openwrt/staging/blogic.git
5 years agomlxsw: spectrum: Add support for traps specific to Spectrum-1
Petr Machata [Sun, 30 Jun 2019 06:04:50 +0000 (09:04 +0300)]
mlxsw: spectrum: Add support for traps specific to Spectrum-1

On Spectrum-1, timestamps for PTP packets are delivered through queues
of ingress and egress timestamps. There are two event traps
corresponding to activity on each of those queues. This mechanism is
absent on Spectrum-2, and therefore the traps should only be registered
on Spectrum-1.

Carry a chip-specific listener array in mlxsw_sp->listeners and
listeners_count. Register listeners from that array in
mlxsw_sp_traps_init(). Add a new listener array for Spectrum-1 traps and
configure the newly-added mlxsw_sp->listeners with this array.

The listener array is empty for now, the events will be added in a later
patch.

Signed-off-by: Petr Machata <petrm@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>
5 years agomlxsw: spectrum: Extract a helper for trap registration
Petr Machata [Sun, 30 Jun 2019 06:04:49 +0000 (09:04 +0300)]
mlxsw: spectrum: Extract a helper for trap registration

On Spectrum-1, timestamps for PTP packets are delivered through queues
of ingress and egress timestamps. There are two event traps
corresponding to activity on each of those queues. This mechanism is
absent on Spectrum-2, and therefore the traps should only be registered
on Spectrum-1.

Extract out of mlxsw_sp_traps_init() a generic helper,
mlxsw_sp_traps_register(), and likewise with _unregister(). The new helpers
will later be called with Spectrum-1-specific traps.

Signed-off-by: Petr Machata <petrm@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>
5 years agomlxsw: reg: Add Monitoring Global Configuration Register
Petr Machata [Sun, 30 Jun 2019 06:04:48 +0000 (09:04 +0300)]
mlxsw: reg: Add Monitoring Global Configuration Register

This register serves to configure global parameters of certain
monitoring operations. The following patches will use it to configure
that when PTP timestamps are delivered through the PTP FIFO traps, the
FIFO in question is cleared as well.

Signed-off-by: Petr Machata <petrm@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>
5 years agomlxsw: reg: Add Time Precision Packet Timestamping Reading
Petr Machata [Sun, 30 Jun 2019 06:04:47 +0000 (09:04 +0300)]
mlxsw: reg: Add Time Precision Packet Timestamping Reading

The MTPPTR is used for reading the per port PTP timestamp FIFO.

Signed-off-by: Petr Machata <petrm@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>
5 years agomlxsw: reg: Add Monitoring Precision Time Protocol Trap Register
Petr Machata [Sun, 30 Jun 2019 06:04:46 +0000 (09:04 +0300)]
mlxsw: reg: Add Monitoring Precision Time Protocol Trap Register

This register is used for configuring under which trap to deliver PTP
packets depending on type of the packet.

Signed-off-by: Petr Machata <petrm@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>
5 years agomlxsw: reg: Add Monitoring Time Precision Packet Port Configuration Register
Petr Machata [Sun, 30 Jun 2019 06:04:45 +0000 (09:04 +0300)]
mlxsw: reg: Add Monitoring Time Precision Packet Port Configuration Register

This register serves for configuration of which PTP messages should be
timestamped. This is a global configuration, despite the register name.

Signed-off-by: Petr Machata <petrm@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>
5 years agosamples: pktgen: allow to specify destination port
Daniel T. Lee [Sat, 29 Jun 2019 13:33:58 +0000 (22:33 +0900)]
samples: pktgen: allow to specify destination port

Currently, kernel pktgen has the feature to specify udp destination port
for sending packet. (e.g. pgset "udp_dst_min 9")

But on samples, each of the scripts doesn't have any option to achieve this.

This commit adds the DST_PORT option to specify the target port(s) in the script.

    -p : ($DST_PORT)  destination PORT range (e.g. 433-444) is also allowed

Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosamples: pktgen: add some helper functions for port parsing
Daniel T. Lee [Sat, 29 Jun 2019 13:33:57 +0000 (22:33 +0900)]
samples: pktgen: add some helper functions for port parsing

This commit adds port parsing and port validate helper function to parse
single or range of port(s) from a given string. (e.g. 1234, 443-444)

Helpers will be used in prior to set target port(s) in samples/pktgen.

Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv6: icmp: allow flowlabel reflection in echo replies
Eric Dumazet [Mon, 1 Jul 2019 13:39:36 +0000 (06:39 -0700)]
ipv6: icmp: allow flowlabel reflection in echo replies

Extend flowlabel_reflect bitmask to allow conditional
reflection of incoming flowlabels in echo replies.

Note this has precedence against auto flowlabels.

Add flowlabel_reflect enum to replace hard coded
values.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5e-updates-2019-06-28' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Mon, 1 Jul 2019 01:41:13 +0000 (18:41 -0700)]
Merge tag 'mlx5e-updates-2019-06-28' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5e-updates-2019-06-28

This series adds some misc updates for mlx5e driver

1) Allow adding the same mac more than once in MPFS table
2) Move to HW checksumming advertising
3) Report netdevice MPLS features
4) Correct physical port name of the PF representor
5) Reduce stack usage in mlx5_eswitch_termtbl_create
6) Refresh TIR improvement for representors
7) Expose same physical switch_id for all representors
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch '10GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Sun, 30 Jun 2019 23:03:35 +0000 (16:03 -0700)]
Merge branch '10GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2019-06-28

This series contains a smorgasbord of updates to many of the Intel
drivers.

Gustavo A. R. Silva updates the ice and iavf drivers to use the
strcut_size() helper where possible.

Miguel increases the pause and refresh time for flow control in the
e1000e driver during reset for certain devices.

Dann Frazier fixes a potential NULL pointer dereference in ixgbe driver
when using non-IPSec enabled devices.

Colin Ian King fixes a potential overflow during a shift in the ixgbe
driver.  Also fixes a potential NULL pointer dereference in the iavf
driver by adding a check.

Venkatesh Srinivas converts the e1000 driver to use dma_wmb() instead of
wmb() for doorbell writes to avoid SFENCEs in the transmit and receive
paths.

Arjan updates the e1000e driver to improve boot time by over 100 msec by
reducing the usleep ranges suring system startup.

Artem updates the igb driver register dump in ethtool, first prepares
the register dump for future additions of registers in the dump, then
secondly, adds the RR2DCDELAY register to the dump.  When dealing with
time-sensitive networks, this register is helpful in determining your
latency from the device to the ring.

Alex fixes the ixgbevf driver to use the current cached link state,
rather than trying to re-check the value from the PF.

Harshitha adds support for MACVLAN offloads in i40e by using channels as
MACVLAN interfaces.

Detlev Casanova updates the e1000e driver to use delayed work instead of
timers to run the watchdog.

Vitaly fixes an issue in e1000e, where when disconnecting and
reconnecting the physical cable connection, the NIC enters a DMoff
state.  This state causes a mismatch in link and duplexing, so check the
PCIm function state and perform a PHY reset when in this state to
resolve the issue.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: remove not needed call to dma_sync_single_for_device
Heiner Kallweit [Thu, 27 Jun 2019 21:19:09 +0000 (23:19 +0200)]
r8169: remove not needed call to dma_sync_single_for_device

DMA_API_HOWTO.txt includes an example explaining when
dma_sync_single_for_device() is not needed, and that example matches
our use case. The buffer isn't changed by the CPU and direction is
DMA_FROM_DEVICE, so we can remove the call to
dma_sync_single_for_device().

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: consider that 32 Bit DMA is the default
Heiner Kallweit [Thu, 27 Jun 2019 21:12:39 +0000 (23:12 +0200)]
r8169: consider that 32 Bit DMA is the default

Documentation/DMA-API-HOWTO.txt states:
By default, the kernel assumes that your device can address 32-bits of
DMA addressing. For a 64-bit capable device, this needs to be increased,
and for a device with limitations, it needs to be decreased.

