When dumping the hardware FDB we should report entries pointing to VLAN
devices with VLAN 0, as packets coming into the bridge are untagged.
Likewise, pass FDB_{ADD,DEL} notifications with VLAN 0 for these
devices.

Fixes: 54a73201 ("mlxsw: spectrum: Adjust switchdev ops for VLAN devices")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When we disable learning on bridge port we should still update the
software bridge's FDB when entry pointing to this bridge port is
aged-out. We can otherwise have an inconsistency between software and
hardware tables.

Fixes: 8a1ab5d7 ("mlxsw: spectrum: Implement FDB add/remove/dump for LAG")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When port is put into LISTENING state it shouldn't populate the FDB, so
set the port's STP state in hardware to DISCARDING instead of LEARNING.
It will therefore keep listening to BPDU packets, but discard other
non-control packets and won't perform any learning.

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When STP state is set to DISABLED the port is assumed to be inactive, but
currently we forward packets ingressing through it.

Instead, set the port's STP state in hardware to DISCARDING, which means
it doesn't forward packets or perform any learning, but it does trap
control packets. However, these packets will be dropped by bridge code,
which results in the expected behavior.

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As explained in previous commit, we should always take care of flushing
the FDB in the driver and not rely on bridge code.

We need to distinguish between two cases with regards to LAG:

1) Port is leaving LAG while LAG is bridged (or VLAN devices on top of
it). In this case don't flush the FDB entries pointing to the LAG ID, as
this will affect other ports still member in the LAG. Only flush the FDB
when the last port in the LAG is leaving the bridge.

2) LAG device is leaving the bridge. In this case the CHANGEUPPER event
is simply propagated to each member port, so make each port flush the
FDB in its turn.

Note that emptying a bridged LAG from ports creates an inconsistency
between hardware and software. A user who later (< ageing_time)
re-populates the LAG won't have any FDB entries pointing to the LAG ID
in hardware, but they will be present in the software bridge's FDB.
Currently there is no good solution to this problem, but this will be
addressed by us in the future.

In order to optimize the flushing process, flush by port or LAG ID if
there are no VLAN interfaces on top of the port. Otherwise, flush using
(Port / LAG ID, FID=VID} for each of the lower 4K FIDs. In the case of
VLAN device simply flush using {Port / LAG ID, vFID} with the vFID to
which the VLAN device is mapped to.

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When removing a net device from a bridge we should flush the FDB entries
associated with this net device. Up until now, we relied upon bridge
code to do that for us, but it is possible for user to prevent hardware
from syncing with the software bridge (learning_sync=0), so we need to
flush overselves.

Add the Switch Filtering DB Flush (SFDF) register that is used to flush
FDB entries according to different parameters (per-port, per-FID etc).

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

It is possible for a user to remove a port from a LAG device, while the
LAG device or VLAN devices on top of it are bridged. In these cases,
bridge's teardown sequence is never issued, so we need to take care of
it ourselves.

When LAG's unlinking event is received by port netdev:

1) Traverse its vPorts list and make those member in a bridge leave it.
   They will be deleted later by LAG code.

2) Make the port netdev itself leave its bridge if member in one.

Fixes: 0d65fc13 ("mlxsw: spectrum: Implement LAG port join/leave")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2016-01-25

This series contains updates to i40e only and so I won't continue receiving
patches to fix the same issue (again).

Arnd fixes the driver from causing the compiler whining about uninitialized
variables, so initialize those variables.

Eric fixes the build errors/warnings which were introduced by Anjali
when she added geneve support to i40e.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

intel/i40e/i40e_txrx.c: In function 'i40e_xmit_frame_ring':
intel/i40e/i40e_txrx.c:2367:20: error: 'oiph' may be used uninitialized in this function [-Werror=maybe-uninitialized]
intel/i40e/i40e_txrx.c:2317:16: note: 'oiph' was declared here
intel/i40e/i40e_txrx.c:2367:17: error: 'oudph' may be used uninitialized in this function [-Werror=maybe-uninitialized]
intel/i40e/i40e_txrx.c:2316:17: note: 'oudph' was declared here
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Fixes following build warnings :

