ice: Update enum ice_flg64_bits to current specification

Currently the VLAN ice_flg64_bits are off by 1. Fix this by
setting the ICE_FLG_EVLAN_x8100 flag to 14, which also updates
ICE_FLG_EVLAN_x9100 to 15 and ICE_FLG_VLAN_x8100 to 16.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: delay less

Shorten the delay for SQ responses, but increase the number of loops.
Max delay time is unchanged, but some operations complete much more
quickly.

In the process, add a new define to make the delay count and delay time
more explicit. Add comments to make things more explicit.

This fixes a problem with VF resets failing on with many VFs.

Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: use pkg_dwnld_status instead of sq_last_status

Since the return value from the Download Package AQ command is stored in
hw->pkg_dwnld_status, use that instead of sq_last_status since that may
have the return value from some other AQ command leading to unexpected
results.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Change max MSI-x vector_id check in cfg_irq_map

Currently we check to make sure the vector_id passed down from iavf
is less than or equal to pf->hw.func_caps.common_caps.num_msix_vectors.
This is incorrect because the vector_id is always 0-based and never
greater than or equal to the ICE_MAX_INTR_PER_VF. Fix this by checking
to make sure the vector_id is less than the max allowed interrupts per
VF (ICE_MAX_INTR_PER_VF).

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Check if VF is disabled for Opcode and other operations

This patch adds code to check if PF or VF is disabled before honoring
mailbox message to configure VF - If it is disabled, and opcode is for
resetting VF, the PF driver simply tell VF that all is set. In addition,
if reset is ongoing, and Admin intend to configure VF on the host, we can
poll the VF enabling bit to make sure it is ready before continue - If
after ~250 milliseconds, VF is not in active state, we can bail out with
invalid error.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: configure software LLDP in ice_init_pf_dcb

Move software LLDP configuration when FW DCBX is disabled to
ice_init_pf_dcb, since that is where the FW DCBX state is determined.
Remove this software LLDP configuration from ice_vsi_setup and
ice_set_priv_flags. Software configuration includes redirecting Rx LLDP
packets up the stack, when FW DCBX is not running.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Fix to change Rx/Tx ring descriptor size via ethtool with DCBx

This patch fixes the call trace caused by the kernel when the Rx/Tx
descriptor size change request is initiated via ethtool when DCB is
configured. ice_set_ringparam() should use vsi->num_txq instead of
vsi->alloc_txq as it represents the queues that are enabled in the
driver when DCB is enabled/disabled. Otherwise, queue index being
used can go out of range.

For example, when vsi->alloc_txq has 104 queues and with 3 TCS enabled
via DCB, each TC gets 34 queues, vsi->num_txq will be 102 and only 102
queues will be enabled.

Signed-off-by: Usha Ketineni <usha.k.ketineni@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: avoid setting features during reset

Certain subsystems behave very badly when called during reset (core
dump). This patch returns -EBUSY when reconfiguring some subsystems
during reset. With this patch some ethtool functions will not core
dump during reset.

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Implement DCBNL support

Implement interface layer for the DCBNL subsystem. These are the functions
to support the callbacks defined in the dcbnl_rtnl_ops struct. These
callbacks are going to be used to interface with the DCB settings of the
device. Implementation of dcb_nl set functions and supporting SW DCB
functions.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Add NDO callback to set the maximum per-queue bitrate

Allow for rate limiting Tx queues. Bitrate is set in
Mbps(megabits per second).

Mbps max-rate is set for the queue via sysfs:
/sys/class/net/<iface>/queues/tx-<queue>/tx_maxrate
ex: echo 100 >/sys/class/net/ens7/queues/tx-0/tx_maxrate
    echo 200 >/sys/class/net/ens7/queues/tx-1/tx_maxrate
Note: A value of zero for tx_maxrate means disabled,
default is disabled.

Signed-off-by: Usha Ketineni <usha.k.ketineni@intel.com>
Co-developed-by: Tarun Singh <tarun.k.singh@intel.com>
Signed-off-by: Tarun Singh <tarun.k.singh@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Use ice_ena_vsi and ice_dis_vsi in DCB configuration flow

DCB configuration flow needs to disable and enable only the PF (main)
VSI, so use ice_ena_vsi and ice_dis_vsi. To avoid the use of ifdef to
control the staticness of these functions, move them to ice_lib.c.

Also replace the allocate and copy of old_cfg to kmemdup() in
ice_pf_dcb_cfg().

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

cxgb4: Use match_string() helper to simplify the code

match_string() returns the array index of a matching string.
Use it instead of the open-coded implementation.

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

net: ethernet: stmmac: Add support for syscfg clock

Add optional support for syscfg clock in dwmac-stm32.c
Now Syscfg clock is activated automatically when syscfg
registers are used

Signed-off-by: Christophe Roullier <christophe.roullier@st.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: eliminate checking netns if node established

Currently, we scan over all network namespaces at each received
discovery message in order to check if the sending peer might be
present in a host local namespaces.

This is unnecessary since we can assume that a peer will not change its
location during an established session.

We now improve the condition for this testing so that we don't perform
any redundant scans.

Fixes: f73b12812a3d ("tipc: improve throughput between nodes in netns")
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add a READ_ONCE() in skb_peek_tail()

skb_peek_tail() can be used without protection of a lock,
as spotted by KCSAN [1]

In order to avoid load-stearing, add a READ_ONCE()

Note that the corresponding WRITE_ONCE() are already there.

[1]
BUG: KCSAN: data-race in sk_wait_data / skb_queue_tail

read to 0xffff8880b36a4118 of 8 bytes by task 20426 on cpu 1:
 skb_peek_tail include/linux/skbuff.h:1784 [inline]
 sk_wait_data+0x15b/0x250 net/core/sock.c:2477
 kcm_wait_data+0x112/0x1f0 net/kcm/kcmsock.c:1103
 kcm_recvmsg+0xac/0x320 net/kcm/kcmsock.c:1130
 sock_recvmsg_nosec net/socket.c:871 [inline]
 sock_recvmsg net/socket.c:889 [inline]
 sock_recvmsg+0x92/0xb0 net/socket.c:885
 ___sys_recvmsg+0x1a0/0x3e0 net/socket.c:2480
 do_recvmmsg+0x19a/0x5c0 net/socket.c:2601
 __sys_recvmmsg+0x1ef/0x200 net/socket.c:2680
 __do_sys_recvmmsg net/socket.c:2703 [inline]
 __se_sys_recvmmsg net/socket.c:2696 [inline]
 __x64_sys_recvmmsg+0x89/0xb0 net/socket.c:2696
 do_syscall_64+0xcc/0x370 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

write to 0xffff8880b36a4118 of 8 bytes by task 451 on cpu 0:
 __skb_insert include/linux/skbuff.h:1852 [inline]
 __skb_queue_before include/linux/skbuff.h:1958 [inline]
 __skb_queue_tail include/linux/skbuff.h:1991 [inline]
 skb_queue_tail+0x7e/0xc0 net/core/skbuff.c:3145
 kcm_queue_rcv_skb+0x202/0x310 net/kcm/kcmsock.c:206
 kcm_rcv_strparser+0x74/0x4b0 net/kcm/kcmsock.c:370
 __strp_recv+0x348/0xf50 net/strparser/strparser.c:309
 strp_recv+0x84/0xa0 net/strparser/strparser.c:343
 tcp_read_sock+0x174/0x5c0 net/ipv4/tcp.c:1639
 strp_read_sock+0xd4/0x140 net/strparser/strparser.c:366
 do_strp_work net/strparser/strparser.c:414 [inline]
 strp_work+0x9a/0xe0 net/strparser/strparser.c:423
 process_one_work+0x3d4/0x890 kernel/workqueue.c:2269
 worker_thread+0xa0/0x800 kernel/workqueue.c:2415
 kthread+0x1d4/0x200 drivers/block/aoe/aoecmd.c:1253
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352

Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 451 Comm: kworker/u4:3 Not tainted 5.4.0-rc3+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: kstrp strp_work

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add annotations on hh->hh_len lockless accesses

KCSAN reported a data-race [1]

While we can use READ_ONCE() on the read sides,
we need to make sure hh->hh_len is written last.

