Depending on when MC connects the DPNI to a MAC, Tx FQIDs may
not be available during probe time.
Read the FQIDs each time the link goes up to avoid using invalid
values. In case an error occurs or an invalid value is retrieved,
fall back to QDID-based enqueueing.
Fixes: 1fa0f68c92 ("dpaa2-eth: Use FQ-based DPIO enqueue API")
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add IRQ for the DPNI endpoint change event, resolving the issue
when a dynamically created DPNI gets a randomly generated hw address
when the endpoint is a DPMAC object.
Signed-off-by: Florin Chiculita <florinlaurentiu.chiculita@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Starting with firmware version MC10.18.0, a new counter for in flight
Tx frames is offered. Use it when bringing down the interface to
determine when all pending Tx frames have been processed by hardware
instead of sleeping a fixed amount of time.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David 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: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever a link state change occurs, we get notified and save
the new link settings in the device's private data. In ethtool
get_link_ksettings, use the stored state instead of interrogating
the firmware each time.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David 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: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Bogdan Purcareata <bogdan.purcareata@nxp.com>
Signed-off-by: David 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: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the code configuring xps on the netdev TX queues to a
separate function. A subsequent patch will need to call
this in another context as well.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver is using netdev_alloc_frag() for allocation in the
->ndo_start_xmit() path. That one is always invoked in a BH disabled
region so we could also use napi_alloc_frag().
Use napi_alloc_frag() for skb allocation.
Cc: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Acked-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to the comment, the preempt_disable() statement is required
due to synchronisation in napi_alloc_frag(). The awful truth is that
local_bh_disable() is required because otherwise the NAPI poll callback
can be invoked while the open function setup buffers. This isn't
unlikely since the dpaa2 provides multiple devices.
The usage of napi_alloc_frag() has been removed in commit
27c874867c ("dpaa2-eth: Use a single page per Rx buffer")
which means that the comment is not accurate and the preempt_disable()
statement is not required.
Remove the outdated comment and the no longer required
preempt_disable().
Cc: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Acked-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
The phylink conflict was between a bug fix by Russell King
to make sure we have a consistent PHY interface mode, and
a change in net-next to pull some code in phylink_resolve()
into the helper functions phylink_mac_link_{up,down}()
On the dp83867 side it's mostly overlapping changes, with
the 'net' side removing a condition that was supposed to
trigger for RGMII but because of how it was coded never
actually could trigger.
Signed-off-by: David S. Miller <davem@davemloft.net>
Use PTR_ERR_OR_ZERO instead of PTR_ERR in cases where
zero is a valid input. Reported by smatch.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit f8b9958534.
The reverted change instructed the QMan hardware block to fetch
RX frame annotation and beginning of frame data to cache before
the core would read them.
It turns out that in rare cases, it's possible that a QMan
stashing transaction is delayed long enough such that, by the time
it gets executed, the frame in question had already been dequeued
by the core and software processing began on it. If the core
manages to unmap the frame buffer _before_ the stashing transaction
is executed, an SMMU exception will be raised.
Unfortunately there is no easy way to work around this while keeping
the performance advantages brought by QMan stashing, so disable
it altogether.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David 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: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce an internal id bitfield to uniquely identify header fields
supported by the Rx distribution keys. For the hash key, add a
conversion from the RXH_* bitmask provided by ethtool to the internal
ids.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set the Rx flow classification enable flag only if key config
operation is successful.
Fixes 3f9b5c9 ("dpaa2-eth: Configure Rx flow classification key")
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Take advantage of the software Rx batching by using
netif_receive_skb_list instead of napi_gro_receive.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It might happen that Tx conf acknowledges a frame before it was
subscribed in bql, as subscribing was previously done after the enqueue
operation.
This patch moves the netdev_tx_sent_queue call before the actual frame
enqueue, so that this can never happen.
Fixes: 569dac6a5a ("dpaa2-eth: bql support")
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make sure we don't try to enqueue XDP_REDIRECT frames to an
inexistent FQ.
While it is guaranteed not to have more than one queue per core,
having fewer queues than CPUs on an interface is a valid
configuration.
Fixes: d678be1dc1 ("dpaa2-eth: add XDP_REDIRECT support")
Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull ARM SoC driver updates from Arnd Bergmann:
"As usual, the drivers/tee and drivers/reset subsystems get merged
here, with the expected set of smaller updates and some new hardware
support. The tee subsystem now supports device drivers to be attached
to a tee, the first example here is a random number driver with its
implementation in the secure world.
