GitHub/LineageOS/android_kernel_motorola_exynos9610.git
8 years agotipc: change node timer unit from jiffies to ms
Jon Paul Maloy [Wed, 8 Jun 2016 16:00:05 +0000 (12:00 -0400)]
tipc: change node timer unit from jiffies to ms

The node keepalive interval is recalculated at each timer expiration
to catch any changes in the link tolerance, and stored in a field in
struct tipc_node. We use jiffies as unit for the stored value.

This is suboptimal, because it makes the calculation unnecessary
complex, including two unit conversions. The conversions also lead to
a rounding error that causes the link "abort limit" to be 3 in the
normal case, instead of 4, as intended. This again leads to unnecessary
link resets when the network is pushed close to its limit, e.g., in an
environment with hundreds of nodes or namesapces.

In this commit, we do instead let the keepalive value be calculated and
stored in milliseconds, so that there is only one conversion and the
rounding error is eliminated.

We also remove a redundant "keepalive" field in struct tipc_link. This
is remnant from the previous implementation.

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>
8 years agotipc: correct error in node fsm
Jon Paul Maloy [Wed, 8 Jun 2016 16:00:04 +0000 (12:00 -0400)]
tipc: correct error in node fsm

commit 88e8ac7000dc ("tipc: reduce transmission rate of reset messages
when link is down") revealed a flaw in the node FSM, as defined in
the log of commit 66996b6c47ed ("tipc: extend node FSM").

We see the following scenario:
1: Node B receives a RESET message from node A before its link endpoint
   is fully up, i.e., the node FSM is in state SELF_UP_PEER_COMING. This
   event will not change the node FSM state, but the (distinct) link FSM
   will move to state RESETTING.
2: As an effect of the previous event, the local endpoint on B will
   declare node A lost, and post the event SELF_DOWN to the its node
   FSM. This moves the FSM state to SELF_DOWN_PEER_LEAVING, meaning
   that no messages will be accepted from A until it receives another
   RESET message that confirms that A's endpoint has been reset. This
   is  wasteful, since we know this as a fact already from the first
   received RESET, but worse is that the link instance's FSM has not
   wasted this information, but instead moved on to state ESTABLISHING,
   meaning that it repeatedly sends out ACTIVATE messages to the reset
   peer A.
3: Node A will receive one of the ACTIVATE messages, move its link FSM
   to state ESTABLISHED, and start repeatedly sending out STATE messages
   to node B.
4: Node B will consistently drop these messages, since it can only accept
   accept a RESET according to its node FSM.
5: After four lost STATE messages node A will reset its link and start
   repeatedly sending out RESET messages to B.
6: Because of the reduced send rate for RESET messages, it is very
   likely that A will receive an ACTIVATE (which is sent out at a much
   higher frequency) before it gets the chance to send a RESET, and A
   may hence quickly move back to state ESTABLISHED and continue sending
   out STATE messages, which will again be dropped by B.
7: GOTO 5.
8: After having repeated the cycle 5-7 a number of times, node A will
   by chance get in between with sending a RESET, and the situation is
   resolved.

Unfortunately, we have seen that it may take a substantial amount of
time before this vicious loop is broken, sometimes in the order of
minutes.

We correct this by making a small correction to the node FSM: When a
node in state SELF_UP_PEER_COMING receives a SELF_DOWN event, it now
moves directly back to state SELF_DOWN_PEER_DOWN, instead of as now
SELF_DOWN_PEER_LEAVING. This is logically consistent, since we don't
need to wait for RESET confirmation from of an endpoint that we alread
know has been reset. It also means that node B in the scenario above
will not be dropping incoming STATE messages, and the link can come up
immediately.

Finally, a symmetry comparison reveals that the  FSM has a similar
error when receiving the event PEER_DOWN in state PEER_UP_SELF_COMING.
Instead of moving to PERR_DOWN_SELF_LEAVING, it should move directly
to SELF_DOWN_PEER_DOWN. Although we have never seen any negative effect
of this logical error, we choose fix this one, too.

The node FSM looks as follows after those changes:

                           +----------------------------------------+
                           |                           PEER_DOWN_EVT|
                           |                                        |
  +------------------------+----------------+                       |
  |SELF_DOWN_EVT           |                |                       |
  |                        |                |                       |
  |              +-----------+          +-----------+               |
  |              |NODE_      |          |NODE_      |               |
  |   +----------|FAILINGOVER|<---------|SYNCHING   |-----------+   |
  |   |SELF_     +-----------+ FAILOVER_+-----------+   PEER_   |   |
  |   |DOWN_EVT   |          A BEGIN_EVT  A         |   DOWN_EVT|   |
  |   |           |          |            |         |           |   |
  |   |           |          |            |         |           |   |
  |   |           |FAILOVER_ |FAILOVER_   |SYNCH_   |SYNCH_     |   |
  |   |           |END_EVT   |BEGIN_EVT   |BEGIN_EVT|END_EVT    |   |
  |   |           |          |            |         |           |   |
  |   |           |          |            |         |           |   |
  |   |           |         +--------------+        |           |   |
  |   |           +-------->|   SELF_UP_   |<-------+           |   |
  |   |   +-----------------|   PEER_UP    |----------------+   |   |
  |   |   |SELF_DOWN_EVT    +--------------+   PEER_DOWN_EVT|   |   |
  |   |   |                    A        A                   |   |   |
  |   |   |                    |        |                   |   |   |
  |   |   |         PEER_UP_EVT|        |SELF_UP_EVT        |   |   |
  |   |   |                    |        |                   |   |   |
  V   V   V                    |        |                   V   V   V
+------------+       +-----------+    +-----------+       +------------+
|SELF_DOWN_  |       |SELF_UP_   |    |PEER_UP_   |       |PEER_DOWN   |
|PEER_LEAVING|       |PEER_COMING|    |SELF_COMING|       |SELF_LEAVING|
+------------+       +-----------+    +-----------+       +------------+
       |               |       A        A       |                |
       |               |       |        |       |                |
       |       SELF_   |       |SELF_   |PEER_  |PEER_           |
       |       DOWN_EVT|       |UP_EVT  |UP_EVT |DOWN_EVT        |
       |               |       |        |       |                |
       |               |       |        |       |                |
       |               |    +--------------+    |                |
       |PEER_DOWN_EVT  +--->|  SELF_DOWN_  |<---+   SELF_DOWN_EVT|
       +------------------->|  PEER_DOWN   |<--------------------+
                            +--------------+

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>
8 years agoMerge branch 'dsa-misc-improvements'
David S. Miller [Wed, 8 Jun 2016 18:23:42 +0000 (11:23 -0700)]
Merge branch 'dsa-misc-improvements'

Florian Fainelli says:

====================
net: dsa: misc improvements

This patch series builds on top of Andrew's "New DSA bind, switches as devices"
patch set and does the following:

- add a few helper functions/goodies for net/dsa/dsa2.c to be as close as possible
  from net/dsa/dsa.c in terms of what drivers can expect, in particular the slave
  MDIO bus and the enabled_port_mask and phy_mii_mask

- fix the CPU port ethtools ops to work in a multiple tree setup since we can
  no longer assume a single tree is supported

- make the bcm_sf2 driver register its own MDIO bus, yet assign it to
  ds->slave_mii_bus for everything to work in net/dsa/slave.c wrt. PHY probing,
  this is a tad cleaner than what we have now

Changes in v2:

Most of the previous patches have been dropped to just keep the relevant ones
now.

Changes in v3:
- split the addition of the slave MII bus as a separate patch
- properly unwind all operations at the right place and right time (ethtool ops,
  slave MDIO bus
- fixed a few typos here and there

Changes in v4:
- removed superfluous dst agrument to dsa_cpu_port_ethtool_{setup,restore}
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: bcm_sf2: Register our slave MDIO bus
Florian Fainelli [Tue, 7 Jun 2016 23:32:43 +0000 (16:32 -0700)]
net: dsa: bcm_sf2: Register our slave MDIO bus

Register a slave MDIO bus which allows us to divert problematic
read/writes towards conflicting pseudo-PHY address (30). Do no longer
rely on DSA's slave_mii_bus, but instead provide our own implementation
which offers more flexibility as to what to do, and when to register it.

We need to register it by the time we are able to get access to our
memory mapped registers, which is not until drv->setup() time. In order
to avoid forward declarations, we need to re-order the function bodies a
bit.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Initialize CPU port ethtool ops per tree
Florian Fainelli [Tue, 7 Jun 2016 23:32:42 +0000 (16:32 -0700)]
net: dsa: Initialize CPU port ethtool ops per tree

Now that we can properly support multiple distinct trees in the system,
using a global variable: dsa_cpu_port_ethtool_ops is getting clobbered
as soon as the second switch tree gets probed, and we don't want that.

We need to move this to be dynamically allocated, and since we can't
really be comparing addresses anymore to determine first time
initialization versus any other times, just move this to dsa.c and
dsa2.c where the remainder of the dst/ds initialization happens.

The operations teardown restores the master netdev's ethtool_ops to its
original ethtool_ops pointer (typically within the Ethernet driver)

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Add initialization helper for CPU port ethtool_ops
Florian Fainelli [Tue, 7 Jun 2016 23:32:41 +0000 (16:32 -0700)]
net: dsa: Add initialization helper for CPU port ethtool_ops

Add a helper function: dsa_cpu_port_ethtool_init() which initializes a
custom ethtool_ops structure with custom DSA ethtool operations for CPU
ports. This is a preliminary change to move the initialization outside
of net/dsa/slave.c.

Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Provide a slave MII bus if needed
Florian Fainelli [Tue, 7 Jun 2016 23:32:40 +0000 (16:32 -0700)]
net: dsa: Provide a slave MII bus if needed

Mimic what net/dsa/dsa.c does and provide a slave MII bus by default
which will be created if the driver implements a phy_read method.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Initialize ds->enabled_port_mask and ds->phys_mii_mask
Florian Fainelli [Tue, 7 Jun 2016 23:32:39 +0000 (16:32 -0700)]
net: dsa: Initialize ds->enabled_port_mask and ds->phys_mii_mask

Some drivers rely on these two bitmasks to contain the correct values
for them to successfully probe and initialize at drv->setup() time,
calculate correct values to put in both masks as early as possible in
dsa_get_ports_dn().

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Provide unique DSA slave MII bus names
Florian Fainelli [Tue, 7 Jun 2016 23:32:38 +0000 (16:32 -0700)]
net: dsa: Provide unique DSA slave MII bus names

In case we have multiples trees and switches with the same index, we
need to add another discriminating id: the switch tree.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: sched: fix missing doc annotations
Eric Dumazet [Wed, 8 Jun 2016 14:22:49 +0000 (07:22 -0700)]
net: sched: fix missing doc annotations

"make htmldocs" complains otherwise:

.//net/core/gen_stats.c:168: warning: No description found for parameter 'running'
.//include/linux/netdevice.h:1867: warning: No description found for parameter 'qdisc_running_key'

Fixes: f9eb8aea2a1e ("net_sched: transform qdisc running bit into a seqcount")
Fixes: edb09eb17ed8 ("net: sched: do not acquire qdisc spinlock in qdisc/class stats dump")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: Reduce queue allocation to one in kdump kernel
Hariprasad Shenai [Wed, 8 Jun 2016 12:39:08 +0000 (18:09 +0530)]
net: Reduce queue allocation to one in kdump kernel

When in kdump kernel, reduce memory usage by only using a single Queue
Set for multiqueue devices. So make netif_get_num_default_rss_queues()
return one, when in kdump kernel.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'qed-dcbnl'
David S. Miller [Wed, 8 Jun 2016 18:11:00 +0000 (11:11 -0700)]
Merge branch 'qed-dcbnl'

Sudarsana Reddy Kalluru says:

====================
qed/qede support for dcbnl.

This series adds the dcbnl functionality to the driver. Patch (1) adds
the qed infrastucture for querying/configuring the dcbx parameters.
Patch (2) adds the qed infrastructure for dcbnl APIs. And patch (3)
adds the qede support for dcbnl.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqede: Add dcbnl support.
Sudarsana Reddy Kalluru [Wed, 8 Jun 2016 10:22:12 +0000 (06:22 -0400)]
qede: Add dcbnl support.

This patch adds the interfaces for ieee/cee dcbnl callbacks and registers
them with the kernel.

Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Add dcbnl support.
Sudarsana Reddy Kalluru [Wed, 8 Jun 2016 10:22:11 +0000 (06:22 -0400)]
qed: Add dcbnl support.

This patch adds the implementation for both cee/ieee dcbnl callbacks by
using the qed query/config APIs.

Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Add support for query/config dcbx.
Sudarsana Reddy Kalluru [Wed, 8 Jun 2016 10:22:10 +0000 (06:22 -0400)]
qed: Add support for query/config dcbx.

Query API reads the dcbx data from the device shared memory and return it
to the caller. The config API configures the user provided dcbx values on
the device, and initiates the dcbx negotiation with the peer.

Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/qe: Do not prefix header guard with CONFIG_
Andreas Ziegler [Wed, 8 Jun 2016 09:36:56 +0000 (11:36 +0200)]
fsl/qe: Do not prefix header guard with CONFIG_

The CONFIG_ prefix should only be used for options which
can be configured through Kconfig and not for guarding headers.

Signed-off-by: Andreas Ziegler <andreas.ziegler@fau.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers/net/fsl_ucc: Do not prefix header guard with CONFIG_
Andreas Ziegler [Wed, 8 Jun 2016 09:40:28 +0000 (11:40 +0200)]
drivers/net/fsl_ucc: Do not prefix header guard with CONFIG_

The CONFIG_ prefix should only be used for options which
can be configured through Kconfig and not for guarding headers.

Signed-off-by: Andreas Ziegler <andreas.ziegler@fau.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoila: Perform only one translation in forwarding path
Tom Herbert [Tue, 7 Jun 2016 23:09:44 +0000 (16:09 -0700)]
ila: Perform only one translation in forwarding path

When setting up ILA in a router we noticed that the the encapsulation
is invoked twice: once in the route input path and again upon route
output. To resolve this we add a flag set_csum_neutral for the
ila_update_ipv6_locator. If this flag is set and the checksum
neutral bit is also set we assume that checksum-neutral translation
has already been performed and take no further action. The
flag is set only in ila_output path. The flag is not set for ila_input and
ila_xlat.

Tested:

Used 3 netns to set to emulate a router and two hosts. The router
translates SIR addresses between the two destinations in other two netns.
Verified ping and netperf are functional.

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotcp: accept RST if SEQ matches right edge of right-most SACK block
Pau Espin Pedrol [Tue, 7 Jun 2016 14:30:34 +0000 (16:30 +0200)]
tcp: accept RST if SEQ matches right edge of right-most SACK block

RFC 5961 advises to only accept RST packets containing a seq number
matching the next expected seq number instead of the whole receive
window in order to avoid spoofing attacks.

However, this situation is not optimal in the case SACK is in use at the
time the RST is sent. I recently run into a scenario in which packet
losses were high while uploading data to a server, and userspace was
willing to frequently terminate connections by sending a RST. In
this case, the ACK sent on the receiver side (rcv_nxt) is frozen waiting
for a lost packet retransmission and SACK blocks are used to let the
client continue uploading data. At some point later on, the client sends
the RST (snd_nxt), which matches the next expected seq number of the
right-most SACK block on the receiver side which is going forward
receiving data.

In this scenario, as RFC 5961 defines, the RST SEQ doesn't match the
frozen main ACK at receiver side and thus gets dropped and a challenge
ACK is sent, which gets usually lost due to network conditions. The main
consequence is that the connection stays alive for a while even if it
made sense to accept the RST. This can get really bad if lots of
connections like this one are created in few seconds, allocating all the
resources of the server easily.

For security reasons, not all SACK blocks are checked (there could be a
big amount of SACK blocks => acceptable SEQ numbers). Furthermore, it
wouldn't make sense to check for RST in blocks other than the right-most
received one because the sender is not expected to be sending new data
after the RST. For simplicity, only up to the 4 most recently updated
SACK blocks (selective_acks[4] field) are compared to find the
right-most block, as usually those are the ones with bigger probability
to contain it.

This patch was tested in a 3.18 kernel and probed to improve the
situation in the scenario described above.

Signed-off-by: Pau Espin Pedrol <pau.espin@tessares.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Tested-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: potential overflow in qed_cxt_src_t2_alloc()
Dan Carpenter [Tue, 7 Jun 2016 12:04:16 +0000 (15:04 +0300)]
qed: potential overflow in qed_cxt_src_t2_alloc()

In the current code "ent_per_page" could be more than "conn_num" making
"conn_num" negative after the subtraction.  In the next iteration
through the loop then the negative is treated as a very high positive
meaning we don't put a limit on "ent_num".  It could lead to memory
corruption.

Fixes: dbb799c39717 ('qed: Initialize hardware for new protocols')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'vrf-local'
David S. Miller [Wed, 8 Jun 2016 07:25:38 +0000 (00:25 -0700)]
Merge branch 'vrf-local'

David Ahern says:

====================
net: vrf: Add support for local traffic to local addresses

Add support for locally originated traffic to VRF-local addresses,
be it addresses on enslaved devices or addresses on the VRF device:

$ ip addr show dev red
33: red: <NOARP,MASTER,UP,LOWER_UP> mtu 65536 qdisc pfifo_fast state UP group default qlen 1000
    link/ether be:00:53:b5:e4:25 brd ff:ff:ff:ff:ff:ff
    inet 1.1.1.1/32 scope global red
       valid_lft forever preferred_lft forever
    inet6 1111:1::1/128 scope global
       valid_lft forever preferred_lft forever

$ ip addr show dev eth1
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
    link/ether 02:e0:f9:79:34:bd brd ff:ff:ff:ff:ff:ff
    inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
       valid_lft forever preferred_lft forever
    inet6 2100:1::1/120 scope global
       valid_lft forever preferred_lft forever
    inet6 fe80::e0:f9ff:fe79:34bd/64 scope link
       valid_lft forever preferred_lft forever

$ ping -c1 -I red 10.100.1.1
    ping: Warning: source address might be selected on device other than red.
    PING 10.100.1.1 (10.100.1.1) from 10.100.1.1 red: 56(84) bytes of data.
    64 bytes from 10.100.1.1: icmp_seq=1 ttl=64 time=0.057 ms

$ ping -c1 -I red 1.1.1.1
PING 1.1.1.1 (1.1.1.1) from 1.1.1.1 red: 56(84) bytes of data.
64 bytes from 1.1.1.1: icmp_seq=1 ttl=64 time=0.136 ms

--- 1.1.1.1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.136/0.136/0.136/0.000 ms

$ ping6 -c1 -I red  2100:1::1
ping6: Warning: source address might be selected on device other than red.
PING 2100:1::1(2100:1::1) from 2100:1::1 red: 56 data bytes
64 bytes from 2100:1::1: icmp_seq=1 ttl=64 time=0.167 ms

--- 2100:1::1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.167/0.167/0.167/0.000 ms

$ ping6 -c1 -I red 1111::1
PING 1111::1(1111::1) from 1111:1::1 red: 56 data bytes
64 bytes from 1111::1: icmp_seq=1 ttl=64 time=0.187 ms