[1]

BUG: KCSAN: data-race in eth_header_cache / neigh_resolve_output

write to 0xffff8880b9dedcb8 of 4 bytes by task 29760 on cpu 0:
 eth_header_cache+0xa9/0xd0 net/ethernet/eth.c:247
 neigh_hh_init net/core/neighbour.c:1463 [inline]
 neigh_resolve_output net/core/neighbour.c:1480 [inline]
 neigh_resolve_output+0x415/0x470 net/core/neighbour.c:1470
 neigh_output include/net/neighbour.h:511 [inline]
 ip6_finish_output2+0x7a2/0xec0 net/ipv6/ip6_output.c:116
 __ip6_finish_output net/ipv6/ip6_output.c:142 [inline]
 __ip6_finish_output+0x2d7/0x330 net/ipv6/ip6_output.c:127
 ip6_finish_output+0x41/0x160 net/ipv6/ip6_output.c:152
 NF_HOOK_COND include/linux/netfilter.h:294 [inline]
 ip6_output+0xf2/0x280 net/ipv6/ip6_output.c:175
 dst_output include/net/dst.h:436 [inline]
 NF_HOOK include/linux/netfilter.h:305 [inline]
 ndisc_send_skb+0x459/0x5f0 net/ipv6/ndisc.c:505
 ndisc_send_ns+0x207/0x430 net/ipv6/ndisc.c:647
 rt6_probe_deferred+0x98/0xf0 net/ipv6/route.c:615
 process_one_work+0x3d4/0x890 kernel/workqueue.c:2269
 worker_thread+0xa0/0x800 kernel/workqueue.c:2415
 kthread+0x1d4/0x200 drivers/block/aoe/aoecmd.c:1253
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352

read to 0xffff8880b9dedcb8 of 4 bytes by task 29572 on cpu 1:
 neigh_resolve_output net/core/neighbour.c:1479 [inline]
 neigh_resolve_output+0x113/0x470 net/core/neighbour.c:1470
 neigh_output include/net/neighbour.h:511 [inline]
 ip6_finish_output2+0x7a2/0xec0 net/ipv6/ip6_output.c:116
 __ip6_finish_output net/ipv6/ip6_output.c:142 [inline]
 __ip6_finish_output+0x2d7/0x330 net/ipv6/ip6_output.c:127
 ip6_finish_output+0x41/0x160 net/ipv6/ip6_output.c:152
 NF_HOOK_COND include/linux/netfilter.h:294 [inline]
 ip6_output+0xf2/0x280 net/ipv6/ip6_output.c:175
 dst_output include/net/dst.h:436 [inline]
 NF_HOOK include/linux/netfilter.h:305 [inline]
 ndisc_send_skb+0x459/0x5f0 net/ipv6/ndisc.c:505
 ndisc_send_ns+0x207/0x430 net/ipv6/ndisc.c:647
 rt6_probe_deferred+0x98/0xf0 net/ipv6/route.c:615
 process_one_work+0x3d4/0x890 kernel/workqueue.c:2269
 worker_thread+0xa0/0x800 kernel/workqueue.c:2415
 kthread+0x1d4/0x200 drivers/block/aoe/aoecmd.c:1253
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 29572 Comm: kworker/1:4 Not tainted 5.4.0-rc6+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events rt6_probe_deferred

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'u64_stats_t'

Eric Dumazet says:

====================
net: introduce u64_stats_t

KCSAN found a data-race in per-cpu u64 stats accounting.

(The stack traces are included in the 8th patch :
 tun: switch to u64_stats_t)

This patch series first consolidate code in five patches.
Then the last three patches address the data-race resolution.
====================

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

net: use u64_stats_t in struct pcpu_lstats

In order to fix the data-race found by KCSAN, we
can use the new u64_stats_t type and its accessors instead
of plain u64 fields. This will still generate optimal code
for both 32 and 64 bit platforms.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tun: switch to u64_stats_t

In order to fix this data-race found by KCSAN [1],
switch to u64_stats_t helpers. They provide all
the needed annotations, without adding extra cost.

[1]
BUG: KCSAN: data-race in tun_get_user / tun_net_get_stats64

read to 0xffffe8ffffd8aca8 of 8 bytes by task 4882 on cpu 0:
 tun_net_get_stats64+0x9b/0x230 drivers/net/tun.c:1171
 dev_get_stats+0x89/0x1e0 net/core/dev.c:9103
 rtnl_fill_stats+0x56/0x370 net/core/rtnetlink.c:1177
 rtnl_fill_ifinfo+0xd3b/0x2100 net/core/rtnetlink.c:1667
 rtmsg_ifinfo_build_skb+0xb0/0x150 net/core/rtnetlink.c:3472
 rtmsg_ifinfo_event.part.0+0x4e/0xb0 net/core/rtnetlink.c:3504
 rtmsg_ifinfo_event net/core/rtnetlink.c:3515 [inline]
 rtmsg_ifinfo+0x85/0x90 net/core/rtnetlink.c:3513
 __dev_notify_flags+0x18b/0x200 net/core/dev.c:7649
 dev_change_flags+0xb8/0xe0 net/core/dev.c:7691
 dev_ifsioc+0x201/0x6a0 net/core/dev_ioctl.c:237
 dev_ioctl+0x149/0x660 net/core/dev_ioctl.c:489
 sock_do_ioctl+0xdb/0x230 net/socket.c:1061
 sock_ioctl+0x3a3/0x5e0 net/socket.c:1189
 vfs_ioctl fs/ioctl.c:46 [inline]
 file_ioctl fs/ioctl.c:509 [inline]
 do_vfs_ioctl+0x991/0xc60 fs/ioctl.c:696

write to 0xffffe8ffffd8aca8 of 8 bytes by task 4883 on cpu 1:
 tun_get_user+0x1d94/0x2ba0 drivers/net/tun.c:2002
 tun_chr_write_iter+0x79/0xd0 drivers/net/tun.c:2022
 call_write_iter include/linux/fs.h:1895 [inline]
 new_sync_write+0x388/0x4a0 fs/read_write.c:483
 __vfs_write+0xb1/0xc0 fs/read_write.c:496
 __kernel_write+0xb8/0x240 fs/read_write.c:515
 write_pipe_buf+0xb6/0xf0 fs/splice.c:794
 splice_from_pipe_feed fs/splice.c:500 [inline]
 __splice_from_pipe+0x248/0x480 fs/splice.c:624
 splice_from_pipe+0xbb/0x100 fs/splice.c:659
 default_file_splice_write+0x45/0x90 fs/splice.c:806
 do_splice_from fs/splice.c:848 [inline]
 direct_splice_actor+0xa0/0xc0 fs/splice.c:1020
 splice_direct_to_actor+0x215/0x510 fs/splice.c:975
 do_splice_direct+0x161/0x1e0 fs/splice.c:1063
 do_sendfile+0x384/0x7f0 fs/read_write.c:1464

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 4883 Comm: syz-executor.1 Not tainted 5.4.0-rc3+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

u64_stats: provide u64_stats_t type

On 64bit arches, struct u64_stats_sync is empty and provides
no help against load/store tearing.

