The TI AM65x/J721E SoCs Gigabit Ethernet Switch subsystem (CPSW2G NUSS) has
two ports - One Ethernet port (port 1) with selectable RGMII and RMII
interfaces and an internal Communications Port Programming Interface (CPPI)
port (Host port 0) and with ALE in between. It also contains
- Management Data Input/Output (MDIO) interface for physical layer device
(PHY) management;
- Updated Address Lookup Engine (ALE) module;
- (TBD) New version of Common platform time sync (CPTS) module.
On the TI am65x/J721E SoCs CPSW NUSS Ethernet subsystem into device MCU
domain named MCU_CPSW0.
Host Port 0 CPPI Packet Streaming Interface interface supports 8 TX
channels and one RX channels operating by TI am654 NAVSS Unified DMA
Peripheral Root Complex (UDMA-P) controller.
Introduced driver provides standard Linux net_device to user space and supports:
- ifconfig up/down
- MAC address configuration
- ethtool operation:
--driver
--change
--register-dump
--negotiate phy
--statistics
--set-eee phy
--show-ring
--show-channels
--set-channels
- net_device ioctl mii-control
- promisc mode
- rx checksum offload for non-fragmented IPv4/IPv6 TCP/UDP packets.
The CPSW NUSS can verify IPv4/IPv6 TCP/UDP packets checksum and fills
csum information for each packet in psdata[2] word:
- BIT(16) CHECKSUM_ERROR - indicates csum error
- BIT(17) FRAGMENT - indicates fragmented packet
- BIT(18) TCP_UDP_N - Indicates TCP packet was detected
- BIT(19) IPV6_VALID, BIT(20) IPV4_VALID - indicates IPv6/IPv4 packet
- BIT(15, 0) CHECKSUM_ADD - This is the value that was summed
during the checksum computation. This value is FFFFh for non fragmented
IPV4/6 UDP/TCP packets with no checksum error.
RX csum offload can be disabled:
ethtool -K <dev> rx-checksum on|off
- tx checksum offload support for IPv4/IPv6 TCP/UDP packets (J721E only).
TX csum HW offload can be enabled/disabled:
ethtool -K <dev> tx-checksum-ip-generic on|off
- multiq and switch between round robin/prio modes for cppi tx queues by
using Netdev private flag "p0-rx-ptype-rrobin" to switch between
Round Robin and Fixed priority modes:
# ethtool --show-priv-flags eth0
Private flags for eth0:
p0-rx-ptype-rrobin: on
# ethtool --set-priv-flags eth0 p0-rx-ptype-rrobin off
Number of TX DMA channels can be changed using "ethtool -L eth0 tx <N>".
- GRO support: the napi_gro_receive() and napi_complete_done() are used.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Document device tree bindings for The TI AM654x/J721E SoC Gigabit Ethernet MAC
(Media Access Controller - CPSW2G NUSS). The CPSW NUSS provides Ethernet packet
communication for the device.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for default thread configuration for AM65x CPSW NUSS ALE to
allow route all ingress packets to one default RX UDMA flow.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The new CPSW ALE version, available on TI K3 AM654/J721E SoCs family,
allows to switch any external port to MAC only mode. When MAC only mode
enabled this port be treated like a MAC port for the host. All traffic
received is only sent to the host. The host must direct traffic to this
port as the lookup engine will not send traffic to the ports with the
p0_maconly bit set and the p0_no_learn also set. If p0_maconly bit is set
and the p0_no_learn is not set, the host can send non-directed packets that
can be sent to the destination of a MacOnly port. It is also possible that
The host can broadcast to all ports including MacOnly ports in this mode.
This patch add ALE supprt for MAC only mode.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On AM65xx MCU CPSW2G NUSS and 66AK2E/L NUSS the unregistered multicast
packets are still can be received with promisc and allmulti disabled.
This happens, because ALE VLAN entries on these SoCs do not contain port
masks for reg/unreg mcast packets, but instead store indexes of
ALE_VLAN_MASK_MUXx_REG registers which intended for store port masks for
reg/unreg mcast packets.
