This patch adds support for driver level TSO in the enetc driver using
the TSO API.

There is not much to say about this specific implementation. We are
using the usual tso_build_hdr(), tso_build_data() to create each data
segment, we create an array of S/G FDs where the first S/G entry is
referencing the header data and the remaining ones the data portion.

For the S/G Table buffer we use the same cache of buffers used on the
other non-GSO cases - dpaa2_eth_sgt_get() and dpaa2_eth_sgt_recycle().

We cannot keep a DMA coherent buffer for all the TSO headers because the
DPAA2 architecture does not work in a ring based fashion so we just
allocate a buffer each time.

Even with these limitations we get the following improvement in TCP
termination on the LX2160A SoC, on a single A72 core running at 2.2GHz.

before: 6.38Gbit/s
after:  8.48Gbit/s
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Up until now, the __dpaa2_eth_tx function used a single FD on the stack
to construct the structure to be enqueued. Since we are now preparing
the ground work to add support for TSO done in software at the driver
level, the same function needs to work with an array of FDs and enqueue
as many as the build_*_fd functions create.

Make the necessary adjustments in order to do this. These include:
keeping an array of FDs in a percpu structure, cleaning up the necessary
FDs before populating it and then, retrying the enqueue process up till
all the generated FDs were enqueued or until we reach the maximum number
retries.

This patch does not change the fact that only a single FD will result
from a __dpaa2_eth_tx call but rather just creates the necessary changes
for the next patch.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of allocating memory for an S/G table each time a nonlinear skb
is processed, and then freeing it on the Tx confirmation path, use the
S/G table cache in order to reuse the memory.

For this to work we have to change the size of the cached buffers so
that it can hold the maximum number of scatterlist entries.

Other than that, each allocate/free call is replaced by a call to the
dpaa2_eth_sgt_get/dpaa2_eth_sgt_recycle functions, introduced in the
previous patch.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Unfortunately, with the blamed commit I also added a side effect in the
ethtool stats shown. Because I added two more fields in the per channel
structure without verifying if its size is used in any way, part of the
ethtool statistics were off by 2.
Fix this by not looking up the size of the structure but instead on a
fixed value kept in a macro.

Fixes: fc398bec ("net: dpaa2: add adaptive interrupt coalescing")
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Link: https://lore.kernel.org/r/20211215105831.290070-1-ioana.ciornei@nxp.comSigned-off-by: NJakub Kicinski <kuba@kernel.org>

Add support for adaptive interrupt coalescing to the dpaa2-eth driver.
First of all, ETHTOOL_COALESCE_USE_ADAPTIVE_RX is defined as a supported
coalesce parameter and the requested state is configured through the
dpio APIs added in the previous patch.

Besides the ethtool API interaction, we keep track of how many bytes and
frames are dequeued per CDAN (Channel Data Availability Notification)
and update the Net DIM instance through the dpaa2_io_update_net_dim()
API.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Move devlink_register to be the last command in the initialization
sequence.
Signed-off-by: NLeon Romanovsky <leonro@nvidia.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's useful, especially for debugging purposes, to have the Rx copybreak
value changeable at runtime. Export it as an ethtool tunable.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

DMA unmapping, allocating a new buffer and DMA mapping it back on the
refill path is really not that efficient. Proper buffer recycling (page
pool, flipping the page and using the other half) cannot be done for
DPAA2 since it's not a ring based controller but it rather deals with
multiple queues which all get their buffers from the same buffer pool on
Rx.

To circumvent these limitations, add support for Rx copybreak. For small
sized packets instead of creating a skb around the buffer in which the
frame was received, allocate a new sk buffer altogether, copy the
contents of the frame and release the initial page back into the buffer
pool.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename the dpaa2_eth_xdp_release_buf function into dpaa2_eth_recycle_buf
since in the next patches we'll be using the same recycle mechanism for
the normal stack path beside for XDP_DROP.

Also, rename the array which holds the buffers to be recycled so that it
does not have any reference to XDP.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add support for backplane link mode, which is, according to discussions
with NXP earlier in the year, is a mode where the OS (Linux) is able to
manage the PCS and Serdes itself.

This commit prepares the ground work for allowing 1G fiber connections
to be used with DPAA2 on the SolidRun CEX7 platforms.
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

If the network interface object is connected to a MAC of TYPE_FIXED, the
link status management is handled exclusively by the firmware. This does
not mean that the driver cannot access the MAC counters and export them
in ethtool.

