It was reported that a considerable amount of cycles were spent on the
expensive indirect calls on fib6_rule_lookup. This patch introduces an
inline helper called pol_route_func that uses the indirect_call_wrappers
to avoid the indirect calls.

This patch saves around 50ns per call.

Performance was measured on the receiver by checking the amount of
syncookies that server was able to generate under a synflood load.

Traffic was generated using trafgen[1] which was pushing around 1Mpps on
a single queue. Receiver was using only one rx queue which help to
create a bottle neck and make the experiment rx-bounded.

These are the syncookies generated over 10s from the different runs:

Whithout the patch:
TcpExtSyncookiesSent            3553749            0.0
TcpExtSyncookiesSent            3550895            0.0
TcpExtSyncookiesSent            3553845            0.0
TcpExtSyncookiesSent            3541050            0.0
TcpExtSyncookiesSent            3539921            0.0
TcpExtSyncookiesSent            3557659            0.0
TcpExtSyncookiesSent            3526812            0.0
TcpExtSyncookiesSent            3536121            0.0
TcpExtSyncookiesSent            3529963            0.0
TcpExtSyncookiesSent            3536319            0.0

With the patch:
TcpExtSyncookiesSent            3611786            0.0
TcpExtSyncookiesSent            3596682            0.0
TcpExtSyncookiesSent            3606878            0.0
TcpExtSyncookiesSent            3599564            0.0
TcpExtSyncookiesSent            3601304            0.0
TcpExtSyncookiesSent            3609249            0.0
TcpExtSyncookiesSent            3617437            0.0
TcpExtSyncookiesSent            3608765            0.0
TcpExtSyncookiesSent            3620205            0.0
TcpExtSyncookiesSent            3601895            0.0

Without the patch the average is 354263 pkt/s or 2822 ns/pkt and with
the patch the average is 360738 pkt/s or 2772 ns/pkt which gives an
estimate of 50 ns per packet.

[1] http://netsniff-ng.org/

Changelog since v1:
 - Change ordering in the ICW (Paolo Abeni)

Cc: Luigi Rizzo <lrizzo@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Brian Vazquez <brianvv@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There are many places where 2 annotations are not enough. This patch
adds INDIRECT_CALL_3 and INDIRECT_CALL_4 to cover such cases.
Signed-off-by: Brian Vazquez <brianvv@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fixes a spelling typo in spectrum_dcb.c
Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Keepalived can set global static ip routes or virtual ip routes dynamically
following VRRP protocol states. Using a dedicated rtm_protocol will help
keeping track of it.

Changes in v2:
 - fix tab/space indenting
Signed-off-by: Alexandre Cassen <acassen@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Igor Russkikh says:

====================
net: atlantic: additional A2 features

This patchset adds more features to A2:
 * half duplex rates;
 * EEE;
 * flow control;
 * link partner capabilities reporting;
 * phy loopback.

Feature-wise A2 is almost on-par with A1 save for WoL and filtering, which
will be submitted as separate follow-up patchset(s).
====================
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds the phy loopback support on A2.
Signed-off-by: Dmitry Bogdanov <dbogdanov@marvell.com>
Signed-off-by: Mark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds link partner capabilities reporting support on A2.
In particular, the following capabilities are available for reporting:
* link rate;
* EEE;
* flow control.
Signed-off-by: Dmitry Bogdanov <dbogdanov@marvell.com>
Signed-off-by: Mark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds flow control support on A2.
Co-developed-by: Dmitry Bogdanov <dbogdanov@marvell.com>
Signed-off-by: Dmitry Bogdanov <dbogdanov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds EEE support on A2.
Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Co-developed-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch removes 2.5G baseX wrong usage/reporting, since it shouldn't have
been mixed with baseT.
Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Mark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for 10M/100M/1G half duplex rates, which are
supported by A2 in additional to full duplex rates supported by A1.
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

arg cannot be NULL since its already being dereferenced
before. Remove the redundant NULL check.
Signed-off-by: Gaurav Singh <gaurav1086@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Convert the mtk_eth_soc driver to use the finalised link parameters in
mac_link_up() rather than the parameters in mac_config().
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Vladimir Oltean says:

====================
Multicast improvement in Ocelot and Felix drivers

This series makes some basic multicast forwarding functionality work for
Felix DSA and for Ocelot switchdev. IGMP/MLD snooping in Felix is still
missing, and there are other improvements to be made in the general area
of multicast address filtering towards the CPU, but let's get these
hardware-specific fixes out of the way first.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The current procedure for installing a multicast address is hardcoded
for IPv4. But, in the ocelot hardware, there are 3 different procedures
for IPv4, IPv6 and for regular L2 multicast.

For IPv6 (33-33-xx-xx-xx-xx), it's the same as for IPv4
(01-00-5e-xx-xx-xx), except that the destination port mask is stuffed
into first 2 bytes of the MAC address except into first 3 bytes.

