Saeed Mahameed says:

====================
mlx5-updates-2020-05-09

This series includes updates to mlx5 netdev driver and bonding updates
to support getting the next active tx slave.

1) merge commit with mlx5-next that includes bonding updates from Maor
   Bonding: Add support to get xmit slave
2) Maxim makes some general code improvements to TX data path
3) Tariq makes some general code improvements to kTLS and mlx5 accel layer
in preparation for mlx5 TLS RX.
====================
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Mark Starovoytov says:

====================
net: atlantic: driver updates

This patch series contains several minor cleanups for the previously
submitted series.

We also add Marvell copyrights on newly touched files.

v2:
 * accommodated review comments related to the last patch in series
   (MAC generation)

v1: https://patchwork.ozlabs.org/cover/1285011/
====================
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

This patch unifies invalid MAC address handling with other drivers.

Basically we've switched to using standard APIs (is_valid_ether_addr /
eth_hw_addr_random) where possible.
It's worth noting that some of engineering Aquantia NICs might be
provisioned with a partially zeroed out MAC, which is still invalid,
but not caught by is_valid_ether_addr(), so we've added a special
handling for this case.

Also adding a warning in case of fallback to random MAC, because
this shouldn't be needed on production NICs, they should all be
provisioned with unique MAC.

NB! Default systemd/udevd configuration is 'MACAddressPolicy=persistent'.
    This causes MAC address to be persisted across driver reloads and
    reboots. We had to change it to 'none' for verification purposes.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

This patch removes unnecessary check for boot code survivability before
reset request.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

No need to call hw_atl_b0_hw_rss_set from hw_atl2_hw_rss_set
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

TPO2 was introduced in B0 only, no reason to check for it in A0 code.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

This patch changes the constant name to a more logical "2G5"
(for 2.5G speeds).
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

This patch fixes a couple of minor merge issues found in macsec_api.c
after corresponding patch series has been applied.

These are not real bugs, so pushing to net-next.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

This patches fixes the review comment made by Jakub Kicinski
in the "net: atlantic: A2 support" patch series.
Signed-off-by: NMark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: NIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

The same WQE opcode might be used in different ICOSQ flows
and WQE types.
To have a better distinguishability, replace it with an enum that
better indicates the WQE type and flow it is used for.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

The include of Ethernet driver header in core is not needed
and actually wrong.
Remove it.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

Struct assignment looks more clean, and implies resetting
the not assigned fields to zero, instead of holding values
from older ring cycles.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

Into the txrx header file.
The mlx5e_sq_wqe_info structure describes WQE info for the ICOSQ,
rename it to better reflect this.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

Every single DUMP WQE resides in a single WQEBB.
As the pi is calculated per each one separately, there is
no real need for a contiguous room for them, allow them to populate
different WQ fragments.
This reduces WQ waste and improves its utilization.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

For the static and progress context params WQEs, do the edge
filling separately.
This improves the WQ utilization, code readability, and reduces
the chance of future bugs.
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Reviewed-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

After previous modifications, the offloads are no longer called one by
one, the pi is calculated and the wqe is cleared on between of TLS and
IPSEC offloads, which doesn't quite fit mlx5e_accel_handle_tx's purpose.

This patch splits mlx5e_accel_handle_tx into two functions that
correspond to two logical phases of running offloads:

1. Before fetching a WQE. Here runs the code that can post WQEs on its
own, before the main WQE is fetched. It's the main part of TLS offload.

2. After fetching a WQE. Here runs the code that updates the WQE's
fields, but can't post other WQEs any more. It's a minor part of TLS
offload that sets the tisn field in the cseg, and eseg-based offloads
(currently IPSEC, and later patches will move GENEVE and checksum
offloads there, too).

