Heiner Kallweit says:

====================
r8169: add EEE support for RTL8168g+

This series adds general EEE support to be used with ethtool.
In addition it implements EEE for chip versions from RTL8168g.
The first patch leaves the default chip settings and the
second enables EEE per default. This allows us to revert patch 2
w/o removing EEE support completely if we should face issues with
EEE on particular chip versions.

Unfortunately Realtek decided not to use the standard EEE MMD
registers but to use proprietary registers. Therefore we can't
use phylib functions like phy_ethtool_set_eee and have to
reimplement the functionality.

Tested on a system with RTL8168g (chip version 40).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Enable EEE per default on chip versions from RTL8168g.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds the general framework to deal with EEE in this driver
plus EEE support for chip versions from RTL8168g. We don't touch the
default chip settings, therefore EEE will usually be disabled and it's
up to the user to enable it.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Peter Oskolkov says:

====================
net: IP defrag: use rbtrees in IPv6 defragmentation

Currently, IPv6 defragmentation code drops non-last fragments that
are smaller than 1280 bytes: see
commit 0ed4229b ("ipv6: defrag: drop non-last frags smaller than min mtu")

This behavior is not specified in IPv6 RFCs and appears to break compatibility
with some IPv6 implementations, as reported here:
https://www.spinics.net/lists/netdev/msg543846.html

This patchset contains four patches:
- patch 1 moves rbtree-related code from IPv4 to files shared b/w
IPv4/IPv6
- patch 2 changes IPv6 defragmenation code to use rbtrees for defrag
queue
- patch 3 changes nf_conntrack IPv6 defragmentation code to use rbtrees
- patch 4 changes ip_defrag selftest to test changes made in the
previous three patches.

Along the way, the 1280-byte restrictions are removed.

I plan to introduce similar changes to 6lowpan defragmentation code
once I figure out how to test it.
====================
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds several changes to the ip_defrag selftest, to cover
new IPv6 defrag behavior:

- min IPv6 frag size is now 8 instead of 1280

- new test cases to cover IPv6 defragmentation in nf_conntrack_reasm.c

- new "permissive" mode in negative (overlap) tests: netfilter
sometimes drops invalid packets without passing them to IPv6
underneath, and thus defragmentation sometimes succeeds when
it is expected to fail; so the permissive mode does not fail the
test if the correct reassembled datagram is received instead of a
timeout.
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, IPv6 defragmentation code drops non-last fragments that
are smaller than 1280 bytes: see
commit 0ed4229b ("ipv6: defrag: drop non-last frags smaller than min mtu")

This behavior is not specified in IPv6 RFCs and appears to break
compatibility with some IPv6 implemenations, as reported here:
https://www.spinics.net/lists/netdev/msg543846.html

This patch re-uses common IP defragmentation queueing and reassembly
code in IP6 defragmentation in nf_conntrack, removing the 1280 byte
restriction.
Signed-off-by: Peter Oskolkov <posk@google.com>
Reported-by: Tom Herbert <tom@herbertland.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, IPv6 defragmentation code drops non-last fragments that
are smaller than 1280 bytes: see
commit 0ed4229b ("ipv6: defrag: drop non-last frags smaller than min mtu")

This behavior is not specified in IPv6 RFCs and appears to break
compatibility with some IPv6 implemenations, as reported here:
https://www.spinics.net/lists/netdev/msg543846.html

This patch re-uses common IP defragmentation queueing and reassembly
code in IPv6, removing the 1280 byte restriction.
Signed-off-by: Peter Oskolkov <posk@google.com>
Reported-by: Tom Herbert <tom@herbertland.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a refactoring patch: without changing runtime behavior,
it moves rbtree-related code from IPv4-specific files/functions
into .h/.c defrag files shared with IPv6 defragmentation code.
Signed-off-by: Peter Oskolkov <posk@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Julian Wiedmann says:

====================
s390/qeth: updates 2019-01-25

please apply a first batch of qeth patches for net-next, primarily touching the
net_device parts of the driver.
In addition to the usual refactoring & code consolidation, patch 7 makes use of
netif_device_detach() to let the stack know when our control plane is down. This
helps quite a bit wrt to overall locking and proper init/shutdown sequencing.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

For recovery purposes, qeth keeps track of all registered VIDs. Replace
this by using the infrastructure introduced in
commit 9daae9bd ("net: Call add/kill vid ndo on vlan filter feature toggling").