For this to happen, we open the attached dpmac device and keep a pointer
to it in priv->mac. Because of this, all the checks in the driver of the
following form 'if (priv->mac)' have to be updated to actually check
the dpmac attribute and not rely on the presence of a non-NULL value.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NJakub Kicinski <kuba@kernel.org>

Add support for the new group of devlink traps - PARSER_ERROR_DROPS.
This consists of registering the array of parser error drops supported,
controlling their action through the .trap_group_action_set() callback
and reporting an erroneous skb received on the error queue
appropriately.
DPAA2 devices do not support controlling the action of independent
parser error traps, thus the .trap_action_set() callback just returns an
EOPNOTSUPP while .trap_group_action_set() actually notifies the hardware
what it should do with a frame marked as having a header error.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add basic support in dpaa2-eth for devlink. For the moment, just
register the device with devlink, add the corresponding devlink port and
implement the .info_get() callback.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch is to add PTP sync packet one-step timestamping support.
Before egress, one-step timestamping enablement needs,

- Enabling timestamp and FAS (Frame Annotation Status) in
  dpni buffer layout.

- Write timestamp to frame annotation and set PTP bit in
  FAS to mark as one-step timestamping event.

- Enabling one-step timestamping by dpni_set_single_step_cfg()
  API, with offset provided to insert correction time on frame.
  The offset must respect all MAC headers, VLAN tags and other
  protocol headers accordingly. The correction field update can
  consider delays up to one second. So PTP frame needs to be
  filtered and parsed, and written timestamp into Sync frame
  originTimestamp field.

The operation of API dpni_set_single_step_cfg() has to be done
when no one-step timestamping frames are in flight. So we have
to make sure the last one-step timestamping frame has already
been transmitted on hardware before starting to send the current
one. The resolution is,

- Utilize skb->cb[0] to mark timestamping request per packet.
  If it is one-step timestamping PTP sync packet, queue to skb queue.
  If not, transmit immediately.

- Schedule a work to transmit skbs in skb queue.

- mutex lock is used to ensure the last one-step timestamping packet
  has already been transmitted on hardware through TX confirmation queue
  before transmitting current packet.
Signed-off-by: NYangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch is a preparation for next hardware one-step timestamping
support. For DPAA2, the one step timestamping configuration on
hardware registers has to be done when there is no one-step timestamping
packet in flight. So we will have to use workqueue and skb queue
for such packets transmitting, to make sure waiting the last packet has
already been sent on hardware, and starting to transmit the current one.

So the tx timestamping flag in private data may not reflect the actual
request for the one-step timestamping packets of skb queue. This also
affects skb headroom allocation. Let's use skb->cb[0] to mark the
timestamping request for each skb.
Signed-off-by: NYangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Define a global ptp_qoriq structure pointer, and export to use.
The ptp clock operations will be used in dpaa2-eth driver.
For example, supporting one step timestamping needs to write
current time to hardware frame annotation before sending and
then hardware inserts the delay time on frame during sending.
So in driver, at least clock gettime operation will be needed
to make sure right time is written to hardware frame annotation
for one step timestamping.
Signed-off-by: NYangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

React to TC_SETUP_QDISC_TBF and configure the egress shaper as
appropriate with the maximum rate and burst size requested by the user.
TBF can only be offloaded on DPAA2 when it's the root qdisc, ie it's a
per port shaper.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With the previous commit, in case of insufficient SKB headroom on the Tx
path instead of reallocing the SKB we now send a S/G frame descriptor.
Export the number of occurences of this case as a per CPU counter (in
debugfs) and a total number in the ethtool statistics - "tx converted sg
frames'.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of realloc-ing the skb on the Tx path when the provided headroom
is smaller than the HW requirements, create a Scatter/Gather frame
descriptor with only one entry.

Remove the '[drv] tx realloc frames' counter exposed previously through
ethtool since it is no longer used.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Leave congestion group taildrop enabled for all traffic classes
when PFC is enabled. Notification threshold is low enough such
that it will be hit first and this also ensures that FQs on
traffic classes which are not PFC enabled won't drain the buffer
pool.

