Doug Berger says:

====================
net: bcmgenet: Turn on offloads by default

This commit stack is based on Florian's commit 4e8aedfe78c7 ("net:
systemport: Turn on offloads by default") and enables the offloads for
the bcmgenet driver by default.

The first commit adds support for the HIGHDMA feature to the driver.

The second converts the Tx checksum implementation to use the generic
hardware logic rather than the deprecated IP centric methods.

The third modifies the Rx checksum implementation to use the hardware
offload to compute the complete checksum rather than filtering out bad
packets detected by the hardware's IP centric implementation. This may
increase processing load by passing bad packets to the network stack,
but it provides for more flexible handling of packets by the network
stack without requiring software computation of the checksum.

The remaining commits mirror the extensions Florian made to the sysport
driver to retain symmetry with that driver and to make the benefits of
the hardware offloads more ubiquitous.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

When inserting the TSB, keep track of how many times we had to do
it and if there was a failure in doing so, this helps profile the
driver for possibly incorrect headroom settings.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

During bcmgenet_put_tx_csum() make sure we differentiate a SKB
headroom re-allocation failure from the normal swap and replace
path.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can turn on the RX/TX checksum offloads and the scatter/gather
features by default and make sure that those are properly reflected
back to e.g: stacked devices such as VLAN.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

During driver resume and open, the HW may have lost its context/state,
utilize bcmgenet_set_features() to make sure we do restore the correct
set of features that were previously configured.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In preparation for unconditionally enabling TX and RX checksum
offloads, refactor bcmgenet_set_features() a bit such that
__netdev_update_features() during register_netdev() can make sure
that features are correctly programmed during network device
registration.

Since we can now be called during register_netdev() with clocks
gated, we need to temporarily turn them on/off in order to have a
successful register programming.

We also move the CRC forward setting read into
bcmgenet_set_features() since priv->crc_fwd_en matters while
turning on RX checksum offload, that way we are guaranteed they
are in sync in case we ever add support for NETIF_F_RXFCS at some
point in the future.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit updates the Rx checksum offload behavior of the driver
to use the more generic CHECKSUM_COMPLETE method that supports all
protocols over the CHECKSUM_UNNECESSARY method that only applies
to some protocols known by the hardware.

This behavior is perceived to be superior.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The GENET hardware should be capable of generating IP checksums
using the NETIF_F_HW_CSUM feature, so switch to using that feature
instead of the depricated NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit configures the DMA masks for the GENET driver and
sets the NETIF_F_HIGHDMA flag to report support of the feature.
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

SYSTEMPORT is capabable of doing up to 40-bit of physical addresses, set
an appropriate DMA mask to permit that.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jakub Kicinski says:

====================
nfp: tls: implement the stream sync RX resync

This small series adds support for using the device
in stream scan RX resync mode which improves the RX
resync success rate. Without stream scan it's pretty
much impossible to successfully resync a continuous
stream.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The simple RX resync strategy controlled by the kernel does not
guarantee as good results as if the device helps by detecting
the potential record boundaries and keeping track of them.

We've called this strategy stream scan in the tls-offload doc.

Implement this strategy for the NFP. The device sends a request
for record boundary confirmation, which is then recorded in
per-TLS socket state and responded to once record is reached.
Because the device keeps track of records passing after the
request was sent the response is not as latency sensitive as
when kernel just tries to tell the device the information
about the next record.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is currently no way for driver to reliably check that
the socket it has looked up is in fact RX offloaded. Add
a helper. This allows drivers to catch misbehaving firmware.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Make nfp_net_parse_meta() take a packet pointer and return
a drop/no drop decision. Right now it returns the end of
metadata and caller compares it to the packet pointer.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

John Hurley says:

====================
Add ipv6 tunnel support to NFP

The following patches add support for IPv6 tunnel offload to the NFP
driver.

Patches 1-2 do some code tidy up and prepare existing code for reuse in
IPv6 tunnels.
Patches 3-4 handle IPv6 tunnel decap (match) rules.
Patches 5-8 handle encap (action) rules.
Patch 9 adds IPv6 support to the merge and pre-tunnel rule functions.

v1->v2:
- fix compiler warning when building without CONFIG_IPV6 set -
  Jakub Kicinski (patch 7)
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Both pre-tunnel match rules and flow merge functions parse compiled
match/action fields for validation.

