platform/kernel/linux-rpi.git
6 years agoudp: avoid refcount_t saturation in __udp_gso_segment()
Eric Dumazet [Fri, 11 May 2018 02:07:13 +0000 (19:07 -0700)]
udp: avoid refcount_t saturation in __udp_gso_segment()

For some reason, Willem thought that the issue we fixed for TCP
in commit 7ec318feeed1 ("tcp: gso: avoid refcount_t warning from
tcp_gso_segment()") was not relevant for UDP GSO.

But syzbot found its way.

refcount_t: saturated; leaking memory.
WARNING: CPU: 0 PID: 10261 at lib/refcount.c:78 refcount_add_not_zero+0x2d4/0x320 lib/refcount.c:78
Kernel panic - not syncing: panic_on_warn set ...

CPU: 0 PID: 10261 Comm: syz-executor5 Not tainted 4.17.0-rc3+ #38
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1b9/0x294 lib/dump_stack.c:113
 panic+0x22f/0x4de kernel/panic.c:184
 __warn.cold.8+0x163/0x1b3 kernel/panic.c:536
 report_bug+0x252/0x2d0 lib/bug.c:186
 fixup_bug arch/x86/kernel/traps.c:178 [inline]
 do_error_trap+0x1de/0x490 arch/x86/kernel/traps.c:296
 do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:315
 invalid_op+0x14/0x20 arch/x86/entry/entry_64.S:992
RIP: 0010:refcount_add_not_zero+0x2d4/0x320 lib/refcount.c:78
RSP: 0018:ffff880196db6b90 EFLAGS: 00010282
RAX: 0000000000000026 RBX: 00000000ffffff01 RCX: ffffc900040d9000
RDX: 0000000000004a29 RSI: ffffffff8160f6f1 RDI: ffff880196db66f0
RBP: ffff880196db6c78 R08: ffff8801b33d6740 R09: 0000000000000002
R10: ffff8801b33d6740 R11: 0000000000000000 R12: 0000000000000000
R13: 00000000ffffffff R14: ffff880196db6c50 R15: 0000000000020101
 refcount_add+0x1b/0x70 lib/refcount.c:102
 __udp_gso_segment+0xaa5/0xee0 net/ipv4/udp_offload.c:272
 udp4_ufo_fragment+0x592/0x7a0 net/ipv4/udp_offload.c:301
 inet_gso_segment+0x639/0x12b0 net/ipv4/af_inet.c:1342
 skb_mac_gso_segment+0x3ad/0x720 net/core/dev.c:2792
 __skb_gso_segment+0x3bb/0x870 net/core/dev.c:2865
 skb_gso_segment include/linux/netdevice.h:4050 [inline]
 validate_xmit_skb+0x54d/0xd90 net/core/dev.c:3122
 __dev_queue_xmit+0xbf8/0x34c0 net/core/dev.c:3579
 dev_queue_xmit+0x17/0x20 net/core/dev.c:3620
 neigh_direct_output+0x15/0x20 net/core/neighbour.c:1401
 neigh_output include/net/neighbour.h:483 [inline]
 ip_finish_output2+0xa5f/0x1840 net/ipv4/ip_output.c:229
 ip_finish_output+0x828/0xf80 net/ipv4/ip_output.c:317
 NF_HOOK_COND include/linux/netfilter.h:277 [inline]
 ip_output+0x21b/0x850 net/ipv4/ip_output.c:405
 dst_output include/net/dst.h:444 [inline]
 ip_local_out+0xc5/0x1b0 net/ipv4/ip_output.c:124
 ip_send_skb+0x40/0xe0 net/ipv4/ip_output.c:1434
 udp_send_skb.isra.37+0x5eb/0x1000 net/ipv4/udp.c:825
 udp_push_pending_frames+0x5c/0xf0 net/ipv4/udp.c:853
 udp_v6_push_pending_frames+0x380/0x3e0 net/ipv6/udp.c:1105
 udp_lib_setsockopt+0x59a/0x600 net/ipv4/udp.c:2403
 udpv6_setsockopt+0x95/0xa0 net/ipv6/udp.c:1447
 sock_common_setsockopt+0x9a/0xe0 net/core/sock.c:3046
 __sys_setsockopt+0x1bd/0x390 net/socket.c:1903
 __do_sys_setsockopt net/socket.c:1914 [inline]
 __se_sys_setsockopt net/socket.c:1911 [inline]
 __x64_sys_setsockopt+0xbe/0x150 net/socket.c:1911
 do_syscall_64+0x1b1/0x800 arch/x86/entry/common.c:287
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

Fixes: ad405857b174 ("udp: better wmem accounting on gso")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: switch pacing timer to softirq based hrtimer
Eric Dumazet [Thu, 10 May 2018 21:59:43 +0000 (14:59 -0700)]
tcp: switch pacing timer to softirq based hrtimer

linux-4.16 got support for softirq based hrtimers.
TCP can switch its pacing hrtimer to this variant, since this
avoids going through a tasklet and some atomic operations.

pacing timer logic looks like other (jiffies based) tcp timers.

v2: use hrtimer_try_to_cancel() in tcp_clear_xmit_timers()
    to correctly release reference on socket if needed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'dsa-Plug-in-PHYLINK-support'
David S. Miller [Fri, 11 May 2018 16:03:06 +0000 (12:03 -0400)]
Merge branch 'dsa-Plug-in-PHYLINK-support'

Florian Fainelli says:

====================
net: dsa: Plug in PHYLINK support

This patch series adds PHYLINK support to DSA which is necessary to support more
complex PHY and pluggable modules setups.

Patch series can be found here:

https://github.com/ffainelli/linux/commits/dsa-phylink-v2

This was tested on:

- dsa-loop
- bcm_sf2
- mv88e6xxx
- b53

With a variety of test cases:
- internal & external MDIO PHYs
- MoCA with link notification through interrupt/MMIO register
- built-in PHYs
- ifconfig up/down for several cycles works
- bind/unbind of the drivers

Changes in v2:

- fixed link configuration for mv88e6xxx (Andrew) after introducing polling

This is technically v2 of what was posted back in March 2018, changes from last
time:

- fixed probe/remove of drivers
- fixed missing gpiod_put() for link GPIOs
- fixed polling of link GPIOs (Russell I would need your SoB on the patch you
  provided offline initially, added some modifications to it)
- tested across a wider set of platforms

And everything should still work as expected. Please be aware of the following:

- switch drivers (like bcm_sf2) which may have user-facing network ports using
  fixed links would need to implement phylink_mac_ops to remain functional.
  PHYLINK does not create a phy_device for fixed links, therefore our
  call to adjust_link() from phylink_mac_link_{up,down} would not be calling
  into the driver. This *should not* affect CPU/DSA ports which are configured
  through adjust_link() but have no network devices

- support for SFP/SFF is now possible, but switch drivers will still need some
  modifications to properly support those, including, but not limited to using
  the correct binding information. This will be submitted on top of this series

Please do test on your respective platforms/switches and let me know if you
find any issues, hopefully everything still works like before.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: bcm_sf2: Get rid of PHYLIB functions
Florian Fainelli [Thu, 10 May 2018 20:17:37 +0000 (13:17 -0700)]
net: dsa: bcm_sf2: Get rid of PHYLIB functions

Now that we have converted the bcm_sf2 driver to implement PHYLINK MAC
operations, we can remove the PHYLIB callbacks: adjust_link() and
fixed_link_update() which are no longer called by DSA.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: Plug in PHYLINK support
Florian Fainelli [Thu, 10 May 2018 20:17:36 +0000 (13:17 -0700)]
net: dsa: Plug in PHYLINK support

Add support for PHYLINK within the DSA subsystem in order to support more
complex devices such as pluggable (SFP) and non-pluggable (SFF) modules, 10G
PHYs, and traditional PHYs. Using PHYLINK allows us to drop some amount of
complexity we had while probing fixed and non-fixed PHYs using Device Tree.

Because PHYLINK separates the Ethernet MAC/port configuration into different
stages, we let switch drivers implement those, and for now, we maintain
functionality by calling dsa_slave_adjust_link() during
phylink_mac_link_{up,down} which provides semantically equivalent steps.

Drivers willing to take advantage of PHYLINK should implement the phylink_mac_*
operations that DSA wraps.

We cannot quite remove the adjust_link() callback just yet, because a number of
drivers rely on that for configuring their "CPU" and "DSA" ports, this is done
dsa_port_setup_phy_of() and dsa_port_fixed_link_register_of() still.

