In order to cook skbs in the same way than Ethernet drivers,
it is probably better to not use GFP_KERNEL, but rather
use the GFP_ATOMIC and PFMEMALLOC mechanisms provided by
netdev_alloc_frag().

This would allow to use tun driver even in memory stress
situations, especially if swap is used over this tun channel.

Fixes: 90e33d45 ("tun: enable napi_gro_frags() for TUN/TAP driver")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Petar Penkov <peterpenkov96@gmail.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Martin Blumenstingl says:

====================
IP101GR: devicetree based configuration of SEL_INTR32

The IP101GR is a 32-pin QFN package variant of the IP101G/IP101GA
Ethernet PHY. Due to it's limited amount of pins the RXER (receive
error) and INTR32 (interrupt) functions share pin 21.

The goal of this series is:
- some small cleanups in patches 3, 4 and 5
- allowing the kernel to detect IRQ floods on boards where the IP101GR
  is configured in RXER mode but the RXER line is configured on the
  host SoC as interrupt line (patch 6)
- configuration of the SEL_INTR32 register so we can use the interrupt
  function on boards where the RXER/INTR32 pin (pin 21) is routed to
  one of the host SoC's interrupt inputs (patches 1, 2, 7)

A use-case where this is needed is the Endless Mini (EC-100). I have
tested my changes on that board. This also confirms that Heiner
Kallweit's recent icplus.c PHY driver changes are working (at least on
my setup).

This series is based on net-next commit 7c460cf9 ("net: aquantia:
fix spelling mistake "specfield" -> "specified"")

Changes since v1 at [0]:
- collected Andrew's Reviewed-by's (thank you!)
- updated description of patch #2 to explain why two properties were
  added instead of adding an "this is a IP101GR" property
- validate that there's no conflicting configuration in patch #7
- rebased on top of latest net-next

[0] https://patchwork.ozlabs.org/cover/999371/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The IP101GR is a 32-pin QFN package variant of the IP101G/IP101GA
Ethernet PHY. Due to it's limited amount of pins the RXER (receive
error) and INTR32 (interrupt) functions share pin 21.
By default the PHY is configured to output the "receive error" status on
pin 21. Depending on the board layout and requirements we may want to
re-configure the PHY to output the interrupt signal there.

The mode of pin 21 can be configured in the "Digital I/O Specific
Control Register" (register 29), bit 2:
- 0 = RXER function
- 1 = INTR(32) function

Depending on the devicetree configuration we will now:
- change the mode to either ther RXER or INTR32 function
- keep the SEL_INTR32 value set by the bootloader (default) if no
  configuration is provided (to ensure that we're not breaking existing
  boards)
- error out if conflicting configuration is given (RXER and INTR32 mode
  are enabled at the same time)
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The IP101A_G_IRQ_CONF_STATUS register has bits to detect which
interrupts have fired. Implement the .did_interrupt callback to let the
PHY core know whether the interrupt was for this specific PHY.

This is useful for debugging interrupt problems with 32-pin IP101GR PHYs
where the interrupt line is shared with the RX_ERR (receive error
status) signal. The default values are:
- RX_ERR is enabled by default (LOW means that there is no receive
  error)
- the PHY's interrupt line is configured "active low" by default

Without any additional changes there is a flood of interrupts if the
RX_ERR/INTR32 signal is configured in RX_ERR mode (which is the
default). Having a did_interrupt ensures that the PHY core returns
IRQ_NONE instead of endlessly triggering the PHY state machine.
Additionally the kernel will report this after a while:
  irq 28: nobody cared (try booting with the "irqpoll" option)
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

The datasheet uses the name "All Mask" for this bit. Change the name of
our #define to be consistent with the datasheet. While here also replace
the tab between the #define and IP101A_G_IRQ_ALL_MASK with a space.
No functional changes.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

This makes the code consistent by using the BIT() macro instead of
manual bit-shifting for some of the fields. No functional changes.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