For plain Ethernet multicast, there's no port-in-address stuffing going
on, instead the DEST_IDX (pointer to PGID) is used there, just as for
unicast. So we have to use one of the nonreserved multicast PGIDs that
the hardware has allocated for this purpose.

This patch classifies the type of multicast address based on its first
bytes, then redirects to one of the 3 different hardware procedures.

Note that this gives us a really better way of redirecting PTP frames
sent at 01-1b-19-00-00-00 to the CPU. Previously, Yangbo Lu tried to add
a trapping rule for PTP EtherType but got a lot of pushback:

https://patchwork.ozlabs.org/project/netdev/patch/20190813025214.18601-5-yangbo.lu@nxp.com/

But right now, that isn't needed at all. The application stack (ptp4l)
does this for the PTP multicast addresses it's interested in (which are
configurable, and include 01-1b-19-00-00-00):

	memset(&mreq, 0, sizeof(mreq));
	mreq.mr_ifindex = index;
	mreq.mr_type = PACKET_MR_MULTICAST;
	mreq.mr_alen = MAC_LEN;
	memcpy(mreq.mr_address, addr1, MAC_LEN);

	err1 = setsockopt(fd, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq,
			  sizeof(mreq));

Into the kernel, this translates into a dev_mc_add on the switch network
interfaces, and our drivers know that it means they should translate it
into a host MDB address (make the CPU port be the destination).
Previously, this was broken because all mdb addresses were treated as
IPv4 (which 01-1b-19-00-00-00 obviously is not).
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The current iterators are impossible to understand at first glance
without switching back and forth between the definitions and their
actual use in the for loops.

So introduce some convenience names to help readability.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This adds the mdb hooks in felix and exports the mdb functions from
ocelot.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When used in DSA mode (as seen in Felix), the DEST_IDX in the MAC table
should point to the PGID for the CPU port (PGID_CPU) and not for the
Ethernet port where the CPU queues are redirected to (also known as Node
Processor Interface - NPI).

Because for Felix this distinction shouldn't really matter (from DSA
perspective, the NPI port _is_ the CPU port), make the ocelot library
act upon the CPU port when NPI mode is enabled. This has no effect for
the mscc_ocelot driver for VSC7514, because that does not use NPI (and
ocelot->npi is -1).
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ocelot hardware designers have made some hacks to support multicast
IPv4 and IPv6 addresses. Normally, the MAC table matches on MAC
addresses and the destination ports are selected through the DEST_IDX
field of the respective MAC table entry. The DEST_IDX points to a Port
Group ID (PGID) which contains the bit mask of ports that frames should
be forwarded to. But there aren't a lot of PGIDs (only 80 or so) and
there are clearly many more IP multicast addresses than that, so it
doesn't scale to use this PGID mechanism, so something else was done.
Since the first portion of the MAC address is known, the hack they did
was to use a single PGID for _flooding_ unknown IPv4 multicast
(PGID_MCIPV4 == 62), but for known IP multicast, embed the destination
ports into the first 3 bytes of the MAC address recorded in the MAC
table.

The VSC7514 datasheet explains it like this:

    3.9.1.5 IPv4 Multicast Entries

    MAC table entries with the ENTRY_TYPE = 2 settings are interpreted
    as IPv4 multicast entries.
    IPv4 multicasts entries match IPv4 frames, which are classified to
    the specified VID, and which have DMAC = 0x01005Exxxxxx, where
    xxxxxx is the lower 24 bits of the MAC address in the entry.
    Instead of a lookup in the destination mask table (PGID), the
    destination set is programmed as part of the entry MAC address. This
    is shown in the following table.

    Table 78: IPv4 Multicast Destination Mask

        Destination Ports            Record Bit Field
        ---------------------------------------------
        Ports 10-0                   MAC[34-24]

    Example: All IPv4 multicast frames in VLAN 12 with MAC 01005E112233 are
    to be forwarded to ports 3, 8, and 9. This is done by inserting the
    following entry in the MAC table entry:
    VALID = 1
    VID = 12
    MAC = 0x000308112233
    ENTRY_TYPE = 2
    DEST_IDX = 0

But this procedure is not at all what's going on in the driver. In fact,
the code that embeds the ports into the MAC address looks like it hasn't
actually been tested. This patch applies the procedure described in the
datasheet.

Since there are many other fixes to be made around multicast forwarding
until it works properly, there is no real reason for this patch to be
backported to stable trees, or considered a real fix of something that
should have worked.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ido Schimmel says:

====================
mlxsw: Offload TC action pedit munge tcp/udp sport/dport

Petr says:

On Spectrum-2 and Spectrum-3, it is possible to overwrite L4 port number of
a TCP or UDP packet in the ACL engine. That corresponds to the pedit munges
of tcp and udp sport resp. dport fields. Offload these munges on the
systems where they are supported.