Drivers that utilize fixed links for user-facing ports (e.g: bcm_sf2) will need
to implement phylink_mac_ops from now on to preserve functionality, since PHYLINK
*does not* create a phy_device instance for fixed links.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: add PHYLINK support
Russell King [Thu, 10 May 2018 20:17:35 +0000 (13:17 -0700)]
net: dsa: mv88e6xxx: add PHYLINK support

Add rudimentary phylink support to mv88e6xxx. This allows the driver
using user ports with fixed links to keep operating normally. User ports
with normal PHYs are not affected since the switch automatically manages
their link parameters. User facing ports which use a SFP/SFF with a
non-fixed link mode might require a call to phylink_mac_change() to
operate properly.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
[Andrew: fixed link setting after adding link polling]
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
[florian: expand commit message]
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: Eliminate dsa_slave_get_link()
Florian Fainelli [Thu, 10 May 2018 20:17:34 +0000 (13:17 -0700)]
net: dsa: Eliminate dsa_slave_get_link()

Since we use PHYLIB to manage the per-port link indication, this will
also be reflected correctly in the network device's carrier state, so we
can use ethtool_op_get_link() instead.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: bcm_sf2: Implement phylink_mac_ops
Florian Fainelli [Thu, 10 May 2018 20:17:33 +0000 (13:17 -0700)]
net: dsa: bcm_sf2: Implement phylink_mac_ops

Make the bcm_sf2 driver implement phylink_mac_ops since it needs to
support a wide variety of network interfaces: internal & external MDIO
PHYs, fixed PHYs, MoCA with MMIO link status.

A large amount of what needs to be done already exists under
bcm_sf2_sw_adjust_link() so we are essentially breaking this down into
the necessary operation for PHYLINK to work: mac_config, mac_link_up,
mac_link_down and validate. We can now entirely get rid of most of what
fixed_link_update() provided because only the link information is actually
necessary. We still have to force DUPLEX_FULL for legacy Device Tree bindings
that did not specify that before.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: Add PHYLINK switch operations
Florian Fainelli [Thu, 10 May 2018 20:17:32 +0000 (13:17 -0700)]
net: dsa: Add PHYLINK switch operations

In preparation for adding support for PHYLINK within DSA, define a number of
operations that we will need and that switch drivers can start implementing.
Proper integration with PHYLINK will follow in subsequent patches.

We start selecting PHYLINK (which implies PHYLIB) in net/dsa/Kconfig
such that drivers can be guaranteed that this dependency is properly
taken care of and can start referencing PHYLINK helper functions without
requiring stubs or anything.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: phylink: Poll link GPIOs
Russell King [Thu, 10 May 2018 20:17:31 +0000 (13:17 -0700)]
net: phy: phylink: Poll link GPIOs

When using a fixed link with a link GPIO, we need to poll that GPIO to
determine link state changes. This is consistent with what fixed_phy.c does.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: phylink: Release link GPIO
Florian Fainelli [Thu, 10 May 2018 20:17:30 +0000 (13:17 -0700)]
net: phy: phylink: Release link GPIO

We are not releasing the link GPIO descriptor with gpiod_put() which results in
subsequent probing to get -EBUSY when calling fwnode_get_named_gpiod(). Fix this
by doing the release in phylink_destroy().

Fixes: 9525ae83959b ("phylink: add phylink infrastructure")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: phylink: Use gpiod_get_value_cansleep()
Florian Fainelli [Thu, 10 May 2018 20:17:29 +0000 (13:17 -0700)]
net: phy: phylink: Use gpiod_get_value_cansleep()

The GPIO provider for the link GPIO line might require the use of the
_cansleep() API, utilize that. This is safe to do since we run in workqueue
context.

Fixes: 9525ae83959b ("phylink: add phylink infrastructure")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agorocker: Postpone filtering of !added_by_user FDB
Petr Machata [Thu, 10 May 2018 13:29:46 +0000 (15:29 +0200)]
rocker: Postpone filtering of !added_by_user FDB

Breaking out of the switch in rocker_switchdev_event() still ends up
scheduling work, except an ill-defined one. This leads to an OOPS cited
below. Fix by postponing the check until rocker_switchdev_event_work().

[   23.148476] BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
[   23.148810] PGD 0 P4D 0
[   23.148982] Oops: 0000 [#1] PREEMPT SMP PTI
[   23.149190] Modules linked in: bridge stp llc iptable_nat nf_nat_ipv4 nf_nat e1000 rocker
[   23.149768] CPU: 0 PID: 239 Comm: kworker/u2:4 Not tainted 4.17.0-rc3-net_next_queue-custom #6
[   23.150298] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-2.fc27 04/01/2014
[   23.150868] Workqueue: rocker rocker_switchdev_event_work [rocker]
[   23.151258] RIP: 0010:ofdpa_port_fdb+0x7b/0x230 [rocker]
[   23.151597] RSP: 0018:ffffc900004b3e18 EFLAGS: 00010246
[   23.151952] RAX: 00000000fffbc68c RBX: 0000000000000000 RCX: 0000000000000000
[   23.152363] RDX: 0000000000000010 RSI: ffff88003b4471e0 RDI: 00000000ffffffff
[   23.152768] RBP: ffff88003b4471c0 R08: ffff88003b4471e0 R09: ffff88003b4471c0
[   23.153141] R10: 0000000000000000 R11: 0000000000000000 R12: ffff880036caf000
[   23.153515] R13: 0000000000000000 R14: 0000000000000000 R15: ffff88003bc00000
[   23.153919] FS:  0000000000000000(0000) GS:ffff88003fc00000(0000) knlGS:0000000000000000
[   23.154444] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   23.154806] CR2: 0000000000000000 CR3: 0000000036eb6000 CR4: 00000000000006f0
[   23.155194] Call Trace:
[   23.155472]  rocker_switchdev_event_work+0x9b/0xd0 [rocker]
[   23.155850]  ? __schedule+0x231/0x700
[   23.156175]  process_one_work+0x1cf/0x3e0
[   23.156490]  worker_thread+0x26/0x3d0
[   23.156795]  ? trace_event_raw_event_workqueue_execute_start+0x80/0x80
[   23.157181]  kthread+0x10e/0x130
[   23.157485]  ? kthread_create_worker_on_cpu+0x40/0x40
[   23.157824]  ret_from_fork+0x35/0x40
[   23.158174] Code: 00 00 c1 e8 02 4c 8d 45 20 bf ff ff ff ff 83 e0 01 4c 89 65 20 88 45 14 48 8b 05 21 da 1f e2 4c 89 c6 4c 89 44 24 10 48 89 45 18 <41> 8b 45 00 89 45 28 41 0f b7 45 04 66 89 45 2c 0f b7 44 24 04
[   23.159346] RIP: ofdpa_port_fdb+0x7b/0x230 [rocker] RSP: ffffc900004b3e18
[   23.159742] CR2: 0000000000000000
[   23.160088] ---[ end trace f9b16d4cb6df0629 ]---

Fixes: 816a3bed9549 ("switchdev: Add fdb.added_by_user to switchdev notifications")
Suggested-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: Fix tls_device initialization
Boris Pismenny [Thu, 10 May 2018 13:27:25 +0000 (16:27 +0300)]
tls: Fix tls_device initialization

Add sg table initialization to fix a BUG_ON encountered when enabling
CONFIG_DEBUG_SG.

Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: fix the wrong conversion of Mbps to Kbps
Ganesh Goudar [Thu, 10 May 2018 10:37:23 +0000 (16:07 +0530)]
cxgb4: fix the wrong conversion of Mbps to Kbps

fix the wrong conversion where 1 Mbps was converted to
1024 Kbps.

Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'mlxsw-Support-VLAN-devices-in-mirroring-offloads'
David S. Miller [Thu, 10 May 2018 21:50:53 +0000 (17:50 -0400)]
Merge branch 'mlxsw-Support-VLAN-devices-in-mirroring-offloads'

Ido Schimmel says:

====================
mlxsw: Support VLAN devices in mirroring offloads

Petr says:

When offloading "tc action mirred mirror", there are several scenarios
where VLAN devices can show up, that mlxsw can offload on Spectrum
machines.

I) A direct mirror to a VLAN device on top of a front-panel port device
   (commonly referred to as "RSPAN")

II) VLAN device in egress path of a packet when resolving a mirror to
    gretap or ip6gretap netdevice.

Specifically in the latter case, the following are the cases that can be
offloaded:

IIa) VLAN device directly above a physical device.
IIb) A VLAN-unaware bridge where the egress device is as in IIa.
IIc) VLAN device on top of a VLAN-aware bridge where the egress device
     is a physical device.

This patch set implements all the above cases.

First, in patch #1, br_vlan_get_info() is extended to allow bridge
master argument.

Case I is then implemented in patches #2 and #3, case II in patch #4.