It allows to make mlx5e_xmit take care of all actions needed to transmit
a packet in the right order, improve the structure of the code and
reduce unnecessary operations. The structure will be further improved in
the following patches (all eseg-based offloads will be moved to a single
place, and reserving space for the main WQE will happen between phase 1
and phase 2 of offloads to eliminate unneeded data movements).
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

mlx5e_udp_gso_handle_tx_skb updates the length field in the UDP header
in case of GSO. It doesn't interfere with other offloads, so do it first
to simplify further restructuring of the code. This way we'll make all
independent modifications to the SKB before starting to work with WQEs.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NRaed Salem <raeds@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

TLS offload may write a 32-bit field (tisn) to the cseg of the WQE. To
do that, it receives pi and wqe pointers. As TLS offload may also send
additional WQEs, it has to update pi and wqe, and in many cases it even
doesn't use pi calculated before and wqe zeroed before and does it
itself. Also, mlx5e_sq_xmit has to copy the whole cseg if it goes to the
mlx5e_fill_sq_frag_edge flow. This all is not efficient.

It's more efficient to do the following:

1. Just return tisn from TLS offload and make the caller fill it in a
more appropriate place.

2. Calculate pi and clear wqe after calling TLS offload.

3. If TLS offload has to send WQEs, calculate pi and clear wqe just
before that. It's already done in all places anyway, so this commit
allows to remove some redundant memsets and calls.

Copying of cseg will be eliminated in one of the following commits, and
all other stuff is done here.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

IPSEC offload needs to modify the eseg of the WQE that is being filled,
but it receives a pointer to the whole WQE. To make the contract
stricter, pass only the pointer to the eseg of that WQE. This commit is
preparation for the following refactoring of offloads in the TX path and
for the MPWQE support.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

mlx5e_sq_xmit and mlx5i_sq_xmit always return NETDEV_TX_OK. Drop the
return value to simplify the code.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

Both INNOVA and ConnectX TLS offloads perform the same checks in the
beginning. Unify them to reduce repeating code. Do WARN_ON_ONCE on
netdev mismatch and finish with an error in both offloads, not only in
the ConnectX one.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

TLS and IPSEC offloads currently return struct sk_buff *, but the value
is either NULL or the same skb that was passed as a parameter. Return
bool instead to provide stronger guarantees to the calling code (it
won't need to support handling a different SKB that could be potentially
returned before this change) and to simplify restructuring this code in
the following commits.
Signed-off-by: NMaxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

This merge includes updates to bonding driver needed for the rdma stack,
to avoid conflicts with the RDMA branch.

Maor Gottlieb Says:

====================
Bonding: Add support to get xmit slave

The following series adds support to get the LAG master xmit slave by
introducing new .ndo - ndo_get_xmit_slave. Every LAG module can
implement it and it first implemented in the bond driver.
This is follow-up to the RFC discussion [1].

The main motivation for doing this is for drivers that offload part
of the LAG functionality. For example, Mellanox Connect-X hardware
implements RoCE LAG which selects the TX affinity when the resources
are created and port is remapped when it goes down.

The first part of this patchset introduces the new .ndo and add the
support to the bonding module.

The second part adds support to get the RoCE LAG xmit slave by building
skb of the RoCE packet based on the AH attributes and call to the new
.ndo.

The third part change the mlx5 driver driver to set the QP's affinity
port according to the slave which found by the .ndo.
====================
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>

The variable tx_done is being assigned with a value that is never read
as the function returns a few statements later.  The assignment is
redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

The variable size is being assigned with a value that is never read,
the assignment is redundant and cab be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

The variable ret is being assigned with a value that is never read,
the assignment is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Reviewed-by: NMichael Chan <michael.chan@broadcom.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Currently the value for 'off' is computed using a multiplication and
a couple of statements later off is being incremented by len and
this value is never read.  Clean up the code by removing the
multiplication and just increment off by len on each iteration.