Update these validation functions to include IPv6 match and action fields.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

FW sends an update of IPv6 tunnels that are active in a given period. Use
this information to update the kernel table so that neighbour entries do
not time out when active on the NIC.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A notifier is used to track route changes in the kernel. If a change is
made to a route that is offloaded to fw then an update is sent to the NIC.
The driver tracks all routes that are offloaded to determine if a kernel
change is of interest.

Extend the notifier to track IPv6 route changes and create a new list that
stores offloaded IPv6 routes. Modify the IPv4 route helper functions to
accept varying address lengths. This way, the same core functions can be
used to handle IPv4 and IPv6.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When fw does not know the next hop for an IPv6 tunnel, it sends a request
to the driver.

Handle this request by doing a route lookup on the IPv6 address and
offloading the next hop to the fw neighbour table.

Similar functions already exist to handle IPv4 no neighbour requests. To
avoid confusion, append these functions with the _ipv4 tag. There is no
change in functionality with this.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The IPv4 set tunnel action allows the setting of tunnel metadata such as
the TTL and ToS values. The pre-tunnel action includes the destination IP
address and is used to calculate the next hop from from the neighbour
table.

Much of the IPv4 tunnel actions can be reused for IPv6 tunnels. Change the
names of associated functions and structs to remove the IPv4 identifier
and make minor modifcations to support IPv6 tunnel actions.

Ensure the pre-tunnel action contains the IPv6 address along with an
identifying flag when an IPv6 tunnel action is required.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fw requires a list of IPv6 addresses that are used as tunnel endpoints to
enable correct decap of tunneled packets.

Store a list of IPv6 endpoints used in rules with a ref counter to track
how many times it is in use. Offload the entire list any time a new IPv6
address is added or when an address is removed (ref count is 0).
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPv6 tunnel matches are now supported by firmware. Modify the NFP driver
to compile these match rules. IPv6 matches are handled similar to IPv4
tunnels with the difference the address length. The type of tunnel is
indicated by the same bitmap that is used in IPv4 with an extra bit
signifying that the IPv6 variation should be used.

Only compile IPv6 tunnel matches when the fw features symbol indicated
that they are compatible with the currently loaded fw.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPv4 UDP and GRE tunnel match rule compile helpers share functions for
compiling fields such as IP addresses. However, they handle fields such
tunnel IDs differently.

Create new helper functions for compiling GRE and UDP tunnel key data.
This is in preparation for supporting IPv6 tunnels where these new
functions can be reused.

This patch does not change functionality.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In kernel 5.1, the flow offload API was introduced along with a helper
function to extract the flow_rule from the TC offload struct. Each of the
match helper functions are passed the offload struct and extract the flow
rule to a local variable.

Simplify the code while also removing the extra compat and local variable
calls by extracting the rule once in the main match handler, and passing
a reference to the rule direct to each helper.

This patch does not change driver functionality.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In hdlcdrv_register, failure to register the driver causes a crash.
The three callers of hdlcdrv_register all pass valid pointers and
do not fail. The patch eliminates the unnecessary BUG_ON assertion.
Signed-off-by: Aditya Pakki <pakki001@umn.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

Presently, at boot time, the comphys are enabled. For firmware
compatibility reasons, the comphy driver does not power down the
comphys at boot. Consequently, the ethernet comphys are left active
until the network interfaces are brought through an up/down cycle.

If the port is never used, the port wastes power needlessly. Arrange
for the ethernet comphys to be cycled by the mvpp2 driver as if the
interface went through an up/down cycle during driver probe, thereby
powering them down.

This saves:
  270mW per 10G SFP+ port on the Macchiatobin Single Shot (eth0/eth1)
  370mW per 10G PHY port on the Macchiatobin Double Shot (eth0/eth1)
  160mW on the SFP port on either Macchiatobin flavour (eth3)
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Acked-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Report a rate-limited error if we fail to read the SFP soft status,
and preserve the current status in that case. This avoids I2C bus
errors from triggering a link flap.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Russell King says:

====================
phylib consolidation

Over the last few releases, there has been a push to clean up and
consolidate the phylib code. Some cases have been missed, and this
series catches those cases.