ALE VLAN entry:UNREG_MCAST_FLOOD_INDEX -> ALE_VLAN_MASK_MUXx
ALE VLAN entry:REG_MCAST_FLOOD_INDEX -> ALE_VLAN_MASK_MUXy
The commit b361da8373 ("net: netcp: ale: add proper ale entry mask bits
for netcp switch ALE") update ALE code to support such ALE entries, it is
always used ALE_VLAN_MASK_MUX0_REG index in ALE VLAN entry for unreg mcast
packets mask configuration, which is read-only, at least for AM65xx MCU
CPSW2G NUSS and 66AK2E/L NUSS. As result unreg mcast packets are allowed
always.
Hence, update ALE code to use ALE_VLAN_MASK_MUX1_REG index for ALE VLAN
entries to configure unreg mcast port mask.
Fixes: b361da8373 ("net: netcp: ale: add proper ale entry mask bits for netcp switch ALE")
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The phy-gmii-sel can be only auto selected in Kconfig and now the pretty
complex Kconfig dependencies are defined for phy-gmii-sel driver, which
also need to be updated every time phy-gmii-sel is re-used for any new
networking driver.
Simplify Kconfig definition for phy-gmii-sel PHY driver - drop all
dependencies and from networking drivers and rely on using 'imply
PHY_TI_GMII_SEL' in Kconfig definitions for networking drivers instead.
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
Tested-by: Murali Karicheri <m-karicheri2@ti.com>
Tested-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Guillaume Nault says:
====================
cls_flower: Use extack in fl_set_key()
Add missing extack messages in fl_set_key(), so that users can get more
meaningfull error messages when netlink attributes are rejected.
Patch 1 also extends extack in tcf_change_indev() (in pkt_cls.h) since
this function is used by fl_set_key().
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass extack down to fl_set_key_flags() and set message on error.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass extack down to fl_set_key_port_range() and set message on error.
Both the min and max ports would qualify as invalid attributes here.
Report the min one as invalid, as it's probably what makes the most
sense from a user point of view.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass extack down to fl_set_key_mpls() and set message on error.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an error message when device wasn't found.
While there, also set the bad attribute's offset in extack.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tobias Waldekranz says:
====================
net: phy: marvell usb to mdio controller
Support for an MDIO controller present on development boards for
Marvell switches from the Link Street (88E6xxx) family.
v3->v4:
- Remove unnecessary dependency on OF_MDIO.
v2->v3:
- Rename driver smi2usb -> mvusb.
- Clean up unused USB references.
v1->v2:
- Reverse christmas tree ordering of local variables.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
An MDIO controller present on development boards for Marvell switches
from the Link Street (88E6xxx) family.
Using this module, you can use the following setup as a development
platform for switchdev and DSA related work.
.-------. .-----------------.
| USB----USB |
| SoC | | 88E6390X-DB ETH1-10
| ETH----ETH0 |
'-------' '-----------------'
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Describe how the USB to MDIO controller can optionally use device tree
bindings to reference attached devices such as switches.
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we have scenarios like
mdiobus_register()
-> loads PHY driver module(s)
-> registers PHY driver(s)
-> may schedule async probe
phydev = mdiobus_get_phy()
<phydev action involving PHY driver>
or
phydev = phy_device_create()
-> loads PHY driver module
-> registers PHY driver
-> may schedule async probe
<phydev action involving PHY driver>
then we expect the PHY driver to be bound to the phydev when triggering
the action. This may not be the case in case of asynchronous probing.
Therefore ensure that PHY drivers are probed synchronously.
Default still is sync probing, except async probing is explicitly
requested. I saw some comments that the intention is to promote
async probing for more parallelism in boot process and want to be
prepared for the case that the default is changed to async probing.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jacob Keller says:
====================
implement DEVLINK_CMD_REGION_NEW
This series adds support for the DEVLINK_CMD_REGION_NEW operation, used to
enable userspace requesting a snapshot of a region on demand.
This can be useful to enable adding regions for a driver for which there is
no trigger to create snapshots. By making this a core part of devlink, there
is no need for the drivers to use a separate channel such as debugfs.
The primary intent for this kind of region is to expose device information
that might be useful for diagnostics and information gathering.
The first few patches refactor regions to support a new ops structure for
extending the available operations that regions can perform. This includes
converting the destructor into an op from a function argument.