drivers/net/ethernet/intel/i40e/i40e_main.c:7057:13: warning:
'i40e_sync_udp_filters_subtask' defined but not used [-Wunused-function]
drivers/net/ethernet/intel/i40e/i40e_main.c:8524:13: warning:
'i40e_add_vxlan_port' defined but not used [-Wunused-function]
drivers/net/ethernet/intel/i40e/i40e_main.c:8569:13: warning:
'i40e_del_vxlan_port' defined but not used [-Wunused-function]
drivers/net/ethernet/intel/i40e/i40e_main.c:8604:13: warning:
'i40e_add_geneve_port' defined but not used [-Wunused-function]
drivers/net/ethernet/intel/i40e/i40e_main.c:8651:13: warning:
'i40e_del_geneve_port' defined but not used [-Wunused-function]

Fixes: 6a899024 ("i40e: geneve tunnel offload support")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Since eliminating send_completion_tid from struct hv_netvsc_packet, we
haven't add proper book keeping for the skb of the batched packet. This
patch fixes this issue and allows the previous skb is properly freed.
Otherwise, a panic may happen.
Thanks to Simon Xiao <sixiao@microsoft.com> for bisecting and analysis.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Recent changes to 'struct flow_keys' (e.g commit d34af823 ("net: Add
VLAN ID to flow_keys")) introduced a performance regression in netvsc
driver. Is problem is, however, not the above mentioned commit but the
fact that netvsc_set_hash() function did some assumptions on the struct
flow_keys data layout and this is wrong.

Get rid of netvsc_set_hash() by switching to skb_get_hash(). This change
will also imply switching to Jenkins hash from the currently used Toeplitz
but it seems there is no good excuse for Toeplitz to stay.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When creating a SIT tunnel with ip tunnel, rtnl_link_ops is not set before
ipip6_tunnel_create is called. When register_netdevice is called, there is
no linkinfo attribute in the NEWLINK message because of that.

Setting rtnl_link_ops before calling register_netdevice fixes that.
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As Arnd Bergmann points out, using CONFIG_ARCH_MXC and/or SOC_IMX28
is wrong if some other ARM platform uses this device - the operation
of the driver would depend on an unrelated ARM platform that might
or might not be set for multi-platform kernels.

Prior to my previous patch, any other platforms using it would have
been broken already due to having the cbd_datlen/cbd_sc fields in
the wrong order, but byte ordering correctly, so no such platforms
can exist and work today.

In any case, it seems likely that only Freescale SoCs use this part,
and those are little-endian on ARM, so CONFIG_ARM is safe for them.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

We are getting many build warnings about:
'bar_start' may be used uninitialized
and
'bar_len' may be used uninitialized

They are not actually uninitialized as dfx_get_bars() will initialize
them properly. But still lets have them initialized just to satisfy the
compiler (gcc 4.8.2).
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Acked-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

We are getting build warning about:
macb.c:2889:13: warning: 'tx_clk' may be used uninitialized in this function
macb.c:2888:11: warning: 'hclk' may be used uninitialized in this function

In reality they are not used uninitialized as clk_init() will initialize
them, this patch will just silence the warning.
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The driver treats the device descriptors as CPU-endian, which appears
to be correct with the default endianness on both ARM (typically LE)
and PowerPC (typically BE) SoCs, indicating that the hardware block
is generated differently. Add endianness annotations and byteswaps as
necessary.

It's not clear that the ifdef there really is correct and shouldn't
just be #ifdef CONFIG_ARM, but I also can't test on anything but the
i.MX6 HummingBoard where this gets it working with a BE kernel.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since commit 76e398a6 ("net: dsa: use switchdev obj for VLAN add/del
ops"), the Marvell 88E6xxx switch has been unable to pass traffic
between ports - any received traffic is discarded by the switch.
Taking a port out of bridge mode and configuring a vlan on it also the
port to start passing traffic.

With the debugfs files re-instated to allow debug of this issue by
comparing the register settings between the working and non-working
case, the reason becomes clear:

     GLOBAL GLOBAL2 SERDES   0    1    2    3    4    5    6
- 7:  1111    707f    2001     2    2    2    2    2    0    2
+ 7:  1111    707f    2001     1    1    1    1    1    0    1

Register 7 for the ports is the default vlan tag register, and in the
non-working setup, it has been set to 2, despite vlan 2 not being
configured.  This causes the switch to drop all packets coming in to
these ports.  The working setup has the default vlan tag register set
to 1, which is the default vlan when none is configured.