By managing NETIF_F_HW_VLAN_CTAG_FILTER as a hw_feature,
netdev_update_features() will select it from dev->wanted_features
and replay all of the netdevice's VIDs to its ndo_vlan_rx_add_vid()
callback.
z/VM NICs strictly require VLAN registration, so don't expose it as
hw_feature there but add a little hack in qeth_enable_hw_features()
to make things work regardless.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When a qeth card is offline, it has no connection to the HW. So none of
our control callbacks can run IO against it, and we can only cache the
input (eg a new MAC address) without providing proper feedback to the
caller. In this context, it seems much more reasonable to simply detach
the netdevice and let the kernel reject any interaction with it.

This also makes all sorts of internal state checks and locking obsolete.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Re-order the code flow a bit so that all initial HW setup is done before
putting the netdevice into play. For a netdevice that hasn't been
registered before, we also don't need to re-enable its HW features or
check for recovery actions.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

At best this is redundant, at worst it papers over a race in the
offline / online code.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 4789a218 ("s390/qeth: fix race when setting MAC address")
resolved a race where our initial programming of dev_addr into the HW
and a call to ndo_set_mac_address() could run concurrently. In this
case, we could end up getting confused about which address was actually
set in the HW.

The quick fix was to introduce additional locking that blocks any
ndo_set_mac_address() while the device is being set online. But the race
primarily originated from the fact that we first register the netdevice,
and only then program its dev_addr. By re-ordering this sequence,
userspace will only be able to change the MAC address _after_ we have
finished with setting the initial dev_addr.

Still, the same MAC address race can also occur during a subsequent call
to qeth_l2_set_online(). So keep around the locking for now, until a
follow-up patch fully resolves this.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The L2 and L3 code for these ops is almost identical, we only need to
provide a custom ndo_validate_addr() for L2 that checks whether
programming the MAC address succeeded.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth_qdio_cq_handler() doesn't replenish the Output Queue(s), and thus
has no reason to wake the txq.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Consolidate the code that marks the current buffer to be flushed, and
let qeth_fill_buffer() advance the Output Queue's buffer cursor.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Recent changes to the phylib API
- removed phy_stop_interrupts
- replaced phy_start_interrupts with phy_request_interrupt
- moved some functionality from phy_connect() and phy_disconnect()
  to phy_start() and phy_stop() respectively.
Reflect these changes in the documentation.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
mlx5-updates-2019-01-25

This series provides some updates to mlx5 driver,

From Tariq,
1) Make sure RX packet header does not cross page boundary
   To avoid page boundary crossing, use stride size that fits
   the maximum possible header. Stride is increased form 64B to 256B.

2) CQ struct cleanup: Take CQ decompress fields into a separate structure

From Moshe,
3) Expand XPS cpumask to cover all online cpus

From Jason Gunthorpe and Tariq:
4) Compilation warning cleanup

From Or,
5) Add trace points for flow tables create/destroy

From Saeed,
6) Software stats update/folding improvements
   this also solves a compilation warning on 32bit systems that was reported
   last release cycle by Arnd and Andrew.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Representors software stats are basic, this patch is reusing the
mlx5e_fold_sw_stats in representors, which sums up the basic stats64 for a
mlx5e netdevice.

Fixes: 8bfaf07f ("net/mlx5e: Present SW stats when state is not opened")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

This behavior is already adopted for all other cases in the cited patch.
The representor's functions were missed, here we modify the them to
behave similarly.

Fixes: 8bfaf07f ("net/mlx5e: Present SW stats when state is not opened")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

mlx5e_grp_sw_update_stats can be called from two threads,
1) ndo_get_stats64
2) get_ethtool_stats

For this reason and to minimize concurrency issue impact on 64bit machines
mlx5e_grp_sw_update_stats folds the software stats into a temporary
variable then copies it to the global driver stats, both ethtool and ndo
statistics callbacks will use the global software stats variable to report
whatever stats they need.

Actually ndo_get_stats64 doesn't need to fold the whole software stats
(mlx5e_grp_sw_update_stats), all it needs is five counters to fill the
rtnl_link_stats64 relevant stats parameter.