Using READ_ONCE()/WRITE_ONCE() would be needed.

But the update side would be slightly more expensive.

local64_t was defined so that we could use regular adds
in a manner which is atomic wrt IRQs.

However the u64_stats infra means we do not have to use
local64_t on 32bit arches since the syncp provides the needed
protection.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dummy: use standard dev_lstats_add() and dev_lstats_read()

This driver can simply use the common infrastructure instead
of duplicating it.

This cleanup will ease u64_stats_t adoption in a single location.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vsockmon: use standard dev_lstats_add() and dev_lstats_read()

This cleanup will ease u64_stats_t adoption in a single location.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

veth: use standard dev_lstats_add() and dev_lstats_read()

This cleanup will ease u64_stats_t adoption in a single location.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: nlmon: use standard dev_lstats_add() and dev_lstats_read()

No need to hand-code the exact same functions.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: provide dev_lstats_add() helper

Many network drivers need it and hand-coded the same function.

In order to ease u64_stats_t adoption, it is time to factorize.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: provide dev_lstats_read() helper

Many network drivers use hand-coded implementation of the same thing,
let's factorize things so that u64_stats_t adoption is done once.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-Demote-MTU-change-prints-to-debug'

Florian Fainelli says:

====================
net: Demote MTU change prints to debug

This patch series demotes several drivers that printed MTU change and
could therefore spam the kernel console if one has a test that it's all
about testing the values. Intel drivers were not also particularly
consistent in how they printed the same message, so now they are.
====================

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

net: qcom/emac: Demote MTU change print to debug

Changing the MTU can be a frequent operation and it is already clear
when (or not) a MTU change is successful, demote prints to debug prints.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Timur Tabi <timur@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: intel: Demote MTU change prints to debug

Changing a network device MTU can be a fairly frequent operation, and
failure to change the MTU is reflected to user-space properly, both by
an appropriate message as well as by looking at whether the device's MTU
matches the configuration.

Demote the prints to debug prints by using netdev_dbg(), making all
Intel wired LAN drivers consistent, since they used a mixture of PCI
device and network device prints before.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ethernet: ti: cpts: use ktime_get_real_ns helper

Update on more short variant for getting real clock in ns.

Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'aquantia-next'

Igor Russkikh says:

====================
Aquantia Marvell atlantic driver updates 11-2019

Here is a bunch of atlantic driver new features and updates.

Shortlist:
- Me adding ethtool private flags for various loopback test modes,
- Nikita is doing some work here on power management, implementing new PM API,
  He also did some checkpatch style cleanup of older driver parts.
- I'm also adding a new UDP GSO offload support and flags for loopback activation
- We are now Marvell, so I am changing email addresses on maintainers list.

v2: styling, ip6 correct handling in udpgso
====================

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

net: atlantic: change email domains to Marvell

Aquantia is now part of Marvell, eventually we'll cease standalone
aquantia.com domain. Thus, change the maintainers file and some other
references to @marvell.com domain

Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: implement UDP GSO offload

atlantic hardware does support UDP hardware segmentation offload.
This allows user to specify one large contiguous buffer with data
which then will be split automagically into multiple UDP packets
of specified size.

Bulk sending of large UDP streams lowers CPU usage and increases
bandwidth.

We did estimations both with udpgso_bench_tx test tool and with modified
iperf3 measurement tool (4 streams, multithread, 200b packet size)
over AQC<->AQC 10G link. Flow control is disabled to prevent RX side
impact on measurements.

No UDP GSO:
iperf3 -c 10.0.1.2 -u -b0 -l 200 -P4 --multithread
UDP GSO:
iperf3 -c 10.0.1.2 -u -b0 -l 12600 --udp-lso 200 -P4 --multithread

Mode          CPU   iperf speed    Line speed   Packets per second
-------------------------------------------------------------
NO UDP GSO    350%   3.07 Gbps      3.8 Gbps     1,919,419
SW UDP GSO    200%   5.55 Gbps      6.4 Gbps     3,286,144
HW UDP GSO    90%    6.80 Gbps      8.4 Gbps     4,273,117

Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: update flow control logic

We now differentiate requested and negotiated flow control
modes. Therefore `ethtool -A` now operates on local requested
FC values, and regular link settings shows the negotiated FC
settings.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: stylistic renames

We are trying to follow the naming of the chip (atlantic), not
company. So replace some old namings.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: code style cleanup

Thats a pure checkpatck walkthrough the code with no functional
changes. Reverse christmas tree, spacing, etc.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: loopback tests via private flags

Here we add a number of ethtool private flags
to allow enabling various loopbacks on HW.

Thats useful for verification and bringup works.

Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: add fw configuration memory area

Device FW has a separate memory area where various
config fields are stored and could be used by the
driver.

Here we modify download/upload infrastructure to
allow accessing this area.

Lateron this will be used to configure various behaviours

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: adding ethtool physical identification

`ethtool -p eth0` will blink leds helping identify
physical port.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: add msglevel configuration

We add ethtool msglevel configuration and change some
printouts to use netdev_info set of functions.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: refactoring pm logic

We now implement .driver.pm callbacks, these
allows driver to work correctly in hibernate
usecases, especially when used in conjunction with
WOL feature.

Before that driver only reacted to legacy .suspend/.resume
callbacks, that was a limitation in some cases.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: implement wake_phy feature

Wake on PHY allows to configure device to wakeup host
as soon as PHY link status is changed to active.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atlantic: update firmware interface

Here we improve FW interface structures layout
and prepare these for the wake phy feature implementation.

Signed-off-by: Nikita Danilov <ndanilov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'mlxsw-Add-layer-3-devlink-trap-support'

Ido Schimmel says:

====================
mlxsw: Add layer 3 devlink-trap support

This patch set from Amit adds support in mlxsw for layer 3 traps that
can report drops and exceptions via devlink-trap.

In a similar fashion to the existing layer 2 traps, these traps can send
packets to the CPU that were not routed as intended by the underlying
device.

The traps are divided between the two types detailed in devlink-trap
documentation: drops and exceptions. Unlike drops, packets received via
exception traps are also injected to the kernel's receive path, as they
are required for the correct functioning of the control plane. For
example, packets trapped due to TTL error must be injected to kernel's
receive path for traceroute to work properly.

Patch set overview:

Patch #1 adds the layer 3 drop traps to devlink along with their
documentation.

Patch #2 adds support for layer 3 drop traps in mlxsw.

Patches #3-#5 add selftests for layer 3 drop traps.

Patch #6 adds the layer 3 exception traps to devlink along with their
documentation.

Patches #7-#9 gradually add support for layer 3 exception traps in
mlxsw.