--- 1111::1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.187/0.187/0.187/0.000 ms

This change also enables use of loopback address on the VRF device:
$ ip addr add dev red 127.0.0.1/8

$ ping -c1 -I red 127.0.0.1
PING 127.0.0.1 (127.0.0.1) from 127.0.0.1 red: 56(84) bytes of data.
64 bytes from 127.0.0.1: icmp_seq=1 ttl=64 time=0.058 ms
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: ipv6 support for local traffic to local addresses
David Ahern [Tue, 7 Jun 2016 03:50:40 +0000 (20:50 -0700)]
net: vrf: ipv6 support for local traffic to local addresses

Add support for locally originated traffic to VRF-local IPv6 addresses.
Similar to IPv4 a local dst is set on the skb and the packet is
reinserted with a call to netif_rx. With this patch, ping, tcp and udp
packets to a local IPv6 address are successfully routed:

    $ ip addr show dev eth1
    4: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
        link/ether 02:e0:f9:1c:b9:74 brd ff:ff:ff:ff:ff:ff
        inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
           valid_lft forever preferred_lft forever
        inet6 2100:1::1/120 scope global
           valid_lft forever preferred_lft forever
        inet6 fe80::e0:f9ff:fe1c:b974/64 scope link
           valid_lft forever preferred_lft forever

    $ ping6 -c1 -I red 2100:1::1
    ping6: Warning: source address might be selected on device other than red.
    PING 2100:1::1(2100:1::1) from 2100:1::1 red: 56 data bytes
    64 bytes from 2100:1::1: icmp_seq=1 ttl=64 time=0.098 ms

ip6_input is exported so the VRF driver can use it for the dst input
function. The dst_alloc function for IPv4 defaults to setting the input and
output functions; IPv6's does not. VRF does not need to duplicate the Rx path
so just export the ipv6 input function.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: ipv4 support for local traffic to local addresses
David Ahern [Tue, 7 Jun 2016 03:50:39 +0000 (20:50 -0700)]
net: vrf: ipv4 support for local traffic to local addresses

Add support for locally originated traffic to VRF-local addresses. If
destination device for an skb is the loopback or VRF device then set
its dst to a local version of the VRF cached dst_entry and call netif_rx
to insert the packet onto the rx queue - similar to what is done for
loopback. This patch handles IPv4 support; follow on patch handles IPv6.

With this patch, ping, tcp and udp packets to a local IPv4 address are
successfully routed:

    $ ip addr show dev eth1
    4: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
        link/ether 02:e0:f9:1c:b9:74 brd ff:ff:ff:ff:ff:ff
        inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
           valid_lft forever preferred_lft forever
        inet6 2100:1::1/120 scope global
           valid_lft forever preferred_lft forever
        inet6 fe80::e0:f9ff:fe1c:b974/64 scope link
           valid_lft forever preferred_lft forever

    $ ping -c1 -I red 10.100.1.1
    ping: Warning: source address might be selected on device other than red.
    PING 10.100.1.1 (10.100.1.1) from 10.100.1.1 red: 56(84) bytes of data.
    64 bytes from 10.100.1.1: icmp_seq=1 ttl=64 time=0.057 ms

This patch also enables use of IPv4 loopback address on the VRF device:
    $ ip addr add dev red 127.0.0.1/8

    $ ping -c1 -I red 127.0.0.1
    PING 127.0.0.1 (127.0.0.1) from 127.0.0.1 red: 56(84) bytes of data.
    64 bytes from 127.0.0.1: icmp_seq=1 ttl=64 time=0.058 ms

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: Minor refactoring for local address patches
David Ahern [Tue, 7 Jun 2016 03:50:38 +0000 (20:50 -0700)]
net: vrf: Minor refactoring for local address patches

Move the stripping of the ethernet header from is_ip_tx_frame into the
ipv4 and ipv6 outbound functions and collapse vrf_send_v4_prep into
vrf_process_v4_outbound.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogue: Implement direction IP encapsulation
Tom Herbert [Mon, 6 Jun 2016 23:06:02 +0000 (16:06 -0700)]
gue: Implement direction IP encapsulation

This patch implements direct encapsulation of IPv4 and IPv6 packets
in UDP. This is done a version "1" of GUE and as explained in I-D
draft-ietf-nvo3-gue-03.

Changes here are only in the receive path, fou with IPxIPx already
supports the transmit side. Both the normal receive path and
GRO path are modified to check for GUE version and check for
IP version in the case that GUE version is "1".

Tested:

IPIP with direct GUE encap
  1 TCP_STREAM
    4530 Mbps
  200 TCP_RR
    1297625 tps
    135/232/444 90/95/99% latencies

IP4IP6 with direct GUE encap
  1 TCP_STREAM
    4903 Mbps
  200 TCP_RR
    1184481 tps
    149/253/473 90/95/99% latencies

IP6IP6 direct GUE encap
  1 TCP_STREAM
   5146 Mbps
  200 TCP_RR
    1202879 tps
    146/251/472 90/95/99% latencies

SIT with direct GUE encap
  1 TCP_STREAM
    6111 Mbps
  200 TCP_RR
    1250337 tps
    139/241/467 90/95/99% latencies

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'net-sched-fast-stats'
David S. Miller [Tue, 7 Jun 2016 23:37:14 +0000 (16:37 -0700)]
Merge branch 'net-sched-fast-stats'

Eric Dumazet says:

====================
net: sched: faster stats gathering

A while back, I sent one RFC patch using lockless stats gathering
on 64bit arches.

This patch series does it more cleanly, using a seqcount.

Since qdisc/class stats are written at dequeue() time,
we can ask the dequeue to change the seqcount, so that
stats readers can avoid taking the root qdisc lock,
and instead the typical read_seqcount_{begin|retry} guarded
loop.

This does not change fast path costs, as the seqcount
increments are not more expensive than the bit manipulation,
and allows readers to not freeze the fast path anymore.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: sched: do not acquire qdisc spinlock in qdisc/class stats dump
Eric Dumazet [Mon, 6 Jun 2016 16:37:16 +0000 (09:37 -0700)]
net: sched: do not acquire qdisc spinlock in qdisc/class stats dump

Large tc dumps (tc -s {qdisc|class} sh dev ethX) done by Google BwE host
agent [1] are problematic at scale :

For each qdisc/class found in the dump, we currently lock the root qdisc
spinlock in order to get stats. Sampling stats every 5 seconds from
thousands of HTB classes is a challenge when the root qdisc spinlock is
under high pressure. Not only the dumps take time, they also slow
down the fast path (queue/dequeue packets) by 10 % to 20 % in some cases.

An audit of existing qdiscs showed that sch_fq_codel is the only qdisc
that might need the qdisc lock in fq_codel_dump_stats() and
fq_codel_dump_class_stats()

In v2 of this patch, I now use the Qdisc running seqcount to provide
consistent reads of packets/bytes counters, regardless of 32/64 bit arches.

I also changed rate estimators to use the same infrastructure
so that they no longer need to lock root qdisc lock.

[1]
http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/43838.pdf

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Kevin Athey <kda@google.com>
Cc: Xiaotian Pei <xiaotian@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet_sched: transform qdisc running bit into a seqcount
Eric Dumazet [Mon, 6 Jun 2016 16:37:15 +0000 (09:37 -0700)]
net_sched: transform qdisc running bit into a seqcount

Instead of using a single bit (__QDISC___STATE_RUNNING)
in sch->__state, use a seqcount.

This adds lockdep support, but more importantly it will allow us
to sample qdisc/class statistics without having to grab qdisc root lock.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'be2net-noncrit-fixes'
David S. Miller [Tue, 7 Jun 2016 23:18:20 +0000 (16:18 -0700)]
Merge branch 'be2net-noncrit-fixes'

Sathya Perla says:

====================
be2net: patch set

Hi David, the following patch set contains three non-critical fixes that
can go into the net-next tree.

Patch 1 fixes the logic for provisioning queue pairs on VFs to take into
account the limit on number of TXQs too as in some profiles the number
of TXQs is less than that of RXQs.

Patch 2 enables WoL support from shutdown on Skyhawk.

Patch 3 enhances the logic for provisioning queue pairs on VFs on
SR-IOV over multi-partition configs. Each PF (partition) on a port has to
compute the number of RSS tables it's VFs can use.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobe2net: Fix provisioning of RSS for VFs in multi-partition configurations
Somnath Kotur [Mon, 6 Jun 2016 11:22:10 +0000 (07:22 -0400)]
be2net: Fix provisioning of RSS for VFs in multi-partition configurations

Currently, we do not distribute queue resources to enable RSS for VFs
in multi-channel/partition configurations.
Fix this by having each PF(SRIOV capable) calculate it's share of the
15 RSS Policy Tables available per port before provisioning resources for
all the VFs.
This  proportional share calculation is done based on division of the
PF's MAX VFs with the Total MAX VFs on that port. It also needs to
learn about the no: of NIC PFs on the port and subtract that from
the 15 RSS Policy Tables on the port.

Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: Sathya Perla <sathya.perla@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobe2net: Enable Wake-On-LAN from shutdown for Skyhawk
Sriharsha Basavapatna [Mon, 6 Jun 2016 11:22:09 +0000 (07:22 -0400)]
be2net: Enable Wake-On-LAN from shutdown for Skyhawk

Skyhawk does support wake-up from ACPI shutdown state - S5, provided the
platform supports it (like Auxiliary power source etc). The changes listed
below are done to fix this.

1) There's no need to defer the HW configuration of WOL to be_suspend().
Remove this in be_suspend() and move it to be_set_wol() ethtool function
so it is configured directly in the context of ethtool. This automatically
takes care of the shutdown case.