FQ taildrop threshold is kept disabled as long as any form of
flow control is on. Since FQ taildrop works with bytes, not number
of frames, we can't guarantee it will not interfere with the
congestion notification mechanism for all frame sizes.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add support in dpaa2-eth for PFC (Priority Flow Control)
through the DCB ops.

Instruct the hardware to respond to received PFC frames.
Current firmware doesn't allow us to selectively enable PFC
on the Rx side for some priorities only, so we will react to
all incoming PFC frames (and stop transmitting on the traffic
classes specified in the frame).

Also, configure the hardware to generate PFC frames based on Rx
congestion notifications. When a certain number of frames accumulate in
the ingress queues corresponding to a traffic class, priority flow
control frames are generated for that TC.

The number of PFC traffic classes available can be queried through
lldptool. Also, which of those traffic classes have PFC enabled is also
controlled through the same dcbnl_rtnl_ops callbacks.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now that we have congestion group taildrop configured at all
times, we can afford to increase the frame queue taildrop
threshold; this will ensure a better response when receiving
bursts of large-sized frames.

Also decouple the buffer pool count from the Rx FQ taildrop
threshold, as above change would increase it too much. Instead,
keep the old count as a hardcoded value.

With the new limits, we try to ensure that:
* we allow enough leeway for large frame bursts (by buffering
enough of them in queues to avoid heavy dropping in case of
bursty traffic, but when overall ingress bandwidth is manageable)
* allow pending frames to be evenly spread between ingress FQs,
regardless of frame size
* avoid dropping frames due to the buffer pool being empty; this
is not a bad behaviour per se, but system overall response is
more linear and predictable when frames are dropped at frame
queue/group level.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The increase in number of ingress frame queues means we now risk
depleting the buffer pool before the FQ taildrop kicks in.

Congestion group taildrop allows us to control the number of frames that
can accumulate on a group of Rx frame queues belonging to the same
traffic class.  This setting coexists with the frame queue based
taildrop: whichever limit gets hit first triggers the frame drop.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add convenient helper functions that determines whether Rx/Tx pause
frames are enabled based on link state flags received from firmware.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Configure static ingress classification based on VLAN PCP field.
If the DPNI doesn't have enough traffic classes to accommodate all
priority levels, the lowest ones end up on TC 0 (default on miss).
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The firmware reserves for each DPNI a number of RX frame queues
equal to the number of configured flows x number of configured
traffic classes.

Current driver configuration directs all incoming traffic to
FQs corresponding to TC0, leaving all other priority levels unused.

Start adding support for multiple ingress traffic classes, by
configuring the FQs associated with all priority levels, not just
TC0. All settings that are per-TC, such as those related to
hashing and flow steering, are also updated.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add driver level bulking to the XDP_TX action.

An array of frame descriptors is held for each Tx frame queue and
populated accordingly when the action returned by the XDP program is
XDP_TX. The frames will be actually enqueued only when the array is
filled. At the end of the NAPI cycle a flush on the queued frames is
performed in order to enqueue the remaining FDs.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Depending on the WRIOP version, the buffer size on the RX path must by a
multiple of 64 or 256. Handle this restriction properly by aligning down
the buffer size to the necessary value. Also, use the new buffer size
dynamically computed instead of the compile time one.

Fixes: 27c87486 ("dpaa2-eth: Use a single page per Rx buffer")
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Create an independent function that takes a particular frame queue and
an array of frame descriptors and tries to enqueue them until it hits
the maximum number fo retries. The same function will be used in the
next patch also on the XDP_TX path.

Also, create the dpaa2_eth_xdp_fds structure to incorporate the array of
FDs as well as the number of FDs already populated.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Compute the average number of frames processed for each CDAN (Channel
Data Availability Notification) and export it to debugfs detailed
channel stats.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Take advantage of the bulk enqueue feature in .ndo_xdp_xmit.
We cannot use the XDP_XMIT_FLUSH since the architecture is not capable
to store all the frames dequeued in a NAPI cycle so we instead are
enqueueing all the frames received in a ndo_xdp_xmit call right away.

After setting up all FDs for the xdp_frames received, enqueue multiple
frames at a time until all are sent or the maximum number of retries is
hit.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Acked-by: NJesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Update the dpaa2-eth driver to use the bulk enqueue function introduced
with the change to QBMAN ring mode. At the moment, no functional changes
are made but rather the driver just transitions to the new interface
while still enqueuing just one frame at a time.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The enqueue dpaa2-eth callback now returns the number of successfully
enqueued frames. This is a preliminary patch necessary for adding
support for bulk ring mode enqueue.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The dpaa2-eth driver now has support for connecting to its associated
PHY device found through standard OF bindings.