Patches #10-#12 add selftests for layer 3 exception traps.
====================

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

selftests: mlxsw: Add test cases for devlink-trap layer 3 exceptions

Test that each supported packet trap exception is triggered under the
right conditions.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: forwarding: tc_common: Add hitting check

Add an option to check that packets hit the tc filter without providing
the exact number of packets that should hit it.

It is useful while sending many packets in background and checking that
at least one of them hit the tc filter.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: forwarding: devlink: Add functionality for trap exceptions test

Add common part of all the tests - check devlink status to ensure that
packets were trapped.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add layer 3 devlink-trap exceptions support

Add the trap IDs used to report layer 3 exceptions.

Trapped packets are first reported to devlink and then injected to the
kernel's receive path. All the packets have 'offload_fwd_mark' set in
order to prevent them from potentially being forwarded by the bridge
again.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add specific trap for packets routed via invalid nexthops

Currently, mlxsw does not differentiate between these two cases of
routes with invalid nexthops:

1. Nexthops whose nexthop device is a mlxsw upper (has a RIF), but whose
neighbour could not be resolved

2. Nexthops whose nexthop device is not a mlxsw upper (e.g., management
interface)

Up until now this did not matter and mlxsw trapped packets for both
cases using the same trap ID. However, packets that should have been
routed in hardware (case 1), but incurred a problem are considered
exceptions and should be reported to the user. The two cases should
therefore be split between two different trap IDs.

Allocate a new adjacency entry during initialization and upon the
insertion of the first route with an invalid mlxsw nexthop, program this
entry to discard packets. Packets hitting this entry will be reported
using new trap ID - "DISCARD_ROUTER3".

In the future, the entry could be written during initialization, but
currently firmware requires a valid RIF, which is not available at this
stage.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add new FIB entry type for reject routes

Currently, packets that cannot be routed in hardware (e.g., nexthop
device is not upper of mlxsw), are trapped to the kernel for forwarding.
Such packets are trapped using "RTR_INGRESS0" trap. This trap also traps
packets that hit reject routes (e.g., "unreachable") so that the kernel
will generate the appropriate ICMP error message for them.

Subsequent patch will need to only report to devlink packets that hit a
reject route, which is impossible as long as "RTR_INGRESS0" is
overloaded like that.

Solve this by using "RTR_INGRESS1" trap for packets that hit reject
routes.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

devlink: Add layer 3 generic packet exception traps

Add layer 3 generic packet exception traps that can report trapped
packets and documentation of the traps.

Unlike drop traps, these exception traps also need to inject the packet
to the kernel's receive path. For example, a packet that was trapped due
to unreachable neighbour need to be injected into the kernel so that it
will trigger an ARP request or a neighbour solicitation message.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: mlxsw: Add test cases for devlink-trap layer 3 drops

Test that each supported packet trap is triggered under the right
conditions and that packets are indeed dropped and not forwarded.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: devlink: Make devlink_trap_cleanup() more generic

Add proto parameter in order to enable the use of devlink_trap_cleanup()
in tests that use IPv6 protocol.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: devlink: Export functions to devlink library

l2_drops_test() is used to check that drop traps are functioning as
intended. Currently it is only used in the layer 2 test, but it is also
useful for the layer 3 test introduced in the subsequent patch.

l2_drops_cleanup() is used to clean configurations and kill mausezahn
proccess.

Export the functions to the common devlink library to allow it to be
re-used by future tests.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add layer 3 devlink-trap support

Add the trap IDs and trap group used to report layer 3 drops. Register
layer 3 packet traps and associated layer 3 trap group with devlink
during driver initialization.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

devlink: Add layer 3 generic packet traps

Add packet traps that can report packets that were dropped during layer
3 forwarding.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

enetc: fix return value for enetc_ioctl()

Return -EOPNOTSUPP instead of -EINVAL if the requested ioctl is not
implemented.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Remove one extra ktime_get_ns() from cookie_init_timestamp

tcp_make_synack() already uses tcp_clock_ns(), and can pass
the value to cookie_init_timestamp() to avoid another call
to ktime_get_ns() helper.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: Add source route tests to fib_tests

Add tests to verify routes with source address set are deleted when
source address is deleted.

Signed-off-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

inetpeer: fix data-race in inet_putpeer / inet_putpeer

We need to explicitely forbid read/store tearing in inet_peer_gc()
and inet_putpeer().

The following syzbot report reminds us about inet_putpeer()
running without a lock held.

BUG: KCSAN: data-race in inet_putpeer / inet_putpeer

write to 0xffff888121fb2ed0 of 4 bytes by interrupt on cpu 0:
 inet_putpeer+0x37/0xa0 net/ipv4/inetpeer.c:240
 ip4_frag_free+0x3d/0x50 net/ipv4/ip_fragment.c:102
 inet_frag_destroy_rcu+0x58/0x80 net/ipv4/inet_fragment.c:228
 __rcu_reclaim kernel/rcu/rcu.h:222 [inline]
 rcu_do_batch+0x256/0x5b0 kernel/rcu/tree.c:2157
 rcu_core+0x369/0x4d0 kernel/rcu/tree.c:2377
 rcu_core_si+0x12/0x20 kernel/rcu/tree.c:2386
 __do_softirq+0x115/0x33f kernel/softirq.c:292
 invoke_softirq kernel/softirq.c:373 [inline]
 irq_exit+0xbb/0xe0 kernel/softirq.c:413
 exiting_irq arch/x86/include/asm/apic.h:536 [inline]
 smp_apic_timer_interrupt+0xe6/0x280 arch/x86/kernel/apic/apic.c:1137
 apic_timer_interrupt+0xf/0x20 arch/x86/entry/entry_64.S:830
 native_safe_halt+0xe/0x10 arch/x86/kernel/paravirt.c:71
 arch_cpu_idle+0x1f/0x30 arch/x86/kernel/process.c:571
 default_idle_call+0x1e/0x40 kernel/sched/idle.c:94
 cpuidle_idle_call kernel/sched/idle.c:154 [inline]
 do_idle+0x1af/0x280 kernel/sched/idle.c:263

write to 0xffff888121fb2ed0 of 4 bytes by interrupt on cpu 1:
 inet_putpeer+0x37/0xa0 net/ipv4/inetpeer.c:240
 ip4_frag_free+0x3d/0x50 net/ipv4/ip_fragment.c:102
 inet_frag_destroy_rcu+0x58/0x80 net/ipv4/inet_fragment.c:228
 __rcu_reclaim kernel/rcu/rcu.h:222 [inline]
 rcu_do_batch+0x256/0x5b0 kernel/rcu/tree.c:2157
 rcu_core+0x369/0x4d0 kernel/rcu/tree.c:2377
 rcu_core_si+0x12/0x20 kernel/rcu/tree.c:2386
 __do_softirq+0x115/0x33f kernel/softirq.c:292
 run_ksoftirqd+0x46/0x60 kernel/softirq.c:603
 smpboot_thread_fn+0x37d/0x4a0 kernel/smpboot.c:165
 kthread+0x1d4/0x200 drivers/block/aoe/aoecmd.c:1253
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 16 Comm: ksoftirqd/1 Not tainted 5.4.0-rc3+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

Fixes: 4b9d9be839fd ("inetpeer: remove unused list")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: add missing dependency on CONFIG_REGULATOR