Inspection of the code reveals why.  The code prior to this commit
was:

-		for (vid = vlan->vid_begin; vid <= vlan->vid_end; ++vid) {
...
-			if (!err && vlan->flags & BRIDGE_VLAN_INFO_PVID)
-				err = ds->drv->port_pvid_set(ds, p->port, vid);

but the new code is:

+	for (vid = vlan->vid_begin; vid <= vlan->vid_end; ++vid) {
...
+	}
...
+	if (pvid)
+		err = _mv88e6xxx_port_pvid_set(ds, port, vid);

This causes the new code to always set the default vlan to one higher
than the old code.

Fix this.

Fixes: 76e398a6 ("net: dsa: use switchdev obj for VLAN add/del ops")
Cc: <stable@vger.kernel.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is an additional patch to the one already submitted recently.
The previous patch was not complete, and the FCC port lock-up scenario
has been reproduced in lab.
I had an opportunity to check the current patch in lab and the FCC
port lock no longer freezes, while the previous patch still locks-up the
FCC port.
The current patch fixes a pointer arithmetic bug (second bug in the same
line), which leads FCC port lock-up during underrun/collision handling.
Within the tx_startup() function in mac-fcc.c, the address of last BD is
not calculated correctly. As a result of wrong calculation of the last BD
address, the next transmitted BD may be set to an area out of the transmit
BD ring. This actually causes to port lock-up and it is not recoverable.
Signed-off-by: Martin Roth <martin.roth@motorolasolutions.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ESP algorithms using CBC mode require echainiv. Hence INET*_ESP have
to select CRYPTO_ECHAINIV in order to work properly. This solves the
issues caused by a misconfiguration as described in [1].
The original approach, patching crypto/Kconfig was turned down by
Herbert Xu [2].

[1] https://lists.strongswan.org/pipermail/users/2015-December/009074.html
[2] http://marc.info/?l=linux-crypto-vger&m=145224655809562&w=2Signed-off-by: Thomas Egerer <hakke_007@gmx.de>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch extends commit b93d6471 ("sctp: implement the sender side
for SACK-IMMEDIATELY extension") as it didn't white list
SCTP_SACK_IMMEDIATELY on sctp_msghdr_parse(), causing it to be
understood as an invalid flag and returning -EINVAL to the application.

Note that the actual handling of the flag is already there in
sctp_datamsg_from_user().

https://tools.ietf.org/html/rfc7053#section-7

Fixes: b93d6471 ("sctp: implement the sender side for SACK-IMMEDIATELY extension")
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The napi_synchronize() function is defined twice: The definition
for SMP builds waits for other CPUs to be done, while the uniprocessor
variant just contains a barrier and ignores its argument.

In the mvneta driver, this leads to a warning about an unused variable
when we lookup the NAPI struct of another CPU and then don't use it:

ethernet/marvell/mvneta.c: In function 'mvneta_percpu_notifier':
ethernet/marvell/mvneta.c:2910:30: error: unused variable 'other_port' [-Werror=unused-variable]

There are no other CPUs on a UP build, so that code never runs, but
gcc does not know this.

The nicest solution seems to be to turn the napi_synchronize() helper
into an inline function for the UP case as well, as that leads gcc to
not complain about the argument being unused. Once we do that, we can
also combine the two cases into a single function definition and use
if(IS_ENABLED()) rather than #ifdef to make it look a bit nicer.

The warning first came up in linux-4.4, but I failed to catch it
earlier.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: f8642885 ("net: mvneta: Statically assign queues to CPUs")
Signed-off-by: David S. Miller <davem@davemloft.net>

Several times already this has been reported as kasan reports caused by
syzkaller and trinity and people always looked at RCU races, but it is
much more simple. :)

In case we bind a pptp socket multiple times, we simply add it to
the callid_sock list but don't remove the old binding. Thus the old
socket stays in the bucket with unused call_id indexes and doesn't get
cleaned up. This causes various forms of kasan reports which were hard
to pinpoint.