Note that handling of VLAN protocol is not included. In case I, mirrored
packets may end up being double-tagged, and it might be reasonable for
the outer tag to be an 802.1ad. However, the protocol type configuration
would have to be put on the same VLAN netdevice that represents normal
VLAN traffic, and mlxsw currently ignores this setting in that case. Thus
this support was left out and the encapsulation always uses 802.1q
protocol.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_span: Support VLAN under mirror-to-gretap
Petr Machata [Thu, 10 May 2018 10:13:06 +0000 (13:13 +0300)]
mlxsw: spectrum_span: Support VLAN under mirror-to-gretap

When mirroring to a gretap or ip6gretap device, allow the underlay
packet path to include VLAN devices. The following configurations are
supported in underlay:

- vlan over phys
- vlan-unaware bridge where the egress device is vlan over phys
- vlan over vlan-aware bridge where the egress device is phys

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_span: Support mirror-to-VLAN
Petr Machata [Thu, 10 May 2018 10:13:05 +0000 (13:13 +0300)]
mlxsw: spectrum_span: Support mirror-to-VLAN

Offload "tc action mirred mirror" to a device that is a vlan device on
top of a front-panel port device. The hardware encapsulates the mirrored
packets in a VLAN tag. That includes the case that the mirrored traffic
is already VLAN-tagged--in that case the monitor traffic will be
double-tagged, just like in the software path.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: reg: Add MLXSW_REG_MPAT_SPAN_TYPE_REMOTE_ETH
Petr Machata [Thu, 10 May 2018 10:13:04 +0000 (13:13 +0300)]
mlxsw: reg: Add MLXSW_REG_MPAT_SPAN_TYPE_REMOTE_ETH

Add MLXSW_REG_MPAT_SPAN_TYPE_REMOTE_ETH to support VLAN-encapsulated
port mirroring.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bridge: Allow bridge master in br_vlan_get_info()
Petr Machata [Thu, 10 May 2018 10:13:03 +0000 (13:13 +0300)]
net: bridge: Allow bridge master in br_vlan_get_info()

Mirroring offload in mlxsw needs to check that a given VLAN is allowed
to ingress the bridge device. br_vlan_get_info() is the function that is
used for this, however currently it only supports bridge port devices.
Extend it to support bridge masters as well.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: Add mark for TIMEWAIT sockets
Jon Maxwell [Thu, 10 May 2018 06:53:51 +0000 (16:53 +1000)]
tcp: Add mark for TIMEWAIT sockets

This version has some suggestions by Eric Dumazet:

- Use a local variable for the mark in IPv6 instead of ctl_sk to avoid SMP
races.
- Use the more elegant "IP4_REPLY_MARK(net, skb->mark) ?: sk->sk_mark"
statement.
- Factorize code as sk_fullsock() check is not necessary.

Aidan McGurn from Openwave Mobility systems reported the following bug:

"Marked routing is broken on customer deployment. Its effects are large
increase in Uplink retransmissions caused by the client never receiving
the final ACK to their FINACK - this ACK misses the mark and routes out
of the incorrect route."

Currently marks are added to sk_buffs for replies when the "fwmark_reflect"
sysctl is enabled. But not for TW sockets that had sk->sk_mark set via
setsockopt(SO_MARK..).

Fix this in IPv4/v6 by adding tw->tw_mark for TIME_WAIT sockets. Copy the the
original sk->sk_mark in __inet_twsk_hashdance() to the new tw->tw_mark location.
Then progate this so that the skb gets sent with the correct mark. Do the same
for resets. Give the "fwmark_reflect" sysctl precedence over sk->sk_mark so that
netfilter rules are still honored.

Signed-off-by: Jon Maxwell <jmaxwell37@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv4: remove define INET_CSK_DEBUG and unnecessary EXPORT_SYMBOL
Joe Perches [Thu, 10 May 2018 06:24:07 +0000 (23:24 -0700)]
net: ipv4: remove define INET_CSK_DEBUG and unnecessary EXPORT_SYMBOL

INET_CSK_DEBUG is always set and only is used for 2 pr_debug calls.

EXPORT_SYMBOL(inet_csk_timer_bug_msg) is only used by these 2
pr_debug calls and is also unnecessary as the exported string can
be used directly by these calls.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: bump up driver version to 1.7.2 to match newer NIC firmware
Felix Manlunas [Wed, 9 May 2018 18:49:38 +0000 (11:49 -0700)]
liquidio: bump up driver version to 1.7.2 to match newer NIC firmware

Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: monitor all of Octeon's cores in watchdog thread
Felix Manlunas [Wed, 9 May 2018 18:31:31 +0000 (11:31 -0700)]
liquidio: monitor all of Octeon's cores in watchdog thread

The liquidio_watchdog kernel thread is watching over only 12 cores of the
Octeon CN23XX; it's neglecting the other 4 cores that are present in the
CN2360.  Fix it by defining LIO_MAX_CORES as 16.

Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Thu, 10 May 2018 21:31:03 +0000 (17:31 -0400)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2018-05-09

This series contains updates to fm10k only.

Jake provides all the changes in the series, starting with adding
support for accelerated MACVLAN devices.  Reduced code duplication by
implementing a macro to be used when setting up the type specific
macros.  Avoided potential bugs with stats by using a macro to calculate
the array size when passing to ensure that the size is correct.

v2: changed macro reference '#' with __stringify() as suggested by
    Joe Perches to patch 2 of the series.  Also made sure the updated
    series of patches is actually pushed to my kernel.org tree
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ipv6: fix lock imbalance in ip6_route_del()
Eric Dumazet [Wed, 9 May 2018 17:05:46 +0000 (10:05 -0700)]
net/ipv6: fix lock imbalance in ip6_route_del()

WARNING: lock held when returning to user space!
4.17.0-rc3+ #37 Not tainted

syz-executor1/27662 is leaving the kernel with locks still held!
1 lock held by syz-executor1/27662:
 #0: 00000000f661aee7 (rcu_read_lock){....}, at: ip6_route_del+0xea/0x13f0 net/ipv6/route.c:3206
BUG: scheduling while atomic: syz-executor1/27662/0x00000002
INFO: lockdep is turned off.
Modules linked in:
Kernel panic - not syncing: scheduling while atomic

CPU: 1 PID: 27662 Comm: syz-executor1 Not tainted 4.17.0-rc3+ #37
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1b9/0x294 lib/dump_stack.c:113
 panic+0x22f/0x4de kernel/panic.c:184
 __schedule_bug.cold.85+0xdf/0xdf kernel/sched/core.c:3290
 schedule_debug kernel/sched/core.c:3307 [inline]
 __schedule+0x139e/0x1e30 kernel/sched/core.c:3412
 schedule+0xef/0x430 kernel/sched/core.c:3549
 exit_to_usermode_loop+0x220/0x310 arch/x86/entry/common.c:152
 prepare_exit_to_usermode arch/x86/entry/common.c:196 [inline]
 syscall_return_slowpath arch/x86/entry/common.c:265 [inline]
 do_syscall_64+0x6ac/0x800 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x455979
RSP: 002b:00007fbf4051dc68 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: 0000000000000000 RBX: 00007fbf4051e6d4 RCX: 0000000000455979
RDX: 00000000200001c0 RSI: 000000000000890c RDI: 0000000000000013
RBP: 000000000072bea0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00000000ffffffff
R13: 00000000000003c8 R14: 00000000006f9b60 R15: 0000000000000000
Dumping ftrace buffer:
   (ftrace buffer empty)
Kernel Offset: disabled
Rebooting in 86400 seconds..

Fixes: 23fb93a4d3f1 ("net/ipv6: Cleanup exception and cache route handling")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Ahern <dsahern@gmail.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'hns3-misc-fixes'
David S. Miller [Thu, 10 May 2018 21:27:11 +0000 (17:27 -0400)]
Merge branch 'hns3-misc-fixes'

Salil Mehta says:

====================
Misc bug fixes for HNS3 Ethernet Driver

Fixes to some of the bugs found during system test, internal review
and clean-up
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: refactor the loopback related function
Yunsheng Lin [Wed, 9 May 2018 16:24:41 +0000 (17:24 +0100)]
net: hns3: refactor the loopback related function

This patch refactors the loopback related function in order
to support the serdes loopback.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: fix for cleaning ring problem
Yunsheng Lin [Wed, 9 May 2018 16:24:40 +0000 (17:24 +0100)]
net: hns3: fix for cleaning ring problem

The head or tail in hardware is not longer valid when resetting,
current hns3_clear_all_ring use them to clean the ring, which
will cause problem during resetting.

This patch fixes it by using next_to_use and next_to_clean in
the ring struct.

Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: remove add/del_tunnel_udp in hns3_enet module
Yunsheng Lin [Wed, 9 May 2018 16:24:39 +0000 (17:24 +0100)]
net: hns3: remove add/del_tunnel_udp in hns3_enet module

The add/del_tunnel_udp is not implemented in hclge_main moulde,
the NETIF_F_RX_UDP_TUNNEL_PORT feature bit is added automatically
by stack when ndo_udp_tunnel_add is not null in dev->netdev_ops.