2) The driver incorrectly uses WOL_CAP field in the FW response to
get_acpi_wol_cap() command, to determine if WOL is enabled. Instead the
driver must rely on the macaddr field in the response to infer WOL state.

3) In be_get_config() during init, if we find that WOL is enabled in FW,
call pci_enable_wake() to enable pmcsr.pme_en bit. This is needed to
support persistent WOL configuration provided by the FW in some platforms.

4) Remove code in be_set_wol() that writes to PCICFG_PM_CONTROL_OFFSET
to set pme_en bit; pci_enable_wake() sets that.

Fixes: 028991e49 ("Enabling Wake-on-LAN is not supported in S5 state")
Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Signed-off-by: Sathya Perla <sathya.perla@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobe2net: use max-TXQs limit too while provisioning VF queue pairs
Suresh Reddy [Mon, 6 Jun 2016 11:22:08 +0000 (07:22 -0400)]
be2net: use max-TXQs limit too while provisioning VF queue pairs

When the PF driver provisions resources for VFs, it currently only looks
at max RSS queues available to calculate the number of VF queue pairs.
This logic breaks when there are less number of TX-queues than RSS-queues.
This patch fixes this problem by using the max-TXQs available in the
PF-pool in the calculations. As a part of this change the
be_calculate_vf_qs() routine is renamed as be_calculate_vf_res() and the
code that calculates limits on other related resources is moved here to
contain all resource calculation code inside one routine.

Signed-off-by: Suresh Reddy <suresh.reddy@broadcom.com>
Signed-off-by: Sathya Perla <sathya.perla@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers/net: support hdlc function for QE-UCC
Zhao Qiang [Mon, 6 Jun 2016 06:30:02 +0000 (14:30 +0800)]
drivers/net: support hdlc function for QE-UCC

The driver add hdlc support for Freescale QUICC Engine.
It support NMSI and TSA mode.

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/qe: Add QE TDM lib
Zhao Qiang [Mon, 6 Jun 2016 06:30:01 +0000 (14:30 +0800)]
fsl/qe: Add QE TDM lib

QE has module to support TDM, some other protocols
supported by QE are based on TDM.
add a qe-tdm lib, this lib provides functions to the protocols
using TDM to configurate QE-TDM.

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/qe: Make regs resouce_size_t
Zhao Qiang [Mon, 6 Jun 2016 06:30:00 +0000 (14:30 +0800)]
fsl/qe: Make regs resouce_size_t

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/qe: setup clock source for TDM mode
Zhao Qiang [Mon, 6 Jun 2016 06:29:59 +0000 (14:29 +0800)]
fsl/qe: setup clock source for TDM mode

Add tdm clock configuration in both qe clock system and ucc
fast controller.

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/qe: add rx_sync and tx_sync for TDM mode
Zhao Qiang [Mon, 6 Jun 2016 06:29:58 +0000 (14:29 +0800)]
fsl/qe: add rx_sync and tx_sync for TDM mode

Rx_sync and tx_sync are used by QE-TDM mode,
add them to struct ucc_fast_info.

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet sched: indentation and other OCD stylistic fixes
Jamal Hadi Salim [Sun, 5 Jun 2016 14:41:32 +0000 (10:41 -0400)]
net sched: indentation and other OCD stylistic fixes

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
8 years agoMerge branch 'sch-action-tstamp'
David S. Miller [Tue, 7 Jun 2016 22:53:44 +0000 (15:53 -0700)]
Merge branch 'sch-action-tstamp'

Jamal Hadi Salim says:

====================
net sched action timestamp improvements

Various aggregations of duplicated code, fixes and introduction of firstused
timestamp

v2: add const for source time info per suggestion from Cong
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet sched actions: aggregate dumping of actions timeinfo
Jamal Hadi Salim [Mon, 6 Jun 2016 10:32:55 +0000 (06:32 -0400)]
net sched actions: aggregate dumping of actions timeinfo

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet sched actions: introduce timestamp for firsttime use
Jamal Hadi Salim [Mon, 6 Jun 2016 10:32:54 +0000 (06:32 -0400)]
net sched actions: introduce timestamp for firsttime use

Useful to know when the action was first used for accounting
(and debugging)

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet sched: actions use tcf_lastuse_update for consistency
Jamal Hadi Salim [Mon, 6 Jun 2016 10:32:53 +0000 (06:32 -0400)]
net sched: actions use tcf_lastuse_update for consistency

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/sched: cls_flower: Introduce support in SKIP SW flag
Amir Vadai [Sun, 5 Jun 2016 14:11:18 +0000 (17:11 +0300)]
net/sched: cls_flower: Introduce support in SKIP SW flag

In order to make a filter processed only by hardware, skip_sw flag
should be supplied. This is an addition to the already existing skip_hw
flag (filter will be processed by software only). If no flag is
specified, filter will be processed by both software and hardware.

If only hardware offloaded filters exist, fl_classify() will return
without doing anything.

A following userspace patch will be sent once kernel patch is accepted.

Example:

tc filter add dev enp0s9 protocol ip prio 20 parent ffff: \
flower \
ip_proto 6 \
indev enp0s9 \
skip_sw \
action skbedit mark 0x1234

Signed-off-by: Amir Vadai <amirva@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'qed-iov-fw-reqs'
David S. Miller [Tue, 7 Jun 2016 22:40:12 +0000 (15:40 -0700)]
Merge branch 'qed-iov-fw-reqs'

Yuval Mintz says:

====================
qed: IOV series - relax firmware requirements

In order for VFs to work, current implementation demands that the VF's
requried storm firmware would be exactly the version that was loaded by
the PF, which is a very harsh requirement.
This patch series is intended to relax this -
the recently submitted firmware is intended to be forward/backward
compatible in its fastpath [slowpath is configured by PF on behalf of VF],
and so VFs would only be required of having the same major faspath HSI in
order to work.

Most of the other patches in this series extend current forward
compatibilty of driver to reduce chance of breaking PF/VF compatibility
in the future. A few are unrelated IOV changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: PF to reply to unknown messages
Yuval Mintz [Sun, 5 Jun 2016 10:11:16 +0000 (13:11 +0300)]
qed: PF to reply to unknown messages

If a future VF would send the PF an unknown message, the PF today would
not send a reply. This would have 2 bad effects:
  a. VF would have to timeout on the request.
  b. If VF were to send an additional message to PF, firmware would mark
     it as malicious.

Instead, if there's some valid reply-address on the message - let the PF
answer and tell the VF it doesn't know the message.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: PF enforce MAC limitation of VFs
Yuval Mintz [Sun, 5 Jun 2016 10:11:15 +0000 (13:11 +0300)]
qed: PF enforce MAC limitation of VFs

The only limitation relating to MACs the PF enforce today on its VFs
is in case it has a forced-unicast MAC address for them, in which case
they can't configure other unicast addresses.
Specifically, the PF isn't enforcing the number of MAC addresse a VF can
configure regardless of the nubmer of such filters agreed upon by PF and
VF during the acquisition process.

PF's shadow-config is now extended to also contain information about its
VFs' unicast addresses configuration, allowing such enforcement.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Move doorbell calculation from VF to PF
Yuval Mintz [Sun, 5 Jun 2016 10:11:14 +0000 (13:11 +0300)]
qed: Move doorbell calculation from VF to PF

Today, the VF is aware of its queues context-ids, and calculates the
doorbell address when opening its queues on its own.
The configuration of doorbells in HW can sometime in the future be changed
by the PF [hw has several configurable features that might affect doorbell
addresses, e.g., dpm support], this would break compatibility with older
VFs as their calculated doorbell addresses would be incorrect for such a
configuration.

In order to avoid such a backward compatibility failure, let the PF make
the calculation of the doorbell offset based on the context-id, and pass
that to the VF.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Make PF more robust against malicious VF
Yuval Mintz [Sun, 5 Jun 2016 10:11:13 +0000 (13:11 +0300)]
qed: Make PF more robust against malicious VF

There are several requests the VF can make toward the PF which the driver
would pass to firmware without checking the validity first - specifically,
opening queues and updating vports. Such configurations might cause the
firmware to assert.

This adds validation of the legality of said configurations on the PF side
before passing it onward via ramrod to firmware.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: PF-VF resource negotiation
Yuval Mintz [Sun, 5 Jun 2016 10:11:12 +0000 (13:11 +0300)]
qed: PF-VF resource negotiation

One of the goals of the vf's first message to the PF [acquire]
is to learn about the number of resources available to it [macs, vlans,
etc.]. This is done via negotiation - the VF requires a set of resources,
which the PF either approves or disaproves and sends a smaller set of
resources as alternative. In this later case, the VF is then expected to
either abort the probe or re-send the acquire message with less
required resources.

While this infrastructure exists since the initial submision of qed
SRIOV support, it's in fact completely inoperational - PF isn't really
looking into the resources the VF has asked for and is never going to
reply to the VF that it lacks resources.

This patch addresses this flow, fixing it and allowing the PF and VF
to actually agree on a set of resources.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Relax VF firmware requirements
Yuval Mintz [Sun, 5 Jun 2016 10:11:11 +0000 (13:11 +0300)]
qed: Relax VF firmware requirements

Current driver require an exact match between VF and PF storm firmware;
Any difference would fail the VF acquire message, causing the VF probe
to be aborted.

While there's still dependencies between the two, the recent FW submission
has relaxed the match requirement - instead of an exact match, there's now
a 'fastpath' HSI major/minor scheme, where VFs and PFs that match in their
major number can co-exist even if their minor is different.