Therefore we don't need the 32 Bit DMA fallback configuration and can
remove it.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: improve handling VLAN tag
Heiner Kallweit [Thu, 27 Jun 2019 21:06:33 +0000 (23:06 +0200)]
r8169: improve handling VLAN tag

The VLAN tag is stored in the descriptor in network byte order.
Using swab16 works on little endian host systems only. Better play safe
and use ntohs or htons respectively.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: rtnetlink: skip ipsec offload tests if netdevsim isn't present
Florian Westphal [Thu, 27 Jun 2019 15:12:42 +0000 (17:12 +0200)]
selftests: rtnetlink: skip ipsec offload tests if netdevsim isn't present

running the script on systems without netdevsim now prints:

SKIP: ipsec_offload can't load netdevsim

instead of error message & failed status.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'em_ipt-add-support-for-addrtype'
David S. Miller [Sat, 29 Jun 2019 18:15:12 +0000 (11:15 -0700)]
Merge branch 'em_ipt-add-support-for-addrtype'

Nikolay Aleksandrov says:

====================
em_ipt: add support for addrtype

We would like to be able to use the addrtype from tc for ACL rules and
em_ipt seems the best place to add support for the already existing xt
match. The biggest issue is that addrtype revision 1 (with ipv6 support)
is NFPROTO_UNSPEC and currently em_ipt can't differentiate between v4/v6
if such xt match is used because it passes the match's family instead of
the packet one. The first 3 patches make em_ipt match only on IP
traffic (currently both policy and addrtype recognize such traffic
only) and make it pass the actual packet's protocol instead of the xt
match family when it's unspecified. They also add support for NFPROTO_UNSPEC
xt matches. The last patch allows to add addrtype rules via em_ipt.
We need to keep the user-specified nfproto for dumping in order to be
compatible with libxtables, we cannot dump NFPROTO_UNSPEC as the nfproto
or we'll get an error from libxtables, thus the nfproto is limited to
ipv4/ipv6 in patch 03 and is recorded.

v3: don't use the user nfproto for matching, only for dumping, more
    information is available in the commit message in patch 03
v2: change patch 02 to set the nfproto only when unspecified and drop
    patch 04 from v1 (Eyal Birger)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: em_ipt: add support for addrtype matching
Nikolay Aleksandrov [Thu, 27 Jun 2019 08:10:47 +0000 (11:10 +0300)]
net: sched: em_ipt: add support for addrtype matching

Allow em_ipt to use addrtype for matching. Restrict the use only to
revision 1 which has IPv6 support. Since it's a NFPROTO_UNSPEC xt match
we use the user-specified nfproto for matching, in case it's unspecified
both v4/v6 will be matched by the rule.

v2: no changes, was patch 5 in v1

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: em_ipt: keep the user-specified nfproto and dump it
Nikolay Aleksandrov [Thu, 27 Jun 2019 08:10:46 +0000 (11:10 +0300)]
net: sched: em_ipt: keep the user-specified nfproto and dump it

If we dump NFPROTO_UNSPEC as nfproto user-space libxtables can't handle
it and would exit with an error like:
"libxtables: unhandled NFPROTO in xtables_set_nfproto"
In order to avoid the error return the user-specified nfproto. If we
don't record it then the match family is used which can be
NFPROTO_UNSPEC. Even if we add support to mask NFPROTO_UNSPEC in
iproute2 we have to be compatible with older versions which would be
also be allowed to add NFPROTO_UNSPEC matches (e.g. addrtype after the
last patch).

v3: don't use the user nfproto for matching, only for dumping the rule,
    also don't allow the nfproto to be unspecified (explained above)
v2: adjust changes to missing patch, was patch 04 in v1

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: em_ipt: set the family based on the packet if it's unspecified
Nikolay Aleksandrov [Thu, 27 Jun 2019 08:10:45 +0000 (11:10 +0300)]
net: sched: em_ipt: set the family based on the packet if it's unspecified

Set the family based on the packet if it's unspecified otherwise
protocol-neutral matches will have wrong information (e.g. NFPROTO_UNSPEC).
In preparation for using NFPROTO_UNSPEC xt matches.

v2: set the nfproto only when unspecified

Suggested-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: em_ipt: match only on ip/ipv6 traffic
Nikolay Aleksandrov [Thu, 27 Jun 2019 08:10:44 +0000 (11:10 +0300)]
net: sched: em_ipt: match only on ip/ipv6 traffic

Restrict matching only to ip/ipv6 traffic and make sure we can use the
headers, otherwise matches will be attempted on any protocol which can
be unexpected by the xt matches. Currently policy supports only ipv4/6.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agohinic: add vlan offload support
Xue Chaojing [Sat, 29 Jun 2019 02:26:27 +0000 (02:26 +0000)]
hinic: add vlan offload support

This patch adds vlan offload support for the HINIC driver.

Signed-off-by: Xue Chaojing <xuechaojing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/mlx5e: Disallow tc redirect offload cases we don't support
Paul Blakey [Mon, 24 Jun 2019 12:04:58 +0000 (15:04 +0300)]
net/mlx5e: Disallow tc redirect offload cases we don't support

After changing the parent_id to be the same for both NICs of same
the hardware device, netdev_port_same_parent_id now returns true for
more cases (all the lower devices in the hierarchy are on the same
hardware device).

If merged eswitch isn't enabled, these cases aren't supported, so disallow
them.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Expose same physical switch_id for all representors
Paul Blakey [Thu, 16 May 2019 12:27:17 +0000 (15:27 +0300)]
net/mlx5e: Expose same physical switch_id for all representors

Report system_image_guid as the E-Switch switch_id, this ensures
that when a NIC contains multiple PCI functions and which
has merged eswitch capability, all representors from
multiple PFs publish same switch_id.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Don't refresh TIRs when updating representor SQs
Gavi Teitz [Thu, 23 May 2019 06:58:56 +0000 (09:58 +0300)]
net/mlx5e: Don't refresh TIRs when updating representor SQs

Refreshing TIRs is done in order to update the TIRs with the current
state of SQs in the transport domain, so that the TIRs can filter out
undesired self-loopback packets based on the source SQ of the packet.

Representor TIRs will only receive packets that originate from their
associated vport, due to dedicated steering, and therefore will never
receive self-loopback packets, whose source vport will be the vport of
the E-Switch manager, and therefore not the vport associated with the
representor. As such, it is not necessary to refresh the representors'
TIRs, since self-loopback packets can't reach them.

Since representors only exist in switchdev mode, and there is no
scenario in which a representor will exist in the transport domain
alongside a non-representor, it is not necessary to refresh the
transport domain's TIRs upon changing the state of a representor's
queues. Therefore, do not refresh TIRs upon such a change. Achieve
this by adding an update_rx callback to the mlx5e_profile, which
refreshes TIRs for non-representors and does nothing for representors,
and replace instances of mlx5e_refresh_tirs() upon changing the state
of the queues with update_rx().

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: reduce stack usage in mlx5_eswitch_termtbl_create
Arnd Bergmann [Tue, 18 Jun 2019 11:15:06 +0000 (13:15 +0200)]
net/mlx5e: reduce stack usage in mlx5_eswitch_termtbl_create

Putting an empty 'mlx5_flow_spec' structure on the stack is a bit
wasteful and causes a warning on 32-bit architectures when building
with clang -fsanitize-coverage:

drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_termtbl.c: In function 'mlx5_eswitch_termtbl_create':
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_termtbl.c:90:1: error: the frame size of 1032 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

Since the structure is never written to, we can statically allocate
it to avoid the stack usage. To be on the safe side, mark all
subsequent function arguments that we pass it into as 'const'
as well.

Fixes: 10caabdaad5a ("net/mlx5e: Use termination table for VLAN push actions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Set drvinfo in generic manner
Parav Pandit [Mon, 27 May 2019 09:47:10 +0000 (04:47 -0500)]
net/mlx5e: Set drvinfo in generic manner

Consider PCI and non PCI device types while setting device name
in get_drvinfo() callback using existing generic device.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Vu Pham <vuhuong@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Correct phys_port_name for PF port
Parav Pandit [Wed, 5 Jun 2019 06:29:05 +0000 (01:29 -0500)]
net/mlx5e: Correct phys_port_name for PF port

Currently PF phys_port_name is named as pfNvf-1 as vport number for PF
vport is 65535.
Correct PF's phys_port name as agreed upon name as pfN.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Vu Pham <vuhuong@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Report netdevice MPLS features
Ariel Levkovich [Wed, 5 Jun 2019 17:01:08 +0000 (20:01 +0300)]
net/mlx5e: Report netdevice MPLS features

Set supported device features in the netdevice MPLS features mask.
This will enable HW checksumming and TSO for MPLS tagged traffic.

Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5e: Move to HW checksumming advertising
Ariel Levkovich [Wed, 5 Jun 2019 16:40:09 +0000 (19:40 +0300)]
net/mlx5e: Move to HW checksumming advertising

This patch changes the way the driver advertises its checksum offload
capabilities within the net device features bit mask.

Instead of advertising protocol specific checksumming capabilities
which are limited today to IPv4 and IPv6, we move to reporing
generic HW checksumming capabilities.

This will allow the network stack to let mlx5 device offload checksum
for cases where the IP header is encapsulated within another protocol
and the skb->protocol doesn't indicate one of the IP versions protocol,
specifically in the case of MPLS label encapsulating the IP header and
the skb->protocol indiciates MPLS ethertype rather than IP.

Moving the HW_CSUM reporting is required in the basic net device hw
features mask and also in the extensions (vlan and encpasulation
features) since the extensions are always multiplied by the basic
features set during the packet's traversal through the stack's tx flow.

Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: MPFS, Allow adding the same MAC more than once
Gavi Teitz [Tue, 11 Jun 2019 09:29:41 +0000 (12:29 +0300)]
net/mlx5: MPFS, Allow adding the same MAC more than once

Remove the limitation preventing adding a vport's MAC address to the
Multi-Physical Function Switch (MPFS) more than once per E-switch, as
there is no difference in the MPFS if an address is being used by an
E-switch more than once.

This allows the E-switch to have multiple vports with the same MAC
address, allowing vports to be classified by VLAN id instead of by MAC
if desired.

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agonet/mlx5: MPFS, Cleanup add MAC flow
Gavi Teitz [Tue, 11 Jun 2019 08:54:36 +0000 (11:54 +0300)]
net/mlx5: MPFS, Cleanup add MAC flow

Unify and isolate the error handling flow in mlx5_mpfs_add_mac(),
removing code duplication.

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agoMerge branch 'mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox...
Saeed Mahameed [Fri, 28 Jun 2019 22:49:59 +0000 (15:49 -0700)]
Merge branch 'mlx5-next' of git://git./linux/kernel/git/mellanox/linux

Misc updates from mlx5-next branch:

1) E-Switch vport metadata support for source vport matching
2) Convert mkey_table to XArray
3) Shared IRQs and to use single IRQ for all async EQs

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
5 years agoe1000e: PCIm function state support
Vitaly Lifshits [Tue, 25 Jun 2019 14:39:11 +0000 (17:39 +0300)]
e1000e: PCIm function state support

Due to commit: 5d8682588605 ("[misc] mei: me: allow runtime
pm for platform with D0i3")
When disconnecting the cable and reconnecting it the NIC
enters DMoff state. This caused wrong link indication
and duplex mismatch. This bug is described in:
https://bugzilla.redhat.com/show_bug.cgi?id=1689436

Checking PCIm function state and performing PHY reset after a
timeout in watchdog task solves this issue.

Signed-off-by: Vitaly Lifshits <vitaly.lifshits@intel.com>
Acked-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoe1000e: Make watchdog use delayed work
Detlev Casanova [Sun, 23 Jun 2019 03:14:37 +0000 (23:14 -0400)]
e1000e: Make watchdog use delayed work

Use delayed work instead of timers to run the watchdog of the e1000e
driver.

Simplify the code with one less middle function.

Signed-off-by: Detlev Casanova <detlev.casanova@gmail.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Add macvlan support on i40e
Harshitha Ramamurthy [Wed, 19 Jun 2019 18:45:40 +0000 (14:45 -0400)]
i40e: Add macvlan support on i40e

This patch enables macvlan offloads for i40e. The idea is to use
channels as macvlan interfaces. The channels are VSIs of
type VMDQ. When the first macvlan is created, the maximum number of
channels possible are created. From then on, as a macvlan interface
is created, a macvlan filter is added to these already created
channels (VSIs).

This patch utilizes subordinate device traffic classes to make queue
groups(channels) available for an upper device like a macvlan.

Steps to configure macvlan offloads:
1. ethtool -K ethx l2-fwd-offload on
2. ip link add link ethx name macvlan1 type macvlan
3. ip addr add <address> dev macvlan1
4. ip link set macvlan1 up

Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoixgbevf: Use cached link state instead of re-reading the value for ethtool
Alexander Duyck [Wed, 19 Jun 2019 16:58:53 +0000 (09:58 -0700)]
ixgbevf: Use cached link state instead of re-reading the value for ethtool

Change the ethtool link settings call to just read the cached state out of
the adapter structure instead of trying to recheck the value from the PF.
Doing this should prevent excessive reading of the mailbox.

Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Reviewed-by: "Guilherme G. Piccoli" <gpiccoli@canonical.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoiavf: fix dereference of null rx_buffer pointer
Colin Ian King [Wed, 19 Jun 2019 14:30:44 +0000 (15:30 +0100)]
iavf: fix dereference of null rx_buffer pointer

A recent commit efa14c3985828d ("iavf: allow null RX descriptors") added
a null pointer sanity check on rx_buffer, however, rx_buffer is being
dereferenced before that check, which implies a null pointer dereference
bug can potentially occur.  Fix this by only dereferencing rx_buffer
until after the null pointer check.

Addresses-Coverity: ("Dereference before null check")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoigb: add RR2DCDELAY to ethtool registers dump
Artem Bityutskiy [Tue, 18 Jun 2019 11:55:13 +0000 (14:55 +0300)]
igb: add RR2DCDELAY to ethtool registers dump

This patch adds the RR2DCDELAY register to the ethtool registers dump.
RR2DCDELAY exists on I210 and I211 Intel Gigabit Ethernet chips and it stands
for "Read Request To Data Completion Delay". Here is how this register is
described in the I210 datasheet:

"This field captures the maximum PCIe split time in 16 ns units, which is the
maximum delay between the read request to the first data completion. This is
giving an estimation of the PCIe round trip time."

In other words, whenever I210 reads from the host memory (e.g., fetches a
descriptor from the ring), the chip measures every PCI DMA read transaction and
captures the maximum value. So it ends up containing the longest DMA
transaction time.

This register is very useful for troubleshooting and research purposes. If you
are dealing with time-sensitive networks, this register can help you get
an idea of your "I210-to-ring" latency. This helps answering questions like
"should I have PCIe ASPM enabled?" or "should I enable deep C-states?" on
my system.

It is safe to read this register at any point, reading it has no effect on
the I210 chip functionality.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoigb: minor ethool regdump amendment
Artem Bityutskiy [Tue, 18 Jun 2019 11:55:12 +0000 (14:55 +0300)]
igb: minor ethool regdump amendment

This patch has no functional impact and it is just a preparation
for the following patch. It removes an early return from the
'igb_get_regs()' function by moving the 82576-only registers
dump into an "if" block. With this preparation, we can dump more
non-82576 registers at the end of this function.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoiavf: Fix up debug print macro
Jeff Kirsher [Mon, 17 Jun 2019 23:10:58 +0000 (16:10 -0700)]
iavf: Fix up debug print macro

This aligns the iavf_debug() macro with the other Intel drivers.

Add the bus number, bus_id field to i40e_bus_info so output shows
each physical port(i.e func) in following format:
  [[[[<domain>]:]<bus>]:][<slot>][.[<func>]]
domains are numbered from 0 to ffff), bus (0-ff), slot (0-1f) and
function (0-7).

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
5 years agoe1000e: Reduce boot time by tightening sleep ranges
Arjan van de Ven [Sat, 15 Jun 2019 00:29:35 +0000 (17:29 -0700)]
e1000e: Reduce boot time by tightening sleep ranges

The e1000e driver is a great user of the usleep_range() API,
and has nice ranges that in principle help power management.

However the ranges that are used only during system startup are
very long (and can add easily 100 msec to the boot time) while
the power savings of such long ranges is irrelevant due to the
one-off, boot only, nature of these functions.

This patch shrinks some of the longest ranges to be shorter
(while still using a power friendly 1 msec range); this saves
100msec+ of boot time on my BDW NUCs

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoiavf: use struct_size() helper
Gustavo A. R. Silva [Fri, 14 Jun 2019 23:23:20 +0000 (16:23 -0700)]
iavf: use struct_size() helper

Make use of the struct_size() helper instead of an open-coded version
in order to avoid any potential type mistakes, in particular in the
context in which this code is being used.

So, replace code of the following form:

sizeof(struct virtchnl_ether_addr_list) + (count * sizeof(struct virtchnl_ether_addr))

with:

struct_size(veal, list, count)

and so on...