This patch removes the add/del_tunnel_udp related function, for
we do not support this feature now.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Fix for setting mac address when resetting
Yunsheng Lin [Wed, 9 May 2018 16:24:38 +0000 (17:24 +0100)]
net: hns3: Fix for setting mac address when resetting

When hns3_init_mac_addr is called during reset process, it will
get the mac address from NCL_CONFIG and set it to hardware. If
user has changed the mac address, then the mac address set by
user is lost during resetting.

This patch fixes it by not getting the mac address from NCL_CONFIG
when resetting.

Fixes: 424eb834a9be ("net: hns3: Unified HNS3 {VF|PF} Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agohv_netvsc: typo in NDIS RSS parameters structure
Stephen Hemminger [Wed, 9 May 2018 16:00:07 +0000 (09:00 -0700)]
hv_netvsc: typo in NDIS RSS parameters structure

Fix simple misspelling kashkey_offset should be hashkey_offset.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agodrivers: net: davinci_mdio: prevent spurious timeout
Sekhar Nori [Wed, 9 May 2018 15:45:15 +0000 (21:15 +0530)]
drivers: net: davinci_mdio: prevent spurious timeout

A well timed kernel preemption in the time_after() loop
in wait_for_idle() can result in a spurious timeout
error to be returned.

Fix it by using readl_poll_timeout() which takes care of
this issue.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'dsa-mv88e6xxx-cleanup-Global-Control-2-register'
David S. Miller [Thu, 10 May 2018 20:12:53 +0000 (16:12 -0400)]
Merge branch 'dsa-mv88e6xxx-cleanup-Global-Control-2-register'

Vivien Didelot says:

====================
net: dsa: mv88e6xxx: cleanup Global Control 2 register

The mv88e6xxx driver still writes arbitrary values in the Global 1
Control 2 register at setup, which layout differs a lot between chips.
This results in an inconsistent configuration, for example with the
Remote Management Unit (RMU).

The first patch adds an operation for the Cascade Port bits, the second
patch sets the device number in the device mapping function and the
third patch adds an operation to correctly disable the RMU.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: add RMU disable op
Vivien Didelot [Wed, 9 May 2018 15:38:51 +0000 (11:38 -0400)]
net: dsa: mv88e6xxx: add RMU disable op

The RMU mode bits moved a lot within the Global Control 2 register of
the Marvell switch families. Add an .rmu_disable op to support at least
3 known alternatives.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: set device number
Vivien Didelot [Wed, 9 May 2018 15:38:50 +0000 (11:38 -0400)]
net: dsa: mv88e6xxx: set device number

All Marvell switches supported by mv88e6xxx have to set their device
number in the Global Control 2 register. Extract this in a read then
write function, called from the device mapping setup code.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: add a cascade port op
Vivien Didelot [Wed, 9 May 2018 15:38:49 +0000 (11:38 -0400)]
net: dsa: mv88e6xxx: add a cascade port op

Only the 88E6185 family has bits 15:12 Cascade Port bits in the Global
Control 2 register. Hence inconsistent values are actually written in
this register for other families.

Add a .set_cascade_port operation to isolate the 88E6185 case, and call
it from the device mapping setup function.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'mlx4-misc-next'
David S. Miller [Thu, 10 May 2018 20:08:26 +0000 (16:08 -0400)]
Merge branch 'mlx4-misc-next'

Tariq Toukan says:

====================
mlx4_core misc for 4.18

This patchset contains misc enhancements from the team
to the mlx4 Core driver.

Patch 1 by Eran adds driver version report in FW.
Patch 2 by Yishai implements suspend/resume PCI callbacks.
Patch 3 extends the range of an existing module param from boolean to numerical.

Series generated against net-next commit:
53a7bdfb2a27 dt-bindings: dsa: Remove unnecessary #address/#size-cells
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx4_core: Use msi_x module param to limit num of MSI-X irqs
Tariq Toukan [Wed, 9 May 2018 15:29:04 +0000 (18:29 +0300)]
net/mlx4_core: Use msi_x module param to limit num of MSI-X irqs

Extend the boolean interpretation of msi_x module parameter
to numerical, as follows:

0   - Don't use MSI-X.
1   - Use MSI-X, driver decides the num of MSI-X irqs.
>=2 - Use MSI-X, limit number of MSI-X irqs to msi_x.
      In SRIOV, this limits the number of MSI-X irqs per VF.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Cc: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Reviewed-by: Ajaykumar Hotchandani <ajaykumar.hotchandani@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx4_core: Add PCI calls for suspend/resume
Yishai Hadas [Wed, 9 May 2018 15:29:03 +0000 (18:29 +0300)]
net/mlx4_core: Add PCI calls for suspend/resume

Implement suspend/resume callbacks in struct pci_driver.

Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx4_core: Report driver version to FW
Eran Ben Elisha [Wed, 9 May 2018 15:29:02 +0000 (18:29 +0300)]
net/mlx4_core: Report driver version to FW

If supported, write a driver version string to FW as part of the
INIT_HCA command.

Example of driver version: "Linux,mlx4_core,4.0-0"

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: fix added_by_user switchdev notification
Vivien Didelot [Wed, 9 May 2018 03:03:12 +0000 (23:03 -0400)]
net: dsa: fix added_by_user switchdev notification

Commit 161d82de1ff8 ("net: bridge: Notify about !added_by_user FDB
entries") causes the below oops when bringing up a slave interface,
because dsa_port_fdb_add is still scheduled, but with a NULL address.

To fix this, keep the dsa_slave_switchdev_event function agnostic of the
notified info structure and handle the added_by_user flag in the
specific dsa_slave_switchdev_event_work function.