Next, patches refactor the snapshot id allocation to use an xarray which
tracks the number of current snapshots using a given id. This is done so
that id lifetime can be determined, and ids can be released when no longer
in use.
Without this change, snapshot ids remain used forever, until the snapshot_id
count rolled over UINT_MAX.
Finally, code to enable the previously unused DEVLINK_CMD_REGION_NEW is
added. This code enforces that the snapshot id is always provided, unlike
previous revisions of this series.
Finally, a patch is added to enable using this new command via the .snapshot
callback in both netdevsim and the ice driver.
For the ice driver, a new "nvm-flash" region is added, which will enable
read access to the NVM flash contents. The intention for this is to allow
diagnostics tools to gather information about the device. By using a
snapshot and gathering the NVM contents all at once, the contents can be
atomic.
Links to previous discussions:
1st RFC - https://lore.kernel.org/netdev/20200130225913.1671982-1-jacob.e.keller@intel.com/
2nd RFC - https://lore.kernel.org/netdev/20200214232223.3442651-1-jacob.e.keller@intel.com/
v1 - https://lore.kernel.org/netdev/20200324223445.2077900-1-jacob.e.keller@intel.com/
v2 - https://lore.kernel.org/netdev/20200326035157.2211090-1-jacob.e.keller@intel.com/
Major changes since RFC:
* use an xarray for tracking snapshot ids, rather than an IDR
* remove support for auto-generated snapshot ids in DEVLINK_CMD_REGION_NEW
See each patch for an individual changelog per-patch
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a devlink region for exposing the device's Non Volatime Memory flash
contents.
Support the recently added .snapshot operation, enabling userspace to
request a snapshot of the NVM contents via DEVLINK_CMD_REGION_NEW.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement the .snapshot region operation for the dummy data region. This
enables a region snapshot to be taken upon request via the new
DEVLINK_CMD_REGION_SNAPSHOT command.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement support for the DEVLINK_CMD_REGION_NEW command for creating
snapshots. This new command parallels the existing
DEVLINK_CMD_REGION_DEL.
In order for DEVLINK_CMD_REGION_NEW to work for a region, the new
".snapshot" operation must be implemented in the region's ops structure.
The desired snapshot id must be provided. This helps avoid confusion on
the purpose of DEVLINK_CMD_REGION_NEW, and keeps the API simpler.
The requested id will be inserted into the xarray tracking the number of
snapshots using each id. If this id is already used by another snapshot
on any region, an error will be returned.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each snapshot created for a devlink region must have an id. These ids
are supposed to be unique per "event" that caused the snapshot to be
created. Drivers call devlink_region_snapshot_id_get to obtain a new id
to use for a new event trigger. The id values are tracked per devlink,
so that the same id number can be used if a triggering event creates
multiple snapshots on different regions.
There is no mechanism for snapshot ids to ever be reused. Introduce an
xarray to store the count of how many snapshots are using a given id,
replacing the snapshot_id field previously used for picking the next id.
The devlink_region_snapshot_id_get() function will use xa_alloc to
insert an initial value of 1 value at an available slot between 0 and
U32_MAX.
The new __devlink_snapshot_id_increment() and
__devlink_snapshot_id_decrement() functions will be used to track how
many snapshots currently use an id.
Drivers must now call devlink_snapshot_id_put() in order to release
their reference of the snapshot id after adding region snapshots.
By tracking the total number of snapshots using a given id, it is
possible for the decrement() function to erase the id from the xarray
when it is not in use.
With this method, a snapshot id can become reused again once all
snapshots that referred to it have been deleted via
DEVLINK_CMD_REGION_DEL, and the driver has finished adding snapshots.
This work also paves the way to introduce a mechanism for userspace to
request a snapshot.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The devlink_snapshot_id_get() function returns a snapshot id. The
snapshot id is a u32, so there is no way to indicate an error code.
A future change is going to possibly add additional cases where this
function could fail. Refactor the function to return the snapshot id in
an argument, so that it can return zero or an error value.
This ensures that snapshot ids cannot be confused with error values, and
aids in the future refactor of snapshot id allocation management.