Simply don't allow multiple binds and correct error handling in
pptp_bind. Also keep sk_state bits in place in pptp_connect.

Fixes: 00959ade ("PPTP: PPP over IPv4 (Point-to-Point Tunneling Protocol)")
Cc: Dmitry Kozlov <xeb@mail.ru>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Reported-by: Dave Jones <davej@codemonkey.org.uk>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

For interrupt controller that doesn't support irq_disable and hardware
with level interrupt, an extra interrupt may be pending. This patch fixes
the issue by setting IRQ_DISABLE_UNLAZY flag for the interrupt line,
as suggested by,

'commit e9849777 ("genirq: Add flag to force mask in
		       disable_irq[_nosync]()")'
Signed-off-by: Iyappan Subramanian <isubramanian@apm.com>
Tested-by: Toan Le <toanle@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Dmitry reported a struct pid leak detected by a syzkaller program.

Bug happens in unix_stream_recvmsg() when we break the loop when a
signal is pending, without properly releasing scm.

Fixes: b3ca9b02 ("net: fix multithreaded signal handling in unix recv routines")
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Rainer Weikusat <rweikusat@mobileactivedefense.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The cgroup methods are no longer used after baac50bb ("net:
tcp_memcontrol: simplify linkage between socket and page counter").
The hunk to delete them was included in the original patch but must
have gotten lost during conflict resolution on the way upstream.

Fixes: baac50bb ("net: tcp_memcontrol: simplify linkage between socket and page counter")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the lan87xx_read_status function is getting called the
energy detect mode is enabled again even if it has been
disabled by device tree.

Added private struct to check the energy detect status.
Signed-off-by: Teresa Remmet <t.remmet@phytec.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jisheng Zhang says:

====================
net: mvneta: support more than one clk

Some platforms may provide more than one clk for the mvneta IP, for
example Marvell BG4CT provides "core" clk for the mac core, and "axi"
clk for the AXI bus logic.

This series tries to addess the "more than one clk" issue. Note: to
support BG4CT, we have lots of refactor work to do, eg. BG4CT doesn't
have mbus concept etc.

Since v2:
 - Name the optional clock as "bus", which is a bit more flexible.

Since v1:
 - Add Thomas Acks to patch1 and patch2.
 - make sure the headers are really sorted (some headers are still
   unsorted in v1).
 - disable axi clk before disabling core clk, Thank Thomas.
 - update dt binding as Thomas suggested.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some platforms may provide more than one clk for the mvneta IP, for
example Marvell BG4CT provides one clk for the mac core, and one
clk for the AXI bus logic. Obviously this bus clk also need to
be enabled. This patch adds this optional "bus" clk support.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some platforms may provide more than one clk for the mvneta IP, for
example Marvell BG4CT provides one clk for the mac core, and one
clk for the AXI bus logic.

To support for more than one clock, we'll need to distinguish between
the clock by name. Change clock probing to first try to get "core"
clock before falling back to unnamed clock.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Sorting the headers in alphabetic order will help to reduce the conflict
when adding new headers in the future.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When s->type is T_REG_64, the high 32bits are lost in val. This patch
fixes this trivial issue.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Fixes: 9b0cdefa ("net: mvneta: add ethtool statistics")
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch replace the assoication between dsaf and enet from string
matching to object reference. It requires the DTS to be updated within
BIOS. Thanks god it can be done for all released boards.
Signed-off-by: Kejian Yan <yankejian@huawei.com>
Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Yisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Neal reported crashes with this stack trace :