Hence this patch introduces a simpler helper function to fold software
stats for ndo_get_stats64 which will work directly on rtnl_link_stats64
stats parameter and not on the global or even temporary mlx5e_sw_stats
variable.

Since now mlx5e_grp_sw_update_stats is not called by ndo_get_stats64 we
can make it static and remove the temp var.

Unlike mlx5e_grp_sw_update_stats the new fold stats function doesn't
need to zero out the output statistics parameter since it is already
done by the stack @dev_get_stats().

This patch is fixing stack usage of mlx5e_grp_sw_update_stats on
x86 gcc-4.9 and higher, the concurrency issue between mlx5's
ndo_get_stats64 and get_ethtool_stats is resolved as well.

Fixes: 8bfaf07f ("net/mlx5e: Present SW stats when state is not opened")
Reported-by: Arnd Bergmann <arnd@arndb.de>
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

We were not tracking flow tables so far, add it up.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

This confusing construction confuses the compiler which can't see
that flow is initialized if !err:

drivers/net/ethernet/mellanox/mlx5/core/en_tc.c: In function `mlx5e_configure_flower`
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c:2727:28: warning:
`flow` may be used uninitialized in this function
[-Wmaybe-uninitialized]

There is no reason for two function outputs, just return the
pointer directly and use ERR_PTR to encode a failure.
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Currently we have one cpu in XPS cpumask per tx queue, this is good
enough for default configuration where there is a tx queue per cpu.
However, once configuration changes to use less tx queues, part of the
cpus are not XPS-mapped and so the select queue decision falls back to
hash calculation and balancing is not guaranteed.

Expand XPS cpumask to enable using all cpus even when number of tx
queues is smaller than number of cpus.
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Only the Receive CQ makes use of these fields. Take them
out into a separate struct and save space in the generic
CQ structure.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

In the non-linear SKB memory scheme of Striding RQ, a packet header
could cross page boundary. This requires special care in fast path
that costs LoC, additional runtime instructions and branches.

It could happen when the header (up to 256B) does not fit in
a single stride. Avoid this by working with a stride size that fits
the maximum possible header. Stride is increased form 64B to 256B.

Performance:
Tested packet rate for UDP streams, single ring, on ConnectX-5.

Configuration:
Set Striding RQ and LRO ON (to enabled the non-linear SKB scheme).
GRO OFF, early drop by TC rule.

64B: 4x worse memory utilization, no page-crossers headers
- No degradation (5,887,305 pps).
- The reduction in memory utilization is compensated by the saving of
  branches tests.

192B: 1.33x worse memory utilization, avoid page-crossers headers
- Before: 5,727,252. After: 5,777,037. ~1% gain.

256B: Same memory util, no page-crossers
- Before: 5,691,885. After: 5,748,007. ~1% gain.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

This reverts the devlink health changes from 9/17/2019,
Jiri wants things to be designed differently and it was
agreed that the easiest way to do this is start from the
beginning again.

Commits reverted:

cb5ccfbe
880ee82f
c7af343b
ff253fed
6f9d5613
fcd852c6
8a66704a
12bd0dce
aba25279
ce019faa
b8c45a03

And the follow-on build fix:

o33a0efa4Signed-off-by: David S. Miller <davem@davemloft.net>

Remove duplicated include.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

max_rcvbuf_size is no longer used since commit "414574a0".
Signed-off-by: Zhaolong Zhang <zhangzl2013@126.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Priyaranjan Jha says:

====================
tcp_bbr: Improving TCP BBR performance for WiFi and cellular networks

Ack aggregation is quite prevalent with wifi, cellular and cable modem
link tchnologies, ACK decimation in middleboxes, and common offloading
techniques such as TSO and GRO, at end hosts. Previously, BBR was often
cwnd-limited in the presence of severe ACK aggregation, which resulted in
low throughput due to insufficient data in flight.

To achieve good throughput for wifi and other paths with aggregation, this
patch series implements an ACK aggregation estimator for BBR, which
estimates the maximum recent degree of ACK aggregation and adapts cwnd
based on it. The algorithm is further described by the following
presentation:
https://datatracker.ietf.org/meeting/101/materials/slides-101-iccrg-an-update-on-bbr-work-at-google-00

(1) A preparatory patch, which refactors bbr_target_cwnd for generic
    inflight provisioning.