In order to accomadate this change some changes in the vf-start init flow
had to be made, as the VF start ramrod now has to be sent only after PF
learns which fastpath HSI its VF is requiring.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: get rid of spin_trylock() in net_tx_action()
Eric Dumazet [Sun, 5 Jun 2016 03:02:28 +0000 (20:02 -0700)]
net: get rid of spin_trylock() in net_tx_action()

Note: Tom Herbert posted almost same patch 3 months back, but for
different reasons.

The reasons we want to get rid of this spin_trylock() are :

1) Under high qdisc pressure, the spin_trylock() has almost no
chance to succeed.

2) We loop multiple times in softirq handler, eventually reaching
the max retry count (10), and we schedule ksoftirqd.

Since we want to adhere more strictly to ksoftirqd being waked up in
the future (https://lwn.net/Articles/687617/), better avoid spurious
wakeups.

3) calls to __netif_reschedule() dirty the cache line containing
q->next_sched, slowing down the owner of qdisc.

4) RT kernels can not use the spin_trylock() here.

With help of busylock, we get the qdisc spinlock fast enough, and
the trylock trick brings only performance penalty.

Depending on qdisc setup, I observed a gain of up to 19 % in qdisc
performance (1016600 pps instead of 853400 pps, using prio+tbf+fq_codel)

("mpstat -I SCPU 1" is much happier now)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tom Herbert <tom@herbertland.com>
Acked-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovhost_net: stop polling socket during rx processing
Jason Wang [Wed, 1 Jun 2016 05:56:33 +0000 (01:56 -0400)]
vhost_net: stop polling socket during rx processing

We don't stop rx polling socket during rx processing, this will lead
unnecessary wakeups from under layer net devices (E.g
sock_def_readable() form tun). Rx will be slowed down in this
way. This patch avoids this by stop polling socket during rx
processing. A small drawback is that this introduces some overheads in
light load case because of the extra start/stop polling, but single
netperf TCP_RR does not notice any change. In a super heavy load case,
e.g using pktgen to inject packet to guest, we get about ~8.8%
improvement on pps:

before: ~1240000 pkt/s
after:  ~1350000 pkt/s

Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: cavium: liquidio: request_manager: Remove create_workqueue
Bhaktipriya Shridhar [Sat, 4 Jun 2016 15:24:00 +0000 (20:54 +0530)]
net: ethernet: cavium: liquidio: request_manager: Remove create_workqueue

alloc_workqueue replaces deprecated create_workqueue().

A dedicated workqueue has been used since the workitem viz
(&db_wq->wk.work which maps to check_db_timeout) is involved
in normal device operation. WQ_MEM_RECLAIM has been set to guarantee
forward progress under memory pressure, which is a requirement here.
Since there are only a fixed number of work items, explicit concurrency
limit is unnecessary.

flush_workqueue is unnecessary since destroy_workqueue() itself calls
drain_workqueue() which flushes repeatedly till the workqueue
becomes empty.

Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: cavium: liquidio: response_manager: Remove create_workqueue
Bhaktipriya Shridhar [Sat, 4 Jun 2016 14:51:40 +0000 (20:21 +0530)]
net: ethernet: cavium: liquidio: response_manager: Remove create_workqueue

alloc_workqueue replaces deprecated create_workqueue().

A dedicated workqueue has been used since the workitem viz
(&cwq->wk.work which maps to oct_poll_req_completion) is involved
in normal device operation. WQ_MEM_RECLAIM has been set to guarantee
forward progress under memory pressure, which is a requirement here.
Since there are only a fixed number of work items, explicit concurrency
limit is unnecessary.

flush_workqueue is unnecessary since destroy_workqueue() itself calls
drain_workqueue() which flushes repeatedly till the workqueue
becomes empty. Hence the call to flush_workqueue() has been dropped.

Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovirtio-net: Add initial MTU advice feature
Aaron Conole [Fri, 3 Jun 2016 20:57:12 +0000 (16:57 -0400)]
virtio-net: Add initial MTU advice feature

This commit adds the feature bit and associated mtu device entry for the
virtio network device.  When a virtio device comes up, it checks the
feature bit for the VIRTIO_NET_F_MTU feature.  If such feature bit is
enabled, the driver will read the advised MTU and use it as the initial
value.

Signed-off-by: Aaron Conole <aconole@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: Revert vrf-local changes.
David S. Miller [Mon, 6 Jun 2016 22:58:34 +0000 (15:58 -0700)]
net: Revert vrf-local changes.

This reverts commit 2fb7ea455d57e22110c54fc2de0656b6f744263c.

It results in build errors because ip6_input is not a
symbol exported to modules.

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'vrf-local'
David S. Miller [Mon, 6 Jun 2016 22:19:07 +0000 (15:19 -0700)]
Merge branch 'vrf-local'

David Ahern says:

====================
net: vrf: Add support for local traffic to local addresses

Add support for locally originated traffic to VRF-local addresses,
be it addresses on enslaved devices or addresses on the VRF device:

$ ip addr show dev red
33: red: <NOARP,MASTER,UP,LOWER_UP> mtu 65536 qdisc pfifo_fast state UP group default qlen 1000
    link/ether be:00:53:b5:e4:25 brd ff:ff:ff:ff:ff:ff
    inet 1.1.1.1/32 scope global red
       valid_lft forever preferred_lft forever
    inet6 1111:1::1/128 scope global
       valid_lft forever preferred_lft forever

$ ip addr show dev eth1
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
    link/ether 02:e0:f9:79:34:bd brd ff:ff:ff:ff:ff:ff
    inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
       valid_lft forever preferred_lft forever
    inet6 2100:1::1/120 scope global
       valid_lft forever preferred_lft forever
    inet6 fe80::e0:f9ff:fe79:34bd/64 scope link
       valid_lft forever preferred_lft forever

$ ping -c1 -I red 10.100.1.1
    ping: Warning: source address might be selected on device other than red.
    PING 10.100.1.1 (10.100.1.1) from 10.100.1.1 red: 56(84) bytes of data.
    64 bytes from 10.100.1.1: icmp_seq=1 ttl=64 time=0.057 ms

$ ping -c1 -I red 1.1.1.1
PING 1.1.1.1 (1.1.1.1) from 1.1.1.1 red: 56(84) bytes of data.
64 bytes from 1.1.1.1: icmp_seq=1 ttl=64 time=0.136 ms

--- 1.1.1.1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.136/0.136/0.136/0.000 ms

$ ping6 -c1 -I red  2100:1::1
ping6: Warning: source address might be selected on device other than red.
PING 2100:1::1(2100:1::1) from 2100:1::1 red: 56 data bytes
64 bytes from 2100:1::1: icmp_seq=1 ttl=64 time=0.167 ms

--- 2100:1::1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.167/0.167/0.167/0.000 ms

$ ping6 -c1 -I red 1111::1
PING 1111::1(1111::1) from 1111:1::1 red: 56 data bytes
64 bytes from 1111::1: icmp_seq=1 ttl=64 time=0.187 ms

--- 1111::1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.187/0.187/0.187/0.000 ms

This change also enables use of loopback address on the VRF device:
$ ip addr add dev red 127.0.0.1/8

$ ping -c1 -I red 127.0.0.1
PING 127.0.0.1 (127.0.0.1) from 127.0.0.1 red: 56(84) bytes of data.
64 bytes from 127.0.0.1: icmp_seq=1 ttl=64 time=0.058 ms
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: ipv6 support for local traffic to local addresses
David Ahern [Thu, 2 Jun 2016 20:15:12 +0000 (13:15 -0700)]
net: vrf: ipv6 support for local traffic to local addresses

Add support for locally originated traffic to VRF-local IPv6 addresses.
Similar to IPv4 a local dst is set on the skb and the packet is
reinserted with a call to netif_rx. With this patch, ping, tcp and udp
packets to a local IPv6 address are successfully routed:

    $ ip addr show dev eth1
    4: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
        link/ether 02:e0:f9:1c:b9:74 brd ff:ff:ff:ff:ff:ff
        inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
           valid_lft forever preferred_lft forever
        inet6 2100:1::1/120 scope global
           valid_lft forever preferred_lft forever
        inet6 fe80::e0:f9ff:fe1c:b974/64 scope link
           valid_lft forever preferred_lft forever

    $ ping6 -c1 -I red 2100:1::1
    ping6: Warning: source address might be selected on device other than red.
    PING 2100:1::1(2100:1::1) from 2100:1::1 red: 56 data bytes
    64 bytes from 2100:1::1: icmp_seq=1 ttl=64 time=0.098 ms

ip6_input is exported so the VRF driver can use it for the dst input
function. The dst_alloc function for IPv4 defaults to setting the input and
output functions; IPv6's does not. VRF does not need to duplicate the Rx path
so just export the ipv6 input function.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: ipv4 support for local traffic to local addresses
David Ahern [Thu, 2 Jun 2016 20:15:11 +0000 (13:15 -0700)]
net: vrf: ipv4 support for local traffic to local addresses

Add support for locally originated traffic to VRF-local addresses. If
destination device for an skb is the loopback or VRF device then set
its dst to a local version of the VRF cached dst_entry and call netif_rx
to insert the packet onto the rx queue - similar to what is done for
loopback. This patch handles IPv4 support; follow on patch handles IPv6.