Because there is no current way to release previously used snapshot ids,
add a simple check ensuring that an error is reported in case the
snapshot_id would over flow.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A future change is going to implement a new devlink command to request
a snapshot on demand. As part of this, the logic for handling the
snapshot ids will be refactored. To simplify the snapshot id allocation
function, move it to a separate function prefixed by `__`. This helper
function will assume the lock is held.
While no other callers will exist, it simplifies refactoring the logic
because there is no need to complicate the function with gotos to handle
unlocking on failure.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The devlink_region_snapshot_create function returns -ENOMEM when the
maximum number of snapshots has been reached. This is confusing because
it is not an issue of being out of memory. Change this to use -ENOSPC
instead.
Reported-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A future change is going to add a new devlink command to request
a snapshot on demand. This function will want to call the
devlink_region_snapshot_create function while already holding the
devlink instance lock.
Extract the logic of this function into a static function prefixed by
`__` to indicate that it is an internal helper function. Modify the
original function to be implemented in terms of the new locked
function.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function documentation comment for devlink_region_snapshot_create
included a literal tab character between 'future analyses' that was
difficult to spot as it happened to only display as one space wide.
Fix the comment to use a space here instead of a stray tab appearing in
the middle of a sentence.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It does not makes sense that two snapshots for a given region would use
different destructors. Simplify snapshot creation by adding
a .destructor op for regions.
This operation will replace the data_destructor for the snapshot
creation, and makes snapshot creation easier.
Noticed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Modify the devlink region code in preparation for adding new operations
on regions.
Create a devlink_region_ops structure, and move the name pointer from
within the devlink_region structure into the ops structure (similar to
the devlink_health_reporter_ops).
This prepares the regions to enable support of additional operations in
the future such as requesting snapshots, or accessing the region
directly without a snapshot.
In order to re-use the constant strings in the mlx4 driver their
declaration must be changed to 'const char * const' to ensure the
compiler realizes that both the data and the pointer cannot change.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Lorenzo Bianconi says:
====================
veth: move ndo_xdp_xmit stats to peer veth_rq
Move ndo_xdp_xmit ethtool stats accounting to peer veth_rq.
Move XDP_TX accounting to veth_xdp_flush_bq routine.
Changes since v1:
- rename xdp_xmit[_err] counters to peer_tq_xdp_xmit[_err]
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Rely on 'remote' veth_rq to account ndo_xdp_xmit ethtool counters.
Move XDP_TX accounting to veth_xdp_flush_bq routine.
Remove 'rx' prefix in rx xdp ethool counters
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Acked-by: Toshiaki Makita <toshiaki.makita1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Substitute net_device point with veth_rq one in veth_xdp_flush_bq,
veth_xdp_flush and veth_xdp_tx signature. This is a preliminary patch
to account xdp_xmit counter on 'receiving' veth_rq
Acked-by: Toshiaki Makita <toshiaki.makita1@gmail.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move away from the deprecated API and return the shiny new ERRPTR where
useful.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move away from the deprecated API and return the shiny new ERRPTR where
useful.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Petr Machata says:
====================
Implement stats_update callback for pedit and skbedit
The stats_update callback is used for adding HW counters to the SW ones.
Both skbedit and pedit actions are actually recognized by flow_offload.h,
but do not implement these callbacks. As a consequence, the reported values
are only the SW ones, even where there is a HW counter available.
Patch #1 adds the callback to action skbedit, patch #2 adds it to action
pedit. Patch #3 tweaks an skbedit selftest with a check that would have
caught this problem.
The pedit test is not likewise tweaked, because the iproute2 pedit action
currently does not support JSON dumping. This will be addressed later.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the test checks the observable effect of skbedit priority:
queueing of packets at the correct qdisc band. It therefore misses the fact
that the counters for offloaded rules are not updated. Add an extra check
for the counter.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement this callback in order to get the offloaded stats added to the
kernel stats.
Reported-by: Alexander Petrovskiy <alexpe@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement this callback in order to get the offloaded stats added to the
kernel stats.
Reported-by: Alexander Petrovskiy <alexpe@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel says:
====================
mlxsw: Offload TC action pedit munge dsfield
Petr says:
The Spectrum switches allow packet prioritization based on DSCP on ingress,
and update of DSCP on egress. This is configured through the DCB APP rules.