This code was detected with the help of Coccinelle.

Signed-off-by: "Gustavo A. R. Silva" <gustavo@embeddedor.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoe1000: Use dma_wmb() instead of wmb() before doorbell writes
Venkatesh Srinivas [Mon, 10 Jun 2019 21:27:50 +0000 (14:27 -0700)]
e1000: Use dma_wmb() instead of wmb() before doorbell writes

e1000 writes to doorbells to post transmit descriptors and fill the
receive ring. After writing descriptors to memory but before
writing to doorbells, use dma_wmb() rather than wmb(). wmb() is more
heavyweight than necessary for a device to see descriptor writes.

On x86, this avoids SFENCEs before doorbell writes in both the
Tx and Rx paths. On ARM, this converts DSB ST -> DMB OSHST.

Tested: 82576EB / x86; QEMU (qemu emulates an 8257x)

Signed-off-by: Venkatesh Srinivas <venkateshs@google.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoixgbe: fix potential u32 overflow on shift
Colin Ian King [Fri, 7 Jun 2019 18:19:20 +0000 (19:19 +0100)]
ixgbe: fix potential u32 overflow on shift

The u32 variable rem is being shifted using u32 arithmetic however
it is being passed to div_u64 that expects the expression to be a u64.
The 32 bit shift may potentially overflow, so cast rem to a u64 before
shifting to avoid this.  Also remove comment about overflow.

Addresses-Coverity: ("Unintentional integer overflow")
Fixes: cd4583206990 ("ixgbe: implement support for SDP/PPS output on X550 hardware")
Fixes: 68d9676fc04e ("ixgbe: fix PTP SDP pin setup on X540 hardware")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoixgbe: Avoid NULL pointer dereference with VF on non-IPsec hw
Dann Frazier [Wed, 22 May 2019 23:22:58 +0000 (17:22 -0600)]
ixgbe: Avoid NULL pointer dereference with VF on non-IPsec hw

An ipsec structure will not be allocated if the hardware does not support
offload. Fixes the following Oops:

[  191.045452] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[  191.054232] Mem abort info:
[  191.057014]   ESR = 0x96000004
[  191.060057]   Exception class = DABT (current EL), IL = 32 bits
[  191.065963]   SET = 0, FnV = 0
[  191.069004]   EA = 0, S1PTW = 0
[  191.072132] Data abort info:
[  191.074999]   ISV = 0, ISS = 0x00000004
[  191.078822]   CM = 0, WnR = 0
[  191.081780] user pgtable: 4k pages, 48-bit VAs, pgdp = 0000000043d9e467
[  191.088382] [0000000000000000] pgd=0000000000000000
[  191.093252] Internal error: Oops: 96000004 [#1] SMP
[  191.098119] Modules linked in: vhost_net vhost tap vfio_pci vfio_virqfd vfio_iommu_type1 vfio xt_CHECKSUM iptable_mangle ipt_MASQUERADE iptable_nat nf_nat_ipv4 nf_nat xt_conntrack nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 ipt_REJECT nf_reject_ipv4 xt_tcpudp bridge stp llc ebtable_filter devlink ebtables ip6table_filter ip6_tables iptable_filter bpfilter ipmi_ssif nls_iso8859_1 input_leds joydev ipmi_si hns_roce_hw_v2 ipmi_devintf hns_roce ipmi_msghandler cppc_cpufreq sch_fq_codel ib_iser rdma_cm iw_cm ib_cm ib_core iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ip_tables x_tables autofs4 ses enclosure btrfs zstd_compress raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor hid_generic usbhid hid raid6_pq libcrc32c raid1 raid0 multipath linear ixgbevf hibmc_drm ttm
[  191.168607]  drm_kms_helper aes_ce_blk aes_ce_cipher syscopyarea crct10dif_ce sysfillrect ghash_ce qla2xxx sysimgblt sha2_ce sha256_arm64 hisi_sas_v3_hw fb_sys_fops sha1_ce uas nvme_fc mpt3sas ixgbe drm hisi_sas_main nvme_fabrics usb_storage hclge scsi_transport_fc ahci libsas hnae3 raid_class libahci xfrm_algo scsi_transport_sas mdio aes_neon_bs aes_neon_blk crypto_simd cryptd aes_arm64
[  191.202952] CPU: 94 PID: 0 Comm: swapper/94 Not tainted 4.19.0-rc1+ #11
[  191.209553] Hardware name: Huawei D06 /D06, BIOS Hisilicon D06 UEFI RC0 - V1.20.01 04/26/2019
[  191.218064] pstate: 20400089 (nzCv daIf +PAN -UAO)
[  191.222873] pc : ixgbe_ipsec_vf_clear+0x60/0xd0 [ixgbe]
[  191.228093] lr : ixgbe_msg_task+0x2d0/0x1088 [ixgbe]
[  191.233044] sp : ffff000009b3bcd0
[  191.236346] x29: ffff000009b3bcd0 x28: 0000000000000000
[  191.241647] x27: ffff000009628000 x26: 0000000000000000
[  191.246946] x25: ffff803f652d7600 x24: 0000000000000004
[  191.252246] x23: ffff803f6a718900 x22: 0000000000000000
[  191.257546] x21: 0000000000000000 x20: 0000000000000000
[  191.262845] x19: 0000000000000000 x18: 0000000000000000
[  191.268144] x17: 0000000000000000 x16: 0000000000000000
[  191.273443] x15: 0000000000000000 x14: 0000000100000026
[  191.278742] x13: 0000000100000025 x12: ffff8a5f7fbe0df0
[  191.284042] x11: 000000010000000b x10: 0000000000000040
[  191.289341] x9 : 0000000000001100 x8 : ffff803f6a824fd8
[  191.294640] x7 : ffff803f6a825098 x6 : 0000000000000001
[  191.299939] x5 : ffff000000f0ffc0 x4 : 0000000000000000
[  191.305238] x3 : ffff000028c00000 x2 : ffff803f652d7600
[  191.310538] x1 : 0000000000000000 x0 : ffff000000f205f0
[  191.315838] Process swapper/94 (pid: 0, stack limit = 0x00000000addfed5a)
[  191.322613] Call trace:
[  191.325055]  ixgbe_ipsec_vf_clear+0x60/0xd0 [ixgbe]
[  191.329927]  ixgbe_msg_task+0x2d0/0x1088 [ixgbe]
[  191.334536]  ixgbe_msix_other+0x274/0x330 [ixgbe]
[  191.339233]  __handle_irq_event_percpu+0x78/0x270
[  191.343924]  handle_irq_event_percpu+0x40/0x98
[  191.348355]  handle_irq_event+0x50/0xa8
[  191.352180]  handle_fasteoi_irq+0xbc/0x148
[  191.356263]  generic_handle_irq+0x34/0x50
[  191.360259]  __handle_domain_irq+0x68/0xc0
[  191.364343]  gic_handle_irq+0x84/0x180
[  191.368079]  el1_irq+0xe8/0x180
[  191.371208]  arch_cpu_idle+0x30/0x1a8
[  191.374860]  do_idle+0x1dc/0x2a0
[  191.378077]  cpu_startup_entry+0x2c/0x30
[  191.381988]  secondary_start_kernel+0x150/0x1e0
[  191.386506] Code: 6b15003f 54000320 f1404a9f 54000060 (79400260)

Fixes: eda0333ac2930 ("ixgbe: add VF IPsec management")
Signed-off-by: Dann Frazier <dann.frazier@canonical.com>
Acked-by: Shannon Nelson <snelson@pensando.io>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoe1000e: Increase pause and refresh time
Miguel Bernal Marin [Mon, 27 Mar 2017 22:01:56 +0000 (16:01 -0600)]
e1000e: Increase pause and refresh time

Suggested-by: Tim Pepper <timothy.c.pepper@linux.intel.com>
Signed-off-by: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com>
Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Acked-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Use struct_size() helper
Gustavo A. R. Silva [Fri, 29 Mar 2019 23:38:47 +0000 (16:38 -0700)]
ice: Use struct_size() helper

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = alloc(size, GFP_KERNEL);

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

size = struct_size(instance, entry, count);

This code was detected with the help of Coccinelle.