With this patch, ping, tcp and udp packets to a local IPv4 address are
successfully routed:

    $ ip addr show dev eth1
    4: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master red state UP group default qlen 1000
        link/ether 02:e0:f9:1c:b9:74 brd ff:ff:ff:ff:ff:ff
        inet 10.100.1.1/24 brd 10.100.1.255 scope global eth1
           valid_lft forever preferred_lft forever
        inet6 2100:1::1/120 scope global
           valid_lft forever preferred_lft forever
        inet6 fe80::e0:f9ff:fe1c:b974/64 scope link
           valid_lft forever preferred_lft forever

    $ ping -c1 -I red 10.100.1.1
    ping: Warning: source address might be selected on device other than red.
    PING 10.100.1.1 (10.100.1.1) from 10.100.1.1 red: 56(84) bytes of data.
    64 bytes from 10.100.1.1: icmp_seq=1 ttl=64 time=0.057 ms

This patch also enables use of IPv4 loopback address on the VRF device:
    $ ip addr add dev red 127.0.0.1/8

    $ ping -c1 -I red 127.0.0.1
    PING 127.0.0.1 (127.0.0.1) from 127.0.0.1 red: 56(84) bytes of data.
    64 bytes from 127.0.0.1: icmp_seq=1 ttl=64 time=0.058 ms

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: Minor refactoring for local address patches
David Ahern [Thu, 2 Jun 2016 20:15:10 +0000 (13:15 -0700)]
net: vrf: Minor refactoring for local address patches

Move the stripping of the ethernet header from is_ip_tx_frame into the
ipv4 and ipv6 outbound functions. If the packet is destined to a local
address the header is retained since the packet is sent back to netif_rx.

Collapse vrf_send_v4_prep into vrf_process_v4_outbound.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'hv_netvsc-cleanups'
David S. Miller [Mon, 6 Jun 2016 03:16:50 +0000 (23:16 -0400)]
Merge branch 'hv_netvsc-cleanups'

Vitaly Kuznetsov says:

====================
hv_netvsc: cleanup after untangling the pointer mess

Changes since v1:
- resend when net-next is open [David Miller]
- rebased to current net-next.

After we made traveling through our internal structures explicit it became
obvious that some functions take arguments they don't need just to do
redundant pointer travel and get to what they really need while their
callers already have the required information.

This is just a cleanup series with no functional changes intended. It
doesn't pretend to be complete, additional cleanup of other functions may
follow.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agohv_netvsc: pass struct net_device to rndis_filter_set_offload_params()
Vitaly Kuznetsov [Fri, 3 Jun 2016 15:51:02 +0000 (17:51 +0200)]
hv_netvsc: pass struct net_device to rndis_filter_set_offload_params()

The only caller rndis_filter_device_add() has 'struct net_device' pointer
already.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agohv_netvsc: pass struct net_device to rndis_filter_set_device_mac()
Vitaly Kuznetsov [Fri, 3 Jun 2016 15:51:01 +0000 (17:51 +0200)]
hv_netvsc: pass struct net_device to rndis_filter_set_device_mac()

We unpack 'struct net_device' in netvsc_set_mac_addr() to get to
'struct hv_device' pointer which we use in rndis_filter_set_device_mac()
to get back to 'struct net_device'.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agohv_netvsc: pass struct netvsc_device to rndis_filter_{open, close}()
Vitaly Kuznetsov [Fri, 3 Jun 2016 15:51:00 +0000 (17:51 +0200)]
hv_netvsc: pass struct netvsc_device to rndis_filter_{open, close}()

Both rndis_filter_open()/rndis_filter_close() use struct hv_device to
reach to struct netvsc_device only and all callers have it already.
While on it, rename net_device to nvdev in rndis_filter_open() as
net_device is misleading.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agohv_netvsc: introduce {net, hv}_device_to_netvsc_device() helpers
Vitaly Kuznetsov [Fri, 3 Jun 2016 15:50:59 +0000 (17:50 +0200)]
hv_netvsc: introduce {net, hv}_device_to_netvsc_device() helpers

Make it easier to get 'struct netvsc_device' from 'struct net_device' and
'struct hv_device' by introducing inline helpers.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agohv_netvsc: remove redundant assignment in netvsc_recv_callback()
Vitaly Kuznetsov [Fri, 3 Jun 2016 15:50:58 +0000 (17:50 +0200)]
hv_netvsc: remove redundant assignment in netvsc_recv_callback()

net_device_ctx is assigned in the very beginning of the function and 'net'
pointer doesn't change.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: disable fragment reassembly if high_thresh is zero
Michal Kubeček [Fri, 27 May 2016 15:53:52 +0000 (17:53 +0200)]
net: disable fragment reassembly if high_thresh is zero

Before commit 6d7b857d541e ("net: use lib/percpu_counter API for
fragmentation mem accounting"), setting the reassembly high threshold
to 0 prevented fragment reassembly as first fragment would be always
evicted before second could be added to the queue. While inefficient,
some users apparently relied on this method.

Since the commit mentioned above, a percpu counter is used for
reassembly memory accounting and high batch size avoids taking slow path
in most common scenarios. As a result, a whole full sized packet can be
reassembled without the percpu counter's main counter changing its value
so that even with high_thresh set to 0, fragmented packets can be still
reassembled and processed.

Add explicit check preventing reassembly if high threshold is zero.

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'hns-acpi'
David S. Miller [Sun, 5 Jun 2016 01:32:46 +0000 (21:32 -0400)]
Merge branch 'hns-acpi'

Kejian Yan says:

====================
net: hns: add support of ACPI

This series adds HNS support of acpi. The routine will call some ACPI
helper functions, like acpi_dev_found() and acpi_evaluate_dsm(), which
are not included in other cases. In order to make system compile
successfully in other cases except ACPI, it needs to add relative stub
functions to linux/acpi.h. And we use device property functions instead
of serial helper functions to suport both DT and ACPI cases. And then
add the supports of ACPI for HNS.

change log:
 v3->v4:
  mii-id gets from dev-name instead of address

 v2->v3:
 1. add Review-by: Andy Shevchenko
 2. fix the potential memory leak

 v1 -> v2:
 1. use acpi_dev_found() instead of acpi_match_device_ids() to check if
it is a acpi node.
 2. use is_of_node() instead of IS_ENABLED() to check if it is a DT node.
 3. split the patch("add support of acpi for hns-mdio") into two patches:
    3.1 Move to use fwnode_handle
    3.2 Add ACPI
 4. add the patch which subject is dsaf misc operation method
 5. fix the comments by Andy Shevchenko
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: net: hns: enet adds support of acpi
Kejian Yan [Fri, 3 Jun 2016 02:55:21 +0000 (10:55 +0800)]
net: hns: net: hns: enet adds support of acpi

Enet needs to get configration parameter by acpi. This patch
adds support of ACPI for enet. The configuration parameter will
be configed in BIOS.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: implement the miscellaneous operation by asl
Kejian Yan [Fri, 3 Jun 2016 02:55:20 +0000 (10:55 +0800)]
net: hns: implement the miscellaneous operation by asl

The miscellaneous operation is implemented in BIOS, the kernel can call
_DSM method help to call the implementation in ACPI case. Here is a patch
to do that.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: register phy device in each mac initial sequence
Kejian Yan [Fri, 3 Jun 2016 02:55:19 +0000 (10:55 +0800)]
net: hns: register phy device in each mac initial sequence

In ACPI case, there is no interface to register phy device to mdio-bus.
Phy device has to be registered itself to mdio-bus, and then enet can
get the phy device's info so that it can config the phy-device to help
to trasmit and receive data.
HNS hardware topology is as below. The MDIO controller may control several
PHY-devices, and each PHY-device connects to a MAC device. PHY-devices
will register when each mac find PHY device in initial sequence.

                       cpu
                        |
                        |
     -------------------------------------------
    |                   |                       |
    |                   |                       |
    |                  dsaf                     |
   MDIO                 |                      MDIO
    |      ---------------------------          |
    |     |         |         |       |         |
    |     |         |         |       |         |
    |    MAC       MAC       MAC     MAC        |
    |     |         |         |       |         |
     ---- |-------- |-------- |       | --------
         ||        ||        ||       ||
         PHY       PHY       PHY     PHY

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: dsaf adds support of acpi
Kejian Yan [Fri, 3 Jun 2016 02:55:18 +0000 (10:55 +0800)]
net: hns: dsaf adds support of acpi

Dsaf needs to get configuration parameter by ACPI, so this patch add
support of ACPI.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: add dsaf misc operation method
Kejian Yan [Fri, 3 Jun 2016 02:55:17 +0000 (10:55 +0800)]
net: hns: add dsaf misc operation method

The misc operation for different hw platform may be different, if using
current implementation, it will add a new branch on each function for
every new hw platform, so we add a method for this operation.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: add uniform interface for phy connection
Kejian Yan [Fri, 3 Jun 2016 02:55:16 +0000 (10:55 +0800)]
net: hns: add uniform interface for phy connection

As device_node is only used by DT case, HNS needs to treat the other
cases including ACPI. It needs to use uniform ways to handle both of
DT and ACPI. This patch chooses phy_device, and of_phy_connect and
of_phy_attach are only used by DT case. It needs to use uniform interface
to handle that sequence by both DT and ACPI.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: enet specify a reference to dsaf by fwnode_handle
Kejian Yan [Fri, 3 Jun 2016 02:55:15 +0000 (10:55 +0800)]
net: hns: enet specify a reference to dsaf by fwnode_handle

As device_node is only used by DT case, it is expected to find uniform
ways. So fwnode_handle is the suitable method.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: use platform_get_irq instead of irq_of_parse_and_map
Kejian Yan [Fri, 3 Jun 2016 02:55:14 +0000 (10:55 +0800)]
net: hns: use platform_get_irq instead of irq_of_parse_and_map

As irq_of_parse_and_map is only used by DT case, it is excepted to use
a uniform interface. So it is used platform_get_irq() instead.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hns: use device_* APIs instead of of_* APIs
Kejian Yan [Fri, 3 Jun 2016 02:55:13 +0000 (10:55 +0800)]
net: hns: use device_* APIs instead of of_* APIs