For some use cases, assigning a custom DSCP value based on an ACL match is
a better tool. To that end, offload FLOW_ACTION_MANGLE to permit changing
of dsfield as a whole, or DSCP and ECN values in isolation.
After fixing a commentary nit in patch #1, and mlxsw naming in patch #2,
patches #3 and #4 add the offload to mlxsw.
Patch #5 adds a forwarding selftest for pedit dsfield, applicable to SW as
well as HW datapaths. Patch #6 adds a mlxsw-specific test to verify DSCP
rewrite due to DCB APP rules is not performed on pedited packets.
The tests only cover IPv4 dsfield setting. We have tests for IPv6 as well,
but would like to postpone their contribution until the corresponding
iproute patches have been accepted.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When DSCP is updated through an offloaded pedit action, DSCP rewrite on
egress should be disabled. Add a test that check that it is so.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a test that runs packets with dsfield set, and test that pedit adjusts
the DSCP or ECN parts or the whole field.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Offload action pedit ex munge when used with a flower classifier. Only
allow setting of DSCP, ECN, or the whole DSField in IPv4 and IPv6 packets.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The QOS_ACTION is used for manipulating the QOS attributes of the packet.
Add the defines and helpers related to DSCP and ECN fields, and dscp_rw.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original idea was to reuse this set of actions for ECN rewrite as well,
but on second look, it's not such a great idea. These two items should each
have its own command. Rename the existing enum to make it obvious that it
belongs to switch_prio_cmd.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This field references FLOW_ACTION_PACKET_EDIT. Such action does not exist
though. Instead the field is used for FLOW_ACTION_MANGLE and _ADD.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) Cleanups from Dan Carpenter and wenxu.
2) Paul and Roi, Some minor updates and fixes to E-Switch to address
issues introduced in the previous reg_c0 updates series.
3) Eli Cohen simplifies and improves flow steering matching group searches
and flow table entries version management.
4) Parav Pandit, improves devlink eswitch mode changes thread safety.
By making devlink rely on driver for thread safety and introducing mlx5
eswitch mode change protection.
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAl58SW8ACgkQSD+KveBX
+j4AxQf8DdrFrBD0NFTcAILS4bnTJC0I3xKRPb/2oYtWLVyJ9G5XAZqHC0DAG7xs
jy8xhIFbeUxgLEdcx0la5vdR1mPlzs4XBHTe99YwzwK/jojrA7YXrlb3kv+RXWVY
uNVAby68wh4EnO61R51ahIBXLPNbiYpo/wAWKvvBKRkOcYMVTKIFiP157AnJWObY
fxnt06I0NFaIX8Va4MEqkrmUYrI4jJcqOJC9FwRBLDhFHcFkLh0Gav3vJJ7M4BCB
ggPJpuZ4pu43qX9TtSOm8V/GlWWN0RB7PdbvliFBEHYG21hf9MfE8bPPKRlB7CO+
B5+9ULhpvbjX7yRrkZ7fd4zlQ1siew==
=Flln
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2020-03-25' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2020-03-25
1) Cleanups from Dan Carpenter and wenxu.
2) Paul and Roi, Some minor updates and fixes to E-Switch to address
issues introduced in the previous reg_c0 updates series.
3) Eli Cohen simplifies and improves flow steering matching group searches
and flow table entries version management.
4) Parav Pandit, improves devlink eswitch mode changes thread safety.
By making devlink rely on driver for thread safety and introducing mlx5
eswitch mode change protection.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/atheros/atlx/atl2.c:40:19: warning: ‘atl2_driver_string’ defined but not used [-Wunused-const-variable=]
static const char atl2_driver_string[] = "Atheros(R) L2 Ethernet Driver";
^~~~~~~~~~~~~~~~~~
commit ea97374214 ("net/atheros: Clean atheros code from driver version")
left behind this, remove it.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the commit f73b12812a
("tipc: improve throughput between nodes in netns"), we're missing a check
to handle TIPC_DIRECT_MSG type, it's still using old sending mechanism for
this message type. So, throughput improvement is not significant as
expected.