Signed-off-by: "Gustavo A. R. Silva" <gustavo@embeddedor.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoMerge branch 'net-sched-Add-txtime-assist-support-for-taprio'
David S. Miller [Fri, 28 Jun 2019 21:45:34 +0000 (14:45 -0700)]
Merge branch 'net-sched-Add-txtime-assist-support-for-taprio'

Vedang Patel says:

====================
net/sched: Add txtime-assist support for taprio.

Changes in v6:
- Use _BITUL() instead of BIT() in UAPI for etf. (patch #1)
- Fix a bug reported by kbuild test bot in length_to_duration(). (patch #6)
- Remove an unused function (get_cycle_start()). (Patch #6)

Changes in v5:
- Commit message improved for the igb patch (patch #1).
- Fixed typo in commit message for etf patch (patch #2).

Changes in v4:
- Remove inline directive from functions in foo.c.
- Fix spacing in pkt_sched.h (for etf patch).

Changes in v3:
- Simplify implementation for taprio flags.
- txtime_delay can only be set if txtime-assist mode is enabled.
- txtime_delay and flags will only be visible in tc output if set by user.
- Minor changes in error reporting.

Changes in v2:
- Txtime-offload has now been renamed to txtime-assist mode.
- Renamed the offload parameter to flags.
- Removed the code which introduced the hardware offloading functionality.

Original Cover letter (with above changes included)
--------------------------------------------------

Currently, we are seeing packets being transmitted outside their
timeslices. We can confirm that the packets are being dequeued at the right
time. So, the delay is induced after the packet is dequeued, because
taprio, without any offloading, has no control of when a packet is actually
transmitted.

In order to solve this, we are making use of the txtime feature provided by
ETF qdisc. Hardware offloading needs to be supported by the ETF qdisc in
order to take advantage of this feature. The taprio qdisc will assign
txtime (in skb->tstamp) for all the packets which do not have the txtime
allocated via the SO_TXTIME socket option. For the packets which already
have SO_TXTIME set, taprio will validate whether the packet will be
transmitted in the correct interval.

In order to support this, the following parameters have been added:
- flags (taprio): This is added in order to support different offloading
  modes which will be added in the future.
- txtime-delay (taprio): This indicates the minimum time it will take for
  the packet to hit the wire after it reaches taprio_enqueue(). This is
  useful in determining whether we can transmit the packet in the remaining
  time if the gate corresponding to the packet is currently open.
- skip_skb_check (ETF): ETF currently drops any packet which does not have
  the SO_TXTIME socket option set. This check can be skipped by specifying
  this option.

Following is an example configuration:

tc qdisc replace dev $IFACE parent root handle 100 taprio \\
    num_tc 3 \\
    map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 \\
    queues 1@0 1@0 1@0 \\
    base-time $BASE_TIME \\
    sched-entry S 01 300000 \\
    sched-entry S 02 300000 \\
    sched-entry S 04 400000 \\
    flags 0x1 \\
    txtime-delay 200000 \\
    clockid CLOCK_TAI

tc qdisc replace dev $IFACE parent 100:1 etf \\
    offload delta 200000 clockid CLOCK_TAI skip_skb_check

Here, the "flags" parameter is indicating that the txtime-assist mode is
enabled. Also, all the traffic classes have been assigned the same queue.
This is to prevent the traffic classes in the lower priority queues from
getting starved. Note that this configuration is specific to the i210
ethernet card. Other network cards where the hardware queues are given the
same priority, might be able to utilize more than one queue.

Following are some of the other highlights of the series:
- Fix a bug where hardware timestamping and SO_TXTIME options cannot be
  used together. (Patch 1)
- Introduces the skip_skb_check option.  (Patch 2)
- Make TxTime assist mode work with TCP packets (Patch 7).

The following changes are recommended to be done in order to get the best
performance from taprio in this mode:
ip link set dev enp1s0 mtu 1514
ethtool -K eth0 gso off
ethtool -K eth0 tso off
ethtool --set-eee eth0 eee off
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotaprio: Adjust timestamps for TCP packets
Vedang Patel [Tue, 25 Jun 2019 22:07:19 +0000 (15:07 -0700)]
taprio: Adjust timestamps for TCP packets

When the taprio qdisc is running in "txtime offload" mode, it will
set the launchtime value (in skb->tstamp) for all the packets which do
not have the SO_TXTIME socket option. But, the TCP packets already have
this value set and it indicates the earliest departure time represented
in CLOCK_MONOTONIC clock.

We need to respect the timestamp set by the TCP subsystem. So, convert
this time to the clock which taprio is using and ensure that the packet
is not transmitted before the deadline set by TCP.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotaprio: make clock reference conversions easier
Vedang Patel [Tue, 25 Jun 2019 22:07:18 +0000 (15:07 -0700)]
taprio: make clock reference conversions easier

Later in this series we will need to transform from
CLOCK_MONOTONIC (used in TCP) to the clock reference used in TAPRIO.

Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotaprio: Add support for txtime-assist mode
Vedang Patel [Tue, 25 Jun 2019 22:07:17 +0000 (15:07 -0700)]
taprio: Add support for txtime-assist mode

Currently, we are seeing non-critical packets being transmitted outside of
their timeslice. We can confirm that the packets are being dequeued at the
right time. So, the delay is induced in the hardware side.  The most likely
reason is the hardware queues are starving the lower priority queues.

In order to improve the performance of taprio, we will be making use of the
txtime feature provided by the ETF qdisc. For all the packets which do not
have the SO_TXTIME option set, taprio will set the transmit timestamp (set
in skb->tstamp) in this mode. TAPrio Qdisc will ensure that the transmit
time for the packet is set to when the gate is open. If SO_TXTIME is set,
the TAPrio qdisc will validate whether the timestamp (in skb->tstamp)
occurs when the gate corresponding to skb's traffic class is open.

Following two parameters added to support this mode:
- flags: used to enable txtime-assist mode. Will also be used to enable
  other modes (like hardware offloading) later.
- txtime-delay: This indicates the minimum time it will take for the packet
  to hit the wire. This is useful in determining whether we can transmit
the packet in the remaining time if the gate corresponding to the packet is
currently open.

An example configuration for enabling txtime-assist:

tc qdisc replace dev eth0 parent root handle 100 taprio \\
      num_tc 3 \\
      map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 \\
      queues 1@0 1@0 1@0 \\
      base-time 1558653424279842568 \\
      sched-entry S 01 300000 \\
      sched-entry S 02 300000 \\
      sched-entry S 04 400000 \\
      flags 0x1 \\
      txtime-delay 40000 \\
      clockid CLOCK_TAI

tc qdisc replace dev $IFACE parent 100:1 etf skip_sock_check \\
      offload delta 200000 clockid CLOCK_TAI

Note that all the traffic classes are mapped to the same queue.  This is
only possible in taprio when txtime-assist is enabled. Also, note that the
ETF Qdisc is enabled with offload mode set.

In this mode, if the packet's traffic class is open and the complete packet
can be transmitted, taprio will try to transmit the packet immediately.
This will be done by setting skb->tstamp to current_time + the time delta
indicated in the txtime-delay parameter. This parameter indicates the time
taken (in software) for packet to reach the network adapter.

If the packet cannot be transmitted in the current interval or if the
packet's traffic is not currently transmitting, the skb->tstamp is set to
the next available timestamp value. This is tracked in the next_launchtime
parameter in the struct sched_entry.

The behaviour w.r.t admin and oper schedules is not changed from what is
present in software mode.

The transmit time is already known in advance. So, we do not need the HR
timers to advance the schedule and wakeup the dequeue side of taprio.  So,
HR timer won't be run when this mode is enabled.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotaprio: Remove inline directive
Vedang Patel [Tue, 25 Jun 2019 22:07:16 +0000 (15:07 -0700)]
taprio: Remove inline directive

Remove inline directive from length_to_duration(). We will let the compiler
make the decisions.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotaprio: calculate cycle_time when schedule is installed
Vedang Patel [Tue, 25 Jun 2019 22:07:15 +0000 (15:07 -0700)]
taprio: calculate cycle_time when schedule is installed

cycle time for a particular schedule is calculated only when it is first
installed. So, it makes sense to just calculate it once right after the
'cycle_time' parameter has been parsed and store it in cycle_time.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoetf: Add skip_sock_check
Vedang Patel [Tue, 25 Jun 2019 22:07:14 +0000 (15:07 -0700)]
etf: Add skip_sock_check

Currently, etf expects a socket with SO_TXTIME option set for each packet
it encounters. So, it will drop all other packets. But, in the future
commits we are planning to add functionality where tstamp value will be set
by another qdisc. Also, some packets which are generated from within the
kernel (e.g. ICMP packets) do not have any socket associated with them.