OF series functions can be used only for DT case. Use unified
device property function instead to support both DT and ACPI.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hisilicon: add support of acpi for hns-mdio
Kejian Yan [Fri, 3 Jun 2016 02:55:12 +0000 (10:55 +0800)]
net: hisilicon: add support of acpi for hns-mdio

hns-mdio needs to register itself to mii-bus. The info of the device can
be read by both DT and ACPI.
HNS tries to call Linux PHY driver to help access PHY-devices, the HNS
hardware topology is as below. The MDIO controller may control several
PHY-devices, and each PHY-device connects to a MAC device. The MDIO will
be registered to mdiobus, then PHY-devices will register when each mac
find PHY device.
                       cpu
                        |
                        |
     -------------------------------------------
    |                   |                       |
    |                   |                       |
    |                  dsaf                     |
   MDIO                 |                      MDIO
    |      ---------------------------          |
    |     |         |         |       |         |
    |     |         |         |       |         |
    |    MAC       MAC       MAC     MAC        |
    |     |         |         |       |         |
     ---- |-------- |-------- |       | --------
         ||        ||        ||       ||
         PHY       PHY       PHY     PHY

And the driver can handle reset sequence by _RST method in DSDT in ACPI
case.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hisilicon: cleanup to prepare for other cases
Kejian Yan [Fri, 3 Jun 2016 02:55:11 +0000 (10:55 +0800)]
net: hisilicon: cleanup to prepare for other cases

Hns-mdio only supports DT case now. do some cleanup to prepare
for introducing other cases later, no functional change.

Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoACPI: bus: add stub acpi_evaluate_dsm() to linux/acpi.h
Kejian Yan [Fri, 3 Jun 2016 02:55:10 +0000 (10:55 +0800)]
ACPI: bus: add stub acpi_evaluate_dsm() to linux/acpi.h

acpi_evaluate_dsm() will be used to handle the _DSM method in ACPI case.
It will be compiled in non-ACPI case, but the function is in acpi_bus.h
and acpi_bus.h can only be used in ACPI case, so this patch add the stub
function to linux/acpi.h to make compiled successfully in non-ACPI cases.

Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoACPI: bus: add stub acpi_dev_found() to linux/acpi.h
Kejian Yan [Fri, 3 Jun 2016 02:55:09 +0000 (10:55 +0800)]
ACPI: bus: add stub acpi_dev_found() to linux/acpi.h

acpi_dev_found() will be used to detect if a given ACPI device is in the
system. It will be compiled in non-ACPI case, but the function is in
acpi_bus.h and acpi_bus.h can only be used in ACPI case, so this patch add
the stub function to linux/acpi.h to make compiled successfully in
non-ACPI cases.

Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'dsa-new-binding'
David S. Miller [Sat, 4 Jun 2016 21:29:55 +0000 (14:29 -0700)]
Merge branch 'dsa-new-binding'

Andrew Lunn says:

====================
New DSA bind, switches as devices

The interesting patches here are the last three. They implement a new
binding for DSA, which removes a few limitations of the current DSA
binding. In particular, it allows switches to be true Linux devices.
These devices can be on any type of bus, unlike the old DSA binding
which assumes MDIO. See the commit log for more details. The second to
last patch modifies an existing boards device tree to use the new
binding, giving a good example of how switches can be true MDIO
devices. The last patch documents the new binding.

Thanks go to Florian and Vivien for reviewing, testing and bug fixing
these patches.

Tested-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Since V1:

* Add lots of reviewed-by's
* Fix rtable comment
* dsa2: Clear cpu port mask in dsa_cpu_port_unapply()
* dsa2: Only set dsa_port_mask when port successfully configured
* dsa: clear {dsa|cpu}_port_mask on destroy

Since RFC:

* Split the mv88e6xxx MDIO refactor into a rename patch and a refactor
  patch.
* Extend commit message with comment about wrong of_node_put()
* Fix destroy of cpu and dsa ports.
* Rename _DSA_TAG_LAST to DSA_TAG_LAST and add a comment.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Document new binding
Andrew Lunn [Sat, 4 Jun 2016 19:17:09 +0000 (21:17 +0200)]
net: dsa: Document new binding

Add the new binding to the documentation of the existing binding.
Mark the old binding as deprecated.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoarm: dt: vf610-zii-devel-b: Make use of new DSA binding
Andrew Lunn [Sat, 4 Jun 2016 19:17:08 +0000 (21:17 +0200)]
arm: dt: vf610-zii-devel-b: Make use of new DSA binding

Hang the three switches of the three MDIO busses using the new DSA
binding. Also, make use of the mdio-bus and explicitly list the phys
on one device. This is not required, but good for testing.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Add new binding implementation
Andrew Lunn [Sat, 4 Jun 2016 19:17:07 +0000 (21:17 +0200)]
net: dsa: Add new binding implementation

The existing DSA binding has a number of limitations and problems. The
main problem is that it cannot represent a switch as a linux device,
hanging off some bus. It is limited to one CPU port. The DSA platform
device is artificial, and does not really represent hardware.

Implement a new binding which can be embedded into any type of node on
a bus to represent one switch device, and its links to other switches.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: Refactor MDIO so driver registers mdio bus
Andrew Lunn [Sat, 4 Jun 2016 19:17:06 +0000 (21:17 +0200)]
net: dsa: mv88e6xxx: Refactor MDIO so driver registers mdio bus

Have the switch driver register its own MDIO bus. This allows for an
mdio property in the device tree, with child nodes for phys, which
can be referenced via phandles, etc.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: Rename _phy_ to _mdio_
Andrew Lunn [Sat, 4 Jun 2016 19:17:05 +0000 (21:17 +0200)]
net: dsa: mv88e6xxx: Rename _phy_ to _mdio_

The switch implements a generic MDIO bus, which could host more than
PHYs. It is conventional to use _mdio_ or _mii_ in the function name,
so rename them. Also postfix make the historically first read/write
function with _direct, to help distinguish it from _indirect and _ppu.

While touching these functions, remove some of the _ prefixes, which
we are deprecating.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Make mdio bus optional
Andrew Lunn [Sat, 4 Jun 2016 19:17:04 +0000 (21:17 +0200)]
net: dsa: Make mdio bus optional

The switch may want to instantiate its own MDIO bus. Only do it
centrally if the switch has not already created one, and the read op
is implemented.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Refactor selection of tag ops into a function
Andrew Lunn [Sat, 4 Jun 2016 19:17:03 +0000 (21:17 +0200)]
net: dsa: Refactor selection of tag ops into a function

Replace the two switch statements with an array lookup, and store the
result in the dsa tree structure. The drivers no longer need to know
the selected tag protocol, so remove it from the dsa switch structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: Only support EDSA tagging
Andrew Lunn [Sat, 4 Jun 2016 19:17:02 +0000 (21:17 +0200)]
net: dsa: mv88e6xxx: Only support EDSA tagging

The merged driver no longer offers the option to use DSA tagging. So
remove the code to setup the switch to do DSA tagging and hard code
the use of EDSA.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>y
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Split up creating/destroying of DSA and CPU ports
Andrew Lunn [Sat, 4 Jun 2016 19:17:01 +0000 (21:17 +0200)]
net: dsa: Split up creating/destroying of DSA and CPU ports

Refactor the code to setup a single DSA/CPU port into a function of
its own, and export it, so it can be used by the new binding.

Similarly, refactor the destroy code into a function.  When destroying
the ports, don't put the of node. They should be released at the end
along with the normal ports.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Copy the routing table into the switch structure
Andrew Lunn [Sat, 4 Jun 2016 19:17:00 +0000 (21:17 +0200)]
net: dsa: Copy the routing table into the switch structure

The new binding will not have a chip data structure, it will place the
routing directly into the switch structure. To enable backwards
compatibility, copy the routing from the chip data into the switch
structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Remove dynamic allocate of routing table
Andrew Lunn [Sat, 4 Jun 2016 19:16:59 +0000 (21:16 +0200)]
net: dsa: Remove dynamic allocate of routing table

With a maximum of four switches, the size of the routing table is the
same as the pointer to it. Removing it makes the code simpler.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Move port device node into port structure
Andrew Lunn [Sat, 4 Jun 2016 19:16:58 +0000 (21:16 +0200)]
net: dsa: Move port device node into port structure

Move the port device node structure into the port structure, from the
chip data. This information is needed in the next step of implementing
the new binding.

The chip data structure is used while parsing the whole old binding,
before the individual switch structures exist. With the new bindings,
this is reversed, the switches exist first, and the interconnections
between the switches is derived from the individual switch
bindings. Thus this chip data structure becomes unneeded.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
eviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: Add a ports structure and use it in the switch structure
Andrew Lunn [Sat, 4 Jun 2016 19:16:57 +0000 (21:16 +0200)]
net: dsa: Add a ports structure and use it in the switch structure

There are going to be more per-port members added to the switch
structure. So add a port structure and move the netdev into it.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: tag_{e}dsa.c: Remove dependency on platform data
Andrew Lunn [Sat, 4 Jun 2016 19:16:56 +0000 (21:16 +0200)]
net: dsa: tag_{e}dsa.c: Remove dependency on platform data

The platform data nr_chips is used when validating a received packet,
to ensure it comes from a know switch chip. The number of possible
switches is limited to DSA_MAX_SWITCHES, so use this as the first
validation step. The new binding allows holes in the dst->ds[] array,
so also ensure ensure there is a valid dsa_switch for this packet.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: slave: Remove MDIO address from switch MDIO bus name
Andrew Lunn [Sat, 4 Jun 2016 19:16:55 +0000 (21:16 +0200)]
net: dsa: slave: Remove MDIO address from switch MDIO bus name

The DSA layer should no longer assume the switch is connected to an
MDIO bus. As a result, we cannot use the address on the MDIO bus when
forming the name of the switches internal MDIO bus for its builtin and
possibly external PHYs. The switch index is sufficient to make the
name unique, so drop the MDIO address.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: fix circular lock in PPU work
Vivien Didelot [Sat, 4 Jun 2016 19:16:54 +0000 (21:16 +0200)]
net: dsa: mv88e6xxx: fix circular lock in PPU work

Lock debugging shows that there is a possible circular lock in the PPU
work code. Switch the lock order of smi_mutex and ppu_mutex to fix this.