Besides that, when sending a large message with that type, we're also
handle wrong receiving queue, it should be enqueued in socket receiving
instead of multicast messages.
Fix this by adding the missing case for TIPC_DIRECT_MSG.
Fixes: f73b12812a ("tipc: improve throughput between nodes in netns")
Reported-by: Tuong Lien <tuong.t.lien@dektech.com.au>
Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently eswitch mode change is occurring from 2 different execution
contexts as below.
1. sriov sysfs enable/disable
2. devlink eswitch set commands
Both of them need to access eswitch related data structures in
synchronized manner.
Without any synchronization below race condition exist.
SR-IOV enable/disable with devlink eswitch mode change:
cpu-0 cpu-1
----- -----
mlx5_device_disable_sriov() mlx5_devlink_eswitch_mode_set()
mlx5_eswitch_disable() esw_offloads_stop()
esw_offloads_disable() mlx5_eswitch_disable()
esw_offloads_disable()
Hence, they are synchronized using a new mode_lock.
eswitch's state_lock is not used as it can lead to a deadlock scenario
below and state_lock is only for vport and fdb exclusive access.
ip link set vf <param>
netlink rcv_msg() - Lock A
rtnl_lock
vfinfo()
esw->state_lock() - Lock B
devlink eswitch_set
devlink_mutex
esw->state_lock() - Lock B
attach_netdev()
register_netdev()
rtnl_lock - Lock A
Alternatives considered:
1. Acquiring rtnl lock before taking esw->state_lock to follow similar
locking sequence as ip link flow during eswitch mode set.
rtnl lock is not good idea for two reasons.
(a) Holding rtnl lock for several hundred device commands is not good
idea.
(b) It leads to below and more similar deadlocks.
devlink eswitch_set
devlink_mutex
rtnl_lock - Lock A
esw->state_lock() - Lock B
eswitch_disable()
reload()
ib_register_device()
ib_cache_setup_one()
rtnl_lock()
2. Exporting devlink lock may lead to undesired use of it in vendor
driver(s) in future.
3. Unloading representors outside of the mode_lock requires
serialization with other process trying to enable the eswitch.
4. Differing the representors life cycle to a different workqueue
requires synchronization with func_change_handler workqueue.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Subsequent patch protects eswitch mode changes across sriov and devlink
interfaces. It is desirable for eswitch to provide thread safe eswitch
enable and disable APIs.
Hence, extend eswitch enable API to optionally update num_vfs when
requested.
In subsequent patch, eswitch num_vfs are updated after all the eswitch
users eswitch drops its reference count.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In order to check eswitch state under a lock, prepare code to split
capability check and eswitch state check into two helper functions.
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
devlink_nl_cmd_eswitch_set_doit() doesn't hold devlink->lock mutex while
invoking driver callback. This is likely due to eswitch mode setting
involves adding/remove devlink ports, health reporters or
other devlink objects for a devlink device.
So it is driver responsiblity to ensure thread safe eswitch state
transition happening via either sriov legacy enablement or via devlink
eswitch set callback.
Therefore, get() callback should also be invoked without holding
devlink->lock mutex.
Vendor driver can use same internal lock which it uses during eswitch
mode set() callback.
This makes get() and set() implimentation symmetric in devlink core and
in vendor drivers.
Hence, remove holding devlink->lock mutex during eswitch get() callback.
Failing to do so results into below deadlock scenario when mlx5_core
driver is improved to handle eswitch mode set critical section invoked
by devlink and sriov sysfs interface in subsequent patch.
devlink_nl_cmd_eswitch_set_doit()
mlx5_eswitch_mode_set()
mutex_lock(esw->mode_lock) <- Lock A
[...]
register_devlink_port()
mutex_lock(&devlink->lock); <- lock B
mutex_lock(&devlink->lock); <- lock B
devlink_nl_cmd_eswitch_get_doit()
mlx5_eswitch_mode_get()
mutex_lock(esw->mode_lock) <- Lock A
In subsequent patch, mlx5_core driver uses its internal lock during
get() and set() eswitch callbacks.
Other drivers have been inspected which returns either constant during
get operations or reads the value from already allocated structure.
Hence it is safe to remove the lock in get( ) callback and let vendor
driver handle it.
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