This happens when the DPNI object (that the driver probes on) gets
connected to a DPMAC. When that happens, the device tree is looked up by
the DPMAC ID, and the associated PHY bindings are found.

The old logic of handling the net device's link state by hand still
needs to be kept, as the DPNI can be connected to other devices on the
bus than a DPMAC: other DPNI, DPSW ports, etc. This logic is only
engaged when there is no DPMAC (and therefore no phylink instance)
attached.

The MC firmware support multiple type of DPMAC links: TYPE_FIXED,
TYPE_PHY. The TYPE_FIXED mode does not require any DPMAC management from
Linux side, and as such, the driver will not handle such a DPMAC.

Although PHYLINK typically handles SFP cages and in-band AN modes, for
the moment the driver only supports the RGMII interfaces found on the
LX2160A. Support for other modes will come later.
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Throughout the driver there are several places where we wait
indefinitely for DPIO portal commands to be executed, while
the portal returns a busy response code.

Even though in theory we are guaranteed the portals become
available eventually, in practice the QBMan hardware module
may become unresponsive in various corner cases.

Make sure we can never get stuck in an infinite while loop
by adding a retry counter for all portal commands.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NIoana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Starting with firmware version MC10.18.0, we have support for
L2 flow control. Asymmetrical configuration (Rx or Tx only) is
supported, but not pause frame autonegotioation.

Pause frame configuration is done via ethtool. By default, we start
with flow control enabled on both Rx and Tx. Changes are propagated
to hardware through firmware commands, using two flags (PAUSE,
ASYM_PAUSE) to specify Rx and Tx pause configuration, as follows:

PAUSE | ASYM_PAUSE | Rx pause | Tx pause
----------------------------------------
  0   |     0      | disabled | disabled
  0   |     1      | disabled | enabled
  1   |     0      | enabled  | enabled
  1   |     1      | enabled  | disabled

The hardware can automatically send pause frames when the number
of buffers in the pool goes below a predefined threshold. Due to
this, flow control is incompatible with Rx frame queue taildrop
(both mechanisms target the case when processing of ingress
frames can't keep up with the Rx rate; for large frames, the number
of buffers in the pool may never get low enough to trigger pause
frames as long as taildrop is enabled). So we set pause frame
generation and Rx FQ taildrop as mutually exclusive.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement mqprio qdisc support by mapping traffic classes to
different hardware enqueue priorities. The maximum number of
supported traffic classes is an attribute of each DPNI object.

The traffic classes map to hardware priorities from highest (0)
to lowest (highest prio number). The skb priority information
received from the stack is used to select the hardware Tx queue
on which to enqueue the frame.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NBogdan Purcareata <bogdan.purcareata@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

DPNI objects can have multiple traffic classes, as reflected by
the num_tc attribute. Until now we ignored its value and only
used traffic class 0.

This patch adds support for multiple Tx traffic classes; we have
num_queues x num_tcs hardware queues available for each interface.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Function dpaa2_eth_cls_key_size() expects a 64bit argument,
but DPAA2_ETH_DIST_ALL is defined as UINT_MAX. Fix this.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On platforms that lack a TCAM (like LS1088A), masking of
flow steering keys is not supported. Until now we didn't
offer flow steering capabilities at all on these platforms,
since our driver implementation configured a "comprehensive"
FS key (containing all supported header fields), with masks
used to ignore the fields not present in the rules provided
by the user.

We now allow ethtool rules that share a common key (i.e. have
the same header fields). The FS key is now kept in the driver
private data and initialized when the first rule is added to
an empty table, rather than at probe time. If a rule with a new
composition key is wanted, the user must first manually delete
all previous rules.

When building a FS table entry to pass to firmware, we still use
the old building algorithm, which assumes an all-supported-fields
key, and later collapse the fields which aren't actually needed.

Masked rules are not supported; if provided, the mask value
will be ignored. For firmware versions older than MC10.7.0
(that only offer the legacy ABIs for configuring distribution
keys) flow steering without masking support remains unavailable.
Signed-off-by: NIoana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>