Three new power domain drivers get added for specific chip families:
- Broadcom BCM283x chips (used in Raspberry Pi)
- Qualcomm Snapdragon phone chips
- Xilinx ZynqMP FPGA SoCs
One new driver is added to talk to the BPMP firmware on NVIDIA
Tegra210
Existing drivers are extended for new SoC variants from NXP, NVIDIA,
Amlogic and Qualcomm"
* tag 'armsoc-drivers' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (113 commits)
tee: optee: update optee_msg.h and optee_smc.h to dual license
tee: add cancellation support to client interface
dpaa2-eth: configure the cache stashing amount on a queue
soc: fsl: dpio: configure cache stashing destination
soc: fsl: dpio: enable frame data cache stashing per software portal
soc: fsl: guts: make fsl_guts_get_svr() static
hwrng: make symbol 'optee_rng_id_table' static
tee: optee: Fix unsigned comparison with less than zero
hwrng: Fix unsigned comparison with less than zero
tee: fix possible error pointer ctx dereferencing
hwrng: optee: Initialize some structs using memset instead of braces
tee: optee: Initialize some structs using memset instead of braces
soc: fsl: dpio: fix memory leak of a struct qbman on error exit path
clk: tegra: dfll: Make symbol 'tegra210_cpu_cvb_tables' static
soc: qcom: llcc-slice: Fix typos
qcom: soc: llcc-slice: Consolidate some code
qcom: soc: llcc-slice: Clear the global drv_data pointer on error
drivers: soc: xilinx: Add ZynqMP power domain driver
firmware: xilinx: Add APIs to control node status/power
dt-bindings: power: Add ZynqMP power domain bindings
...
Pull crypto update from Herbert Xu:
"API:
- Add helper for simple skcipher modes.
- Add helper to register multiple templates.
- Set CRYPTO_TFM_NEED_KEY when setkey fails.
- Require neither or both of export/import in shash.
- AEAD decryption test vectors are now generated from encryption
ones.
- New option CONFIG_CRYPTO_MANAGER_EXTRA_TESTS that includes random
fuzzing.
Algorithms:
- Conversions to skcipher and helper for many templates.
- Add more test vectors for nhpoly1305 and adiantum.
Drivers:
- Add crypto4xx prng support.
- Add xcbc/cmac/ecb support in caam.
- Add AES support for Exynos5433 in s5p.
- Remove sha384/sha512 from artpec7 as hardware cannot do partial
hash"
[ There is a merge of the Freescale SoC tree in order to pull in changes
required by patches to the caam/qi2 driver. ]
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (174 commits)
crypto: s5p - add AES support for Exynos5433
dt-bindings: crypto: document Exynos5433 SlimSSS
crypto: crypto4xx - add missing of_node_put after of_device_is_available
crypto: cavium/zip - fix collision with generic cra_driver_name
crypto: af_alg - use struct_size() in sock_kfree_s()
crypto: caam - remove redundant likely/unlikely annotation
crypto: s5p - update iv after AES-CBC op end
crypto: x86/poly1305 - Clear key material from stack in SSE2 variant
crypto: caam - generate hash keys in-place
crypto: caam - fix DMA mapping xcbc key twice
crypto: caam - fix hash context DMA unmap size
hwrng: bcm2835 - fix probe as platform device
crypto: s5p-sss - Use AES_BLOCK_SIZE define instead of number
crypto: stm32 - drop pointless static qualifier in stm32_hash_remove()
crypto: chelsio - Fixed Traffic Stall
crypto: marvell - Remove set but not used variable 'ivsize'
crypto: ccp - Update driver messages to remove some confusion
crypto: adiantum - add 1536 and 4096-byte test vectors
crypto: nhpoly1305 - add a test vector with len % 16 != 0
crypto: arm/aes-ce - update IV after partial final CTR block
...
Implement support for the XDP_REDIRECT action.
The redirected frame is transmitted and confirmed on the regular Tx/Tx
conf queues. Frame is marked with the "XDP" type in the software
annotation, since it requires special treatment.
We don't have good hardware support for TX batching, so the
XDP_XMIT_FLUSH flag doesn't make a difference for now; ndo_xdp_xmit
performs the actual Tx operation on the spot.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We write different metadata information in the software annotation
area of Tx frames, depending on frame type. Make this more explicit
by introducing a type field and separate structures for single buffer
and scatter-gather frames.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Configure the amount of 64 bytes of frame, annotation and context data
that will be cache stashed for a specific frame queue. Since the frame
context is not used, configure that only 64 bytes of frame data and 64
bytes of annotation will be stashed.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Li Yang <leoyang.li@nxp.com>
Starting with MC10.14.0, dpaa2_io_service_enqueue_fq() API is
functional. Since there are a number of cases where it offers
better performance compared to the currently used enqueue
function, switch to it for firmware versions that support it.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While in NAPI context, free skbs by calling napi_consume_skb()
instead of dev_kfree_skb(), to take advantage of the bulk freeing
mechanism.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of allocating page fragments via the network stack,
use the page allocator directly. For now, we consume one page
for each Rx buffer.
With the new memory model we are free to consider adding more
XDP support.
Performance decreases slightly in some IP forwarding cases.
