec_stop() to stop the transmit ring while the
transmit queue is still active.  This can lead to the transmit ring
being restarted by an intervening packet queued for transmission, or
by the tx quirk timer expiring.

Fix this by disabling NAPI (which ensures that the NAPI handlers are
not running), and then take the transmit lock while we stop and
restart the adapter (which prevents new packets being queued).

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: quiesce packet processing before stopping device in fec_suspend()
Russell King [Tue, 8 Jul 2014 11:40:23 +0000 (12:40 +0100)]
net: fec: quiesce packet processing before stopping device in fec_suspend()

fec_suspend() calls fec_stop() to stop the transmit ring while the
transmit packet processing is still active.  This can lead to the
transmit queue being restarted by an intervening packet queued for
transmission, or by the tx quirk timer expiring.

Fix this by disabling NAPI first, which will ensure that the NAPI
handlers are not running.  Then, take the transmit lock before
detaching the netif device.  This ensures that there are no races
with the transmit path - and also ensures that the watchdog won't
fire.

We can then safely stop the ethernet device itself, knowing that the
rest of the driver is safely shut down.

On resume, we bring the device back up in reverse order - we restart
the device, reattach the device (under the tx lock), and then enable
the NAPI handlers.

We also need to adjust the close function to cope with this new
sequence, so that it's possible to cleanly close down the driver
after the hardware fails to resume (eg, due to the regulator_enable()
or pinctrl calls in the resume path returning an error.)

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: remove inappropriate calls around fec_restart()
Russell King [Tue, 8 Jul 2014 11:40:18 +0000 (12:40 +0100)]
net: fec: remove inappropriate calls around fec_restart()

This is the second stage to "move calls to quiesce/resume packet
processing out of fec_restart()", where we remove calls which are not
appropriate to the call site.

In the majority of cases, there is no need to detach and reattach the
interface as we are holding the queue xmit lock across the reset.  The
exception to that is in fec_resume(), where we are already detached by
the suspend function.  Here, we can remove the call to detach the
interface.

We also do not need to stop the transmit queue.  Holding the xmit lock
is enough to ensure that the transmit packet processing is not running
while we perform our task.  However, since fec_restart() always cleans
the rings, we call netif_wake_queue() (or netif_device_attach() in the
case of resume) just before dropping the xmit lock.  This prevents the
watchdog firing.

Lastly, always call napi_enable() after the device has been reattached
in the resume path so that we know that the transmit packet processing
is already in an enabled state, so we don't call netif_wake_queue()
while detached.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: move calls to quiesce/resume packet processing out of fec_restart()
Russell King [Tue, 8 Jul 2014 11:40:12 +0000 (12:40 +0100)]
net: fec: move calls to quiesce/resume packet processing out of fec_restart()

Move the calls to quiesce and resume packet processing out of
fec_restart() to its call sites.  This is the first step in a two stage
clean up of this code, where we just move the calls out of fec_restart()
without changing them.  Not everywhere needs to issue these calls, and
not everywhere needs all of these calls to be issued.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: only restart or stop the device if it is present and running
Russell King [Tue, 8 Jul 2014 11:40:07 +0000 (12:40 +0100)]
net: fec: only restart or stop the device if it is present and running

Avoid calling fec_restart() or fec_stop() while the device is down
or not present (iow suspended.)

Although the ndo_timeout method will only be called if the device is
present and running, we defer this to a work queue.  The work queue
can run independently, and so needs to repeat these checks to ensure
that a restart doesn't occur after the device has been taken down or
detached for suspend.  In this case, we call fec_restart() in the
resume path, so nothing is lost.

For fec_set_features, we add a call to fec_restart() in fec_enet_open()
to ensure that the hardware is appropriate programmed when the interface
is opened.  fec_set_features() call should not occur while we're
suspended, so we don't have to worry about that case.

The adjust_link needs similar treatment - this also is called from a
work queue, which may be run independently after we have taken the
device down and detached it.  In this case, we just mark the link
down and take no further action.  We will reset things appropriately
once the device is up and running again, at which point we will receive
another adjust_link callback.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: ensure fec_enet_close() copes with resume failure
Russell King [Tue, 8 Jul 2014 11:40:02 +0000 (12:40 +0100)]
net: fec: ensure fec_enet_close() copes with resume failure

When the FEC is suspended, the device is detached.  Upon resume failure,
the device is left in detached mode, possibly with some of the required
clocks not running.  We don't want to be poking the device in that state
because as it may cause bus errors.

If the device is marked detached, avoid calling fec_stop().

This depends upon: "net:fec: improve safety of suspend/resume paths"

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: improve safety of suspend/resume/transmit timeout paths
Russell King [Tue, 8 Jul 2014 11:39:57 +0000 (12:39 +0100)]
net: fec: improve safety of suspend/resume/transmit timeout paths

We should hold the rtnl lock while suspending, resuming or processing
the transmit timeout to ensure that nothing will interfere while we
bring up, take down or restart the hardware.  The transmit timeout
could run if we're preempted during suspend.

Acked-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'mlx4-next'
David S. Miller [Wed, 9 Jul 2014 02:58:54 +0000 (19:58 -0700)]
Merge branch 'mlx4-next'

Amir Vadai says:

====================
Mellanox driver update Jul-08-2014

This patch set introduce some small bug fixes.
Most of the patches are small fixes to cornet case bugs.
The patch by Noa ("Fix mac_hash database inconsistency") was sent in the past
[1] and was droped because a fix to the bonding code was supposed to make it
unnecessary. After a second look on the patch, it is still needed even
after the direct access to dev_addr by the bonding will be fixed.

Patches were applied and tested over commit bd4578b
("drivers/net/hyperv/netvsc.c: remove unnecessary null test before kfree")

[1] - http://permalink.gmane.org/gmane.linux.network/315900
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/mlx4_en: Fix mac_hash database inconsistency
Noa Osherovich [Tue, 8 Jul 2014 08:25:24 +0000 (11:25 +0300)]
net/mlx4_en: Fix mac_hash database inconsistency

Using a local copy of dev_addr in mlx4_en_set_mac() to prevent dev_addr
from being modified during error flow or when dev_addr is modified in
another context (which is another problem that is being discussed over
the mailing list [1]).
Also fixing bad naming of priv->prev_mac into priv->current_mac.

[1] - http://patchwork.ozlabs.org/patch/351489/

Reviewed-by: Eyal Perry <eyalpe@mellanox.com>
Signed-off-by: Noa Osherovich <noaos@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>