Compilation fails on PPC targets as CONFIG_REGULATOR is not set and
drivers/regulator/devres.c is not compiled in while functions exported
there are used by drivers/net/phy/at803x.c. Here's the error log:

  LD      .tmp_vmlinux1
drivers/net/phy/at803x.o: In function `at803x_rgmii_reg_set_voltage_sel':
drivers/net/phy/at803x.c:294: undefined reference to `.rdev_get_drvdata'
drivers/net/phy/at803x.o: In function `at803x_rgmii_reg_get_voltage_sel':
drivers/net/phy/at803x.c:306: undefined reference to `.rdev_get_drvdata'
drivers/net/phy/at803x.o: In function `at8031_register_regulators':
drivers/net/phy/at803x.c:359: undefined reference to `.devm_regulator_register'
drivers/net/phy/at803x.c:365: undefined reference to `.devm_regulator_register'
drivers/net/phy/at803x.o:(.data.rel+0x0): undefined reference to `regulator_list_voltage_table'
linux/Makefile:1074: recipe for target 'vmlinux' failed
make[1]: *** [vmlinux] Error 1

Fixes: 2f664823a470 ("net: phy: at803x: add device tree binding")
Signed-off-by: Madalin Bucur <madalin.bucur@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dpaa2-eth: add ethtool MAC counters

When a DPNI is connected to a MAC, export its associated counters.
Ethtool related functions are added in dpaa2_mac for returning the
number of counters, their strings and also their values.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

enetc: ethtool: add wake-on-lan callbacks

If there is an external PHY, pass the wake-on-lan request to the PHY.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

enetc: add ioctl() support for PHY-related ops

If there is an attached PHY try to handle the requested ioctl with its
handler, which allows the userspace to access PHY registers, for
example. This will make mii-diag and similar tools work.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: spectrum: Fix error return code in mlxsw_sp_port_module_info_init()

Fix to return negative error code -ENOMEM from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: 4a7f970f1240 ("mlxsw: spectrum: Replace port_to_module array with array of structs")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'cxgb4-add-support-for-TC-MQPRIO-Qdisc-Offload'

Rahul Lakkireddy says:

====================
cxgb4: add support for TC-MQPRIO Qdisc Offload

This series of patches add support for offloading TC-MQPRIO Qdisc
to Chelsio T5/T6 NICs. Offloading QoS traffic shaping and pacing
requires using Ethernet Offload (ETHOFLD) resources available on
Chelsio NICs. The ETHOFLD resources are configured by firmware
and taken from the resource pool shared with other Chelsio Upper
Layer Drivers. Traffic flowing through ETHOFLD region requires a
software netdev Tx queue (EOSW_TXQ) exposed to networking stack,
and an underlying hardware Tx queue (EOHW_TXQ) used for sending
packets through hardware.

ETHOFLD region is addressed using EOTIDs, which are per-connection
resource. Hence, EOTIDs are capable of storing only a very small
number of packets in flight. To allow more connections to share
the the QoS rate limiting configuration, multiple EOTIDs must be
allocated to reduce packet drops. EOTIDs are 1-to-1 mapped with
software EOSW_TXQ. Several software EOSW_TXQs can post packets to
a single hardware EOHW_TXQ.

The series is broken down as follows:

Patch 1 queries firmware for maximum available traffic classes,
as well as, start and maximum available indices (EOTID) into ETHOFLD
region, supported by the underlying device.

Patch 2 reworks queue configuration and simplifies MSI-X allocation
logic in preparation for ETHOFLD queues support.

Patch 3 adds skeleton for validating and configuring TC-MQPRIO Qdisc
offload. Also, adds support for software EOSW_TXQs and exposes them
to network stack. Updates Tx queue selection to use fallback NIC Tx
path for unsupported traffic that can't go through ETHOFLD queues.

Patch 4 adds support for managing hardware queues to rate limit
traffic flowing through them. The queues are allocated/removed based
on enabling/disabling TC-MQPRIO Qdisc offload, respectively.

Patch 5 adds Tx path for traffic flowing through software EOSW_TXQ
and EOHW_TXQ. Also, adds Rx path to handle Tx completions.

Patch 6 updates exisiting SCHED API to configure FLOWC based QoS
offload. In the existing QUEUE based rate limiting, multiple queues
sharing a traffic class get the aggreagated max rate limit value.
On the other hand, in FLOWC based rate limiting, multiple queues
sharing a traffic class get their own individual max rate limit
value. For example, if 2 queues are bound to class 0, which is rate
limited to 1 Gbps, then in QUEUE based rate limiting, both the
queues get the aggregate max output of 1 Gbps only. In FLOWC based
rate limiting, each queue gets its own output of max 1 Gbps each;
i.e. 2 queues * 1 Gbps rate limit = 2 Gbps max output.
====================

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

cxgb4: add FLOWC based QoS offload

Rework SCHED API to allow offloading TC-MQPRIO QoS configuration.
The existing QUEUE based rate limiting throttles all queues sharing
a traffic class, to the specified max rate limit value. So, if
multiple queues share a traffic class, then all the queues get
the aggregate specified max rate limit.

So, introduce the new FLOWC based rate limiting, where multiple
queues can share a traffic class with each queue getting its own
individual specified max rate limit.

For example, if 2 queues are bound to class 0, which is rate limited
to 1 Gbps, then 2 queues using QUEUE based rate limiting, get the
aggregate output of 1 Gbps only. In FLOWC based rate limiting, each
queue gets its own output of max 1 Gbps each; i.e. 2 queues * 1 Gbps
rate limit = 2 Gbps.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: add Tx and Rx path for ETHOFLD traffic

Implement Tx path for traffic flowing through software EOSW_TXQ
and EOHW_TXQ. Since multiple EOSW_TXQ can post packets to a single
EOHW_TXQ, protect the hardware queue with necessary spinlock. Also,
move common code used to generate TSO work request to a common
function.

Implement Rx path to handle Tx completions for successfully
transmitted packets.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: add ETHOFLD hardware queue support

Add support for configuring and managing ETHOFLD hardware queues.
Keep the queue count and MSI-X allocation scheme same as NIC queues.
ETHOFLD hardware queues are dynamically allocated/destroyed as
TC-MQPRIO Qdisc offload is enabled/disabled on the corresponding
interface, respectively.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: parse and configure TC-MQPRIO offload

Add logic for validation and configuration of TC-MQPRIO Qdisc
offload. Also, add support to manage EOSW_TXQ, which have 1-to-1
mapping with EOTIDs, and expose them to network stack.

Move common skb validation in Tx path to a separate function and
add minimal Tx path for ETHOFLD. Update Tx queue selection to return
normal NIC Txq to send traffic pattern that can't go through ETHOFLD
Tx path.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: rework queue config and MSI-X allocation

Simplify queue configuration and MSI-X allocation logic. Use a single
MSI-X information table for both NIC and ULDs. Remove hard-coded
MSI-X indices for firmware event queue and non data interrupts.
Instead, use the MSI-X bitmap to obtain a free MSI-X index
dynamically. Save each Rxq's index into the MSI-X information table,
within the Rxq structures themselves, for easier cleanup.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: query firmware for QoS offload resources

QoS offload needs Ethernet Offload (ETHOFLD) resources present in the
NIC. These resources are shared with other ULDs. So, query firmware
for the available number of traffic classes, as well as, start and
end indices (EOTID) of the ETHOFLD region.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: gen_estimator: extend packet counter to 64bit

I forgot to change last_packets field in struct net_rate_estimator.