This simply moves ip101a_g_config_init right above
ip101a_g_config_intr so all functions for the ICPlus IP101A/G PHYs are
grouped together.
No functional changes.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

The IP101A and IP101G series both have various models. Depending on the
board implementation we need a special property for the IP101GR (32-pin
LQFP package) PHY:
pin 21 ("RXER/INTR_32") outputs the "receive error" signal by default
(LOW means "normal operation", HIGH means that there's either a decoding
error of the received signal or that the PHY is receiving LPI). This pin
can also be switched to INTR32 mode, where the interrupt signal is
routed to this pin. The other PHYs don't need this special handling
because they have more pins available so the interrupt function gets a
dedicated pin.

This adds two properties to either select the "receive error" or
"interrupt" function of pin 21. Not specifying any function means that
the default set by the bootloader is used. This is required because the
IP101GR cannot be differentiated between other IP101 PHYs as the PHY
identification registers on all of these is 0x02430c54.

The IP101G (sold as die only, without package) may suffer from the same
issue depending on how it's integrated into a multi chip package by
another manufacturer. If only the RXER/INTR_32 pin is routed then the
users of the die-only variant may also have to explicitly configure the
mode of hte RXER/INTR_32 pin. This is the reason why no "is-ip101gr"
property was added. I have no evidence though which would confirm this
theory - so the binding itself is independent of that.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

IC Plus Corp. has various Ethernet related products such as Ethernet
transceivers, Ethernet controllers, Ethernet switches, etc.
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This function will try to determine the mac address via the devicetree,
or via an architecture-specific method (e.g. a PROM on SPARC).

The SPARC-specific code in this driver (#ifdef SPARC) did exactly this,
and is therefore removed.

Note that you can now specify the tg3 mac address via the devicetree,
on any platform, not just SPARC:

Devicetree example:
(see Documentation/devicetree/bindings/pci/pci.txt)

&pcie {
	host@0 {
		#address-cells = <3>;
		#size-cells = <2>;
		reg = <0 0 0 0 0>;
		bcm5778: bcm5778@0 {
			reg = <0 0 0 0 0>;
			mac-address = [CA 11 AB 1E 10 01];
		};
	};
};
Signed-off-by: Sven Van Asbroeck <svendev@arcx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add explanations of some generic TCP counters, fast open
related counters and TCP abort related counters and several
examples.
Signed-off-by: yupeng <yupeng0921@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Doug Berger says:

====================
net: bcmgenet: fix aborted suspend

It is not enough to return an error code from the driver suspend
routine. The driver must also restore the device functionality.

This commit corrects the issue introduced by commit 0db55093
("net: bcmgenet: return correct value 'ret' from bcmgenet_power_down")
by calling the driver resume function if the suspend function returns
an error.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

If an error occurs during suspension of the driver the driver should
restore the hardware configuration and return an error to force the
system to resume.

Fixes: 0db55093 ("net: bcmgenet: return correct value 'ret' from bcmgenet_power_down")
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 switches the order of bcmgenet_suspend and bcmgenet_resume
in the file to prevent the need for a forward declaration in the next
commit and to make the review of that commit easier.
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>

Clang warns:

drivers/net/geneve.c:428:29: error: suggest braces around initialization
of subobject [-Werror,-Wmissing-braces]
                struct in6_addr addr6 = { 0 };
                                          ^
                                          {}

Rather than trying to appease the various compilers that support the
kernel, use memset, which is unambiguous.

Fixes: a0796644 ("geneve: ICMP error lookup handler")
Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Salil Mehta says:

====================
net: hns3: Add vf mtu support

This patchset adds vf mtu support to HNS3 driver.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently netdev is down in enet module, and it is before
mtu range checking in hclge module, which may be cause
netdev being down unnecessarily.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The patch adds mtu setting support for vf, currently
vf and pf share the same hardware mtu setting. Mtu set
by vf must be less than or equal to pf' mtu, and mtu
set by pf must be greater than or equal to vf' mtu.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently there is no way for pf to know if a vf device is
alive or not, so PF does not know which vf to notify when
reset happens, or which vf's mtu is invalid when vf and pf
share the same hardware mtu setting.