    [   75.512263] Unable to handle kernel NULL pointer dereference at virtual address 00000000
    [   75.519063] pgd = (ptrval)
    [   75.520545] [00000000] *pgd=00000000
    [   75.522839] Internal error: Oops: 17 [#1] ARM
    [   75.525898] Modules linked in:
    [   75.527673] CPU: 0 PID: 9 Comm: kworker/u2:1 Not tainted 4.17.0-rc2 #78
    [   75.532988] Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
    [   75.538153] Workqueue: dsa_ordered dsa_slave_switchdev_event_work
    [   75.542970] PC is at mv88e6xxx_port_db_load_purge+0x60/0x1b0
    [   75.547341] LR is at mdiobus_read_nested+0x6c/0x78
    [   75.550833] pc : [<804cd5c0>]    lr : [<804bba84>]    psr: 60070013
    [   75.555796] sp : 9f54bd78  ip : 9f54bd87  fp : 9f54bddc
    [   75.559719] r10: 00000000  r9 : 0000000e  r8 : 9f6a6010
    [   75.563643] r7 : 00000000  r6 : 81203048  r5 : 9f6a6010  r4 : 9f6a601c
    [   75.568867] r3 : 00000000  r2 : 00000000  r1 : 0000000d  r0 : 00000000
    [   75.574094] Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none
    [   75.579933] Control: 10c53c7d  Table: 9de20059  DAC: 00000051
    [   75.584384] Process kworker/u2:1 (pid: 9, stack limit = 0x(ptrval))
    [   75.589349] Stack: (0x9f54bd78 to 0x9f54c000)
    [   75.592406] bd60:                                                       00000000 00000000
    [   75.599295] bd80: 00000391 9f299d10 9f299d68 8014317c 9f7f0000 8120af00 00006dc2 00000000
    [   75.606186] bda0: 8120af00 00000000 9f54bdec 1c9f5d92 8014317c 9f6a601c 9f6a6010 00000000
    [   75.613076] bdc0: 00000000 00000000 9dd1141c 8125a0b4 9f54be0c 9f54bde0 804cd8a8 804cd56c
    [   75.619966] bde0: 0000000e 80143680 00000001 9dce9c1c 81203048 9dce9c10 00000003 00000000
    [   75.626858] be00: 9f54be5c 9f54be10 806abcac 804cd864 9f54be54 80143664 8014317c 80143054
    [   75.633748] be20: ffcaa81d 00000000 812030b0 1c9f5d92 00000000 81203048 9f54beb4 00000003
    [   75.640639] be40: ffffffff 00000000 9dd1141c 8125a0b4 9f54be84 9f54be60 80138e98 806abb18
    [   75.647529] be60: 81203048 9ddc4000 9dce9c54 9f72a300 00000000 00000000 9f54be9c 9f54be88
    [   75.654420] be80: 801390bc 80138e50 00000000 9dce9c54 9f54beac 9f54bea0 806a9524 801390a0
    [   75.661310] bea0: 9f54bedc 9f54beb0 806a9c7c 806a950c 9f54becc 00000000 00000000 00000000
    [   75.668201] bec0: 9f540000 1c9f5d92 805fe604 9ddffc00 9f54befc 9f54bee0 806ab228 806a9c38
    [   75.675092] bee0: 806ab178 9ddffc00 9f4c1900 9f40d200 9f54bf34 9f54bf00 80131e30 806ab184
    [   75.681983] bf00: 9f40d214 9f54a038 9f40d200 9f40d200 9f4c1918 812119a0 9f40d214 9f54a038
    [   75.688873] bf20: 9f40d200 9f4c1900 9f54bf7c 9f54bf38 80132124 80131d1c 9f5f2dd8 00000000
    [   75.695764] bf40: 812119a0 9f54a038 812119a0 81259c5b 9f5f2dd8 9f5f2dc0 9f53dbc0 00000000
    [   75.702655] bf60: 9f4c1900 801320b4 9f5f2dd8 9f4f7e88 9f54bfac 9f54bf80 80137ad0 801320c0
    [   75.709544] bf80: 9f54a000 9f53dbc0 801379a0 00000000 00000000 00000000 00000000 00000000
    [   75.716434] bfa0: 00000000 9f54bfb0 801010e8 801379ac 00000000 00000000 00000000 00000000
    [   75.723324] bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
    [   75.730206] bfe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
    [   75.737083] Backtrace:
    [   75.738252] [<804cd560>] (mv88e6xxx_port_db_load_purge) from [<804cd8a8>] (mv88e6xxx_port_fdb_add+0x50/0x68)
    [   75.746795]  r10:8125a0b4 r9:9dd1141c r8:00000000 r7:00000000 r6:00000000 r5:9f6a6010
    [   75.753323]  r4:9f6a601c
    [   75.754570] [<804cd858>] (mv88e6xxx_port_fdb_add) from [<806abcac>] (dsa_switch_event+0x1a0/0x660)
    [   75.762238]  r8:00000000 r7:00000003 r6:9dce9c10 r5:81203048 r4:9dce9c1c
    [   75.767655] [<806abb0c>] (dsa_switch_event) from [<80138e98>] (notifier_call_chain+0x54/0x94)
    [   75.774893]  r10:8125a0b4 r9:9dd1141c r8:00000000 r7:ffffffff r6:00000003 r5:9f54beb4
    [   75.781423]  r4:81203048
    [   75.782672] [<80138e44>] (notifier_call_chain) from [<801390bc>] (raw_notifier_call_chain+0x28/0x30)
    [   75.790514]  r9:00000000 r8:00000000 r7:9f72a300 r6:9dce9c54 r5:9ddc4000 r4:81203048
    [   75.796982] [<80139094>] (raw_notifier_call_chain) from [<806a9524>] (dsa_port_notify+0x24/0x38)
    [   75.804483] [<806a9500>] (dsa_port_notify) from [<806a9c7c>] (dsa_port_fdb_add+0x50/0x6c)
    [   75.811371] [<806a9c2c>] (dsa_port_fdb_add) from [<806ab228>] (dsa_slave_switchdev_event_work+0xb0/0x10c)
    [   75.819635]  r4:9ddffc00
    [   75.820885] [<806ab178>] (dsa_slave_switchdev_event_work) from [<80131e30>] (process_one_work+0x120/0x3a4)
    [   75.829241]  r6:9f40d200 r5:9f4c1900 r4:9ddffc00 r3:806ab178
    [   75.833612] [<80131d10>] (process_one_work) from [<80132124>] (worker_thread+0x70/0x574)
    [   75.840415]  r10:9f4c1900 r9:9f40d200 r8:9f54a038 r7:9f40d214 r6:812119a0 r5:9f4c1918
    [   75.846945]  r4:9f40d200
    [   75.848191] [<801320b4>] (worker_thread) from [<80137ad0>] (kthread+0x130/0x160)
    [   75.854300]  r10:9f4f7e88 r9:9f5f2dd8 r8:801320b4 r7:9f4c1900 r6:00000000 r5:9f53dbc0
    [   75.860830]  r4:9f5f2dc0
    [   75.862076] [<801379a0>] (kthread) from [<801010e8>] (ret_from_fork+0x14/0x2c)
    [   75.867999] Exception stack(0x9f54bfb0 to 0x9f54bff8)
    [   75.871753] bfa0:                                     00000000 00000000 00000000 00000000
    [   75.878640] bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
    [   75.885519] bfe0: 00000000 00000000 00000000 00000000 00000013 00000000
    [   75.890844]  r10:00000000 r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:801379a0
    [   75.897377]  r4:9f53dbc0 r3:9f54a000
    [   75.899663] Code: e3a02000 e3a03000 e14b26f4 e24bc055 (e5973000)
    [   75.904575] ---[ end trace fbca818a124dbf0d ]---

Fixes: 816a3bed9549 ("switchdev: Add fdb.added_by_user to switchdev notifications")
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: clean up removal of binding table items
Jon Maloy [Wed, 9 May 2018 00:59:41 +0000 (02:59 +0200)]
tipc: clean up removal of binding table items

In commit be47e41d77fb ("tipc: fix use-after-free in tipc_nametbl_stop")
we fixed a problem caused by premature release of service range items.

That fix is correct, and solved the problem. However, it doesn't address
the root of the problem, which is that we don't lookup the tipc_service
 -> service_range -> publication items in the correct hierarchical
order.

In this commit we try to make this right, and as a side effect obtain
some code simplification.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotun: Do SIOCGSKNS out of rtnl_lock()
Kirill Tkhai [Tue, 8 May 2018 16:21:34 +0000 (19:21 +0300)]
tun: Do SIOCGSKNS out of rtnl_lock()

Since net ns of tun device is assigned on the device creation,
and it never changes, we do not need to use any lock to get it
from alive tun.

Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'net-Update-static-keys-to-modern-api'
David S. Miller [Thu, 10 May 2018 19:13:35 +0000 (15:13 -0400)]
Merge branch 'net-Update-static-keys-to-modern-api'

Davidlohr Bueso says:

====================
net: Update static keys to modern api

The following patches update pretty much all core net static key users
to the modern api. Changes are mostly trivial conversion without affecting
any semantics. The motivation is a resend of patches 1 and 2 from a while[1]
back, and the rest are added patches, specific for -net.

Applies against today's linux-next. Compile tested only.

[1] lkml.kernel.org/r/20180326210929.5244-1-dave@stgolabs.net
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/udp: Update udp_encap_needed static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:07:03 +0000 (09:07 -0700)]
net/udp: Update udp_encap_needed static key to modern api

No changes in refcount semantics -- key init is false; replace

static_key_enable         with   static_branch_enable
static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Added a '_key' suffix to udp and udpv6 encap_needed, for better
self documentation.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Update generic_xdp_needed static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:07:02 +0000 (09:07 -0700)]
net: Update generic_xdp_needed static key to modern api

No changes in refcount semantics -- key init is false; replace

static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Added a '_key' suffix to generic_xdp_needed, for better self
documentation.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Update netstamp_needed static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:07:01 +0000 (09:07 -0700)]
net: Update netstamp_needed static key to modern api

No changes in refcount semantics -- key init is false; replace

static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Added a '_key' suffix to netstamp_needed, for better self
documentation.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Update [e/in]gress_needed static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:07:00 +0000 (09:07 -0700)]
net: Update [e/in]gress_needed static key to modern api

No changes in semantics -- key init is false; replace

static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Added a '_key' suffix to both ingress_needed and egress_needed,
for better self documentation.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/sock: Update memalloc_socks static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:06:59 +0000 (09:06 -0700)]
net/sock: Update memalloc_socks static key to modern api

No changes in refcount semantics -- key init is false; replace

static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Added a '_key' suffix to memalloc_socks, for better self
documentation.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ipv4: Update ip_tunnel_metadata_cnt static key to modern api
Davidlohr Bueso [Tue, 8 May 2018 16:06:58 +0000 (09:06 -0700)]
net/ipv4: Update ip_tunnel_metadata_cnt static key to modern api

No changes in refcount semantics -- key init is false; replace

static_key_slow_inc|dec   with   static_branch_inc|dec
static_key_false          with   static_branch_unlikely

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: 88E6141/6341 SERDES support
Marek Behún [Fri, 4 May 2018 17:26:10 +0000 (19:26 +0200)]
net: dsa: mv88e6xxx: 88E6141/6341 SERDES support

The 88E6141/6341 switches (also known as Topaz) have 1 SGMII lane,
which can be configured the same way as the SERDES lane on 88E6390.

Signed-off-by: Marek Behun <marek.behun@nic.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: stmmac: Add support for U32 TC filter using Flexible RX Parser
Jose Abreu [Fri, 4 May 2018 09:01:38 +0000 (10:01 +0100)]
net: stmmac: Add support for U32 TC filter using Flexible RX Parser

This adds support for U32 filter by using an HW only feature called
Flexible RX Parser. This allow us to match any given packet field with a
pattern and accept/reject or even route the packet to a specific DMA
channel.

Right now we only support acception or rejection of frame and we only
support simple rules. Though, the Parser has the flexibility of jumping to
specific rules as an if condition so complex rules can be established.

This is only supported in GMAC5.10+.