So, this commit adds support for skip_sock_check. When this option is set,
etf will skip checking for a socket and other associated options for all
skbs.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoetf: Don't use BIT() in UAPI headers.
Vedang Patel [Tue, 25 Jun 2019 22:07:13 +0000 (15:07 -0700)]
etf: Don't use BIT() in UAPI headers.

The BIT() macro isn't exported as part of the UAPI interface. So, the
compile-test to ensure they are self contained fails. So, use _BITUL()
instead.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoigb: clear out skb->tstamp after reading the txtime
Vedang Patel [Tue, 25 Jun 2019 22:07:12 +0000 (15:07 -0700)]
igb: clear out skb->tstamp after reading the txtime

If a packet which is utilizing the launchtime feature (via SO_TXTIME socket
option) also requests the hardware transmit timestamp, the hardware
timestamp is not delivered to the userspace. This is because the value in
skb->tstamp is mistaken as the software timestamp.

Applications, like ptp4l, request a hardware timestamp by setting the
SOF_TIMESTAMPING_TX_HARDWARE socket option. Whenever a new timestamp is
detected by the driver (this work is done in igb_ptp_tx_work() which calls
igb_ptp_tx_hwtstamps() in igb_ptp.c[1]), it will queue the timestamp in the
ERR_QUEUE for the userspace to read. When the userspace is ready, it will
issue a recvmsg() call to collect this timestamp.  The problem is in this
recvmsg() call. If the skb->tstamp is not cleared out, it will be
interpreted as a software timestamp and the hardware tx timestamp will not
be successfully sent to the userspace. Look at skb_is_swtx_tstamp() and the
callee function __sock_recv_timestamp() in net/socket.c for more details.

Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mirred-recurse'
David S. Miller [Fri, 28 Jun 2019 21:36:25 +0000 (14:36 -0700)]
Merge branch 'mirred-recurse'

John Hurley says:

====================
Track recursive calls in TC act_mirred

These patches aim to prevent act_mirred causing stack overflow events from
recursively calling packet xmit or receive functions. Such events can
occur with poor TC configuration that causes packets to travel in loops
within the system.

Florian Westphal advises that a recursion crash and packets looping are
separate issues and should be treated as such. David Miller futher points
out that pcpu counters cannot track the precise skb context required to
detect loops. Hence these patches are not aimed at detecting packet loops,
rather, preventing stack flows arising from such loops.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: protect against stack overflow in TC act_mirred
John Hurley [Mon, 24 Jun 2019 22:13:36 +0000 (23:13 +0100)]
net: sched: protect against stack overflow in TC act_mirred

TC hooks allow the application of filters and actions to packets at both
ingress and egress of the network stack. It is possible, with poor
configuration, that this can produce loops whereby an ingress hook calls
a mirred egress action that has an egress hook that redirects back to
the first ingress etc. The TC core classifier protects against loops when
doing reclassifies but there is no protection against a packet looping
between multiple hooks and recursively calling act_mirred. This can lead
to stack overflow panics.

Add a per CPU counter to act_mirred that is incremented for each recursive
call of the action function when processing a packet. If a limit is passed
then the packet is dropped and CPU counter reset.

Note that this patch does not protect against loops in TC datapaths. Its
aim is to prevent stack overflow kernel panics that can be a consequence
of such loops.

Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: refactor reinsert action
John Hurley [Mon, 24 Jun 2019 22:13:35 +0000 (23:13 +0100)]
net: sched: refactor reinsert action

The TC_ACT_REINSERT return type was added as an in-kernel only option to
allow a packet ingress or egress redirect. This is used to avoid
unnecessary skb clones in situations where they are not required. If a TC
hook returns this code then the packet is 'reinserted' and no skb consume
is carried out as no clone took place.

This return type is only used in act_mirred. Rather than have the reinsert
called from the main datapath, call it directly in act_mirred. Instead of
returning TC_ACT_REINSERT, change the type to the new TC_ACT_CONSUMED
which tells the caller that the packet has been stolen by another process
and that no consume call is required.

Moving all redirect calls to the act_mirred code is in preparation for
tracking recursion created by act_mirred.

Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv4: enable route flushing in network namespaces
Christian Brauner [Mon, 24 Jun 2019 13:29:23 +0000 (15:29 +0200)]
ipv4: enable route flushing in network namespaces

Tools such as vpnc try to flush routes when run inside network
namespaces by writing 1 into /proc/sys/net/ipv4/route/flush. This
currently does not work because flush is not enabled in non-initial
network namespaces.
Since routes are per network namespace it is safe to enable
/proc/sys/net/ipv4/route/flush in there.

Link: https://github.com/lxc/lxd/issues/4257
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'batadv-next-for-davem-20190627v2' of git://git.open-mesh.org/linux-merge
David S. Miller [Fri, 28 Jun 2019 16:48:24 +0000 (09:48 -0700)]
Merge tag 'batadv-next-for-davem-20190627v2' of git://git.open-mesh.org/linux-merge

Simon Wunderlich says:

====================
This feature/cleanup patchset includes the following patches:

 - bump version strings, by Simon Wunderlich

 - fix includes for _MAX constants, atomic functions and fwdecls,
   by Sven Eckelmann (3 patches)

 - shorten multicast tt/tvlv worker spinlock section, by Linus Luessing

 - routeable multicast preparations: implement MAC multicast filtering,
   by Linus Luessing (2 patches, David Millers comments integrated)

 - remove return value checks for debugfs_create, by Greg Kroah-Hartman

 - add routable multicast optimizations, by Linus Luessing (2 patches)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'hns3-next'
David S. Miller [Fri, 28 Jun 2019 16:39:50 +0000 (09:39 -0700)]
Merge branch 'hns3-next'

Huazhong Tan says:

====================
net: hns3: some code optimizations & cleanups & bugfixes

[patch 01/12] fixes a TX timeout issue.

[patch 02/12 - 04/12] adds some patch related to TM module.

[patch 05/12] fixes a compile warning.

[patch 06/12] adds Asym Pause support for autoneg

[patch 07/12] optimizes the error handler for VF reset.

[patch 08/12] deals with the empty interrupt case.

[patch 09/12 - 12/12] adds some cleanups & optimizations.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: optimize the CSQ cmd error handling
Peng Li [Fri, 28 Jun 2019 11:50:18 +0000 (19:50 +0800)]
net: hns3: optimize the CSQ cmd error handling

If CMDQ ring is full, hclge_cmd_send may return directly, but IMP still
working and HW pointer changed, SW ring pointer do not match the HW
pointer. This patch update the SW pointer every time when the space is
full, so it can work normally next time if IMP and HW still working.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: remove RXD_VLD check in hns3_handle_bdinfo
Yunsheng Lin [Fri, 28 Jun 2019 11:50:17 +0000 (19:50 +0800)]
net: hns3: remove RXD_VLD check in hns3_handle_bdinfo

The HNS3_RXD_VLD_B bit has already been checked in hns3_add_frag
or hns3_handle_rx_bd before calling hns3_handle_bdinfo, so when
hns3_handle_bdinfo is called, the HNS3_RXD_VLD_B bit is always
set, which makes the checking in hns3_handle_bdinfo unnecessary.

This patch removes the RXD_VLD_B checking in hns3_handle_bdinfo.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: remove unused linkmode definition
Jian Shen [Fri, 28 Jun 2019 11:50:16 +0000 (19:50 +0800)]
net: hns3: remove unused linkmode definition

This patch removes unused linkmode definition.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix a statistics issue about l3l4 checksum error
Yufeng Mo [Fri, 28 Jun 2019 11:50:15 +0000 (19:50 +0800)]
net: hns3: fix a statistics issue about l3l4 checksum error

The frame column is based on rx_crc_errors and rx_frame_errors. So
l3l4 checksum error should not be counted by rx_crc_errors. Instead,
l3l4 checksum error should be counted in ifconfig error column.