This patch adds vport alive state checking support, in order
to support the above scenario.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch refactors mac mtu setting related functions,
normalizes the use of mps and mtu.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds supports for two vlan header when setting mtu.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remove directly accessing device_node.type pointer and use the accessors
instead. This will eventually allow removing the type pointer.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

In preparation to remove the node name pointer from struct device_node,
convert printf users to use the %pOFn format specifier.

Cc: Chas Williams <3chas3@gmail.com>
Cc: linux-atm-general@lists.sourceforge.net
Cc: netdev@vger.kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

This structure is small (12 or 16 bytes depending on 64bit
or 32bit kernels), but we do not want it spanning two cache lines.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Do not risk spanning these small structures on two cache lines,
it is absolutely not worth it.

For 32bit arches, the hint might not be enough, but we do not
really care anymore.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is a spelling mistake in a netdev_err message. Fix this.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

'err' always be 0 in the two places.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Samuel Mendoza-Jonas says:

====================
net/ncsi: Allow enabling multiple packages & channels

This series extends the NCSI driver to configure multiple packages
and/or channels simultaneously. Since the RFC series this includes a few
extra changes to fix areas in the driver that either made this harder or
were roadblocks due to deviations from the NCSI specification.

Patches 1 & 2 fix two issues where the driver made assumptions about the
capabilities of the NCSI topology.
Patches 3 & 4 change some internal semantics slightly to make multi-mode
easier.
Patch 5 introduces a cleaner way of reconfiguring the NCSI configuration
and keeping track of channel states.
Patch 6 implements the main multi-package/multi-channel configuration,
configured via the Netlink interface.

Readers who have an interesting NCSI setup - especially multi-package
with HWA - please test! I think I've covered all permutations but I
don't have infinite hardware to test on.

Changes in v2:
- Updated use of the channel lock in ncsi_reset_dev(), making the
channel invisible and leaving the monitor check to
ncsi_stop_channel_monitor().
- Fixed ncsi_channel_is_tx() to consider the state of channels in other
packages.
Changes in v3:
- Fixed bisectability bug in patch 1
- Consider channels on all packages in a few places when multi-package
is enabled.
- Avoid doubling up reset operations, and check the current driver state
before reset to let any running operations complete.
- Reorganise the LSC handler slightly to avoid enabling Tx twice.
Changes in v4:
- Fix failover in the single-channel case
- Better handle ncsi_reset_dev() entry during a current config/suspend
operation.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch extends the ncsi-netlink interface with two new commands and
three new attributes to configure multiple packages and/or channels at
once, and configure specific failover modes.

NCSI_CMD_SET_PACKAGE mask and NCSI_CMD_SET_CHANNEL_MASK set a whitelist
of packages or channels allowed to be configured with the
NCSI_ATTR_PACKAGE_MASK and NCSI_ATTR_CHANNEL_MASK attributes
respectively. If one of these whitelists is set only packages or
channels matching the whitelist are considered for the channel queue in
ncsi_choose_active_channel().

These commands may also use the NCSI_ATTR_MULTI_FLAG to signal that
multiple packages or channels may be configured simultaneously. NCSI
hardware arbitration (HWA) must be available in order to enable
multi-package mode. Multi-channel mode is always available.

If the NCSI_ATTR_CHANNEL_ID attribute is present in the
NCSI_CMD_SET_CHANNEL_MASK command the it sets the preferred channel as
with the NCSI_CMD_SET_INTERFACE command. The combination of preferred
channel and channel whitelist defines a primary channel and the allowed
failover channels.
If the NCSI_ATTR_MULTI_FLAG attribute is also present then the preferred
channel is configured for Tx/Rx and the other channels are enabled only
for Rx.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the NCSI driver is stopped with ncsi_stop_dev() the channel
monitors are stopped and the state set to "inactive". However the
channels are still configured and active from the perspective of the
network controller. We should suspend each active channel but in the
context of ncsi_stop_dev() the transmit queue has been or is about to be
stopped so we won't have time to do so.