The following commands can be used to test this code:

1) Setup an ingress qdisk:
# tc qdisc add dev eth0 handle ffff: ingress

2) Setup a filter (e.g. filter by IP):
# tc filter add dev eth0 parent ffff: protocol ip u32 match ip \
src 192.168.0.3 skip_sw action drop

In every tests performed we always used the "skip_sw" flag to make sure
only the RX Parser was involved.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Vitor Soares <soares@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Cc: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add support of hardware rx-vlan-offload to HNS3 VF driver
Yunsheng Lin [Thu, 3 May 2018 16:28:11 +0000 (17:28 +0100)]
net: hns3: Add support of hardware rx-vlan-offload to HNS3 VF driver

This patch adds support of hardware rx-vlan-offload to VF driver.
VF uses mailbox to convey PF to configure the hardware.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomicrochip_t1: Add driver for Microchip LAN87XX T1 PHYs
Nisar Sayed [Wed, 2 May 2018 15:39:17 +0000 (21:09 +0530)]
microchip_t1: Add driver for Microchip LAN87XX T1 PHYs

Add driver for Microchip LAN87XX T1 PHYs

This patch support driver for Microchp T1 PHYs.
There will be followup patches to this driver to support T1 PHY
features such as cable diagnostics, signal quality indicator(SQI),
sleep and wakeup (TC10) support.

Signed-off-by: Nisar Sayed <Nisar.Sayed@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'mlx5-updates-2018-05-07' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Thu, 10 May 2018 12:17:55 +0000 (08:17 -0400)]
Merge tag 'mlx5-updates-2018-05-07' of git://git./linux/kernel/git/mellanox/linux

Saeed Mahameed says:

====================
mlx5-updates-2018-05-07

mlx5 core driver misc cleanups and updates:
 - fix spelling mistake: "modfiy" -> "modify"
 - Cleanup unused field in Work Queue parameters
 - dump_command mailbox length printed
 - Refactor num of blocks in mailbox calculation
 - Decrease level of prints about non-existent MKEY
 - remove some extraneous spaces in indentations
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agofm10k: don't protect fm10k_queue_mac_request by fm10k_host_mbx_ready
Jacob Keller [Thu, 12 Apr 2018 18:15:59 +0000 (11:15 -0700)]
fm10k: don't protect fm10k_queue_mac_request by fm10k_host_mbx_ready

We don't actually need to check if the host mbx is ready when queuing
MAC requests, because these are not handled by a special queue which
queues up requests until the mailbox is capable of handling them.

Pull these requests outside the fm10k_host_mbx_ready() check, as it is
not necessary.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agofm10k: warn if the stat size is unknown
Jacob Keller [Thu, 12 Apr 2018 18:15:58 +0000 (11:15 -0700)]
fm10k: warn if the stat size is unknown

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agofm10k: use macro to avoid passing the array and size separately
Jacob Keller [Thu, 12 Apr 2018 18:15:57 +0000 (11:15 -0700)]
fm10k: use macro to avoid passing the array and size separately

Avoid potential bugs with fm10k_add_stat_strings and
fm10k_add_ethtool_stats by using a macro to calculate the ARRAY_SIZE
when passing. This helps ensure that the size is always correct.

Note that it assumes we only pass static const fm10k_stat arrays, and
that evaluation of the argument won't have side effects.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agofm10k: use variadic arguments to fm10k_add_stat_strings
Jacob Keller [Thu, 12 Apr 2018 18:15:56 +0000 (11:15 -0700)]
fm10k: use variadic arguments to fm10k_add_stat_strings

Instead of using a fixed prefix string we setup before each call to
fm10k_add_stat_strings, modify the helper to take variadic arguments and
pass them to vsnprintf. This requires changing the fm10k_stat strings to
take % format specifiers where necessary, but the resulting code is much
simpler.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agofm10k: reduce duplicate fm10k_stat macro code
Jacob Keller [Thu, 12 Apr 2018 18:15:55 +0000 (11:15 -0700)]
fm10k: reduce duplicate fm10k_stat macro code

Share some of the code for setting up fm10k_stat macros by implementing
an FM10K_STAT_FIELDS macro which we can use when setting up the type
specific macros.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agofm10k: setup VLANs for l2 accelerated macvlan interfaces
Jacob Keller [Thu, 12 Apr 2018 18:15:54 +0000 (11:15 -0700)]
fm10k: setup VLANs for l2 accelerated macvlan interfaces

We have support for accelerating macvlan devices via the
.ndo_dfwd_add_station() netdev op. These accelerated macvlan MAC
addresses are stored in the l2_accel structure, separate from the
unicast or multicast address lists.

If a VLAN is added on top of the macvlan device by the stack, traffic
will not properly flow to the macvlan. This occurs because we fail to
setup the VLANs for l2_accel MAC addresses.

In the non-offloaded case the MAC address is added to the unicast
address list, and thus the normal setup for enabling VLANs works as
expected.

We also need to add VLANs marked from .ndo_vlan_rx_add_vid() into the
l2_accel MAC addresses. Otherwise, VLAN traffic will not properly be
received by the VLAN devices attached to the offloaded macvlan devices.

Fix this by adding necessary logic to setup VLANs not only for the
unicast and multicast addresses, but also the l2_accel list. We need
similar logic in dfwd_add_station, dfwd_del_station, fm10k_update_vid,
and fm10k_restore_rx_state.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoMerge branch 'udp-gso-cleanups'
David S. Miller [Wed, 9 May 2018 02:30:06 +0000 (22:30 -0400)]
Merge branch 'udp-gso-cleanups'

Alexander Duyck says:

====================
UDP GSO Segmentation clean-ups

This patch set addresses a number of issues I found while sorting out
enabling UDP GSO Segmentation support for ixgbe/ixgbevf. Specifically there
were a number of issues related to the checksum and such that seemed to
cause either minor irregularities or kernel panics in the case of the
offload request being allowed to traverse between name spaces.

With this set applied I am was able to get UDP GSO traffic to pass over
vxlan tunnels in both offloaded modes and non-offloaded modes for ixgbe and
ixgbevf.

I submitted the driver specific patches earlier as an RFC:
https://patchwork.ozlabs.org/project/netdev/list/?series=42477&archive=both&state=*

v2: Updated patches based on feedback from Eric Dumazet
    Split first patch into several patches based on feedback from Eric
v3: Drop patch that was calling pskb_may_pull as it was redundant.
    Added code to use MANGLED_0 in case of UDP checksum
    Drop patch adding NETIF_F_GSO_UDP_L4 to list of GSO software offloads
    Added Acked-by for patches reviewed by Willem and not changed
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Do not copy destructor if one is not present
Alexander Duyck [Mon, 7 May 2018 18:08:52 +0000 (11:08 -0700)]
udp: Do not copy destructor if one is not present

This patch makes it so that if a destructor is not present we avoid trying
to update the skb socket or any reference counting that would be associated
with the NULL socket and/or descriptor. By doing this we can support
traffic coming from another namespace without any issues.

Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Add support for software checksum and GSO_PARTIAL with GSO offload
Alexander Duyck [Mon, 7 May 2018 18:08:46 +0000 (11:08 -0700)]
udp: Add support for software checksum and GSO_PARTIAL with GSO offload

This patch adds support for a software provided checksum and GSO_PARTIAL
segmentation support. With this we can offload UDP segmentation on devices
that only have partial support for tunnels.

Since we are no longer needing the hardware checksum we can drop the checks
in the segmentation code that were verifying if it was present.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Partially unroll handling of first segment and last segment
Alexander Duyck [Mon, 7 May 2018 18:08:40 +0000 (11:08 -0700)]
udp: Partially unroll handling of first segment and last segment

This patch allows us to take care of unrolling the first segment and the
last segment of the loop for processing the segmented skb. Part of the
motivation for this is that it makes it easier to process the fact that the
first fame and all of the frames in between should be mostly identical
in terms of header data, and the last frame has differences in the length
and partial checksum.

In addition I am dropping the header length calculation since we don't
really need it for anything but the last frame and it can be easily
obtained by just pulling the data_len and offset of tail from the transport
header.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Do not pass checksum as a parameter to GSO segmentation
Alexander Duyck [Mon, 7 May 2018 18:08:34 +0000 (11:08 -0700)]
udp: Do not pass checksum as a parameter to GSO segmentation

This patch is meant to allow us to avoid having to recompute the checksum
from scratch and have it passed as a parameter.

Instead of taking that approach we can take advantage of the fact that the
length that was used to compute the existing checksum is included in the
UDP header.

Finally to avoid the need to invert the result we can just call csum16_add
and csum16_sub directly. By doing this we can avoid a number of
instructions in the loop that is handling segmentation.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Do not pass MSS as parameter to GSO segmentation
Alexander Duyck [Mon, 7 May 2018 18:08:28 +0000 (11:08 -0700)]
udp: Do not pass MSS as parameter to GSO segmentation

There is no point in passing MSS as a parameter for for the GSO
segmentation call as it is already available via the shared info for the
skb itself.

Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoudp: Record gso_segs when supporting UDP segmentation offload
Alexander Duyck [Mon, 7 May 2018 18:08:22 +0000 (11:08 -0700)]
udp: Record gso_segs when supporting UDP segmentation offload

We need to record the number of segments that will be generated when this
frame is segmented. The expectation is that if gso_size is set then
gso_segs is set as well. Without this some drivers such as ixgbe get
confused if they attempt to offload this as they record 0 segments for the
entire packet instead of the correct value.

Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agodt-bindings: dsa: Remove unnecessary #address/#size-cells
Fabio Estevam [Mon, 7 May 2018 12:17:51 +0000 (09:17 -0300)]
dt-bindings: dsa: Remove unnecessary #address/#size-cells

If the example binding is used on a real dts file, the following DTC
warning is seen with W=1:

arch/arm/boot/dts/imx6q-b450v3.dtb: Warning (avoid_unnecessary_addr_size): /mdio-gpio/switch@0: unnecessary #address-cells/#size-cells without "ranges" or child "reg" property

Remove unnecessary #address-cells/#size-cells to improve the binding
document examples.

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: sfp: handle cases where neither BR, min nor BR, max is given
Antoine Tenart [Fri, 4 May 2018 15:21:03 +0000 (17:21 +0200)]
net: phy: sfp: handle cases where neither BR, min nor BR, max is given

When computing the bitrate using values read from an SFP module EEPROM,
we use the nominal BR plus BR,min and BR,max to determine the
boundaries. But in some cases BR,min and BR,max aren't provided, which
led the SFP code to end up having the nominal value for both the minimum
and maximum bitrate values. When using a passive cable, the nominal
value should be used as the maximum one, and there is no minimum one
so we should use 0.

Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bnxt_en-Fixes-for-net-next'
David S. Miller [Tue, 8 May 2018 14:14:22 +0000 (10:14 -0400)]
Merge branch 'bnxt_en-Fixes-for-net-next'

Michael Chan says:

====================
bnxt_en: Fixes for net-next.

This series includes a bug fix for a regression in firmware message polling
introduced recently on net-next.  There are 3 additional minor fixes for
unsupported link speed checking, VF MAC address handling, and setting
PHY eeprom length.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobnxt_en: Always forward VF MAC address to the PF.
Michael Chan [Tue, 8 May 2018 07:18:41 +0000 (03:18 -0400)]
bnxt_en: Always forward VF MAC address to the PF.

The current code already forwards the VF MAC address to the PF, except
in one case.  If the VF driver gets a valid MAC address from the firmware
during probe time, it will not forward the MAC address to the PF,
incorrectly assuming that the PF already knows the MAC address.  This
causes "ip link show" to show zero VF MAC addresses for this case.

This assumption is not correct.  Newer firmware remembers the VF MAC
address last used by the VF and provides it to the VF driver during
probe.  So we need to always forward the VF MAC address to the PF.

The forwarded MAC address may now be the PF assigned MAC address and so we
need to make sure we approve it for this case.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobnxt_en: Read phy eeprom A2h address only when optical diagnostics is supported.
Vasundhara Volam [Tue, 8 May 2018 07:18:40 +0000 (03:18 -0400)]
bnxt_en: Read phy eeprom A2h address only when optical diagnostics is supported.

For SFP+ modules, 0xA2 page is available only when Diagnostic Monitoring
Type [Address A0h, Byte 92] is implemented. Extend bnxt_get_module_info(),
to read optical diagnostics support at offset 92(0x5c) and set eeprom_len
length to ETH_MODULE_SFF_8436_LEN (to exclude A2 page), if dianostics is
not supported.

Also in bnxt_get_module_info(), module id is read from offset 0x5e which
is not correct. It was working by accident, as offset was not effective
without setting enables flag in the firmware request. SFP module id is
present at location 0. Fix this by removing the offset and read it
from location 0.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobnxt_en: Check unsupported speeds in bnxt_update_link() on PF only.
Michael Chan [Tue, 8 May 2018 07:18:39 +0000 (03:18 -0400)]
bnxt_en: Check unsupported speeds in bnxt_update_link() on PF only.

Only non-NPAR PFs need to actively check and manage unsupported link
speeds.  NPAR functions and VFs do not control the link speed and
should skip the unsupported speed detection logic, to avoid warning
messages from firmware rejecting the unsupported firmware calls.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobnxt_en: Fix firmware message delay loop regression.
Michael Chan [Tue, 8 May 2018 07:18:38 +0000 (03:18 -0400)]
bnxt_en: Fix firmware message delay loop regression.

A recent change to reduce delay granularity waiting for firmware
reponse has caused a regression.  With a tighter delay loop,
the driver may see the beginning part of the response faster.
The original 5 usec delay to wait for the rest of the message
is not long enough and some messages are detected as invalid.

Increase the maximum wait time from 5 usec to 20 usec.  Also, fix
the debug message that shows the total delay time for the response
when the message times out.  With the new logic, the delay time
is not fixed per iteration of the loop, so we define a macro to
show the total delay time.

Fixes: 9751e8e71487 ("bnxt_en: reduce timeout on initial HWRM calls")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet-next/hinic: add pci device ids for 25ge and 100ge card
Zhao Chen [Mon, 7 May 2018 13:21:57 +0000 (09:21 -0400)]
net-next/hinic: add pci device ids for 25ge and 100ge card

This patch adds PCI device IDs to support 25GE and 100GE card:

1. Add device id 0x0201 for HINIC 100GE dual port card.
2. Add device id 0x0200 for HINIC 25GE dual port card.
3. Macro of device id 0x1822 is modified for HINIC 25GE quad port card.

Signed-off-by: Zhao Chen <zhaochen6@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoflow_dissector: do not rely on implicit casts
Paolo Abeni [Mon, 7 May 2018 10:06:03 +0000 (12:06 +0200)]
flow_dissector: do not rely on implicit casts

This change fixes a couple of type mismatch reported by the sparse
tool, explicitly using the requested type for the offending arguments.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: core: rework basic flow dissection helper
Paolo Abeni [Fri, 4 May 2018 09:32:59 +0000 (11:32 +0200)]
net: core: rework basic flow dissection helper

When the core networking needs to detect the transport offset in a given
packet and parse it explicitly, a full-blown flow_keys struct is used for
storage.
This patch introduces a smaller keys store, rework the basic flow dissect
helper to use it, and apply this new helper where possible - namely in
skb_probe_transport_header(). The used flow dissector data structures
are renamed to match more closely the new role.

The above gives ~50% performance improvement in micro benchmarking around
skb_probe_transport_header() and ~30% around eth_get_headlen(), mostly due
to the smaller memset. Small, but measurable improvement is measured also
in macro benchmarking.

v1 -> v2: use the new helper in eth_get_headlen() and skb_get_poff(),
  as per DaveM suggestion

Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
David S. Miller [Tue, 8 May 2018 03:56:32 +0000 (23:56 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec-next

Minor conflict in ip_output.c, overlapping changes to
the body of an if() statement.

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'ipv6-misc'
David S. Miller [Tue, 8 May 2018 03:50:28 +0000 (23:50 -0400)]
Merge branch 'ipv6-misc'

Tariq Toukan says:

====================
net/ipv6 misc

This patchset contains two patches for net/ipv6.

Patch 1 is a trivial typo fix in documentation.
Patch 2 by Eran is a re-spin. It adds GRO support for IPv6 GRE tunnel,
this significantly improves performance in case GRO in native interface
is disabled.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv6/gre: Add GRO support
Eran Ben Elisha [Mon, 7 May 2018 07:45:27 +0000 (10:45 +0300)]
net: ipv6/gre: Add GRO support

Add GRO capability for IPv6 GRE tunnel and ip6erspan tap, via gro_cells
infrastructure.

Performance testing: 55% higher badwidth.
Measuring bandwidth of 1 thread IPv4 TCP traffic over IPv6 GRE tunnel
while GRO on the physical interface is disabled.
CPU: Intel Xeon E312xx (Sandy Bridge)
NIC: Mellanox Technologies MT27700 Family [ConnectX-4]
Before (GRO not working in tunnel) : 2.47 Gbits/sec
After  (GRO working in tunnel)     : 3.85 Gbits/sec

Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv6: Fix typo in ipv6_find_hdr() documentation
Tariq Toukan [Mon, 7 May 2018 07:45:26 +0000 (10:45 +0300)]
net: ipv6: Fix typo in ipv6_find_hdr() documentation

Fix 'an' into 'and', and use a comma instead of a period.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'qed-Add-support-for-new-multi-partitioning-modes'
David S. Miller [Tue, 8 May 2018 03:46:11 +0000 (23:46 -0400)]
Merge branch 'qed-Add-support-for-new-multi-partitioning-modes'

Sudarsana Reddy Kalluru says:

====================
qed*: Add support for new multi partitioning modes.