Fixes: d3ec4ef66937 ("net: hns3: refactor the statistics updating for netdev")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: handle empty unknown interrupt
Huazhong Tan [Fri, 28 Jun 2019 11:50:14 +0000 (19:50 +0800)]
net: hns3: handle empty unknown interrupt

Since some MSI-X interrupt's status may be cleared by hardware,
so when the driver receives the interrupt, reading
HCLGE_VECTOR0_PF_OTHER_INT_STS_REG register will get an empty
unknown interrupt. For this case, the irq handler should enable
vector0 interrupt. This patch also use dev_info() instead of
dev_dbg() in the hclge_check_event_cause(), since this information
will be useful for normal usage.

Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: re-schedule reset task while VF reset fail
Huazhong Tan [Fri, 28 Jun 2019 11:50:13 +0000 (19:50 +0800)]
net: hns3: re-schedule reset task while VF reset fail

The VF reset may fail for some probabilistic reasons,
such as wait for hardware reset timeout, wait for mailbox
response timeout, so this patch tries to re-schedule the
reset task when the number of reset failing is under
HCLGEVF_RESET_MAX_FAIL_CNT. This patch also add a function
hclgevf_reset_err_handle() to handle the reset failing.

Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: add Asym Pause support to fix autoneg problem
Yonglong Liu [Fri, 28 Jun 2019 11:50:12 +0000 (19:50 +0800)]
net: hns3: add Asym Pause support to fix autoneg problem

Local device and link partner config auto-negotiation on both,
local device config pause frame use as: rx on/tx off,
link partner config pause frame use as: rx off/tx on.

We except the result is:
Local device:
Autonegotiate:  on
RX:             on
TX:             off
RX negotiated:  on
TX negotiated:  off

Link partner:
Autonegotiate:  on
RX:             off
TX:             on
RX negotiated:  off
TX negotiated:  on

But actually, the result of Local device and link partner is both:
Autonegotiate:  on
RX:             off
TX:             off
RX negotiated:  off
TX negotiated:  off

The root cause is that the supported flag is has only Pause,
reference to the function genphy_config_advert():
static int genphy_config_advert(struct phy_device *phydev)
{
...
linkmode_and(phydev->advertising, phydev->advertising,
     phydev->supported);
...
}
The pause frame use of link partner is rx off/tx on, so its
advertising only set the bit Asym_Pause, and the supported is
only set the bit Pause, so the result of linkmode_and(), is
rx off/tx off.

This patch adds Asym_Pause to the supported flag to fix it.

Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix a -Wformat-nonliteral compile warning
Yonglong Liu [Fri, 28 Jun 2019 11:50:11 +0000 (19:50 +0800)]
net: hns3: fix a -Wformat-nonliteral compile warning

When setting -Wformat=2, there is a compiler warning like this:

hclge_main.c:xxx:x: warning: format not a string literal and no
format arguments [-Wformat-nonliteral]
strs[i].desc);
^~~~

This patch adds missing format parameter "%s" to snprintf() to
fix it.

Fixes: 46a3df9f9718 ("Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: add some error checking in hclge_tm module
Yunsheng Lin [Fri, 28 Jun 2019 11:50:10 +0000 (19:50 +0800)]
net: hns3: add some error checking in hclge_tm module

When hdev->tx_sch_mode is HCLGE_FLAG_VNET_BASE_SCH_MODE, the
hclge_tm_schd_mode_vnet_base_cfg calls hclge_tm_pri_schd_mode_cfg
with vport->vport_id as pri_id, which is used as index for
hdev->tm_info.tc_info, it will cause out of bound access issue
if vport_id is equal to or larger than HNAE3_MAX_TC.

Also hardware only support maximum speed of HCLGE_ETHER_MAX_RATE.

So this patch adds two checks for above cases.

Fixes: 848440544b41 ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: change SSU's buffer allocation according to UM
Yunsheng Lin [Fri, 28 Jun 2019 11:50:09 +0000 (19:50 +0800)]
net: hns3: change SSU's buffer allocation according to UM

Currently when there is share buffer in the SSU(storage
switching unit), the low waterline for RX private buffer is
too low to keep the hardware running. Hardware may have
processed all the packet stored in the private buffer of the
low waterline before the new packet comes, because hardware
only tell the peer send packet again when the private buffer
is under the low waterline.

So this patch only allocate RX private buffer if there is
enough buffer according to hardware user manual.

This patch also reserve some buffer for reusing when TC num
is less than or equal to 2, and change PAUSE_TRANS_GAP &
HCLGE_NON_DCB_ADDITIONAL_BUF according to hardware user
manual.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: enable DCB when TC num is one and pfc_en is non-zero
Yunsheng Lin [Fri, 28 Jun 2019 11:50:08 +0000 (19:50 +0800)]
net: hns3: enable DCB when TC num is one and pfc_en is non-zero

Currently when TC num is one, the DCB will be disabled no matter if
pfc_en is non-zero or not.

This patch enables the DCB if pfc_en is non-zero, even when TC num
is one.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix __QUEUE_STATE_STACK_XOFF not cleared issue
Huazhong Tan [Fri, 28 Jun 2019 11:50:07 +0000 (19:50 +0800)]
net: hns3: fix __QUEUE_STATE_STACK_XOFF not cleared issue

When change MTU or other operations, which just calling .reset_notify
to do HNAE3_DOWN_CLIENT and HNAE3_UP_CLIENT, then
the netdev_tx_reset_queue() in the hns3_clear_all_ring() will be
ignored. So the dev_watchdog() may misdiagnose a TX timeout.

This patch separates netdev_tx_reset_queue() from
hns3_clear_all_ring(), and unifies hns3_clear_all_ring() and
hns3_force_clear_all_ring into one, since they are doing
similar things.

Fixes: 3a30964a2eef ("net: hns3: delay ring buffer clearing during reset")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'Better-PHYLINK-compliance-for-SJA1105-DSA'
David S. Miller [Fri, 28 Jun 2019 16:31:31 +0000 (09:31 -0700)]
Merge branch 'Better-PHYLINK-compliance-for-SJA1105-DSA'

Vladimir Oltean says:

====================
Better PHYLINK compliance for SJA1105 DSA

After discussing with Russell King, it appears this driver is making a
few confusions and not performing some checks for consistent operation.

Changes in v2:
- Removed redundant print in the phylink_validate callback (in 2/3).
====================

Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: sja1105: Mark in-band AN modes not supported for PHYLINK
Vladimir Oltean [Thu, 27 Jun 2019 21:46:37 +0000 (00:46 +0300)]
net: dsa: sja1105: Mark in-band AN modes not supported for PHYLINK

We need a better way to signal this, perhaps in phylink_validate, but
for now just print this error message as guidance for other people
looking at this driver's code while trying to rework PHYLINK.

Cc: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: sja1105: Check for PHY mode mismatches with what PHYLINK reports
Vladimir Oltean [Thu, 27 Jun 2019 21:46:36 +0000 (00:46 +0300)]
net: dsa: sja1105: Check for PHY mode mismatches with what PHYLINK reports

PHYLINK being designed with PHYs in mind that can change MII protocol,
for correct operation it is necessary to ensure that the PHY interface
mode stays the same (otherwise clear the supported bit mask, as
required).

Because this is just a hypothetical situation for now, we don't bother
to check whether we could actually support the new PHY interface mode.
Actually we could modify the xMII table, reset the switch and send an
updated static configuration, but adding that would just be dead code.

Cc: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: sja1105: Don't check state->link in phylink_mac_config
Vladimir Oltean [Thu, 27 Jun 2019 21:46:35 +0000 (00:46 +0300)]
net: dsa: sja1105: Don't check state->link in phylink_mac_config

It has been pointed out that PHYLINK can call mac_config only to update
the phy_interface_type and without knowing what the AN results are.

Experimentally, when this was observed to happen, state->link was also
unset, and therefore was used as a proxy to ignore this call. However it
is also suggested that state->link is undefined for this callback and
should not be relied upon.

So let the previously-dead codepath for SPEED_UNKNOWN be called, and
update the comment to make sure the MAC's behavior is sane.

Cc: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agohinic: reduce rss_init stack usage
Arnd Bergmann [Fri, 28 Jun 2019 10:31:44 +0000 (12:31 +0200)]
hinic: reduce rss_init stack usage

On 32-bit architectures, putting an array of 256 u32 values on the
stack uses more space than the warning limit:

drivers/net/ethernet/huawei/hinic/hinic_main.c: In function 'hinic_rss_init':
drivers/net/ethernet/huawei/hinic/hinic_main.c:286:1: error: the frame size of 1068 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

I considered changing the code to use u8 values here, since that's
all the hardware supports, but dynamically allocating the array is
a more isolated fix here.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'stmmac-10GbE-using-XGMAC'
David S. Miller [Fri, 28 Jun 2019 16:23:39 +0000 (09:23 -0700)]
Merge branch 'stmmac-10GbE-using-XGMAC'

Jose Abreu says:

====================
net: stmmac: 10GbE using XGMAC

Support for 10Gb Link using XGMAC core plus some performance tweaks.