Without this fix, rate estimators would misbehave after more
than 2^32 packets have been sent.

Another solution would be to be careful and only use the
32 least significant bits of packets counters, but we have
a hole in net_rate_estimator structure and this looks
easier to read/maintain.

Fixes: d0083d98f685 ("net_sched: extend packet counter to 64bit")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dpaa2-ptp: fix compile error

phylink_set_port_modes will be compiled if CONFIG_PHYLINK enabled,
dpaa2_mac_validate will be compiled if CONFIG_FSL_DPAA2_ETH enabled,
it should select CONFIG_PHYLINK when dpaa2_mac_validate call
phylink_set_port_modes

drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.o: In function `dpaa2_mac_validate':
dpaa2-mac.c:(.text+0x3a1): undefined reference to `phylink_set_port_modes'
drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.o: In function `dpaa2_mac_connect':
dpaa2-mac.c:(.text+0x91a): undefined reference to `phylink_create'
dpaa2-mac.c:(.text+0x94e): undefined reference to `phylink_of_phy_connect'
dpaa2-mac.c:(.text+0x97f): undefined reference to `phylink_destroy'
drivers/net/ethernet/freescale/dpaa2/dpaa2-mac.o: In function `dpaa2_mac_disconnect':
dpaa2-mac.c:(.text+0xa9f): undefined reference to `phylink_disconnect_phy'
dpaa2-mac.c:(.text+0xab0): undefined reference to `phylink_destroy'
make: *** [vmlinux] Error 1

Fixes: 719479230893 ("dpaa2-eth: add MAC/PHY support through phylink")
Signed-off-by: Chenwandun <chenwandun@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2019-11-06

This series contains updates to ice driver only.

Scott adds ethtool -m support so that we can read eeprom data on SFP/OSFP
modules.

Anirudh updates the return value to properly reflect when SRIOV is not
supported.

Md Fahad updates the driver to handle a change in the NVM, where the
boot configuration section was moved to the Preserved Field Area (PFA)
of the NVM.

Paul resolves an issue when DCBx requests non-contiguous TCs, transmit
hangs could occur, so configure a default traffic class (TC0) in these
cases to prevent traffic hangs.  Adds a print statement to notify the
user when unsupported modules are inserted.

Bruce fixes up the driver unload code flow to ensure we do not clear the
interrupt scheme until the reset is complete, otherwise a hardware error
may occur.

Dave updates the DCB initialization to set is_sw_lldp boolean when the
firmware has been detected to be in an untenable state.  This will
ensure that the firmware is in a known state.

Michal saves off the PCI state and I/O BARs address after PCI bus reset
so that after the reset, device registers can be read.  Also adds a NULL
pointer check to prevent a potential kernel panic.

Mitch resolves an issue where VF's on PF's other than 0 were not seeing
resets by using the per-PF VF ID instead of the absolute VF ID.

Krzysztof does some code cleanup to remove a unneeded wrapper and
reduces the code complexity.

Brett reduces confusion by changing the name of ice_vc_dis_vf() to
ice_vc_reset_vf() to better describe what the function is actually
doing.

v2: dropped patch 3 "ice: Add support for FW recovery mode detection"
    from the origin al series, while Ani makes changes based on
    community feedback to implement devlink into the changes.
v3: dropped patch 1 "ice: implement set_eeprom functionality" due to a
    bug found and additional changes will be needed when Ani implements
    devlink in the driver.
====================

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

net: dsa: mv8e6xxx: Fix stub function parameters

mv88e6xxx_g2_atu_stats_get() takes two parameters. Make the stub
function also take two, otherwise we get compile errors.

Fixes: c5f299d59261 ("net: dsa: mv88e6xxx: global1_atu: Add helper for get next")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-phy-at803x-device-tree-binding'

Michael Walle says:

====================
net: phy: at803x device tree binding

Adds a device tree binding to configure the clock and the RGMII voltage.

Changes since v1:
 - rebased to latest net-next
 - renamed "Atheros" to "Qualcomm Atheros"
 - add a new patch to remove config_init() from AR9331

Changes since the RFC:
 - renamed the Kconfig entry to "Qualcomm Atheros.." and reordered the
   item
 - renamed the prefix from atheros to qca
 - use the correct name AR803x (instead of AT803x) in new files and
   dt-bindings.
 - listed the PHY maintainers in the new schema. Hopefully, thats ok.
 - fixed a typo in the bindings schema
 - run dtb_checks and dt_binding_check and fixed the schema
 - dropped the rgmii-io-1v8 property; instead provide two regulators vddh
   and vddio, add one consumer vddio-supply
 - fix the clock settings for the AR8030/AR8035
 - only the AR8031 supports chaning the LDO and the PLL mode in software.
   Check if we have the correct PHY.
 - new patch to mention the AR8033 which is the same as the AR8031 just
   without PTP support
 - new patch which corrects any displayed PHY names and comments. Be
   consistent.
====================

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

net: phy: at803x: remove config_init for AR9331

According to its datasheet, the internal PHY doesn't have debug
registers nor MMDs. Since config_init() only configures delays and
clocks and so on in these registers it won't be needed on this PHY.
Remove it.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: fix the PHY names

Fix at least the displayed strings. The actual name of the chip is
AR803x.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: mention AR8033 as same as AR8031

The AR8033 is the AR8031 without PTP support. All other registers are
the same. Unfortunately, they share the same PHY ID. Therefore, we
cannot distinguish between the one with PTP support and the one without.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: add device tree binding

Add support for configuring the CLK_25M pin as well as the RGMII I/O
voltage by the device tree.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: phy: Add support for AT803X

Document the Atheros AR803x PHY bindings.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: fix Kconfig description

The name of the PHY is actually AR803x not AT803x. Additionally, add the
name of the vendor and mention the AR8031 support.

Signed-off-by: Michael Walle <michael@walle.cc>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: fix data-race in tcp_recvmsg()

Reading tp->recvmsg_inq after socket lock is released
raises a KCSAN warning [1]

Replace has_tss & has_cmsg by cmsg_flags and make
sure to not read tp->recvmsg_inq a second time.

[1]
BUG: KCSAN: data-race in tcp_chrono_stop / tcp_recvmsg

write to 0xffff888126adef24 of 2 bytes by interrupt on cpu 0:
 tcp_chrono_set net/ipv4/tcp_output.c:2309 [inline]
 tcp_chrono_stop+0x14c/0x280 net/ipv4/tcp_output.c:2338
 tcp_clean_rtx_queue net/ipv4/tcp_input.c:3165 [inline]
 tcp_ack+0x274f/0x3170 net/ipv4/tcp_input.c:3688
 tcp_rcv_established+0x37e/0xf50 net/ipv4/tcp_input.c:5696
 tcp_v4_do_rcv+0x381/0x4e0 net/ipv4/tcp_ipv4.c:1561
 tcp_v4_rcv+0x19dc/0x1bb0 net/ipv4/tcp_ipv4.c:1942
 ip_protocol_deliver_rcu+0x4d/0x420 net/ipv4/ip_input.c:204
 ip_local_deliver_finish+0x110/0x140 net/ipv4/ip_input.c:231
 NF_HOOK include/linux/netfilter.h:305 [inline]
 NF_HOOK include/linux/netfilter.h:299 [inline]
 ip_local_deliver+0x133/0x210 net/ipv4/ip_input.c:252
 dst_input include/net/dst.h:442 [inline]
 ip_rcv_finish+0x121/0x160 net/ipv4/ip_input.c:413
 NF_HOOK include/linux/netfilter.h:305 [inline]
 NF_HOOK include/linux/netfilter.h:299 [inline]
 ip_rcv+0x18f/0x1a0 net/ipv4/ip_input.c:523
 __netif_receive_skb_one_core+0xa7/0xe0 net/core/dev.c:5010
 __netif_receive_skb+0x37/0xf0 net/core/dev.c:5124
 netif_receive_skb_internal+0x59/0x190 net/core/dev.c:5214
 napi_skb_finish net/core/dev.c:5677 [inline]
 napi_gro_receive+0x28f/0x330 net/core/dev.c:5710