The patch series simplifies the multi function (MF) mode implementation of
qed/qede drivers, and adds support for new MF modes.

Please consider applying it to net-next branch.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Add support for Unified Fabric Port.
Sudarsana Reddy Kalluru [Sun, 6 May 2018 01:43:02 +0000 (18:43 -0700)]
qed: Add support for Unified Fabric Port.

This patch adds driver changes for supporting the Unified Fabric Port
(UFP). This is a new paritioning mode wherein MFW provides the set of
parameters to be used by the device such as traffic class, outer-vlan
tag value, priority type etc. Drivers receives this info via notifications
from mfw and configures the hardware accordingly.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Add support for multi function mode with 802.1ad tagging.
Sudarsana Reddy Kalluru [Sun, 6 May 2018 01:43:01 +0000 (18:43 -0700)]
qed: Add support for multi function mode with 802.1ad tagging.

The patch adds support for new Multi function mode wherein the traffic
classification is done based on the 802.1ad tagging and the outer vlan tag
provided by the management firmware.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Remove unused data member 'is_mf_default'.
Sudarsana Reddy Kalluru [Sun, 6 May 2018 01:43:00 +0000 (18:43 -0700)]
qed: Remove unused data member 'is_mf_default'.

The data member 'is_mf_default' is not used by the qed/qede drivers,
removing the same.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed*: Refactor mf_mode to consist of bits.
Sudarsana Reddy Kalluru [Sun, 6 May 2018 01:42:59 +0000 (18:42 -0700)]
qed*: Refactor mf_mode to consist of bits.

`mf_mode' field indicates the multi-partitioning mode the device is
configured to. This method doesn't scale very well, adding a new MF mode
requires going over all the existing conditions, and deciding whether those
are needed for the new mode or not.
The patch defines a set of bit-fields for modes which are derived according
to the mode info shared by the MFW and all the configuration would be made
according to those. To add a new mode, there would be a single place where
we'll need to go and choose which bits apply and which don't.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/9p: correct the variable name in v9fs_get_trans_by_name() comment
Sun Lianwen [Sat, 5 May 2018 03:29:16 +0000 (11:29 +0800)]
net/9p: correct the variable name in v9fs_get_trans_by_name() comment

The v9fs_get_trans_by_name(char *s) variable name is not "name" but "s".

Signed-off-by: Sun Lianwen <sunlw.fnst@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agovlan: correct the file path in vlan_dev_change_flags() comment
Sun Lianwen [Sat, 5 May 2018 01:08:18 +0000 (09:08 +0800)]
vlan: correct the file path in vlan_dev_change_flags() comment

The vlan_flags enum is defined in include/uapi/linux/if_vlan.h file.
not in include/linux/if_vlan.h file.

Signed-off-by: Sun Lianwen <sunlw.fnst@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
David S. Miller [Tue, 8 May 2018 03:35:08 +0000 (23:35 -0400)]
Merge git://git./linux/kernel/git/bpf/bpf-next

Minor conflict, a CHECK was placed into an if() statement
in net-next, whilst a newline was added to that CHECK
call in 'net'.  Thanks to Daniel for the merge resolution.

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: support use of ethtool to set link speed of CN23XX-225 cards
Weilin Chang [Fri, 4 May 2018 18:07:19 +0000 (11:07 -0700)]
liquidio: support use of ethtool to set link speed of CN23XX-225 cards

Support setting the link speed of CN23XX-225 cards (which can do 25Gbps or
10Gbps) via ethtool_ops.set_link_ksettings.

Also fix the function assigned to ethtool_ops.get_link_ksettings to use the
new link_ksettings api completely (instead of partially via
ethtool_convert_legacy_u32_to_link_mode).

Signed-off-by: Weilin Chang <weilin.chang@cavium.com>
Acked-by: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch '3c59x-patches-and-the-removal-of-an-unused-function'
David S. Miller [Tue, 8 May 2018 03:25:25 +0000 (23:25 -0400)]
Merge branch '3c59x-patches-and-the-removal-of-an-unused-function'

Sebastian Andrzej Siewior says:

====================
3c59x patches and the removal of an unused function

The first patch removes an unused function. The goal of remaining three
patches is to get rid of the local_irq_save() usage in the driver which
benefits -RT.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: 3com: 3c59x: irq save variant of ISR
Anna-Maria Gleixner [Fri, 4 May 2018 15:17:49 +0000 (17:17 +0200)]
net: 3com: 3c59x: irq save variant of ISR

When vortex_boomerang_interrupt() is invoked from vortex_tx_timeout() or
poll_vortex() interrupts must be disabled. This detaches the interrupt
disable logic from locking which requires patching for PREEMPT_RT.

The advantage of avoiding spin_lock_irqsave() in the interrupt handler is
minimal, but converting it removes all the extra code for callers which
come not from interrupt context.

Cc: Steffen Klassert <klassert@mathematik.tu-chemnitz.de>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: 3com: 3c59x: Pull locking out of ISR
Anna-Maria Gleixner [Fri, 4 May 2018 15:17:48 +0000 (17:17 +0200)]
net: 3com: 3c59x: Pull locking out of ISR

Locking is done in the same way in _vortex_interrupt() and
_boomerang_interrupt(). To prevent duplication, move the locking into the
calling vortex_boomerang_interrupt() function.

No functional change.

Cc: Steffen Klassert <klassert@mathematik.tu-chemnitz.de>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: 3com: 3c59x: Move boomerang/vortex conditional into function
Anna-Maria Gleixner [Fri, 4 May 2018 15:17:47 +0000 (17:17 +0200)]
net: 3com: 3c59x: Move boomerang/vortex conditional into function

If vp->full_bus_master_tx is set, vp->full_bus_master_rx is set as well
(see vortex_probe1()). Therefore the conditionals for the decision if
boomerang or vortex ISR is executed have the same result. Instead of
repeating the explicit conditional execution of the boomerang/vortex ISR,
move it into an own function.

No functional change.

Cc: Steffen Klassert <klassert@mathematik.tu-chemnitz.de>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: u64_stats_sync: Remove functions without user
Anna-Maria Gleixner [Fri, 4 May 2018 15:17:46 +0000 (17:17 +0200)]
net: u64_stats_sync: Remove functions without user

Commit 67db3e4bfbc9 ("tcp: no longer hold ehash lock while calling
tcp_get_info()") removes the only users of u64_stats_update_end/begin_raw()
without removing the function in header file.

Remove no longer used functions.

Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests: net: add udpgso* to TEST_GEN_FILES
Anders Roxell [Fri, 4 May 2018 09:17:25 +0000 (11:17 +0200)]
selftests: net: add udpgso* to TEST_GEN_FILES

The generated files udpgso* shouldn't be part of TEST_PROGS, they are
used by udpgso.sh and udpgsp_bench.sh. They should be added to the
TEST_GEN_FILES to get installed without being added to the main
run_kselftest.sh script.

Fixes: 3a687bef148d ("selftests: udp gso benchmark")
Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Mon, 7 May 2018 01:51:37 +0000 (21:51 -0400)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
Netfilter/IPVS updates for net-next

The following patchset contains Netfilter/IPVS updates for your net-next
tree, more relevant updates in this batch are:

1) Add Maglev support to IPVS. Moreover, store lastest server weight in
   IPVS since this is needed by maglev, patches from from Inju Song.

2) Preparation works to add iptables flowtable support, patches
   from Felix Fietkau.

3) Hand over flows back to conntrack slow path in case of TCP RST/FIN
   packet is seen via new teardown state, also from Felix.

4) Add support for extended netlink error reporting for nf_tables.

5) Support for larger timeouts that 23 days in nf_tables, patch from
   Florian Westphal.

6) Always set an upper limit to dynamic sets, also from Florian.

7) Allow number generator to make map lookups, from Laura Garcia.

8) Use hash_32() instead of opencode hashing in IPVS, from Vicent Bernat.

9) Extend ip6tables SRH match to support previous, next and last SID,
   from Ahmed Abdelsalam.

10) Move Passive OS fingerprint nf_osf.c, from Fernando Fernandez.

11) Expose nf_conntrack_max through ctnetlink, from Florent Fourcot.

12) Several housekeeping patches for xt_NFLOG, x_tables and ebtables,
   from Taehee Yoo.

13) Unify meta bridge with core nft_meta, then make nft_meta built-in.
   Make rt and exthdr built-in too, again from Florian.

14) Missing initialization of tbl->entries in IPVS, from Cong Wang.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonetfilter: nft_dynset: fix timeout updates on 32bit
Florian Westphal [Fri, 27 Apr 2018 20:37:43 +0000 (22:37 +0200)]
netfilter: nft_dynset: fix timeout updates on 32bit

This must now use a 64bit jiffies value, else we set
a bogus timeout on 32bit.

Fixes: 8e1102d5a1596 ("netfilter: nf_tables: support timeouts larger than 23 days")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>