 RIP: 0010:[<ffffffff8c57231b>] tcp_v4_send_ack+0x41/0x20f
...
 CR2: 0000000000000018 CR3: 000000044005c000 CR4: 00000000001427e0
...
  [<ffffffff8c57258e>] tcp_v4_reqsk_send_ack+0xa5/0xb4
  [<ffffffff8c1a7caa>] tcp_check_req+0x2ea/0x3e0
  [<ffffffff8c19e420>] tcp_rcv_state_process+0x850/0x2500
  [<ffffffff8c1a6d21>] tcp_v4_do_rcv+0x141/0x330
  [<ffffffff8c56cdb2>] sk_backlog_rcv+0x21/0x30
  [<ffffffff8c098bbd>] tcp_recvmsg+0x75d/0xf90
  [<ffffffff8c0a8700>] inet_recvmsg+0x80/0xa0
  [<ffffffff8c17623e>] sock_aio_read+0xee/0x110
  [<ffffffff8c066fcf>] do_sync_read+0x6f/0xa0
  [<ffffffff8c0673a1>] SyS_read+0x1e1/0x290
  [<ffffffff8c5ca262>] system_call_fastpath+0x16/0x1b

The problem here is the skb we provide to tcp_v4_send_ack() had to
be parked in the backlog of a new TCP fastopen child because this child
was owned by the user at the time an out of window packet arrived.

Before queuing a packet, TCP has to set skb->dev to NULL as the device
could disappear before packet is removed from the queue.

Fix this issue by using the net pointer provided by the socket (being a
timewait or a request socket).

IPv6 is immune to the bug : tcp_v6_send_response() already gets the net
pointer from the socket if provided.

Fixes: 168a8f58 ("tcp: TCP Fast Open Server - main code path")
Reported-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jerry Chu <hkchu@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When configuring checksums on UDP tunnels, the flags are different
for IPv4 vs. IPv6 (and reversed). However, when lightweight tunnels
are enabled the flags used are always the IPv4 versions, which are
ignored in the IPv6 code paths. This uses the correct IPv6 flags, so
checksums can be controlled appropriately.

Fixes: a725e514 ("vxlan: metadata based tunneling for IPv6")
Fixes: abe492b4 ("geneve: UDP checksum configuration via netlink")
Signed-off-by: Jesse Gross <jesse@kernel.org>
Acked-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Florian Fainelli says:

====================
net: phy: Finally fix PHY_IGNORE_INTERRUPTS

This patch series finally fixes how PHY_IGNORE_INTERRUPTS are treated by
avoiding to poll the PHY *and* getting notified from link state changes by the
Ethernet MAC interrupt service routine.

Tested with bcmgenet since this is the HW that I have access to.

Targetting the "net" tree since these are bugfixes, but I would like Woojun and
Andrew to take a look and test that on their respective HW setups as well.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

By the time we execute bcmgenet_mii_probe(), the MDIO bus structure has
long been allocated and registered. Overirring the PHY interrupt using
the MDIO bus structure has no chance to work anymore, because
of_mdiobus_register() has call phy_device_create() for use, which copied
the MDIO bus address's irq for the PHY into the PHY device "irq" member.

Since we do have a proper reference to a PHY device in
bcmgenet_mii_probe(), just assign the desired IRQ value here.

Fixes: aa09677c ("net: bcmgenet: add MDIO routines")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 5ea94e76 ("phy: add phy_mac_interrupt()") to use with
PHY_IGNORE_INTERRUPT added a cancel_work_sync() into phy_mac_interrupt()
which is allowed to sleep, whereas phy_mac_interrupt() is expected to be
callable from interrupt context.

Now that we have fixed how the PHY state machine treats
PHY_IGNORE_INTERRUPT with respect to state changes, we can just set the
new link state, and queue the PHY state machine for execution so it is
going to read the new link state.

For that to work properly, we need to update phy_change() not to try to
invoke any interrupt callbacks if we have configured the PHY device for
PHY_IGNORE_INTERRUPT, because that PHY device and its driver are not
required to implement those.

Fixes: 5ea94e76 ("phy: add phy_mac_interrupt() to use with PHY_IGNORE_INTERRUPT")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 2c7b4921 ("phy: fix the use of PHY_IGNORE_INTERRUPT") changed
a hunk in phy_state_machine() in the PHY_RUNNING case which was not
needed. The change essentially makes the PHY library treat PHY devices
with PHY_IGNORE_INTERRUPT to keep polling for the PHY device, even
though the intent is not to do it.

Fix this by reverting that specific hunk, which makes the PHY state
machine wait for state changes, and stay in the PHY_RUNNING state for as
long as needed.

Fixes: 2c7b4921 ("phy: fix the use of PHY_IGNORE_INTERRUPT")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>