read to 0xffff888126adef25 of 1 bytes by task 7275 on cpu 1:
 tcp_recvmsg+0x77b/0x1a30 net/ipv4/tcp.c:2187
 inet_recvmsg+0xbb/0x250 net/ipv4/af_inet.c:838
 sock_recvmsg_nosec net/socket.c:871 [inline]
 sock_recvmsg net/socket.c:889 [inline]
 sock_recvmsg+0x92/0xb0 net/socket.c:885
 sock_read_iter+0x15f/0x1e0 net/socket.c:967
 call_read_iter include/linux/fs.h:1889 [inline]
 new_sync_read+0x389/0x4f0 fs/read_write.c:414
 __vfs_read+0xb1/0xc0 fs/read_write.c:427
 vfs_read fs/read_write.c:461 [inline]
 vfs_read+0x143/0x2c0 fs/read_write.c:446
 ksys_read+0xd5/0x1b0 fs/read_write.c:587
 __do_sys_read fs/read_write.c:597 [inline]
 __se_sys_read fs/read_write.c:595 [inline]
 __x64_sys_read+0x4c/0x60 fs/read_write.c:595
 do_syscall_64+0xcc/0x370 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 7275 Comm: sshd Not tainted 5.4.0-rc3+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

Fixes: b75eba76d3d7 ("tcp: send in-queue bytes in cmsg upon read")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: silence data-races on sk_backlog.tail

sk->sk_backlog.tail might be read without holding the socket spinlock,
we need to add proper READ_ONCE()/WRITE_ONCE() to silence the warnings.

KCSAN reported :

BUG: KCSAN: data-race in tcp_add_backlog / tcp_recvmsg

write to 0xffff8881265109f8 of 8 bytes by interrupt on cpu 1:
 __sk_add_backlog include/net/sock.h:907 [inline]
 sk_add_backlog include/net/sock.h:938 [inline]
 tcp_add_backlog+0x476/0xce0 net/ipv4/tcp_ipv4.c:1759
 tcp_v4_rcv+0x1a70/0x1bd0 net/ipv4/tcp_ipv4.c:1947
 ip_protocol_deliver_rcu+0x4d/0x420 net/ipv4/ip_input.c:204
 ip_local_deliver_finish+0x110/0x140 net/ipv4/ip_input.c:231
 NF_HOOK include/linux/netfilter.h:305 [inline]
 NF_HOOK include/linux/netfilter.h:299 [inline]
 ip_local_deliver+0x133/0x210 net/ipv4/ip_input.c:252
 dst_input include/net/dst.h:442 [inline]
 ip_rcv_finish+0x121/0x160 net/ipv4/ip_input.c:413
 NF_HOOK include/linux/netfilter.h:305 [inline]
 NF_HOOK include/linux/netfilter.h:299 [inline]
 ip_rcv+0x18f/0x1a0 net/ipv4/ip_input.c:523
 __netif_receive_skb_one_core+0xa7/0xe0 net/core/dev.c:4929
 __netif_receive_skb+0x37/0xf0 net/core/dev.c:5043
 netif_receive_skb_internal+0x59/0x190 net/core/dev.c:5133
 napi_skb_finish net/core/dev.c:5596 [inline]
 napi_gro_receive+0x28f/0x330 net/core/dev.c:5629
 receive_buf+0x284/0x30b0 drivers/net/virtio_net.c:1061
 virtnet_receive drivers/net/virtio_net.c:1323 [inline]
 virtnet_poll+0x436/0x7d0 drivers/net/virtio_net.c:1428
 napi_poll net/core/dev.c:6311 [inline]
 net_rx_action+0x3ae/0xa90 net/core/dev.c:6379
 __do_softirq+0x115/0x33f kernel/softirq.c:292
 invoke_softirq kernel/softirq.c:373 [inline]
 irq_exit+0xbb/0xe0 kernel/softirq.c:413
 exiting_irq arch/x86/include/asm/apic.h:536 [inline]
 do_IRQ+0xa6/0x180 arch/x86/kernel/irq.c:263
 ret_from_intr+0x0/0x19
 native_safe_halt+0xe/0x10 arch/x86/kernel/paravirt.c:71
 arch_cpu_idle+0x1f/0x30 arch/x86/kernel/process.c:571
 default_idle_call+0x1e/0x40 kernel/sched/idle.c:94
 cpuidle_idle_call kernel/sched/idle.c:154 [inline]
 do_idle+0x1af/0x280 kernel/sched/idle.c:263
 cpu_startup_entry+0x1b/0x20 kernel/sched/idle.c:355
 start_secondary+0x208/0x260 arch/x86/kernel/smpboot.c:264
 secondary_startup_64+0xa4/0xb0 arch/x86/kernel/head_64.S:241

read to 0xffff8881265109f8 of 8 bytes by task 8057 on cpu 0:
 tcp_recvmsg+0x46e/0x1b40 net/ipv4/tcp.c:2050
 inet_recvmsg+0xbb/0x250 net/ipv4/af_inet.c:838
 sock_recvmsg_nosec net/socket.c:871 [inline]
 sock_recvmsg net/socket.c:889 [inline]
 sock_recvmsg+0x92/0xb0 net/socket.c:885
 sock_read_iter+0x15f/0x1e0 net/socket.c:967
 call_read_iter include/linux/fs.h:1889 [inline]
 new_sync_read+0x389/0x4f0 fs/read_write.c:414
 __vfs_read+0xb1/0xc0 fs/read_write.c:427
 vfs_read fs/read_write.c:461 [inline]
 vfs_read+0x143/0x2c0 fs/read_write.c:446
 ksys_read+0xd5/0x1b0 fs/read_write.c:587
 __do_sys_read fs/read_write.c:597 [inline]
 __se_sys_read fs/read_write.c:595 [inline]
 __x64_sys_read+0x4c/0x60 fs/read_write.c:595
 do_syscall_64+0xcc/0x370 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 8057 Comm: syz-fuzzer Not tainted 5.4.0-rc6+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dpaa2-eth: fix an always true condition in dpaa2_mac_get_if_mode

Convert the phy_mode() function to return the if_mode through an
argument, similar to the new form of of_get_phy_mode().
This will help with handling errors in a common manner and also will fix
an always true condition.