No visible effect on termination traffic. The driver memory
footprint increases as a result of this change, but it is
still small enough to not really matter.
Another side effect is that now Rx buffer alignment requirements
are naturally satisfied without any additional actions needed.
Remove alignment related code, except in the buffer layout
information conveyed to MC, as hardware still needs to know the
alignment value we guarantee.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Export detailed driver counters through debugfs.
Statistics already available in ethtool are presented in a
structured manner. Includes per-core, per-FQ and per-channel statistics.
Also transition from module_fsl_mc_driver to explicit module_init/exit
in order to create the debugfs directory besides registering the driver.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the current implementation, on interface down we disabled NAPI and
then manually drained any remaining ingress frames. This could lead
to a situation when, under heavy traffic, the data availability
notification for some of the channels would not get rearmed correctly.
Change the implementation such that we let all remaining ingress frames
be processed as usual and only disable NAPI once the hardware queues
are empty.
We also add a wait on the Tx side, to allow hardware time to process
all in-flight Tx frames before issueing the disable command.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Automatically add a device link between the actual device requesting the
dpaa2_io_service_register and the underlying dpaa2_io used. This link
will ensure that when a DPIO device, which is indirectly used by other
devices, is unbound any consumer devices will be also unbound from their
drivers.
For example, any DPNI, bound to the dpaa2-eth driver, which is using
DPIO devices will be unbound before its supplier device.
Also, add a new parameter to the dpaa2_io_service_[de]register functions
to specify the requesting device (ie the consumer).
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Horia Geanta <horia.geanta@nxp.com>
Reviewed-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Li Yang <leoyang.li@nxp.com>
Add comments in the switch statement for XDP action to indicate
fallthrough is intended.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove unused counter. Reorder fields in channel stats structure
to match the ethtool strings order and make it easier to print them
with ethtool -S.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Send frames back on the same port for XDP_TX action.
Since the frame buffers have been allocated by us, we can recycle
them directly into the Rx buffer pool instead of requesting a
confirmation frame upon transmission complete.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to support enqueueing Rx FDs back to hardware, we need to
DMA map Rx buffers as bidirectional.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of freeing the RX buffers, release them back into the pool.
We wait for the maximum number of buffers supported by a single
release command to accumulate before issuing the command.
Also, don't unmap the Rx buffers at the beginning of the Rx routine
anymore, since that would require remapping them before release.
Instead, just do a DMA sync at first and only unmap if the frame is
meant for the stack.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We'll use function free_bufs() on the XDP path as well, so move
it higher in order to avoid a forward declaration.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reserve XDP_PACKET_HEADROOM bytes in Rx buffers to allow XDP
programs to increase frame header size.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We keep one XDP program reference per channel. The only actions
supported for now are XDP_DROP and XDP_PASS.
Until now we didn't enforce a maximum size for Rx frames based
on MTU value. Change that, since for XDP mode we must ensure no
scatter-gather frames can be received.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for byte queue limit.
On NAPI poll, we save the total number of Tx confirmed frames/bytes
and register them with bql at the end of the poll function.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the frame consume callback signature:
* the entire FQ structure is passed to the callback instead
of just the queue index
* the NAPI structure can be easily obtained from the channel
it is associated to, so we don't need to pass it explicitly
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The DPNI object on which we build a network interface has a
certain number of {Rx, Tx, Tx confirmation} frame queues as
resources. The default hardware setup offers one queue of each
type, as well as one DPCON channel, for each core available
in the system.
There are however cases where the number of queues is greater
than the number of cores or channels. Until now, we configured
and used all the frame queues associated with a DPNI, even if it
meant assigning multiple queues of one type to the same channel.
Update the driver to only use a number of queues equal to the
number of channels, ensuring each channel will contain exactly
one Rx and one Tx confirmation queue.
>From the user viewpoint, this change is completely transparent.
Performance wise there is no impact in most scenarios. In case
the number of queues is larger than and not a multiple of the
number of channels, Rx hash distribution offers now better load
balancing between cores, which can have a positive impact on
overall system performance.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The fsl_mc_object_allocate function can fail because not all allocatable
objects are probed by the fsl_mc_allocator at the call time. Defer the
dpaa2-eth probe when this happens.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to the hardware ArchDef, the PTV1 field in FD[CTRL]
is ignored by WRIOP, so setting it for Tx FDs is pointless.
Remove all references to it from the code.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ch parameter is never used in the dpaa2_eth_tx_conf function but
since its prototype must match the type defined in the consume field of
struct dpaa2_eth_fq, just mark it as __always_unused.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The priv parameter is never used in the build_linear_skb and
drain_channel function. Remove it from the function definitions.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All 3 cases of possible uninitialized variables are false
positives since they are used only as output parameters.
Nonetheless, fix the warnings.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The dpaa2_eth_set_dist_key function is only used in a single file.
Make it static.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