Addresses-Coverity: ("Unused value")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Heiner Kallweit says:

====================
r8169: sync few functionalities with vendor driver

Add few helpers (with names copied from vendor drivers) to make clearer
what the respective code is doing. In addition improve reset preparation
for chips from RTL8168g.
====================
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Sync the reset preparation for chips from RTL8168g with the r8168 and
r8125 vendor drivers.
Signed-off-by: NHeiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Add a helper for waiting for FIFO's to be empty, again the name is
borrowed from the vendor driver.
Signed-off-by: NHeiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Add a helper for setting RXDV_GATED_EN, the 2ms delay is copied from
the vendor driver.
Signed-off-by: NHeiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Create a helper for this waiting function, name of the helper is
borrowed from the vendor driver. In addition don't return in the two
hw_init functions if the first wait runs into a timeout, there's no
benefit in doing so.
Signed-off-by: NHeiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

When building arm64 allyesconfig:

drivers/net/ipa/ipa_endpoint.c: In function 'ipa_endpoint_stop_rx_dma':
drivers/net/ipa/ipa_endpoint.c:1274:13: error: 'IPA_ENDPOINT_STOP_RX_SIZE' undeclared (first use in this function)
drivers/net/ipa/ipa_endpoint.c:1274:13: note: each undeclared identifier is reported only once for each function it appears in
drivers/net/ipa/ipa_endpoint.c:1289:2: error: implicit declaration of function 'ipa_cmd_dma_task_32b_addr_add' [-Werror=implicit-function-declaration]
drivers/net/ipa/ipa_endpoint.c:1291:45: error: 'ENDPOINT_STOP_DMA_TIMEOUT' undeclared (first use in this function)
drivers/net/ipa/ipa_endpoint.c: In function 'ipa_endpoint_stop':
drivers/net/ipa/ipa_endpoint.c:1309:16: error: 'IPA_ENDPOINT_STOP_RX_RETRIES' undeclared (first use in this function)

These functions were removed in a series, merged in as
commit 33395f4a ("Merge branch 'net-ipa-kill-endpoint-stop-workaround'").

Remove them again so that the build works properly.

Fixes: 3793faad ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net")
Signed-off-by: NNathan Chancellor <natechancellor@gmail.com>
Reviewed-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

We currently have to adjust ipv6 route gc_thresh/max_size depending
on number of cpus on a server, this makes very little sense.

If the kernels sets /proc/sys/net/ipv6/route/gc_thresh to 1024
and /proc/sys/net/ipv6/route/max_size to 4096, then we better
not track the percpu dst that our implementation uses.

Only routes not added (directly or indirectly) by the admin
should be tracked and limited.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: David Ahern <dsahern@kernel.org>
Cc: Maciej Żenczykowski <maze@google.com>
Acked-by: NWei Wang <weiwan@google.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/dsa/sja1105/sja1105_vl.c:468:6: warning: variable ‘prev_time’ set but not used [-Wunused-but-set-variable]
  u32 prev_time = 0;
      ^~~~~~~~~
Reported-by: NHulk Robot <hulkci@huawei.com>
Signed-off-by: NSamuel Zou <zou_wei@huawei.com>
Reviewed-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: NVladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Use the helper function that wraps the calls to platform_get_resource()
and devm_ioremap_resource() together.
Signed-off-by: NWei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

The type of dispatch is u8 which is always '<=' 0xff, so the
dispatch <= 0xff is always true, we can remove this comparison.
Reported-by: NHulk Robot <hulkci@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>
Signed-off-by: NStefan Schmidt <stefan@datenfreihafen.org>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

percpu_counter_add() uses a default batch size which is quite big
on platforms with 256 cpus. (2*256 -> 512)

This means dst_entries_get_fast() can be off by +/- 2*(nr_cpus^2)
(131072 on servers with 256 cpus)

Reduce the batch size to something more reasonable, and
add logic to ip6_dst_gc() to call dst_entries_get_slow()
before calling the _very_ expensive fib6_run_gc() function.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Now sch_fq has horizon feature, we want to allow QUIC/UDP applications
to use EDT model so that pacing can be offloaded to the kernel (sch_fq)
or the NIC.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Acked-by: NWillem de Bruijn <willemb@google.com>
Acked-by: NVinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
bonding: report transmit status to callers

First patches cleanup netpoll, and make sure it provides tx status to its users.

Last patch changes bonding to not pretend packets were sent without error.

By providing more accurate status, TCP stack can avoid adding more
packets if the slave qdisc is already full.

This came while testing latest horizon feature in sch_fq, with
very low pacing rate flows, but should benefit hosts under stress.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>