(2) Implements BBR ack aggregation estimator and adapts cwnd based
    on measured degree of ACK aggregation.
====================
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Aggregation effects are extremely common with wifi, cellular, and cable
modem link technologies, ACK decimation in middleboxes, and LRO and GRO
in receiving hosts. The aggregation can happen in either direction,
data or ACKs, but in either case the aggregation effect is visible
to the sender in the ACK stream.

Previously BBR's sending was often limited by cwnd under severe ACK
aggregation/decimation because BBR sized the cwnd at 2*BDP. If packets
were acked in bursts after long delays (e.g. one ACK acking 5*BDP after
5*RTT), BBR's sending was halted after sending 2*BDP over 2*RTT, leaving
the bottleneck idle for potentially long periods. Note that loss-based
congestion control does not have this issue because when facing
aggregation it continues increasing cwnd after bursts of ACKs, growing
cwnd until the buffer is full.

To achieve good throughput in the presence of aggregation effects, this
algorithm allows the BBR sender to put extra data in flight to keep the
bottleneck utilized during silences in the ACK stream that it has evidence
to suggest were caused by aggregation.

A summary of the algorithm: when a burst of packets are acked by a
stretched ACK or a burst of ACKs or both, BBR first estimates the expected
amount of data that should have been acked, based on its estimated
bandwidth. Then the surplus ("extra_acked") is recorded in a windowed-max
filter to estimate the recent level of observed ACK aggregation. Then cwnd
is increased by the ACK aggregation estimate. The larger cwnd avoids BBR
being cwnd-limited in the face of ACK silences that recent history suggests
were caused by aggregation. As a sanity check, the ACK aggregation degree
is upper-bounded by the cwnd (at the time of measurement) and a global max
of BW * 100ms. The algorithm is further described by the following
presentation:
https://datatracker.ietf.org/meeting/101/materials/slides-101-iccrg-an-update-on-bbr-work-at-google-00

In our internal testing, we observed a significant increase in BBR
throughput (measured using netperf), in a basic wifi setup.
- Host1 (sender on ethernet) -> AP -> Host2 (receiver on wifi)
- 2.4 GHz -> BBR before: ~73 Mbps; BBR after: ~102 Mbps; CUBIC: ~100 Mbps
- 5.0 GHz -> BBR before: ~362 Mbps; BBR after: ~593 Mbps; CUBIC: ~601 Mbps

Also, this code is running globally on YouTube TCP connections and produced
significant bandwidth increases for YouTube traffic.

This is based on Ian Swett's max_ack_height_ algorithm from the
QUIC BBR implementation.
Signed-off-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Because bbr_target_cwnd() is really a general-purpose BBR helper for
computing some volume of inflight data as a function of the estimated
BDP, refactor it into following helper functions:
- bbr_bdp()
- bbr_quantization_budget()
- bbr_inflight()
Signed-off-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Few chip versions use the same sequence to adjust 10M and ALDPS, so
let's factor it out. This patch also fixes a (most likely) typo in
rtl8168g_1_hw_phy_config. There bit 8 in reg 0x14 on page 0x0bcc
was set and not cleared. According to the vendor driver this bit
needs to be cleared in all cases.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Chip versions from RTL8168g onward use the same sequence to disable
ALDPS (Advanced Link-Down Power Saving). So let's factor this out.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

I made a dumb mistake when I summed up the slave stats, obviously slaves
can come and go which would make the master stats unreliable.
Count and export the master stats separately.

Fixes: a258aeac ("bonding: add support for xstats and export 3ad stats")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Heiner Kallweit says:

====================
net: phy: improve starting PHY

This patch series improves few aspects of starting the PHY.

v2:
- improve a warning in patch 4
v3:
- extend commit message for patch 2
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Now that we enable the interrupts in phy_start() we don't have to do it
before. Therefore remove enabling interrupts from phy_start_interrupts()
and rename this function to reflect the changed functionality.

v2:
- improve warning to clearly state that we fall back to polling
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Interrupts don't have to be enabled before calling phy_start().
Therefore let's enable them in phy_start(). In a subsequent step
we'll remove enabling interrupts from phy_connect_direct().
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>