The current offloading code assumes that all systems support the same set
of fields. This now changes, so in patch #1 first split handling of pedit
munges by chip type. The analysis of which packet field a given munge
describes is kept generic.

Patch #2 introduces the new flexible action fields. Patch #3 then adds the
new pedit fields, and dispatches on them on Spectrum>1.

Patch #4 adds a forwarding selftest for pedit dsfield, applicable to SW as
well as HW datapaths.
====================
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a test that checks that pedit adjusts port numbers of tcp and udp
packets.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Spectrum-2 supports an ACL action L4_PORT, which allows TCP and UDP source
and destination port number change. Offload suitable mangles to this
action.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add fields related to L4_PORT_ACTION, which is used for changing of TCP and
UDP port numbers.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Certain ACL actions are only available on some Spectrum revisions. In
particular, L4_PORT_ACTION is not available on Spectrum-1. Introduce a
new ops struct intended to hold these differences, mlxsw_sp_rulei_ops.
Prime it with a sole member, act_mangle_field, meant for handling of
pedit mangles.

Create two ops structures, one for Spectrum-1, the other for Spectrum-2
and above. Add callbacks for act_mangle_field and dispatch to the common
handler.

Invoke mlxsw_sp_rulei_ops.act_mangle_field from the field mangler
instead of calling the common handler directly.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Maxim Kochetkov says:

====================
Add Marvell 88E1340S, 88E1548P support

This patch series add new PHY id support.
Russell King asked to use single style for referencing functions.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for this new phy ID.
Signed-off-by: Maxim Kochetkov <fido_max@inbox.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for this new phy ID.
Signed-off-by: Maxim Kochetkov <fido_max@inbox.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

The kernel in general does not use &func referencing format.
Signed-off-by: Maxim Kochetkov <fido_max@inbox.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

Heiner Kallweit says:

====================
r8169: mark device as detached in PCI D3 and improve locking

Mark the netdevice as detached whenever parent is in PCI D3hot and not
accessible. This mainly applies to runtime-suspend state.
In addition take RTNL lock in suspend calls, this allows to remove
the driver-specific mutex and improve PM callbacks in general.
====================
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Simplify rtl8169_runtime_resume() by calling rtl8169_resume().
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Now that the critical sections are protected with RTNL lock, we don't
need a separate mutex any longer.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Most relevant ops (open, close, ethtool ops) are protected with RTNL
lock by net core. Make sure that such ops can't be interrupted by
e.g. (runtime-)suspending by taking the RTNL lock in suspend ops
and the PCI error handler.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Factor out bringing device up to a new function rtl8169_up(), similar
to rtl8169_down() for bringing the device down.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Because the netdevice is marked as detached now when parent is not
accessible we can remove quite some checks.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Mark the netdevice as detached whenever we go into PCI D3hot.
This allows to remove some checks e.g. from ethtool ops because
dev_ethtool() checks for netif_device_present() in the beginning.

In this context move waking up the queue out of rtl_reset_work()
because in cases where netif_device_attach() is called afterwards
the queue should be woken up by the latter function only.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A netdevice may be marked as detached because the parent is
runtime-suspended and not accessible whilst interface or link is down.
An example are PCI network devices that go into PCI D3hot, see e.g.
__igc_shutdown() or rtl8169_net_suspend().
If netdevice is down and marked as detached we can only open it if
we runtime-resume it before __dev_open() calls netif_device_present().

Therefore, if netdevice is detached, try to runtime-resume the parent
and only return with an error if it's still detached.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Martin Blumenstingl says:

====================
prepare dwmac-meson8b for G12A specific initialization

Some users are reporting that RGMII (and sometimes also RMII) Ethernet
is not working for them on G12A/G12B/SM1 boards. Upon closer inspection
of the vendor code for these SoCs new register bits are found.

It's not clear yet how these registers work. Add a new compatible string
as the first preparation step to improve Ethernet support on these SoCs.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Amlogic Meson G12A, G12B and SM1 have the same (at least as far as we
know at the time of writing) PRG_ETHERNET glue register implementation.
This implementation however is slightly different from AXG as it now has
an undocument "auto cali idx val" register in PRG_ETH1[17:16] which
seems to be related to RGMII Ethernet.

Add a new compatible string for G12A SoCs so the logic for this new
register can be implemented in the future.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Amlogic Meson G12A, G12B and SM1 have the same (at least as far as we
know at the time of writing) PRG_ETHERNET glue register implementation.
This implementation however is slightly different from AXG as it now has
an undocument "auto cali idx val" register in PRG_ETH1[17:16] which
seems to be related to RGMII Ethernet.

Add a compatible string for G12A and newer so the new registers can be
used.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>