Fixes: 0c65b2b90d13 ("net: of_get_phy_mode: Change API to solve int/unit warnings")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: openvswitch: select vport upcall portid directly

The commit 69c51582ff786 ("dpif-netlink: don't allocate per
thread netlink sockets"), in Open vSwitch ovs-vswitchd, has
changed the number of allocated sockets to just one per port
by moving the socket array from a per handler structure to
a per datapath one. In the kernel datapath, a vport will have
only one socket in most case, if so select it directly in
fast-path.

Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Acked-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: axienet: Fix error return code in axienet_probe()

In the DMA memory resource get failed case, the error is not
set and 0 will be returned. Fix it by removing redundant check
since devm_ioremap_resource() will handle it.

Fixes: 28ef9ebdb64c ("net: axienet: make use of axistream-connected attribute optional")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: aquantia: fix return value check in aq_ptp_init()

Function ptp_clock_register() returns ERR_PTR() and never returns
NULL. The NULL test should be removed.

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: ptp_clockmatrix: Fix missing unlock on error in idtcm_probe()

Add the missing unlock before return from function idtcm_probe()
in the error handling case.

Fixes: 3a6ba7dc7799 ("ptp: Add a ptp clock driver for IDT ClockMatrix.")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Vincent Cheng <vincent.cheng.xh@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: eliminate the dummy packet in link synching

When preparing tunnel packets for the link failover or synchronization,
as for the safe algorithm, we added a dummy packet on the pair link but
never sent it out. In the case of failover, the pair link will be reset
anyway. But for link synching, it will always result in retransmission
of the dummy packet after that.
We have also observed that such the retransmission at the early stage
when a new node comes in a large cluster will take some time and hard
to be done, leading to the repeated retransmit failures and the link is
reset.

Since in commit 4929a932be33 ("tipc: optimize link synching mechanism")
we have already built a dummy 'TUNNEL_PROTOCOL' message on the new link
for the synchronization, there's no need for the dummy on the pair one,
this commit will skip it when the new mechanism takes in place. In case
nothing exists in the pair link's transmq, the link synching will just
start and stop shortly on the peer side.

The patch is backward compatible.

Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Tested-by: Hoang Le <hoang.h.le@dektech.com.au>
Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'lwtunnel-add-ip-and-ip6-options-setting-and-dumping'

Xin Long says:

====================
lwtunnel: add ip and ip6 options setting and dumping

With this patchset, users can configure options by ip route encap
for geneve, vxlan and ersapn lwtunnel, like:

  # ip r a 1.1.1.0/24 encap ip id 1 geneve class 0 type 0 \
    data "1212121234567890" dst 10.1.0.2 dev geneve1

  # ip r a 1.1.1.0/24 encap ip id 1 vxlan gbp 456 \
    dst 10.1.0.2 dev erspan1

  # ip r a 1.1.1.0/24 encap ip id 1 erspan ver 1 idx 123 \
    dst 10.1.0.2 dev erspan1

iproute side patch is attached on the reply of this mail.
====================

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

lwtunnel: add options setting and dumping for erspan

Based on the code framework built on the last patch, to
support setting and dumping for vxlan, we only need to
add ip_tun_parse_opts_erspan() for .build_state and
ip_tun_fill_encap_opts_erspan() for .fill_encap and
if (tun_flags & TUNNEL_ERSPAN_OPT) for .get_encap_size.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lwtunnel: add options setting and dumping for vxlan

Based on the code framework built on the last patch, to
support setting and dumping for vxlan, we only need to
add ip_tun_parse_opts_vxlan() for .build_state and
ip_tun_fill_encap_opts_vxlan() for .fill_encap and
if (tun_flags & TUNNEL_VXLAN_OPT) for .get_encap_size.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lwtunnel: add options setting and dumping for geneve

To add options setting and dumping, .build_state(), .fill_encap() and
.get_encap_size() in ip_tun_lwt_ops needs to be extended:

ip_tun_build_state():
  ip_tun_parse_opts():
    ip_tun_parse_opts_geneve()

ip_tun_fill_encap_info():
  ip_tun_fill_encap_opts():
    ip_tun_fill_encap_opts_geneve()

ip_tun_encap_nlsize()
   ip_tun_opts_nlsize():
     if (tun_flags & TUNNEL_GENEVE_OPT)

ip_tun_parse_opts(), ip_tun_fill_encap_opts() and ip_tun_opts_nlsize()
processes LWTUNNEL_IP_OPTS.

ip_tun_parse_opts_geneve(), ip_tun_fill_encap_opts_geneve() and
if (tun_flags & TUNNEL_GENEVE_OPT) processes LWTUNNEL_IP_OPTS_GENEVE.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lwtunnel: add options process for cmp_encap

When comparing two tun_info, dst_cache member should have been skipped,
as dst_cache is a per cpu pointer and they are always different values
even in two tun_info with the same keys.

So this patch is to skip dst_cache member and compare the key, mode and
options_len only. For the future opts setting support, also to compare
options.

Fixes: 2d79849903e0 ("lwtunnel: ip tunnel: fix multiple routes with different encap")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lwtunnel: add options process for arp request

Without options copied to the dst tun_info in iptunnel_metadata_reply()
called by arp_process for handling arp_request, the generated arp_reply
packet may be dropped or sent out with wrong options for some tunnels
like erspan and vxlan, and the traffic will break.

Fixes: 63d008a4e9ee ("ipv4: send arp replies to the correct tunnel")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: reduce sensitive to retransmit failures

With huge cluster (e.g >200nodes), the amount of that flow:
gap -> retransmit packet -> acked will take time in case of STATE_MSG
dropped/delayed because a lot of traffic. This lead to 1.5 sec tolerance
value criteria made link easy failure around 2nd, 3rd of failed
retransmission attempts.

Instead of re-introduced criteria of 99 faled retransmissions to fix the
issue, we increase failure detection timer to ten times tolerance value.

Fixes: 77cf8edbc0e7 ("tipc: simplify stale link failure criteria")
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au>
Acked-by: Jon
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: update cluster capabilities if node deleted

There are two improvements when re-calculate cluster capabilities:

- When deleting a specific down node, need to re-calculate.
- In tipc_node_cleanup(), do not need to re-calculate if node
is still existing in cluster.

Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au>
Acked-by: Jon
Signed-off-by: David S. Miller <davem@davemloft.net>

selftest: net: add some traceroute tests

Added the following traceroute tests.

IPV6:
Verify that in this scenario

       ------------------------ N2
        |                    |
      ------              ------  N3  ----
      | R1 |              | R2 |------|H2|
      ------              ------      ----
        |                    |
       ------------------------ N1
                 |
                ----
                |H1|
                ----

where H1's default route goes through R1 and R1's default route goes
through R2 over N2, traceroute6 from H1 to H2 reports R2's address
on N2 and not N1.

IPV4:
Verify that traceroute from H1 to H2 shows 1.0.1.1 in this scenario

                   1.0.3.1/24
---- 1.0.1.3/24    1.0.1.1/24 ---- 1.0.2.1/24    1.0.2.4/24 ----
|H1|--------------------------|R1|--------------------------|H2|
----            N1            ----            N2            ----

where net.ipv4.icmp_errors_use_inbound_ifaddr is set on R1 and
1.0.3.1/24 and 1.0.1.1/24 are respectively R1's primary and secondary
address on N1.

v2: fixed some typos, and have bridge in R1 instead of R2 in IPV6 test.

Signed-off-by: Francesco Ruggeri <fruggeri@arista.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>