1. Remove redundant .aneg_done initialisers; calling genphy_aneg_done()
   for clause 22 PHYs is the default when .aneg_done is not set.

2. Some PHY drivers manually set phydev->pause and phydev->asym_pause,
   but we have a helper for this - phy_resolve_aneg_pause(), introduced
   in 2d880b87 ("net: phy: extract pause mode").  Use this in the
   lxt, marvell and uPD60620 drivers.

   Incidentally, this brings up the question whether marvell fiber mode
   is correctly interpreting and advertising the pause parameters.

3. Add a genphy_check_and_restart_aneg() helper, which complements the
   clause 45 version of this. This will be useful for PHY drivers that
   open code this logic (e.g. marvell.c)

4. Add a genphy_read_status_fixed() helper to read the fixed-mode
   status from a clause 22 PHY.  lxt and marvell both contain copies
   of this code, so convert them over.

5. Arrange marvell driver to use genphy_read_lpa() for copper mode.
   This needs some rearrangement of the code in
   marvell_read_status_page_an(), but preserves using the PHY specific
   status register to derive the current negotiation results.

6. Simplify the marvell driver so we can use the
   genphy_read_status_fixed() helper directly rather than
   marvell_read_status_page_fixed().

7. Use positive logic in the marvell driver to determine the link
   state, and get rid of the REGISTER_LINK_STATUS definition; we
   already have a definition for this.

8. The marvell driver reads the PHY specific status register multiple
   times when determining the status: once in marvell_update_link()
   and again in marvell_read_status_page_an(). This is a waste;
   rearrange to read the status register once, and pass its value into
   marvell_read_status_page_an().  We preserve using
   genphy_update_link() for the copper side.

9. The marvell driver was using private clause 37 definitions, but we
   have clause 37 definitions in uapi/linux/mii.h. Use the generic
   definitions.

10. Switch the marvell driver to use phy_modify_changed() to modify
    the fiber advertisement.

11. Switch the marvell driver to use genphy_check_and_restart_aneg()
    introduced above rather than open-coding this functionality.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Use the helper to check and restart autonegotiation for the marvell
fiber page negotiation setting.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use phy_modify_changed() to change the fiber advertisement register
rather than open coding this functionality.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use existing clause 37 advertising/link partner definitions rather than
private ones for the advertisement registers.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

marvell_read_status_page_an() always reads the PHY status register, but
marvell_update_link() has already done this.  Rather than wastefully
reading the register twice in quick succession, read it once in
marvell_read_status_page() and use the result for both.

This makes marvell_update_link() rather pointless, so move it into
marvell_read_status_page().
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Rather than using negative logic:

	if (there is no link)
		set link = 0
	else
		set link = 1

use the more natural positive logic:

	if (there is link)
		set link = 1
	else
		set link = 0
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move the initialisation of the link partner state earlier, inside
marvell_read_status_page(), so we don't have the same initialisation
scattered amongst the other files.  This is in a similar place to
the genphy implementation, so would result in the same behaviour if
a PHY read error occurs.

This allows us to get rid of marvell_read_status_page_fixed(), which
became a pointless wrapper around genphy_read_status_fixed().
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Rearrange the Marvell PHY driver to use genphy_read_lpa() rather than
open-coding this functionality.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There are two drivers and generic code which contain exactly the same
code to read the status of a PHY operating without autonegotiation
enabled. Rather than duplicate this code, provide a helper to read
this information.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a helper for restarting autonegotiation(), similar to the clause 45
variant.  Use it in __genphy_config_aneg()
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Several drivers code their own version of this, working from the LPA
register, after setting the ethtool link partner advertisement bitmask.
Use the generic function instead.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remove initialisers that set .aneg_done to genphy_aneg_done - this is
the default for clause 22 PHYs, so the initialiser is redundant.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For ARCHs that don't support 64 bits division we need to use the
helpers.

Fixes: b60189e0 ("net: stmmac: Integrate EST with TAPRIO scheduler API")
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>