Here's the full trace:

    [    4.341325] ======================================================
    [    4.347519] [ INFO: possible circular locking dependency detected ]
    [    4.353800] 4.6.0 #4 Not tainted
    [    4.357039] -------------------------------------------------------
    [    4.363315] kworker/0:1/328 is trying to acquire lock:
    [    4.368463]  (&ps->smi_mutex){+.+.+.}, at: [<8049c758>] mv88e6xxx_reg_read+0x30/0x54
    [    4.376313]
    [    4.376313] but task is already holding lock:
    [    4.382160]  (&ps->ppu_mutex){+.+...}, at: [<8049cac0>] mv88e6xxx_ppu_reenable_work+0x28/0xd4
    [    4.390772]
    [    4.390772] which lock already depends on the new lock.
    [    4.390772]
    [    4.398963]
    [    4.398963] the existing dependency chain (in reverse order) is:
    [    4.406461]
    [    4.406461] -> #1 (&ps->ppu_mutex){+.+...}:
    [    4.410897]        [<806d86bc>] mutex_lock_nested+0x54/0x360
    [    4.416606]        [<8049a800>] mv88e6xxx_ppu_access_get+0x28/0x100
    [    4.422906]        [<8049b778>] mv88e6xxx_phy_read+0x90/0xdc
    [    4.428599]        [<806a4534>] dsa_slave_phy_read+0x3c/0x40
    [    4.434300]        [<804943ec>] mdiobus_read+0x68/0x80
    [    4.439481]        [<804939d4>] get_phy_device+0x58/0x1d8
    [    4.444914]        [<80493ed0>] mdiobus_scan+0x24/0xf4
    [    4.450078]        [<8049409c>] __mdiobus_register+0xfc/0x1ac
    [    4.455857]        [<806a40b0>] dsa_probe+0x860/0xca8
    [    4.460934]        [<8043246c>] platform_drv_probe+0x5c/0xc0
    [    4.466627]        [<804305a0>] driver_probe_device+0x118/0x450
    [    4.472589]        [<80430b00>] __device_attach_driver+0xac/0x128
    [    4.478724]        [<8042e350>] bus_for_each_drv+0x74/0xa8
    [    4.484235]        [<804302d8>] __device_attach+0xc4/0x154
    [    4.489755]        [<80430cec>] device_initial_probe+0x1c/0x20
    [    4.495612]        [<8042f620>] bus_probe_device+0x98/0xa0
    [    4.501123]        [<8042fbd0>] deferred_probe_work_func+0x4c/0xd4
    [    4.507328]        [<8013a794>] process_one_work+0x1a8/0x604
    [    4.513030]        [<8013ac54>] worker_thread+0x64/0x528
    [    4.518367]        [<801409e8>] kthread+0xec/0x100
    [    4.523201]        [<80108f30>] ret_from_fork+0x14/0x24
    [    4.528462]
    [    4.528462] -> #0 (&ps->smi_mutex){+.+.+.}:
    [    4.532895]        [<8015ad5c>] lock_acquire+0xb4/0x1dc
    [    4.538154]        [<806d86bc>] mutex_lock_nested+0x54/0x360
    [    4.543856]        [<8049c758>] mv88e6xxx_reg_read+0x30/0x54
    [    4.549549]        [<8049cad8>] mv88e6xxx_ppu_reenable_work+0x40/0xd4
    [    4.556022]        [<8013a794>] process_one_work+0x1a8/0x604
    [    4.561707]        [<8013ac54>] worker_thread+0x64/0x528
    [    4.567053]        [<801409e8>] kthread+0xec/0x100
    [    4.571878]        [<80108f30>] ret_from_fork+0x14/0x24
    [    4.577139]
    [    4.577139] other info that might help us debug this:
    [    4.577139]
    [    4.585159]  Possible unsafe locking scenario:
    [    4.585159]
    [    4.591093]        CPU0                    CPU1
    [    4.595631]        ----                    ----
    [    4.600169]   lock(&ps->ppu_mutex);
    [    4.603693]                                lock(&ps->smi_mutex);
    [    4.609742]                                lock(&ps->ppu_mutex);
    [    4.615790]   lock(&ps->smi_mutex);
    [    4.619314]
    [    4.619314]  *** DEADLOCK ***
    [    4.619314]
    [    4.625256] 3 locks held by kworker/0:1/328:
    [    4.629537]  #0:  ("events"){.+.+..}, at: [<8013a704>] process_one_work+0x118/0x604
    [    4.637288]  #1:  ((&ps->ppu_work)){+.+...}, at: [<8013a704>] process_one_work+0x118/0x604
    [    4.645653]  #2:  (&ps->ppu_mutex){+.+...}, at: [<8049cac0>] mv88e6xxx_ppu_reenable_work+0x28/0xd4
    [    4.654714]
    [    4.654714] stack backtrace:
    [    4.659098] CPU: 0 PID: 328 Comm: kworker/0:1 Not tainted 4.6.0 #4
    [    4.665286] Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
    [    4.671748] Workqueue: events mv88e6xxx_ppu_reenable_work
    [    4.677174] Backtrace:
    [    4.679674] [<8010d354>] (dump_backtrace) from [<8010d5a0>] (show_stack+0x20/0x24)
    [    4.687252]  r6:80fb3c88 r5:80fb3c88 r4:80fb4728 r3:00000002
    [    4.693003] [<8010d580>] (show_stack) from [<803b45e8>] (dump_stack+0x24/0x28)
    [    4.700246] [<803b45c4>] (dump_stack) from [<80157398>] (print_circular_bug+0x208/0x32c)
    [    4.708361] [<80157190>] (print_circular_bug) from [<8015a630>] (__lock_acquire+0x185c/0x1b80)
    [    4.716982]  r10:9ec22a00 r9:00000060 r8:8164b6bc r7:00000040 r6:00000003 r5:8163a5b4
    [    4.724905]  r4:00000003 r3:9ec22de8
    [    4.728537] [<80158dd4>] (__lock_acquire) from [<8015ad5c>] (lock_acquire+0xb4/0x1dc)
    [    4.736378]  r10:60000013 r9:00000000 r8:00000000 r7:00000000 r6:9e5e9c50 r5:80e618e0
    [    4.744301]  r4:00000000
    [    4.746879] [<8015aca8>] (lock_acquire) from [<806d86bc>] (mutex_lock_nested+0x54/0x360)
    [    4.754976]  r10:9e5e9c1c r9:80e616c4 r8:9f685ea0 r7:0000001b r6:9ec22a00 r5:8163a5b4
    [    4.762899]  r4:9e5e9c1c
    [    4.765477] [<806d8668>] (mutex_lock_nested) from [<8049c758>] (mv88e6xxx_reg_read+0x30/0x54)
    [    4.774008]  r10:80e60c5b r9:80e616c4 r8:9f685ea0 r7:0000001b r6:00000004 r5:9e5e9c10
    [    4.781930]  r4:9e5e9c1c
    [    4.784507] [<8049c728>] (mv88e6xxx_reg_read) from [<8049cad8>] (mv88e6xxx_ppu_reenable_work+0x40/0xd4)
    [    4.793907]  r7:9ffd5400 r6:9e5e9c68 r5:9e5e9cb0 r4:9e5e9c10
    [    4.799659] [<8049ca98>] (mv88e6xxx_ppu_reenable_work) from [<8013a794>] (process_one_work+0x1a8/0x604)
    [    4.809059]  r9:80e616c4 r8:9f685ea0 r7:9ffd5400 r6:80e0a1c8 r5:9f5f2e80 r4:9e5e9cb0
    [    4.816910] [<8013a5ec>] (process_one_work) from [<8013ac54>] (worker_thread+0x64/0x528)
    [    4.825010]  r10:9f5f2e80 r9:00000008 r8:80e0dc80 r7:80e0a1fc r6:80e0a1c8 r5:9f5f2e98
    [    4.832933]  r4:80e0a1c8
    [    4.835510] [<8013abf0>] (worker_thread) from [<801409e8>] (kthread+0xec/0x100)
    [    4.842827]  r10:00000000 r9:00000000 r8:00000000 r7:8013abf0 r6:9f5f2e80 r5:9ec15740
    [    4.850749]  r4:00000000
    [    4.853327] [<801408fc>] (kthread) from [<80108f30>] (ret_from_fork+0x14/0x24)
    [    4.860557]  r7:00000000 r6:00000000 r5:801408fc r4:9ec15740

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: slave: chip data is optional, don't dereference NULL
Andrew Lunn [Sat, 4 Jun 2016 19:16:52 +0000 (21:16 +0200)]
net: dsa: slave: chip data is optional, don't dereference NULL

The new binding does not make use of dsa_chip_data, a.k.a cd.  When
retrieving the size of the EEPROM attached to a switch, don't assume
there is a cd attached to the switch structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: Add docbook description for 'mtu' arg to skb_gso_validate_mtu()
David S. Miller [Sat, 4 Jun 2016 05:56:28 +0000 (22:56 -0700)]
net: Add docbook description for 'mtu' arg to skb_gso_validate_mtu()

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