Instead when ncsi_start_dev() is called if the NCSI topology has already
been probed then call ncsi_reset_dev() to suspend any channels that were
previously active. This resets the network controller to a known state,
provides an up to date view of channel link state, and makes sure that
mode flags such as NCSI_MODE_TX_ENABLE are properly reset.

In addition to ncsi_start_dev() use ncsi_reset_dev() in ncsi-netlink.c
to update the channel configuration more cleanly.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The concepts of a channel being 'active' and it having link are slightly
muddled in the NCSI driver. Tweak this slightly so that
NCSI_CHANNEL_ACTIVE represents a channel that has been configured and
enabled, and NCSI_CHANNEL_INACTIVE represents a de-configured channel.
This distinction is important because a channel can be 'active' but have
its link down; in this case the channel may still need to be configured
so that it may receive AEN link-state-change packets.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When a package is deselected all channels of that package cease
communication. If there are other channels active on the package of the
suspended channel this will disable them as well, so only send a
deselect-package command if no other channels are active.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently the NCSI driver sends a select-package command to all possible
packages simultaneously to discover what packages are available. However
at this stage in the probe process the driver does not know if
hardware arbitration is available: if it isn't then this process could
cause collisions on the RMII bus when packages try to respond.

Update the probe loop to probe each package one by one, and once
complete check if HWA is universally supported.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

NCSI hardware arbitration allows multiple packages to be enabled at once
and share the same wiring. If the NCSI driver recognises that HWA is
available it unconditionally enables all packages and channels; but that
is a configuration decision rather than something required by HWA.
Additionally the current implementation will not failover on link events
which can cause connectivity to be lost unless the interface is manually
bounced.

Retain basic HWA support but remove the separate configuration path to
enable all channels, leaving this to be handled by a later
implementation.
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The SGE Host Page Size has nothing to do with the actual
Host Page Size. It's the SGE's BAR2 Doorbell/GTS Page Size
for interpreting the SGE Ingress/Egress Queue per Page values.
Firmware reads all of these things and makes all the
subsequent changes necessary. The Host Driver uses the SGE
Host Page Size in order to properly calculate BAR2 Offsets.
Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Arjun Vynipadath <arjun@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add TCP_NLA_SRTT to SCM_TIMESTAMPING_OPT_STATS that reports the smoothed
round trip time in microseconds (tcp_sock.srtt_us >> 3).
Signed-off-by: Yousuk Seung <ysseung@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Trivial spelling errors found by codespell.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Instead of constantly playing with the struct initializer
syntax trying to make gcc and CLang both happy, just clear
it out using memset().

>> drivers/net/tun.c:2503:42: warning: Using plain integer as NULL pointer
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Thanks to the batched XDP buffs through msg_control. Instead of
calling put_page() for each page which involves a atomic operation,
let's batch them by record the last page that needs to be freed and
its refcnt count and free them in a batch.

Testpmd(virtio-user + vhost_net) + XDP_DROP shows 3.8% improvement.

Before: 4.71Mpps
After : 4.89Mpps
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We do a get_page() which involves a atomic operation. This patch tries
to mitigate a per packet atomic operation by maintaining a reference
bias which is initially USHRT_MAX. Each time a page is got, instead of
calling get_page() we decrease the bias and when we find it's time to
use a new page we will decrease the bias at one time through
__page_cache_drain_cache().

Testpmd(virtio_user + vhost_net) + XDP_DROP on TAP shows about 1.6%
improvement.

Before: 4.63Mpps
After:  4.71Mpps
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>