Tested in a PCI based setup.

iperf3 TCP results:
TSO ON, MTU=1500, TX Queues = 1, RX Queues = 1, Flow Control ON
Pinned CPU (-A), Zero-Copy (-Z)

[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-600.00 sec   643 GBytes  9.21 Gbits/sec    1             sender
[  5]   0.00-600.00 sec   643 GBytes  9.21 Gbits/sec                  receiver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Update Kconfig entry
Jose Abreu [Fri, 28 Jun 2019 07:29:21 +0000 (09:29 +0200)]
net: stmmac: Update Kconfig entry

We support more speeds now. Update the Kconfig entry.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Only disable interrupts if NAPI is scheduled
Jose Abreu [Fri, 28 Jun 2019 07:29:20 +0000 (09:29 +0200)]
net: stmmac: Only disable interrupts if NAPI is scheduled

Only disable the interrupts if RX NAPI gets to be scheduled. Also,
schedule the TX NAPI only when the interrupts are disabled.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Update RX Tail Pointer to last free entry
Jose Abreu [Fri, 28 Jun 2019 07:29:19 +0000 (09:29 +0200)]
net: stmmac: Update RX Tail Pointer to last free entry

Update the RX Tail Pointer to the last available SKB entry.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Enable support for > 32 Bits addressing in XGMAC
Jose Abreu [Fri, 28 Jun 2019 07:29:18 +0000 (09:29 +0200)]
net: stmmac: Enable support for > 32 Bits addressing in XGMAC

Currently, stmmac only supports 32 bits addressing for SKB. Enable the
support for upto 48 bits addressing in XGMAC core.

This avoids the use of bounce buffers and increases performance.

Changes from v1:
- Fallback to 32 bits in failure (Andrew)

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Do not disable interrupts when cleaning TX
Jose Abreu [Fri, 28 Jun 2019 07:29:17 +0000 (09:29 +0200)]
net: stmmac: Do not disable interrupts when cleaning TX

This is a performance killer and anyways the interrupts are being
disabled by RX NAPI so no need to disable them again.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Add the missing speeds that XGMAC supports
Jose Abreu [Fri, 28 Jun 2019 07:29:16 +0000 (09:29 +0200)]
net: stmmac: Add the missing speeds that XGMAC supports

XGMAC supports following speeds:
- 10G XGMII
- 5G XGMII
- 2.5G XGMII
- 2.5G GMII
- 1G GMII
- 100M MII
- 10M MII

Add them to the stmmac driver.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: dwxgmac: Fix the undefined burst setting
Jose Abreu [Fri, 28 Jun 2019 07:29:15 +0000 (09:29 +0200)]
net: stmmac: dwxgmac: Fix the undefined burst setting

Undefined burst shall only be set if pdata asks to.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Decrease default RX Watchdog value
Jose Abreu [Fri, 28 Jun 2019 07:29:14 +0000 (09:29 +0200)]
net: stmmac: Decrease default RX Watchdog value

For performance reasons decrease the default RX Watchdog value for the
minimum allowed.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Do not try to enable PHY EEE if MAC does not support it
Jose Abreu [Fri, 28 Jun 2019 07:29:13 +0000 (09:29 +0200)]
net: stmmac: Do not try to enable PHY EEE if MAC does not support it

Do not enable EEE feature in the PHY if MAC does not support it.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: dwxgmac: Enable EDMA by default
Jose Abreu [Fri, 28 Jun 2019 07:29:12 +0000 (09:29 +0200)]
net: stmmac: dwxgmac: Enable EDMA by default

Enable the EDMA feature by default which gives higher performance.

Changes from v1:
- Do not use magic values (David)

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: Fix case when PHY handle is not present
Jose Abreu [Fri, 28 Jun 2019 07:25:07 +0000 (09:25 +0200)]
net: stmmac: Fix case when PHY handle is not present

Some DT bindings do not have the PHY handle. Let's fallback to manually
discovery in case phylink_of_phy_connect() fails.

Changes from v1:
- Fixup comment style (Sergei)

Fixes: 74371272f97f ("net: stmmac: Convert to phylink and remove phylib logic")
Reported-by: Katsuhiro Suzuki <katsuhiro@katsuster.net>
Tested-by: Katsuhiro Suzuki <katsuhiro@katsuster.net>
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosis900: remove TxIDLE
Sergej Benilov [Mon, 24 Jun 2019 21:21:02 +0000 (23:21 +0200)]
sis900: remove TxIDLE

Before "sis900: fix TX completion" patch, TX completion was done on TxIDLE interrupt.
TX completion also was the only thing done on TxIDLE interrupt.
Since "sis900: fix TX completion", TX completion is done on TxDESC interrupt.
So it is not necessary any more to set and to check for TxIDLE.

Eliminate TxIDLE from sis900.
Correct some typos, too.

Signed-off-by: Sergej Benilov <sergej.benilov@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: add dst_cache support for udp media
Xin Long [Thu, 20 Jun 2019 11:03:41 +0000 (19:03 +0800)]
tipc: add dst_cache support for udp media

As other udp/ip tunnels do, tipc udp media should also have a
lockless dst_cache supported on its tx path.

Here we add dst_cache into udp_replicast to support dst cache
for both rmcast and rcast, and rmcast uses ub->rcast and each
rcast uses its own node in ub->rcast.list.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Fri, 28 Jun 2019 04:06:39 +0000 (21:06 -0700)]
Merge git://git./linux/kernel/git/davem/net

The new route handling in ip_mc_finish_output() from 'net' overlapped
with the new support for returning congestion notifications from BPF
programs.

In order to handle this I had to take the dev_loopback_xmit() calls
out of the switch statement.

The aquantia driver conflicts were simple overlapping changes.

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'nfp-extend-flower-capabilities-for-GRE-tunnel-offload'
David S. Miller [Fri, 28 Jun 2019 02:47:36 +0000 (19:47 -0700)]
Merge branch 'nfp-extend-flower-capabilities-for-GRE-tunnel-offload'

Jakub Kicinski says:

====================
nfp: extend flower capabilities for GRE tunnel offload

Pieter says:

This set extends the flower match and action components to offload
GRE decapsulation with classification and encapsulation actions. The
first 3 patches are refactor and cleanup patches for improving
readability and reusability. Patch 4 and 5 implement GRE decap and
encap functionality respectively.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add GRE encap action support
Pieter Jansen van Vuuren [Thu, 27 Jun 2019 23:12:43 +0000 (16:12 -0700)]
nfp: flower: add GRE encap action support

Add new GRE encapsulation support, which allows offload of filters
using tunnel_key set action in combination with actions that egress
to GRE type ports.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add GRE decap classification support
Pieter Jansen van Vuuren [Thu, 27 Jun 2019 23:12:42 +0000 (16:12 -0700)]
nfp: flower: add GRE decap classification support

Extend the existing tunnel matching support to include GRE decap
classification. Specifically matching existing tunnel fields for
NVGRE (GRE with protocol field set to TEB).

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: rename tunnel related functions in action offload
Pieter Jansen van Vuuren [Thu, 27 Jun 2019 23:12:41 +0000 (16:12 -0700)]
nfp: flower: rename tunnel related functions in action offload

Previously tunnel related functions in action offload only applied
to UDP tunnels. Rename these functions in preparation for new
tunnel types.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add helper functions for tunnel classification
Pieter Jansen van Vuuren [Thu, 27 Jun 2019 23:12:40 +0000 (16:12 -0700)]
nfp: flower: add helper functions for tunnel classification

Adds IPv4 address and TTL/TOS helper functions, which is done in
preparation for compiling new tunnel types.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: refactor tunnel key layer calculation
Pieter Jansen van Vuuren [Thu, 27 Jun 2019 23:12:39 +0000 (16:12 -0700)]
nfp: flower: refactor tunnel key layer calculation

Refactor the key layer calculation function, in particular the tunnel
key layer calculation by introducing helper functions. This is done
in preparation for supporting GRE tunnel offloads.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>